e5ca2112a65a874dfb394a0e8cc3f35d Environmental_Microbiology_-_2026_-_Romano_Gude_-_Spatial.pdf 6e1118ca73f471d0efae5b4dfaa3e63e80bfd7bc Environmental_Microbiology_-_2026_-_Romano_Gude_-_Spatial.pdf 6e17de77be6e8cf23ce4fc0e336b756003f446ceae141ed8359bb8e15c0f6fb0 Environmental_Microbiology_-_2026_-_Romano_Gude_-_Spatial.pdf Title: Spatial Partitioning of Microbial Functional Potential Suggests Complex Nitrogen Cycling in a Subterranean Estuary Subject: Environmental Microbiology 2026.28:e70400 Keywords: Keywords: functional potential; microbial communities; nitrogen cycling; PICRUSt2; qPCR; subterranean estuary Author: Daniel Romano-Gude, Nestor Arandia-Gorostidi, Catalina Mena, Marc Diego-Feliu, Bella Almillategui, Anna J. Székely, Stefan Bertilsson, Andrea G. Bravo, Jesus Carrera, Cristina Romera-Castillo, Albert Folch, Jordi Garcia-Orellana, Valentí Rodellas, Clara Ruiz-González Creator: Adobe InDesign 19.5 (Windows) Producer: Adobe PDF Library 17.0; modified using iTextSharp 5.5.13.4 ©2000-2024 iText Group NV (AGPL-version) CreationDate: Wed Aug 12 03:59:18 2026 CEST ModDate: Tue Sep 1 15:30:15 2026 CEST Custom Metadata: yes Metadata Stream: yes Tagged: yes UserProperties: no Suspects: no Form: none JavaScript: no Pages: 21 Encrypted: no Page size: 595.276 x 782.362 pts Page rot: 0 File size: 3002481 bytes Optimized: no PDF version: 1.7 name type encoding emb sub uni object ID ------------------------------------ ----------------- ---------------- --- --- --- --------- EHPHJF+STIXTwoText Type 1C Custom yes yes yes 72 0 SFMIXD+MyriadPro-Bold Type 1C WinAnsi yes yes no 3794 0 VZCANX+STIXTwoText-Italic Type 1C WinAnsi yes yes yes 74 0 AZLZFN+STIXTwoText-Bold Type 1C Custom yes yes yes 73 0 SFMIXD+STIXTwoText-BoldItalic Type 1C WinAnsi yes yes yes 239 0 UYMPZZ+STIX-Regular Type 1C Custom yes yes yes 221 0 VCWVNX+ArialRoundedMTBold Type 1C WinAnsi yes yes no 3795 0 Times-Roman Type 1 WinAnsi no no no 5 0 Jhove (Rel. 1.28.0, 2023-05-18) Date: 2026-09-05 03:56:37 CEST RepresentationInformation: Environmental_Microbiology_-_2026_-_Romano_Gude_-_Spatial.pdf ReportingModule: PDF-hul, Rel. 1.12.4 (2023-03-16) LastModified: 2026-09-04 07:01:44 CEST Size: 3002481 Format: PDF Version: 1.7 Status: Well-Formed and valid SignatureMatches: PDF-hul MIMEtype: application/pdf PDFMetadata: Objects: 0 FreeObjects: 2 IncrementalUpdates: 0 DocumentCatalog: ViewerPreferences: HideToolbar: false HideMenubar: false HideWindowUI: false FitWindow: false CenterWindow: false DisplayDocTitle: true NonFullScreenPageMode: UseNone Direction: L2R ViewArea: CropBox ViewClip: CropBox PrintArea: CropBox PageClip: CropBox PageLayout: SinglePage PageMode: UseNone Language: en-US Outlines: Item: Title: Spatial Partitioning of Microbial Functional Potential Suggests Complex Nitrogen Cycling in a Subterranean Estuary Children: Item: Title: ABSTRACT Item: Title: 1   |   Introduction Item: Title: 2   |   Material and Methods Children: Item: Title: 2.1   |   Study Area Item: Title: 2.2   |   Sampling and Physicochemical Parameters Item: Title: 2.3   |   Estimates of Deviations From Conservative Mixing of Nutrient Item: Title: 2.4   |   Prokaryotic Abundances and Heterotrophic Activity Item: Title: 2.5   |   DNA Extraction and 16S rRNA Gene Sequencing Item: Title: 2.6   |   Prediction of Genes Involved in the Nitrogen Cycle Item: Title: 2.7   |   Quantification of nirK, nirS and amoA Genes Through Quantitative Polymerase Chain Reaction (qPCR) Item: Title: 2.8   |   Statistical Analyses Item: Title: 3   |   Results Children: Item: Title: 3.1   |   Physicochemical Variability Along the STE Item: Title: 3.2   |   Spatio-Temporal Variations in Prokaryotic Abundance and Heterotrophic Activity Item: Title: 3.3   |   Microbial Community Structure and Taxonomic Composition Item: Title: 3.4   |   Functional Prediction of Nitrogen Cycling Potential Item: Title: 3.5   |   Distribution of Ammonia Oxidizers and Denitrifiers Using qPCR Item: Title: 3.6   |   Links Between Microbial Functional Potential and Groundwater Physico-Chemistry Item: Title: 4   |   Discussion Children: Item: Title: 4.1   |   Large Spatial Heterogeneity and Seasonal Stability of STE Microbial Communities Driven by Physicochemical Gradients Item: Title: 4.2   |   Spatial Segregation of Microbial Functional Groups Suggests Complex Regulation of the Aquifer N Cycling Item: Title: 4.3   |   Microbial Information Improves Understanding of Nitrogen Concentrations in the STE Item: Title: 5   |   Conclusions Item: Title: Author Contributions Item: Title: Acknowledgements Item: Title: Funding Item: Title: Ethics Statement Item: Title: Conflicts of Interest Item: Title: Data Availability Statement Item: Title: References Filters: FilterPipeline: FlateDecode FilterPipeline: DCTDecode Images: Image: NisoImageMetadata: FormatName: image/png ImageWidth: 228 ImageHeight: 45 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 265 ImageHeight: 79 ColorSpace: CMYK BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 228 ImageHeight: 45 ColorSpace: black is zero BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Image: NisoImageMetadata: FormatName: image/png ImageWidth: 293 ImageHeight: 137 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 95 ImageHeight: 72 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 109 ImageHeight: 239 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 19 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 39 ImageHeight: 35 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 109 ImageHeight: 104 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 45 ImageHeight: 39 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 30 ImageHeight: 8 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 26 ImageHeight: 7 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 43 ImageHeight: 37 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 48 ImageHeight: 35 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 51 ImageHeight: 34 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 346 ImageHeight: 319 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 346 ImageHeight: 319 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 51 ImageHeight: 34 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 59 ImageHeight: 319 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 346 ImageHeight: 266 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 922 ImageHeight: 11 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 347 ImageHeight: 319 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 461 ImageHeight: 266 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 175 ImageHeight: 91 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 31 ImageHeight: 9 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 198 ImageHeight: 97 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 48 ImageHeight: 35 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 51 ImageHeight: 34 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 4 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 51 ImageHeight: 34 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 42 ImageHeight: 36 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 30 ImageHeight: 8 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 40 ImageHeight: 35 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 26 ImageHeight: 7 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 42 ImageHeight: 38 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 32 ImageHeight: 9 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 39 ImageHeight: 35 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 27 ImageHeight: 8 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 40 ImageHeight: 36 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/jpg CompressionScheme: JPEG ImageWidth: 28 ImageHeight: 8 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 275 ImageHeight: 159 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 286 ImageHeight: 167 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 59 ImageHeight: 70 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 3 ImageHeight: 3 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 42 ImageHeight: 37 ColorSpace: palette color BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Decode: 0, 255 Name: X Image: NisoImageMetadata: FormatName: image/png ImageWidth: 998 ImageHeight: 622 ColorSpace: RGB BitsPerSample: 8 BitsPerSampleUnit: integer Filter: FlateDecode Intent: RelativeColorimetric Name: X Fonts: Type1: Font: BaseFont: SFMIXD+MyriadPro-Bold FontSubset: true FirstChar: 32 LastChar: 89 FontDescriptor: FontName: SFMIXD+MyriadPro-Bold Flags: Nonsymbolic FontBBox: -163, -250, 1256, 989 FontFile3: true Encoding: WinAnsiEncoding Font: BaseFont: VCWVNX+ArialRoundedMTBold FontSubset: true FirstChar: 32 LastChar: 83 FontDescriptor: FontName: VCWVNX+ArialRoundedMTBold Flags: Nonsymbolic FontBBox: -174, -211, 1185, 946 FontFile3: true Encoding: WinAnsiEncoding Font: BaseFont: EHPHJF+STIXTwoText FontSubset: true FirstChar: 29 LastChar: 252 FontDescriptor: FontName: EHPHJF+STIXTwoText Flags: Serif, Nonsymbolic FontBBox: -948, -373, 1297, 1047 FontFile3: true EncodingDictionary: BaseEncoding: WinAnsiEncoding Differences: true ToUnicode: true Font: BaseFont: AZLZFN+STIXTwoText-Bold FontSubset: true FirstChar: 24 LastChar: 237 FontDescriptor: FontName: AZLZFN+STIXTwoText-Bold Flags: Serif, Nonsymbolic FontBBox: -948, -434, 1429, 1047 FontFile3: true EncodingDictionary: BaseEncoding: WinAnsiEncoding Differences: true ToUnicode: true Font: BaseFont: VZCANX+STIXTwoText-Italic FontSubset: true FirstChar: 32 LastChar: 232 FontDescriptor: FontName: VZCANX+STIXTwoText-Italic Flags: Serif, Nonsymbolic, Italic FontBBox: -893, -362, 1298, 1047 FontFile3: true Encoding: WinAnsiEncoding ToUnicode: true Font: BaseFont: UYMPZZ+STIX-Regular FontSubset: true FirstChar: 31 LastChar: 31 FontDescriptor: FontName: UYMPZZ+STIX-Regular Flags: Serif, Nonsymbolic FontBBox: -970, -486, 2208, 1023 FontFile3: true EncodingDictionary: BaseEncoding: WinAnsiEncoding Differences: true ToUnicode: true Font: BaseFont: SFMIXD+STIXTwoText-BoldItalic FontSubset: true FirstChar: 32 LastChar: 121 FontDescriptor: FontName: SFMIXD+STIXTwoText-BoldItalic Flags: Serif, Nonsymbolic, Italic FontBBox: -893, -427, 1451, 1047 FontFile3: true Encoding: WinAnsiEncoding ToUnicode: true XMP: application/pdfdoi:10.1111/1462-2920.70400 10.1111/1462-2920.70400https://doi.org/10.1111/1462-2920.70400 2026-08-12trueenviromicro-journals.onlinelibrary.wiley.com10.1111/1462-2920.70400 enviromicro-journals.onlinelibrary.wiley.comtrue10.1111/1462-2920.704002026-08-12 VoR Pages: Page: Sequence: 1 Annotations: Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.70400 for your more information Rect: 45, 25, 173, 14 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.70400 for your more information Rect: 162, 676, 217, 661 Annotation: Subtype: Link Contents: Institut de Ciències del Mar (ICMCSIC)BarcelonaSpain Rect: 129, 565, 137, 552 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0003-2630-9316 for your more information Rect: 137, 566, 149, 554 Annotation: Subtype: Link Contents: Institut de Ciències del Mar (ICMCSIC)BarcelonaSpain Rect: 259, 565, 262, 552 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-9370-2065 for your more information Rect: 262, 566, 275, 554 Annotation: Subtype: Link Contents: Centre Oceanogràfic de BalearsInstituto Español de Oceanografía (IEOCSIC)PalmaSpain Rect: 343, 565, 346, 552 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-9001-5163 for your more information Rect: 346, 566, 359, 554 Annotation: Subtype: Link Contents: Department of Civil and Environmental EngineeringUniversitat Politècnica de CatalunyaBarcelonaSpain Rect: 437, 565, 445, 552 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-4782-3210 for your more information Rect: 445, 566, 458, 554 Annotation: Subtype: Link Contents: Department of Civil and Environmental EngineeringUniversitat Politècnica de CatalunyaBarcelonaSpain Rect: 116, 554, 124, 541 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-0212-9006 for your more information Rect: 124, 555, 136, 543 Annotation: Subtype: Link Contents: Department of Aquatic Sciences and Assessment and Science for Life LaboratorySwedish University of Agricultural SciencesUppsalaSweden Rect: 208, 554, 212, 541 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0001-8063-7156 for your more information Rect: 212, 555, 224, 543 Annotation: Subtype: Link Contents: Department of Aquatic Sciences and Assessment and Science for Life LaboratorySwedish University of Agricultural SciencesUppsalaSweden Rect: 300, 554, 304, 541 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-4265-1835 for your more information Rect: 304, 555, 316, 543 Annotation: Subtype: Link Contents: Institut de Ciències del Mar (ICMCSIC)BarcelonaSpain Rect: 391, 554, 395, 541 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-8341-3462 for your more information Rect: 395, 555, 407, 543 Annotation: Subtype: Link Contents: Associated UnitGroundwater Hydrology Group (GHSUPCCSIC)BarcelonaSpain Rect: 471, 554, 478, 541 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-8054-4352 for your more information Rect: 478, 555, 491, 543 Annotation: Subtype: Link Contents: Institut de Ciències del Mar (ICMCSIC)BarcelonaSpain Rect: 141, 543, 144, 530 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-4888-1340 for your more information Rect: 144, 544, 156, 532 Annotation: Subtype: Link Contents: Department of Civil and Environmental EngineeringUniversitat Politècnica de CatalunyaBarcelonaSpain Rect: 216, 543, 224, 530 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-8490-1038 for your more information Rect: 224, 544, 236, 532 Annotation: Subtype: Link Contents: Institut de Ciència i Tecnologia Ambientals (ICTAUAB)Universitat Autònoma de BarcelonaBellaterraBarcelonaSpain Rect: 331, 543, 334, 530 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-0543-2641 for your more information Rect: 334, 544, 347, 532 Annotation: Subtype: Link Contents: Institut de Ciència i Tecnologia Ambientals (ICTAUAB)Universitat Autònoma de BarcelonaBellaterraBarcelonaSpain Rect: 421, 543, 424, 530 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0002-5896-9987 for your more information Rect: 424, 544, 437, 532 Annotation: Subtype: Link Contents: Institut de Ciències del Mar (ICMCSIC)BarcelonaSpain Rect: 528, 543, 531, 530 ActionDest: 643 Annotation: Subtype: Link Contents: Please visit https://orcid.org/0000-0003-3568-4943 for your more information Rect: 531, 544, 544, 532 Annotation: Subtype: Link Contents: Please visit mailto:romano@icm.csic.es for your more information Rect: 182, 444, 248, 433 Annotation: Subtype: Link Contents: Please visit mailto:clararg@icm.csic.es for your more information Rect: 332, 444, 396, 433 Annotation: Subtype: Link Contents: Duque, C., H. A.Michael, and A. M.Wilson. 2020. “The Subterranean Estuary: Technical Term, Simple Analogy, or Source of Confusion?” Water Resources Research56, no. 2: e2019WR026554. https://doi.org/10.1029/2019WR026554. Rect: 214, 124, 233, 111 ActionDest: 465 Annotation: Subtype: Link Contents: Moore, W. S.1999. “The Subterranean Estuary: A Reaction Zone of Ground Water and Sea Water.” Marine Chemistry65, no. 1 2: 111 125. https://doi.org/10.1016/S03044203(99)000146. Rect: 266, 124, 283, 111 ActionDest: 510 Annotation: Subtype: Link Contents: Moore, W. S., and S. B.Joye. 2021. “Saltwater Intrusion and Submarine Groundwater Discharge: Acceleration of Biogeochemical Reactions in Changing Coastal Aquifers.” Frontiers in Earth Science9: 600710. https://doi.org/10.3389/feart.2021.600710. Rect: 428, 158, 446, 146 ActionDest: 510 Annotation: Subtype: Link Contents: Wilson, S. J., A.Moody, T.McKenzie, etal. 2024. “Global Subterranean Estuaries Modify Groundwater Nutrient Loading to the Ocean.” Limnology and Oceanography Letters9, no. 4: 411 422. https://doi.org/10.1002/lol2.10390. Rect: 503, 158, 521, 146 ActionDest: 603 Annotation: Subtype: Link Contents: Lecher, A. L., K.Mackey, R.Kudela, etal. 2015. “Nutrient Loading Through Submarine Groundwater Discharge and Phytoplankton Growth in Monterey Bay, CA.” Environmental Science & Technology49, no. 11: 6665 6673. https://doi.org/10.1021/acs.est.5b00909. Rect: 529, 135, 547, 123 ActionDest: 510 Annotation: Subtype: Link Contents: Lecher, A. L., and K. R. M.Mackey. 2018. “Synthesizing the Effects of Submarine Groundwater Discharge on Marine Biota.” Hydrology5, no. 4: 60. https://doi.org/10.3390/hydrology5040060. Rect: 387, 124, 404, 111 ActionDest: 510 Annotation: Subtype: Link Contents: Rodellas, V., J.GarciaOrellana, P.Masqué, M.Feldman, and Y.Weinstein. 2015. “Submarine Groundwater Discharge as a Major Source of Nutrients to the Mediterranean Sea.” Proceedings of the National Academy of Sciences112, no. 13: 3926 3930. https://doi.org/10.1073/pnas.1419049112. Rect: 470, 124, 488, 111 ActionDest: 559 Annotation: Subtype: Link Contents: RomanoGude, D., J.DordalSoriano, J.RodriguezPuig, etal. 2025. “Impact of Submarine Groundwater Discharge on Microbial Communities in a Highly Anthropized Coastal Lagoon (Mar Menor, SE Spain).” Molecular Ecology34, no. 24: e70201. https://doi.org/10.1111/mec.70201. Rect: 328, 112, 346, 100 ActionDest: 559 Annotation: Subtype: Link Contents: Santos, I. R., X.Chen, A. L.Lecher, etal. 2021. “Submarine Groundwater Discharge Impacts on Coastal Nutrient Biogeochemistry.” Nature Reviews Earth & Environment2, no. 5: 307 323. https://doi.org/10.1038/s43017021001520. Rect: 403, 112, 420, 100 ActionDest: 603 Annotation: Subtype: Link Contents: Please visit http://creativecommons.org/licenses/by-nc-nd/4.0/ for your more information Rect: 202, 78, 380, 68 Annotation: Subtype: Link Rect: 519, 782, 595, 767 Page: Sequence: 2 Annotations: Annotation: Subtype: Link Contents: Hwang, D.W., Y.W.Lee, and G.Kim. 2005. “Large Submarine Groundwater Discharge and Benthic Eutrophication in Bangdu Bay on Volcanic Jeju Island, Korea.” Limnology and Oceanography50, no. 5: 1393 1403. https://doi.org/10.4319/lo.2005.50.5.1393. Rect: 265, 751, 283, 739 ActionDest: 510 Annotation: Subtype: Link Contents: Montiel, D., A.Lamore, J.Stewart, and N.Dimova. 2019. “Is Submarine Groundwater Discharge (SGD) Important for the Historical Fish Kills and Harmful Algal Bloom Events of Mobile Bay?” Estuaries and Coasts42, no. 2: 470 493. https://doi.org/10.1007/s1223701804855. Rect: 79, 728, 96, 716 ActionDest: 510 Annotation: Subtype: Link Contents: AlordaKleinglass, A., V.Rodellas, M.DiegoFeliu, N.Marbà, C.Morell, and J.GarciaOrellana. 2024. “The Connection Between Submarine Groundwater Discharge and Seawater Quality: The Threat of Treated Wastewater Injected Into Coastal Aquifers.” Science of the Total Environment922: 170940. https://doi.org/10.1016/j.scitotenv.2024.170940. Rect: 169, 717, 188, 704 ActionDest: 465 Annotation: Subtype: Link Contents: Amato, D. W., J. M.Bishop, C. R.Glenn, H.Dulai, and C. M.Smith. 2016. “Impact of Submarine Groundwater Discharge on Marine Water Quality and Reef Biota of Maui.” PLoS One11, no. 11: e0165825. https://doi.org/10.1371/journal.pone.0165825. Rect: 241, 717, 259, 704 ActionDest: 465 Annotation: Subtype: Link Contents: Archana, A., C. A.Francis, and A. B.Boehm. 2021. “The Beach Aquifer Microbiome: Research Gaps and Data Needs.” Frontiers in Environmental Science9: 653568. https://doi.org/10.3389/fenvs.2021.653568. Rect: 268, 671, 286, 658 ActionDest: 465 Annotation: Subtype: Link Contents: RuizGonzález, C., V.Rodellas, and J.GarciaOrellana. 2021. “The Microbial Dimension of Submarine Groundwater Discharge: Current Challenges and Future Directions.” FEMS Microbiology Reviews45, no. 5: fuab010. https://doi.org/10.1093/femsre/fuab010. Rect: 132, 659, 149, 647 ActionDest: 559 Annotation: Subtype: Link Contents: Abarike, G. A., S.Brick, B.Engelen, and J.Niggemann. 2024. “Different Dissolved Organic Matter Sources Sustain Microbial Life in a Sandy Beach Subterranean Estuary   An Incubation Study.” Frontiers in Marine Science11: 1501781. https://doi.org/10.3389/fmars.2024.1501781. Rect: 70, 613, 88, 601 ActionDest: 465 Annotation: Subtype: Link Contents: CalvoMartin, E., X. A.ÁlvarezSalgado, M. J.Pazó Fernández, V.Vieitez Dos Santos, and J. S. P.Ibánhez. 2025. “Drivers of Dissolved Organic Matter Processing in Subterranean Estuaries.” Limnology and Oceanography Letters10: 536 546. https://doi.org/10.1002/lol2.70026. Rect: 176, 613, 194, 601 ActionDest: 465 Annotation: Subtype: Link Contents: Archana, A., C. A.Francis, and A. B.Boehm. 2021. “The Beach Aquifer Microbiome: Research Gaps and Data Needs.” Frontiers in Environmental Science9: 653568. https://doi.org/10.3389/fenvs.2021.653568. Rect: 66, 544, 84, 532 ActionDest: 465 Annotation: Subtype: Link Contents: RuizGonzález, C., V.Rodellas, and J.GarciaOrellana. 2021. “The Microbial Dimension of Submarine Groundwater Discharge: Current Challenges and Future Directions.” FEMS Microbiology Reviews45, no. 5: fuab010. https://doi.org/10.1093/femsre/fuab010. Rect: 168, 544, 186, 532 ActionDest: 559 Annotation: Subtype: Link Contents: Couturier, M., G.TommiMorin, M.Sirois, A.Rao, C.Nozais, and G.Chaillou. 2017. “Nitrogen Transformations Along a Shallow Subterranean Estuary.” Biogeosciences14, no. 13: 3321 3336. https://doi.org/10.5194/bg1433212017. Rect: 220, 487, 237, 474 ActionDest: 465 Annotation: Subtype: Link Contents: Lecher, A. L., and K. R. M.Mackey. 2018. “Synthesizing the Effects of Submarine Groundwater Discharge on Marine Biota.” Hydrology5, no. 4: 60. https://doi.org/10.3390/hydrology5040060. Rect: 78, 475, 96, 463 ActionDest: 510 Annotation: Subtype: Link Contents: Santoro, A. E.2010. “Microbial Nitrogen Cycling at the Saltwater Freshwater Interface.” Hydrogeology Journal18, no. 1: 187 202. https://doi.org/10.1007/s100400090526z. Rect: 134, 475, 151, 463 ActionDest: 559 Annotation: Subtype: Link Contents: Wilson, S. J., A.Moody, T.McKenzie, etal. 2024. “Global Subterranean Estuaries Modify Groundwater Nutrient Loading to the Ocean.” Limnology and Oceanography Letters9, no. 4: 411 422. https://doi.org/10.1002/lol2.10390. Rect: 209, 475, 227, 463 ActionDest: 603 Annotation: Subtype: Link Contents: Galloway, J. N., A. M.Leach, A.Bleeker, and J. W.Erisman. 2013. “A Chronology of Human Understanding of the Nitrogen Cycle†.” Philosophical Transactions of the Royal Society, B: Biological Sciences368, no. 1621: 20130120. https://doi.org/10.1098/rstb.2013.0120. Rect: 123, 429, 140, 417 ActionDest: 465 Annotation: Subtype: Link Contents: Gao, H., G.Wang, Y.Fan, etal. 2024. “Tracing Groundwater Nitrate Sources in an Intensive Agricultural Region Integrated of a SelfOrganizing Map and EndMember Mixing Model Tool.” Scientific Reports14, no. 1: 16873. https://doi.org/10.1038/s4159802467735x. Rect: 186, 429, 204, 417 ActionDest: 465 Annotation: Subtype: Link Contents: Morrissy, J. G., M. J.Currell, S. M.Reichman, etal. 2021. “Nitrogen Contamination and Bioremediation in Groundwater and the Environment: A Review.” EarthScience Reviews222: 103816. https://doi.org/10.1016/j.earscirev.2021.103816. Rect: 268, 429, 286, 417 ActionDest: 510 Annotation: Subtype: Link Contents: Richardson, C. M., B.PeuckerEhrenbrink, S.Wyatt, etal. 2025. “Effects of Climate Change on River and Groundwater Nutrient Inputs to the Coastal Ocean.” Communications Earth & Environment6, no. 1: 761. https://doi.org/10.1038/s43247025025946. Rect: 112, 418, 130, 405 ActionDest: 559 Annotation: Subtype: Link Contents: Santoro, A. E.2010. “Microbial Nitrogen Cycling at the Saltwater Freshwater Interface.” Hydrogeology Journal18, no. 1: 187 202. https://doi.org/10.1007/s100400090526z. Rect: 104, 383, 122, 371 ActionDest: 559 Annotation: Subtype: Link Contents: Santos, I. R., X.Chen, A. L.Lecher, etal. 2021. “Submarine Groundwater Discharge Impacts on Coastal Nutrient Biogeochemistry.” Nature Reviews Earth & Environment2, no. 5: 307 323. https://doi.org/10.1038/s43017021001520. Rect: 172, 383, 190, 371 ActionDest: 603 Annotation: Subtype: Link Contents: Euler, S., L. C.Jeffrey, D. T.Maher, S. G.Johnston, R.Sugimoto, and D. R.Tait. 2024. “Microbiome Mediating Methane and Nitrogen Transformations in a Subterranean Estuary.” Environmental Microbiology26, no. 1: e16558. https://doi.org/10.1111/14622920.16558. Rect: 148, 349, 166, 336 ActionDest: 465 Annotation: Subtype: Link Contents: Chen, X., I.Santos, J.Du, etal. 2025. “Sandy Subterranean Estuaries Minimize Groundwater Nitrogen Pollution Impacts on Coastal Waters.” Geophysical Research Letters52, no. 3: e2024GL109621. https://doi.org/10.1029/2024GL109621. Rect: 265, 326, 283, 313 ActionDest: 465 Annotation: Subtype: Link Contents: Robinson, C. E., P.Xin, I. R.Santos, M. A.Charette, L.Li, and D. A.Barry. 2018. “Groundwater Dynamics in Subterranean Estuaries of Coastal Unconfined Aquifers: Controls on Submarine Groundwater Discharge and Chemical Inputs to the Ocean.” Advances in Water Resources115: 315 331. https://doi.org/10.1016/j.advwatres.2017.10.041. Rect: 108, 291, 126, 279 ActionDest: 559 Annotation: Subtype: Link Contents: Santos, I. R., W. C.Burnett, T.Dittmar, I. G. N. A.Suryaputra, and J.Chanton. 2009. “Tidal Pumping Drives Nutrient and Dissolved Organic Matter Dynamics in a Gulf of Mexico Subterranean Estuary.” Geochimica et Cosmochimica Acta73, no. 5: 1325 1339. https://doi.org/10.1016/j.gca.2008.11.029. Rect: 179, 291, 198, 279 ActionDest: 603 Annotation: Subtype: Link Contents: Brick, S., J.Niggemann, A.Reckhardt, M.Könneke, and B.Engelen. 2025. “Interstitial Microbial Communities of Coastal Sediments Are Dominated by Nanoarchaeota.” Frontiers in Microbiology16: 1532193. https://doi.org/10.3389/fmicb.2025.1532193. Rect: 244, 245, 262, 233 ActionDest: 465 Annotation: Subtype: Link Contents: RuizGonzález, C., C.Mena, F. M.CornejoCastillo, D.RomanoGude, N.ArandiaGorostidi, and J. M.Gasol. 2025. “Diverse Patescibacteria Assemblages and Prevalence of UltraSmall FreeLiving Parcubacteria Along a Subterranean Estuary.” mSystems10, no. 11: e0112525. https://doi.org/10.1128/msystems.0112525. Rect: 104, 234, 122, 221 ActionDest: 559 Annotation: Subtype: Link Contents: Mosley, O. E., E.Gios, M.Close, L.Weaver, C.Daughney, and K. M.Handley. 2022. “Nitrogen Cycling and Microbial Cooperation in the Terrestrial Subsurface.” ISME Journal16, no. 11: 2561 2573. https://doi.org/10.1038/s41396022013000. Rect: 68, 199, 86, 187 ActionDest: 510 Annotation: Subtype: Link Contents: RuizGonzález, C., V.Rodellas, and J.GarciaOrellana. 2021. “The Microbial Dimension of Submarine Groundwater Discharge: Current Challenges and Future Directions.” FEMS Microbiology Reviews45, no. 5: fuab010. https://doi.org/10.1093/femsre/fuab010. Rect: 181, 176, 199, 164 ActionDest: 559 Annotation: Subtype: Link Contents: RuizGonzález, C., C.Mena, F. M.CornejoCastillo, D.RomanoGude, N.ArandiaGorostidi, and J. M.Gasol. 2025. “Diverse Patescibacteria Assemblages and Prevalence of UltraSmall FreeLiving Parcubacteria Along a Subterranean Estuary.” mSystems10, no. 11: e0112525. https://doi.org/10.1128/msystems.0112525. Rect: 203, 176, 221, 164 ActionDest: 559 Annotation: Subtype: Link Contents: Santoro, A. E., C. A.Francis, N. R.De Sieyes, and A. B.Boehm. 2008. “Shifts in the Relative Abundance of AmmoniaOxidizing Bacteria and Archaea Across Physicochemical Gradients in a Subterranean Estuary.” Environmental Microbiology10, no. 4: 1068 1079. https://doi.org/10.1111/j.14622920.2007.01547.x. Rect: 69, 119, 88, 106 ActionDest: 559 Annotation: Subtype: Link Contents: Wilson, S. J., A.Moody, T.McKenzie, etal. 2024. “Global Subterranean Estuaries Modify Groundwater Nutrient Loading to the Ocean.” Limnology and Oceanography Letters9, no. 4: 411 422. https://doi.org/10.1002/lol2.10390. Rect: 148, 119, 167, 106 ActionDest: 603 Annotation: Subtype: Link Contents: Adyasari, D., N. T.Dimova, S. M.Ní Chadhain, and H.Waska. 2024. “Microbial Assemblages and Metabolic Activity in OrganicRich Subterranean Estuaries: Impact of Climate and Land Use Changes.” Journal of Geophysical Research: Biogeosciences129, no. 3: e2023JG007660. https://doi.org/10.1029/2023JG007660. Rect: 149, 96, 167, 83 ActionDest: 465 Annotation: Subtype: Link Contents: Jiang, Z., L.Qian, M.Cui, etal. 2023. “Bacterial Sulfate Reduction Facilitates Iodine Mobilization in the Deep Confined Aquifer of the North China Plain.” Environmental Science & Technology57, no. 40: 15277 15287. https://doi.org/10.1021/acs.est.3c05513. Rect: 220, 96, 239, 83 ActionDest: 510 Annotation: Subtype: Link Contents: Rogers, D. R., and K. L.Casciotti. 2010. “Abundance and Diversity of Archaeal Ammonia Oxidizers in a Coastal Groundwater System.” Applied and Environmental Microbiology76, no. 24: 7938 7948. https://doi.org/10.1128/AEM.0205609. Rect: 82, 84, 100, 72 ActionDest: 559 Annotation: Subtype: Link Contents: Santoro, A. E., A. B.Boehm, and C. A.Francis. 2006. “Denitrifier Community Composition Along a Nitrate and Salinity Gradient in a Coastal Aquifer.” Applied and Environmental Microbiology72, no. 3: 2102 2109. https://doi.org/10.1128/AEM.72.3.21022109.2006. Rect: 161, 84, 180, 72 ActionDest: 559 Annotation: Subtype: Link Contents: Santoro, A. E., C. A.Francis, N. R.De Sieyes, and A. B.Boehm. 2008. “Shifts in the Relative Abundance of AmmoniaOxidizing Bacteria and Archaea Across Physicochemical Gradients in a Subterranean Estuary.” Environmental Microbiology10, no. 4: 1068 1079. https://doi.org/10.1111/j.14622920.2007.01547.x. Rect: 185, 84, 204, 72 ActionDest: 559 Annotation: Subtype: Link Contents: Sun, D., J. S. P.Ibánhez, J.Zhang, G.Zhang, and S.Jiang. 2024. “Phosphorus Cycling in a Subterranean Estuary Seepage Face: A Seasonal View on Inventory Composition and Linkage With Nitrate Transformations.” Science of the Total Environment924: 171473. https://doi.org/10.1016/j.scitotenv.2024.171473. Rect: 250, 84, 268, 72 ActionDest: 603 Annotation: Subtype: Link Contents: Sun, W., L.Jiao, J.Wu, M.Wei, and Y.Hong. 2022. “Distribution, Diversity, and Abundance of Nitrite Oxidizing Bacteria in the Subterranean Estuary of the Daya Bay.” Frontiers in Marine Science9: 822939. https://doi.org/10.3389/fmars.2022.822939. Rect: 66, 73, 85, 60 ActionDest: 603 Annotation: Subtype: Link Contents: Wilson, S. J., B.Song, and I. C.Anderson. 2023. “Geochemical Factors Impacting Nitrifying Communities in Sandy Sediments.” Environmental Microbiology25, no. 12: 3180 3191. https://doi.org/10.1111/14622920.16504. Rect: 141, 73, 159, 60 ActionDest: 603 Annotation: Subtype: Link Contents: Wu, Y., Y.Wei, J. S. P.Ibánhez, etal. 2023. “Microbiota of a Seepage Face at the Mouth of a Subterranean Estuary: Diversity, Distribution and Substrate Dependence.” Acta Oceanologica Sinica42, no. 8: 147 157. https://doi.org/10.1007/s1313102322578. Rect: 201, 73, 219, 60 ActionDest: 603 Annotation: Subtype: Link Contents: Xiao, K., J.Wu, H.Li, etal. 2018. “Nitrogen Fate in a Subtropical Mangrove Swamp: Potential Association With SeawaterGroundwater Exchange.” Science of the Total Environment635: 586 597. https://doi.org/10.1016/j.scitotenv.2018.04.143. Rect: 266, 73, 283, 60 ActionDest: 603 Annotation: Subtype: Link Contents: Rogers, D. R., and K. L.Casciotti. 2010. “Abundance and Diversity of Archaeal Ammonia Oxidizers in a Coastal Groundwater System.” Applied and Environmental Microbiology76, no. 24: 7938 7948. https://doi.org/10.1128/AEM.0205609. Rect: 415, 694, 433, 681 ActionDest: 559 Annotation: Subtype: Link Contents: Santoro, A. E., C. A.Francis, N. R.De Sieyes, and A. B.Boehm. 2008. “Shifts in the Relative Abundance of AmmoniaOxidizing Bacteria and Archaea Across Physicochemical Gradients in a Subterranean Estuary.” Environmental Microbiology10, no. 4: 1068 1079. https://doi.org/10.1111/j.14622920.2007.01547.x. Rect: 497, 694, 516, 681 ActionDest: 559 Annotation: Subtype: Link Contents: Wilson, S. J., B.Song, and I. C.Anderson. 2023. “Geochemical Factors Impacting Nitrifying Communities in Sandy Sediments.” Environmental Microbiology25, no. 12: 3180 3191. https://doi.org/10.1111/14622920.16504. Rect: 326, 682, 344, 670 ActionDest: 603 Annotation: Subtype: Link Contents: Sun, W., L.Jiao, J.Wu, M.Wei, and Y.Hong. 2022. “Distribution, Diversity, and Abundance of Nitrite Oxidizing Bacteria in the Subterranean Estuary of the Daya Bay.” Frontiers in Marine Science9: 822939. https://doi.org/10.3389/fmars.2022.822939. Rect: 526, 659, 544, 647 ActionDest: 603 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 432, 613, 450, 601 ActionDest: 465 Annotation: Subtype: Link Contents: Massmann, G., G.Abarike, K.Amoako, etal. 2023. “The DynaDeep Observatory   A Unique Approach to Study HighEnergy Subterranean Estuaries.” Frontiers in Marine Science10: 1189281. https://doi.org/10.3389/fmars.2023.1189281. Rect: 526, 613, 544, 601 ActionDest: 510 Annotation: Subtype: Link Contents: Santoro, A. E., C. A.Francis, N. R.De Sieyes, and A. B.Boehm. 2008. “Shifts in the Relative Abundance of AmmoniaOxidizing Bacteria and Archaea Across Physicochemical Gradients in a Subterranean Estuary.” Environmental Microbiology10, no. 4: 1068 1079. https://doi.org/10.1111/j.14622920.2007.01547.x. Rect: 388, 590, 406, 578 ActionDest: 559 Annotation: Subtype: Link Contents: Sun, W., L.Jiao, J.Wu, M.Wei, and Y.Hong. 2022. “Distribution, Diversity, and Abundance of Nitrite Oxidizing Bacteria in the Subterranean Estuary of the Daya Bay.” Frontiers in Marine Science9: 822939. https://doi.org/10.3389/fmars.2022.822939. Rect: 451, 590, 469, 578 ActionDest: 603 Annotation: Subtype: Link Contents: Wilson, S. J., B.Song, and I. C.Anderson. 2023. “Geochemical Factors Impacting Nitrifying Communities in Sandy Sediments.” Environmental Microbiology25, no. 12: 3180 3191. https://doi.org/10.1111/14622920.16504. Rect: 526, 590, 544, 578 ActionDest: 603 Annotation: Subtype: Link Contents: DiegoFeliu, M., V.Rodellas, A.AlordaKleinglass, M.Saaltink, A.Folch, and J.GarciaOrellana. 2022. “Extreme Precipitation Events Induce High Fluxes of Groundwater and Associated Nutrients to Coastal Ocean.” Hydrology and Earth System Sciences26, no. 18: 4619 4635. https://doi.org/10.5194/hess2646192022. Rect: 465, 441, 484, 428 ActionDest: 465 Annotation: Subtype: Link Contents: MartínezPérez, L., L.Luquot, J.Carrera, etal. 2022. “A Multidisciplinary Approach to Characterizing Coastal Alluvial Aquifers to Improve Understanding of Seawater Intrusion and Submarine Groundwater Discharge.” Journal of Hydrology607: 127510. https://doi.org/10.1016/j.jhydrol.2022.127510. Rect: 330, 429, 348, 417 ActionDest: 510 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 529, 383, 547, 371 ActionDest: 465 Annotation: Subtype: Link Contents: GutiérrezMartín, D., R.GilSolsona, M. W.Saaltink, etal. 2023. “Chemicals of Emerging Concern in Coastal Aquifers: Assessment Along the LandOcean Interface.” Journal of Hazardous Materials448: 130876. https://doi.org/10.1016/j.jhazmat.2023.130876. Rect: 395, 372, 413, 359 ActionDest: 510 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 419, 130, 437, 118 ActionDest: 465 Annotation: Subtype: Link Contents: Palacios, A., J. J.Ledo, N.Linde, etal. 2020. “TimeLapse CrossHole Electrical Resistivity Tomography (CHERT) for Monitoring Seawater Intrusion Dynamics in a Mediterranean Aquifer.” Hydrology and Earth System Sciences24, no. 4: 2121 2139. https://doi.org/10.5194/hess2421212020. Rect: 444, 107, 462, 95 ActionDest: 559 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 331, 73, 350, 60 ActionDest: 465 Annotation: Subtype: Link Contents: MartínezPérez, L., L.Luquot, J.Carrera, etal. 2022. “A Multidisciplinary Approach to Characterizing Coastal Alluvial Aquifers to Improve Understanding of Seawater Intrusion and Submarine Groundwater Discharge.” Journal of Hydrology607: 127510. https://doi.org/10.1016/j.jhydrol.2022.127510. Rect: 446, 73, 464, 60 ActionDest: 510 Page: Sequence: 3 Annotations: Annotation: Subtype: Link Contents: DiegoFeliu, M., V.Rodellas, M. W.Saaltink, etal. 2021. “New Perspectives on the Use of 224Ra/228Ra and 222Rn/226Ra Activity Ratios in Groundwater Studies.” Journal of Hydrology596: 126043. https://doi.org/10.1016/j.jhydrol.2021.126043. Rect: 244, 728, 263, 716 ActionDest: 465 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 66, 717, 84, 704 ActionDest: 465 Annotation: Subtype: Link Contents: Palacios, A., J. J.Ledo, N.Linde, etal. 2020. “TimeLapse CrossHole Electrical Resistivity Tomography (CHERT) for Monitoring Seawater Intrusion Dynamics in a Mediterranean Aquifer.” Hydrology and Earth System Sciences24, no. 4: 2121 2139. https://doi.org/10.5194/hess2421212020. Rect: 143, 717, 161, 704 ActionDest: 559 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 197, 717, 206, 704 ActionDest: 220 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 239, 717, 247, 704 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 337, 717, 345, 704 ActionDest: 220 Page: Sequence: 4 Annotations: Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 142, 601, 151, 588 ActionDest: 220 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 121, 578, 140, 565 ActionDest: 465 Annotation: Subtype: Link Contents: GutiérrezMartín, D., R.GilSolsona, M. W.Saaltink, etal. 2023. “Chemicals of Emerging Concern in Coastal Aquifers: Assessment Along the LandOcean Interface.” Journal of Hazardous Materials448: 130876. https://doi.org/10.1016/j.jhazmat.2023.130876. Rect: 234, 578, 253, 565 ActionDest: 510 Annotation: Subtype: Link Contents: Palacios, A., J. J.Ledo, N.Linde, etal. 2020. “TimeLapse CrossHole Electrical Resistivity Tomography (CHERT) for Monitoring Seawater Intrusion Dynamics in a Mediterranean Aquifer.” Hydrology and Earth System Sciences24, no. 4: 2121 2139. https://doi.org/10.5194/hess2421212020. Rect: 66, 567, 85, 554 ActionDest: 559 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 122, 567, 130, 554 ActionDest: 220 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 76, 532, 85, 519 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 148, 463, 157, 450 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 163, 406, 172, 393 ActionDest: 220 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 119, 141, 127, 128 ActionDest: 220 Annotation: Subtype: Link Contents: RomeraCastillo, C., S.Birnstiel, X. A.ÁlvarezSalgado, and M.Sebastián. 2022. “Aged Plastic Leaching of Dissolved Organic Matter Is Two Orders of Magnitude Higher Than Virgin Plastic Leading to a Strong Uplift in Marine Microbial Activity.” Frontiers in Marine Science9: 861557. https://doi.org/10.3389/fmars.2022.861557. Rect: 424, 521, 442, 508 ActionDest: 559 Annotation: Subtype: Link Contents: Coble, P. G.1996. “Characterization of Marine and Terrestrial DOM in Seawater Using ExcitationEmission Matrix Spectroscopy.” Marine Chemistry51, no. 4: 325 346. https://doi.org/10.1016/03044203(95)000623. Rect: 485, 486, 503, 473 ActionDest: 465 Annotation: Subtype: Link Contents: Data S1: Supplementary Methods. Rect: 306, 268, 397, 255 ActionDest: 639 Annotation: Subtype: Link Contents: Pernice, M. C., and J. M.Gasol. 2023. “Automated Flow Cytometry as a Tool to Obtain a FineGrain Picture of Marine Prokaryote Community Structure Along an Entire Oceanographic Cruise.” Frontiers in Microbiology13: 1064112. https://doi.org/10.3389/fmicb.2022.1064112. Rect: 494, 164, 512, 151 ActionDest: 559 Annotation: Subtype: Link Contents: Kirchman, D., E.K'nees, and R.Hodson. 1985. “Leucine Incorporation and Its Potential as a Measure of Protein Synthesis by Bacteria in Natural Aquatic Systems.” Applied and Environmental Microbiology49, no. 3: 599 607. https://doi.org/10.1128/aem.49.3.599607.1985. Rect: 528, 61, 546, 48 ActionDest: 510 Page: Sequence: 5 Annotations: Annotation: Subtype: Link Contents: Smith, D. C., and F.Azam. 1992. “A Simple, Economical Method for Measuring Bacterial Protein Synthesis Rates in Seawater Using 3HLeucine.”https://www.semanticscholar.org/paper/Asimple%2CeconomicalmethodformeasuringbacterialSmithAzam/de4d1a5b9a27b6ed7f982976c60c85175a4e08b7. Rect: 212, 751, 230, 739 ActionDest: 603 Annotation: Subtype: Link Contents: Leizeaga, A., M.Estrany, I.Forn, and M.Sebastián. 2017. “Using ClickChemistry for Visualizing InSitu Changes of Translational Activity in Planktonic Marine Bacteria.” Frontiers in Microbiology8: 2360. https://doi.org/10.3389/fmicb.2017.02360. Rect: 183, 694, 201, 681 ActionDest: 510 Annotation: Subtype: Link Contents: Parada, A. E., D. M.Needham, and J. A.Fuhrman. 2016. “Every Base Matters: Assessing Small Subunit rRNA Primers for Marine Microbiomes With Mock Communities, Time Series and Global Field Samples.” Environmental Microbiology18, no. 5: 1403 1414. https://doi.org/10.1111/14622920.13023. Rect: 266, 395, 283, 382 ActionDest: 559 Annotation: Subtype: Link Contents: Please visit http://www.crg.eu for your more information Rect: 169, 372, 215, 359 Annotation: Subtype: Link Contents: R Core Team. 2022. “R: A Language and Environment for Statistical Computing (Version 4.2.2) [Computer Software].” R Foundation for Statistical Computing. https://www.Rproject.org/. Rect: 212, 303, 230, 290 ActionDest: 559 Annotation: Subtype: Link Contents: Callahan, B. J., P. J.McMurdie, M. J.Rosen, A. W.Han, A. J. A.Johnson, and S. P.Holmes. 2016. “DADA2: HighResolution Sample Inference From Illumina Amplicon Data.” Nature Methods13, no. 7: 581 583. https://doi.org/10.1038/nmeth.3869. Rect: 132, 291, 149, 279 ActionDest: 465 Annotation: Subtype: Link Contents: Auladell, A., and A.Obiol. 2022. DADA2 Pipeline. GitHub. https://github.com/adriaaula/dada2_guidelines. Rect: 268, 291, 286, 279 ActionDest: 465 Annotation: Subtype: Link Contents: Martin, M.2011. “Cutadapt Removes Adapter Sequences From HighThroughput Sequencing Reads.” EMBnet.Journal17, no. 1: 1. https://doi.org/10.14806/ej.17.1.200. Rect: 170, 268, 187, 256 ActionDest: 510 Annotation: Subtype: Link Contents: Wang, Q., G. M.Garrity, J. M.Tiedje, and J. R.Cole. 2007. “Naïve Bayesian Classifier for Rapid Assignment of rRNA Sequences Into the New Bacterial Taxonomy.” Applied and Environmental Microbiology73, no. 16: 5261 5267. https://doi.org/10.1128/AEM.0006207. Rect: 233, 211, 252, 198 ActionDest: 603 Annotation: Subtype: Link Contents: McLaren, M. R., and B. J.Callahan. 2021. Silva 138.1 Prokaryotic SSU Taxonomic Training Data Formatted for DADA2 [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.4587955. Rect: 196, 199, 213, 187 ActionDest: 510 Annotation: Subtype: Link Contents: Oksanen, J., G.Simpson, F. G.Blanchet, etal. 2022. “Vegan Community Ecology Package Version 2.62 April 2022.” Rect: 138, 153, 156, 141 ActionDest: 559 Annotation: Subtype: Link Contents: Douglas, G. M., V. J.Maffei, J. R.Zaneveld, etal. 2020. “PICRUSt2 for Prediction of Metagenome Functions.” Nature Biotechnology38, no. 6: 685 688. https://doi.org/10.1038/s4158702005486. Rect: 167, 61, 185, 49 ActionDest: 465 Annotation: Subtype: Link Contents: Wright, R. J., and M. G. I.Langille. 2025. “PICRUSt2SC: An Update to the Reference Database Used for Functional Prediction Within PICRUSt2.” Bioinformatics41, no. 5: btaf269. https://doi.org/10.1093/bioinformatics/btaf269. Rect: 267, 61, 286, 49 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 458, 487, 467, 474 ActionDest: 603 Annotation: Subtype: Link Contents: Please visit https://www.kegg.jp/kegg/pathway.html for your more information Rect: 468, 406, 549, 394 Annotation: Subtype: Link Contents: Please visit https://www.kegg.jp/kegg/pathway.html for your more information Rect: 306, 395, 380, 382 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 366, 176, 376, 164 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 348, 107, 357, 95 ActionDest: 603 Annotation: Subtype: Link Contents: Saghaï, A., O. C.Moore, C. M.Jones, and S.Hallin. 2025. “Microbial Controls of Nitrogen Retention and N2O Production in Cropping Systems Supporting Soil Carbon Accrual.” Soil Biology and Biochemistry208: 109858. https://doi.org/10.1016/j.soilbio.2025.109858. Rect: 418, 107, 437, 95 ActionDest: 559 Page: Sequence: 6 Annotations: Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 106, 751, 115, 739 ActionDest: 603 Annotation: Subtype: Link Contents: Persson, S., J. E.Varg, S.Coker, etal. 2026. “From Food to sewer: Dietary Inputs Support the Temporal and PopulationWide Stability of Pepper Mild Mottle Virus in Wastewater.” Water Research297: 125629. https://doi.org/10.1016/j.watres.2026.125629. Rect: 68, 717, 87, 704 ActionDest: 559 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 259, 694, 268, 681 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 236, 671, 244, 658 ActionDest: 603 Annotation: Subtype: Link Contents: Bustin, S. A., V.Benes, J. A.Garson, etal. 2009. “The MIQE Guidelines: Minimum Information for Publication of Quantitative RealTime PCR Experiments.” Clinical Chemistry55, no. 4: 611 622. https://doi.org/10.1373/clinchem.2008.112797. Rect: 258, 579, 277, 566 ActionDest: 465 Annotation: Subtype: Link Contents: R Core Team. 2022. “R: A Language and Environment for Statistical Computing (Version 4.2.2) [Computer Software].” R Foundation for Statistical Computing. https://www.Rproject.org/. Rect: 170, 418, 189, 405 ActionDest: 559 Annotation: Subtype: Link Contents: MartínezPérez, L., L.Luquot, J.Carrera, etal. 2022. “A Multidisciplinary Approach to Characterizing Coastal Alluvial Aquifers to Improve Understanding of Seawater Intrusion and Submarine Groundwater Discharge.” Journal of Hydrology607: 127510. https://doi.org/10.1016/j.jhydrol.2022.127510. Rect: 148, 303, 166, 290 ActionDest: 510 Annotation: Subtype: Link Contents: Please see "figure 1b" for your more reference Rect: 274, 176, 283, 164 ActionDest: 220 Annotation: Subtype: Link Contents: Oksanen, J., G.Simpson, F. G.Blanchet, etal. 2022. “Vegan Community Ecology Package Version 2.62 April 2022.” Rect: 114, 73, 132, 60 ActionDest: 559 Annotation: Subtype: Link Contents: Cáceres, M. D., and P.Legendre. 2009. “Associations Between Species and Groups of Sites: Indices and Statistical Inference.” Ecology90, no. 12: 3566 3574. https://doi.org/10.1890/081823.1. Rect: 400, 510, 419, 497 ActionDest: 465 Annotation: Subtype: Link Contents: Please see "figure 1c" for your more reference Rect: 431, 372, 439, 359 ActionDest: 220 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 536, 349, 544, 336 ActionDest: 220 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 384, 326, 393, 313 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 1c" for your more reference Rect: 383, 199, 391, 187 ActionDest: 220 Annotation: Subtype: Link Contents: Please see "figure 1c" for your more reference Rect: 334, 130, 341, 118 ActionDest: 220 Page: Sequence: 7 Annotations: Annotation: Subtype: Link Contents: Please see "figure 2" for your more reference Rect: 77, 305, 81, 292 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 2a" for your more reference Rect: 250, 178, 259, 166 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 2c" for your more reference Rect: 218, 155, 226, 143 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 2b" for your more reference Rect: 78, 132, 87, 120 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 2e" for your more reference Rect: 110, 109, 118, 97 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 2d" for your more reference Rect: 189, 98, 198, 85 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 2f" for your more reference Rect: 240, 75, 248, 62 ActionDest: 278 Annotation: Subtype: Link Contents: Please see "figure 3" for your more reference Rect: 451, 350, 455, 337 ActionDest: 296 Annotation: Subtype: Link Contents: Please see "figure 3a,b" for your more reference Rect: 338, 281, 353, 268 ActionDest: 296 Annotation: Subtype: Link Contents: Please see "figure 3c,d" for your more reference Rect: 422, 270, 438, 257 ActionDest: 296 Annotation: Subtype: Link Contents: Please see "figure 3e,f" for your more reference Rect: 421, 258, 436, 245 ActionDest: 296 Annotation: Subtype: Link Contents: Please see "figure 3c f" for your more reference Rect: 463, 224, 479, 211 ActionDest: 296 Annotation: Subtype: Link Contents: Please see "figure 3a,b" for your more reference Rect: 334, 178, 349, 165 ActionDest: 296 Annotation: Subtype: Link Contents: Please see "figure 4a" for your more reference Rect: 536, 74, 544, 61 ActionDest: 306 Page: Sequence: 8 Annotations: Annotation: Subtype: Link Contents: Please see "figure 4a" for your more reference Rect: 243, 291, 252, 278 ActionDest: 306 Annotation: Subtype: Link Contents: Please see "figure 4b" for your more reference Rect: 114, 222, 122, 209 ActionDest: 306 Annotation: Subtype: Link Contents: Please see "figure 4c" for your more reference Rect: 78, 153, 86, 140 ActionDest: 306 Annotation: Subtype: Link Contents: Please see "figure 4c" for your more reference Rect: 371, 302, 379, 289 ActionDest: 306 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 479, 210, 492, 197 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 474, 141, 487, 128 ActionDest: 603 Page: Sequence: 9 Annotations: Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 217, 103, 221, 90 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 256, 103, 265, 90 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 393, 103, 406, 90 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 436, 91, 448, 78 ActionDest: 603 Page: Sequence: 10 Annotations: Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 126, 164, 135, 151 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 279, 118, 283, 105 ActionDest: 314 Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 125, 95, 129, 82 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 188, 72, 197, 59 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 375, 130, 379, 117 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 414, 130, 423, 117 ActionDest: 603 Page: Sequence: 11 Annotations: Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 76, 648, 80, 635 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 243, 613, 251, 601 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 235, 533, 239, 520 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 272, 533, 281, 520 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 76, 464, 81, 451 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 114, 464, 123, 451 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 187, 360, 196, 348 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 5" for your more reference Rect: 282, 337, 286, 325 ActionDest: 314 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 76, 326, 86, 313 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 104, 326, 113, 313 ActionDest: 603 Annotation: Subtype: Link Contents: Ren, Z.Q., H.Wang, L.G.Zhang, X.N.Du, B.C.Huang, and R.C.Jin. 2022. “A Review of AnammoxBased Nitrogen Removal Technology: From Microbial Diversity to Engineering Applications.” Bioresource Technology363: 127896. https://doi.org/10.1016/j.biortech.2022.127896. Rect: 265, 268, 283, 256 ActionDest: 559 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 253, 257, 262, 244 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 119, 188, 128, 175 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 100, 165, 113, 152 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 102, 119, 122, 106 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 212, 61, 226, 49 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 6" for your more reference Rect: 543, 693, 547, 680 ActionDest: 357 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 333, 682, 342, 669 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 6a,b" for your more reference Rect: 363, 647, 378, 634 ActionDest: 357 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 413, 636, 421, 623 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 6c f" for your more reference Rect: 335, 567, 351, 554 ActionDest: 357 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 386, 567, 395, 554 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 6g,h" for your more reference Rect: 338, 532, 354, 519 ActionDest: 357 Annotation: Subtype: Link Contents: Please see "figure 6" for your more reference Rect: 338, 463, 342, 450 ActionDest: 357 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 376, 463, 385, 450 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 6c f" for your more reference Rect: 452, 325, 468, 312 ActionDest: 357 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 503, 325, 512, 312 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 535, 279, 544, 266 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 6h" for your more reference Rect: 450, 245, 459, 232 ActionDest: 357 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 433, 130, 450, 117 ActionDest: 603 Page: Sequence: 12 Annotations: Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 266, 187, 283, 174 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 221, 95, 238, 82 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 115, 61, 132, 48 ActionDest: 603 Annotation: Subtype: Link Contents: Folch, A., L.Del Val, L.Luquot, etal. 2020. “Combining Fiber Optic DTS, CrossHole ERT and TimeLapse Induction Logging to Characterize and Monitor a Coastal Aquifer.” Journal of Hydrology588: 125050. https://doi.org/10.1016/j.jhydrol.2020.125050. Rect: 375, 72, 394, 59 ActionDest: 465 Annotation: Subtype: Link Contents: RuizGonzález, C., L.RodríguezPie, O.Maister, etal. 2022. “High Spatial Heterogeneity and Low Connectivity of Bacterial Communities Along a Mediterranean Subterranean Estuary.” Molecular Ecology31, no. 22: 5745 5764. https://doi.org/10.1111/mec.16695. Rect: 482, 72, 500, 59 ActionDest: 559 Page: Sequence: 13 Annotations: Annotation: Subtype: Link Contents: Goyetche, T., L.Luquot, J.Carrera, L.MartínezPérez, and A.Folch. 2022. “Identification and Quantification of Chemical Reactions in a Coastal Aquifer to Assess Submarine Groundwater Discharge Composition.” Science of the Total Environment838: 155978. https://doi.org/10.1016/j.scitotenv.2022.155978. Rect: 268, 648, 286, 635 ActionDest: 510 Annotation: Subtype: Link Contents: RufíSalís, M., J.GarciaOrellana, G.Cantero, etal. 2019. “Influence of Land Use Changes on Submarine Groundwater Discharge.” Environmental Research Communications1, no. 3: 31005. https://doi.org/10.1088/25157620/ab1695. Rect: 68, 579, 86, 566 ActionDest: 559 Annotation: Subtype: Link Contents: Chen, X., I.Santos, J.Du, etal. 2025. “Sandy Subterranean Estuaries Minimize Groundwater Nitrogen Pollution Impacts on Coastal Waters.” Geophysical Research Letters52, no. 3: e2024GL109621. https://doi.org/10.1029/2024GL109621. Rect: 65, 544, 84, 532 ActionDest: 465 Annotation: Subtype: Link Contents: Santos, I. R., X.Chen, A. L.Lecher, etal. 2021. “Submarine Groundwater Discharge Impacts on Coastal Nutrient Biogeochemistry.” Nature Reviews Earth & Environment2, no. 5: 307 323. https://doi.org/10.1038/s43017021001520. Rect: 135, 544, 153, 532 ActionDest: 603 Annotation: Subtype: Link Contents: Couturier, M., G.TommiMorin, M.Sirois, A.Rao, C.Nozais, and G.Chaillou. 2017. “Nitrogen Transformations Along a Shallow Subterranean Estuary.” Biogeosciences14, no. 13: 3321 3336. https://doi.org/10.5194/bg1433212017. Rect: 189, 475, 207, 463 ActionDest: 465 Annotation: Subtype: Link Contents: Rodellas, V., T. C.Stieglitz, A.Andrisoa, etal. 2018. “GroundwaterDriven Nutrient Inputs to Coastal Lagoons: The Relevance of Lagoon Water Recirculation as a Conveyor of Dissolved Nutrients.” Science of the Total Environment642: 764 780. https://doi.org/10.1016/j.scitotenv.2018.06.095. Rect: 266, 475, 283, 463 ActionDest: 559 Annotation: Subtype: Link Contents: Daims, H., S.Lücker, and M.Wagner. 2016. “A New Perspective on Microbes Formerly Known as NitriteOxidizing Bacteria.” Trends in Microbiology24, no. 9: 699 712. https://doi.org/10.1016/j.tim.2016.05.004. Rect: 266, 441, 283, 428 ActionDest: 465 Annotation: Subtype: Link Contents: Morán, X. A. G., A.Bode, L. Á.Suárez, and E.Nogueira. 2007. “Assessing the Relevance of Nucleic Acid Content as an Indicator of Marine Bacterial Activity.” Aquatic Microbial Ecology46: 141 152. https://doi.org/10.3354/ame046141. Rect: 203, 326, 222, 313 ActionDest: 510 Annotation: Subtype: Link Contents: RuizGonzález, C., L.RodríguezPie, O.Maister, etal. 2022. “High Spatial Heterogeneity and Low Connectivity of Bacterial Communities Along a Mediterranean Subterranean Estuary.” Molecular Ecology31, no. 22: 5745 5764. https://doi.org/10.1111/mec.16695. Rect: 67, 280, 85, 267 ActionDest: 559 Annotation: Subtype: Link Contents: MartínezPérez, L., L.Luquot, J.Carrera, etal. 2022. “A Multidisciplinary Approach to Characterizing Coastal Alluvial Aquifers to Improve Understanding of Seawater Intrusion and Submarine Groundwater Discharge.” Journal of Hydrology607: 127510. https://doi.org/10.1016/j.jhydrol.2022.127510. Rect: 206, 119, 224, 106 ActionDest: 510 Annotation: Subtype: Link Contents: Overholt, W. A., S.Trumbore, X.Xu, etal. 2022. “Carbon Fixation Rates in Groundwater Similar to Those in Oligotrophic Marine Systems.” Nature Geoscience15, no. 7: 561 567. https://doi.org/10.1038/s41561022009685. Rect: 119, 73, 138, 60 ActionDest: 559 Annotation: Subtype: Link Contents: RuizGonzález, C., V.Rodellas, and J.GarciaOrellana. 2021. “The Microbial Dimension of Submarine Groundwater Discharge: Current Challenges and Future Directions.” FEMS Microbiology Reviews45, no. 5: fuab010. https://doi.org/10.1093/femsre/fuab010. Rect: 225, 73, 243, 60 ActionDest: 559 Annotation: Subtype: Link Contents: Sebastián, M., M.Estrany, C.RuizGonzález, etal. 2019. “High Growth Potential of LongTerm Starved Deep Ocean Opportunistic Heterotrophic Bacteria.” Frontiers in Microbiology10: 760. https://doi.org/10.3389/fmicb.2019.00760. Rect: 473, 751, 490, 739 ActionDest: 603 Annotation: Subtype: Link Contents: Brick, S., J.Niggemann, A.Reckhardt, M.Könneke, and B.Engelen. 2025. “Interstitial Microbial Communities of Coastal Sediments Are Dominated by Nanoarchaeota.” Frontiers in Microbiology16: 1532193. https://doi.org/10.3389/fmicb.2025.1532193. Rect: 502, 625, 520, 612 ActionDest: 465 Annotation: Subtype: Link Contents: Hong, Y., J.Wu, S.Wilson, and B.Song. 2019. “Vertical Stratification of Sediment Microbial Communities Along Geochemical Gradients of a Subterranean Estuary Located at the Gloucester Beach of Virginia, United States.” Frontiers in Microbiology9: 3343. https://doi.org/10.3389/fmicb.2018.03343. Rect: 328, 613, 346, 601 ActionDest: 510 Annotation: Subtype: Link Contents: Sang, S., H.Dai, B. X.Hu, Y.Hao, T.Zhou, and J.Zhang. 2019. “The Study of Hydrogeochemical Environments and Microbial Communities Along a Groundwater Salinity Gradient in the Pearl River Delta, China.” Water11, no. 4: 4. https://doi.org/10.3390/w11040804. Rect: 395, 613, 412, 601 ActionDest: 559 Annotation: Subtype: Link Contents: RuizGonzález, C., C.Mena, F. M.CornejoCastillo, D.RomanoGude, N.ArandiaGorostidi, and J. M.Gasol. 2025. “Diverse Patescibacteria Assemblages and Prevalence of UltraSmall FreeLiving Parcubacteria Along a Subterranean Estuary.” mSystems10, no. 11: e0112525. https://doi.org/10.1128/msystems.0112525. Rect: 395, 579, 413, 566 ActionDest: 559 Annotation: Subtype: Link Contents: RuizGonzález, C., L.RodríguezPie, O.Maister, etal. 2022. “High Spatial Heterogeneity and Low Connectivity of Bacterial Communities Along a Mediterranean Subterranean Estuary.” Molecular Ecology31, no. 22: 5745 5764. https://doi.org/10.1111/mec.16695. Rect: 505, 579, 523, 566 ActionDest: 559 Annotation: Subtype: Link Contents: Palacios, A., J. J.Ledo, N.Linde, etal. 2020. “TimeLapse CrossHole Electrical Resistivity Tomography (CHERT) for Monitoring Seawater Intrusion Dynamics in a Mediterranean Aquifer.” Hydrology and Earth System Sciences24, no. 4: 2121 2139. https://doi.org/10.5194/hess2421212020. Rect: 327, 452, 345, 440 ActionDest: 559 Annotation: Subtype: Link Contents: RuizGonzález, C., L.RodríguezPie, O.Maister, etal. 2022. “High Spatial Heterogeneity and Low Connectivity of Bacterial Communities Along a Mediterranean Subterranean Estuary.” Molecular Ecology31, no. 22: 5745 5764. https://doi.org/10.1111/mec.16695. Rect: 430, 452, 448, 440 ActionDest: 559 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 490, 234, 502, 221 ActionDest: 603 Annotation: Subtype: Link Contents: Chen, X., I.Santos, J.Du, etal. 2025. “Sandy Subterranean Estuaries Minimize Groundwater Nitrogen Pollution Impacts on Coastal Waters.” Geophysical Research Letters52, no. 3: e2024GL109621. https://doi.org/10.1029/2024GL109621. Rect: 526, 165, 544, 152 ActionDest: 465 Annotation: Subtype: Link Contents: Montiel, D., A. F.Lamore, J.Stewart, etal. 2019. “Natural Groundwater Nutrient Fluxes Exceed Anthropogenic Inputs in an Ecologically Impacted Estuary: Lessons Learned From Mobile Bay, Alabama.” Biogeochemistry145, no. 1: 1 33. https://doi.org/10.1007/s10533019005870. Rect: 440, 119, 457, 106 ActionDest: 510 Annotation: Subtype: Link Contents: Liao, Y., T.He, C.Wang, C.Zheng, and M.Zhang. 2025. “Dissimilatory Nitrate Reduction to Ammonium Has a Competitive Advantage Over Denitrification Under NitrateLimited Conditions.” Reviews in Environmental Science and Bio/Technology24, no. 1: 97 114. https://doi.org/10.1007/s11157025097195. Rect: 485, 73, 503, 60 ActionDest: 510 Page: Sequence: 14 Annotations: Annotation: Subtype: Link Contents: Santoro, A. E.2010. “Microbial Nitrogen Cycling at the Saltwater Freshwater Interface.” Hydrogeology Journal18, no. 1: 187 202. https://doi.org/10.1007/s100400090526z. Rect: 102, 740, 119, 727 ActionDest: 559 Annotation: Subtype: Link Contents: Wu, J., Y.Hong, S. J.Wilson, and B.Song. 2021. “Microbial Nitrogen Loss by Coupled Nitrification to Denitrification and Anammox in a Permeable Subterranean Estuary at Gloucester Point, Virginia.” Marine Pollution Bulletin168: 112440. https://doi.org/10.1016/j.marpolbul.2021.112440. Rect: 69, 682, 87, 670 ActionDest: 603 Annotation: Subtype: Link Contents: Jiao, L., J.Wu, X.He, X.Wen, Y.Li, and Y.Hong. 2018. “Significant Microbial Nitrogen Loss From Denitrification and Anammox in the LandSea Interface of Low Permeable Sediments.” International Biodeterioration & Biodegradation135: 80 89. https://doi.org/10.1016/j.ibiod.2018.10.002. Rect: 149, 659, 167, 647 ActionDest: 510 Annotation: Subtype: Link Contents: Jiao, L., J.Wu, X.He, X.Wen, Y.Li, and Y.Hong. 2018. “Significant Microbial Nitrogen Loss From Denitrification and Anammox in the LandSea Interface of Low Permeable Sediments.” International Biodeterioration & Biodegradation135: 80 89. https://doi.org/10.1016/j.ibiod.2018.10.002. Rect: 248, 625, 265, 612 ActionDest: 510 Annotation: Subtype: Link Contents: Ibánhez, J. S. P., X. A.ÁlvarezSalgado, and C.Rocha. 2021. “Does Nitrate Enrichment Accelerate Organic Matter Turnover in Subterranean Estuaries?” Frontiers in Marine Science8: 661201. https://doi.org/10.3389/fmars.2021.661201. Rect: 103, 510, 121, 497 ActionDest: 510 Annotation: Subtype: Link Contents: Peñarroya, X., A.Lupon, X.TriadóMargarit, etal. 2023. “Dissolved Organic Carbon Bioreactivity and DOC:DIN Stoichiometry Control Ammonium Uptake in an Intermittent Mediterranean Stream.” Freshwater Biology68, no. 9: 1572 1587. https://doi.org/10.1111/fwb.14152. Rect: 193, 441, 212, 428 ActionDest: 559 Annotation: Subtype: Link Contents: Peñarroya, X., N.Catalán, A.Freixa, etal. 2024. “Organic Matter Composition and Water Stoichiometry Are Main Drivers of Heterotrophic Nitrate Uptake in Mediterranean Headwater Streams.” Journal of Geophysical Research: Biogeosciences129, no. 12: e2024JG008346. https://doi.org/10.1029/2024JG008346. Rect: 216, 441, 234, 428 ActionDest: 559 Annotation: Subtype: Link Contents: Couturier, M., G.TommiMorin, M.Sirois, A.Rao, C.Nozais, and G.Chaillou. 2017. “Nitrogen Transformations Along a Shallow Subterranean Estuary.” Biogeosciences14, no. 13: 3321 3336. https://doi.org/10.5194/bg1433212017. Rect: 111, 337, 128, 325 ActionDest: 465 Annotation: Subtype: Link Contents: Wu, J., Y.Hong, S. J.Wilson, and B.Song. 2021. “Microbial Nitrogen Loss by Coupled Nitrification to Denitrification and Anammox in a Permeable Subterranean Estuary at Gloucester Point, Virginia.” Marine Pollution Bulletin168: 112440. https://doi.org/10.1016/j.marpolbul.2021.112440. Rect: 173, 337, 191, 325 ActionDest: 603 Annotation: Subtype: Link Contents: Burgin, A. J., and S. K.Hamilton. 2007. “Have We Overemphasized the Role of Denitrification in Aquatic Ecosystems? A Review of Nitrate Removal Pathways.” Frontiers in Ecology and the Environment5, no. 2: 89 96. https://doi.org/10.1890/15409295(2007)5[89:HWOTRO]2.0.CO;2. Rect: 244, 245, 263, 233 ActionDest: 465 Annotation: Subtype: Link Contents: Kelso, B., R. V.Smith, R. J.Laughlin, and S. D.Lennox. 1997. “Dissimilatory Nitrate Reduction in Anaerobic Sediments Leading to River Nitrite Accumulation.” Applied and Environmental Microbiology63, no. 12: 4679 4685. https://doi.org/10.1128/aem.63.12.46794685.1997. Rect: 66, 234, 84, 221 ActionDest: 510 Annotation: Subtype: Link Contents: Lu, J., Y.Hong, Y.Wei, etal. 2021. “Nitrification Mainly Driven by AmmoniaOxidizing Bacteria and NitriteOxidizing Bacteria in an AnammoxInoculated Wastewater Treatment System.” AMB Express11, no. 1: 158. https://doi.org/10.1186/s13568021013216. Rect: 144, 188, 162, 175 ActionDest: 510 Annotation: Subtype: Link Contents: Wu, J., Y.Hong, S. J.Wilson, and B.Song. 2021. “Microbial Nitrogen Loss by Coupled Nitrification to Denitrification and Anammox in a Permeable Subterranean Estuary at Gloucester Point, Virginia.” Marine Pollution Bulletin168: 112440. https://doi.org/10.1016/j.marpolbul.2021.112440. Rect: 203, 188, 220, 175 ActionDest: 603 Annotation: Subtype: Link Contents: Biggs, C. R., L. A.Yeager, D. G.Bolser, etal. 2020. “Does Functional Redundancy Affect Ecological Stability and Resilience? A Review and MetaAnalysis.” Ecosphere11, no. 7: e03184. https://doi.org/10.1002/ecs2.3184. Rect: 67, 73, 86, 60 ActionDest: 465 Annotation: Subtype: Link Contents: Louca, S., M. F.Polz, F.Mazel, etal. 2018. “Function and Functional Redundancy in Microbial Systems.” Nature Ecology & Evolution2, no. 6: 936 943. https://doi.org/10.1038/s4155901805191. Rect: 139, 73, 157, 60 ActionDest: 510 Annotation: Subtype: Link Contents: Yachi, S., and M.Loreau. 1999. “Biodiversity and Ecosystem Productivity in a Fluctuating Environment: The Insurance Hypothesis.” Proceedings of the National Academy of Sciences96, no. 4: 1463 1468. https://doi.org/10.1073/pnas.96.4.1463. Rect: 234, 73, 252, 60 ActionDest: 603 Annotation: Subtype: Link Contents: Mosley, O. E., E.Gios, M.Close, L.Weaver, C.Daughney, and K. M.Handley. 2022. “Nitrogen Cycling and Microbial Cooperation in the Terrestrial Subsurface.” ISME Journal16, no. 11: 2561 2573. https://doi.org/10.1038/s41396022013000. Rect: 464, 682, 482, 670 ActionDest: 510 Annotation: Subtype: Link Contents: CalvoMartin, E., E.Teira, X. A.ÁlvarezSalgado, etal. 2022. “On the Hidden Diversity and Niche Specialization of the Microbial Realm of Subterranean Estuaries.” Environmental Microbiology24, no. 12: 5859 5881. https://doi.org/10.1111/14622920.16160. Rect: 442, 648, 460, 635 ActionDest: 465 Annotation: Subtype: Link Contents: Wu, J., Y.Hong, S. J.Wilson, and B.Song. 2021. “Microbial Nitrogen Loss by Coupled Nitrification to Denitrification and Anammox in a Permeable Subterranean Estuary at Gloucester Point, Virginia.” Marine Pollution Bulletin168: 112440. https://doi.org/10.1016/j.marpolbul.2021.112440. Rect: 414, 613, 432, 601 ActionDest: 603 Annotation: Subtype: Link Contents: Goyetche, T., L.Luquot, J.Carrera, L.MartínezPérez, and A.Folch. 2022. “Identification and Quantification of Chemical Reactions in a Coastal Aquifer to Assess Submarine Groundwater Discharge Composition.” Science of the Total Environment838: 155978. https://doi.org/10.1016/j.scitotenv.2022.155978. Rect: 419, 556, 437, 543 ActionDest: 510 Annotation: Subtype: Link Contents: MartínezPérez, L., L.Luquot, J.Carrera, etal. 2022. “A Multidisciplinary Approach to Characterizing Coastal Alluvial Aquifers to Improve Understanding of Seawater Intrusion and Submarine Groundwater Discharge.” Journal of Hydrology607: 127510. https://doi.org/10.1016/j.jhydrol.2022.127510. Rect: 526, 556, 544, 543 ActionDest: 510 Annotation: Subtype: Link Contents: Roumelis, C., F.Willert, M.Scaccia, etal. 2025. “Water Table Fluctuations Control Nitrate and Ammonium Fate in Coastal Aquifers.” Water Resources Research61, no. 1: e2024WR038087. https://doi.org/10.1029/2024WR038087. Rect: 370, 475, 388, 463 ActionDest: 559 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 505, 222, 518, 210 ActionDest: 603 Annotation: Subtype: Link Contents: Tourna, M., T. E.Freitag, G. W.Nicol, and J. I.Prosser. 2008. “Growth, Activity and Temperature Responses of AmmoniaOxidizing Archaea and Bacteria in Soil Microcosms.” Environmental Microbiology10, no. 5: 1357 1364. https://doi.org/10.1111/j.14622920.2007.01563.x. Rect: 394, 211, 412, 198 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 526, 188, 544, 175 ActionDest: 603 Annotation: Subtype: Link Contents: Throbäck, I. N., K.Enwall, A.Jarvis, and S.Hallin. 2004. “Reassessing PCR Primers Targeting nirS, nirK and nosZ Genes for Community Surveys of Denitrifying Bacteria With DGGE.” FEMS Microbiology Ecology49, no. 3: 401 417. https://doi.org/10.1016/j.femsec.2004.04.011. Rect: 452, 96, 470, 83 ActionDest: 603 Page: Sequence: 15 Annotations: Annotation: Subtype: Link Contents: Henry, S., E.Baudoin, J. C.LópezGutiérrez, F.MartinLaurent, A.Brauman, and L.Philippot. 2004. “Quantification of Denitrifying Bacteria in Soils by nirK Gene Targeted RealTime PCR.” Journal of Microbiological Methods59, no. 3: 327 335. https://doi.org/10.1016/j.mimet.2004.07.002. Rect: 198, 486, 217, 473 ActionDest: 510 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 211, 360, 228, 347 ActionDest: 603 Annotation: Subtype: Link Contents: Adyasari, D., N. T.Dimova, S. M.Ní Chadhain, and H.Waska. 2024. “Microbial Assemblages and Metabolic Activity in OrganicRich Subterranean Estuaries: Impact of Climate and Land Use Changes.” Journal of Geophysical Research: Biogeosciences129, no. 3: e2023JG007660. https://doi.org/10.1029/2023JG007660. Rect: 106, 233, 124, 220 ActionDest: 465 Annotation: Subtype: Link Contents: Hong, Y., J.Wu, S.Wilson, and B.Song. 2019. “Vertical Stratification of Sediment Microbial Communities Along Geochemical Gradients of a Subterranean Estuary Located at the Gloucester Beach of Virginia, United States.” Frontiers in Microbiology9: 3343. https://doi.org/10.3389/fmicb.2018.03343. Rect: 172, 233, 190, 220 ActionDest: 510 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 219, 187, 228, 174 ActionDest: 603 Annotation: Subtype: Link Contents: Rogers, D. R., and K. L.Casciotti. 2010. “Abundance and Diversity of Archaeal Ammonia Oxidizers in a Coastal Groundwater System.” Applied and Environmental Microbiology76, no. 24: 7938 7948. https://doi.org/10.1128/AEM.0205609. Rect: 269, 141, 286, 128 ActionDest: 559 Annotation: Subtype: Link Contents: Santoro, A. E., C. A.Francis, N. R.De Sieyes, and A. B.Boehm. 2008. “Shifts in the Relative Abundance of AmmoniaOxidizing Bacteria and Archaea Across Physicochemical Gradients in a Subterranean Estuary.” Environmental Microbiology10, no. 4: 1068 1079. https://doi.org/10.1111/j.14622920.2007.01547.x. Rect: 98, 130, 116, 117 ActionDest: 559 Annotation: Subtype: Link Contents: Wilson, S. J., B.Song, and I. C.Anderson. 2023. “Geochemical Factors Impacting Nitrifying Communities in Sandy Sediments.” Environmental Microbiology25, no. 12: 3180 3191. https://doi.org/10.1111/14622920.16504. Rect: 171, 130, 189, 117 ActionDest: 603 Annotation: Subtype: Link Contents: Xiao, K., J.Wu, H.Li, etal. 2018. “Nitrogen Fate in a Subtropical Mangrove Swamp: Potential Association With SeawaterGroundwater Exchange.” Science of the Total Environment635: 586 597. https://doi.org/10.1016/j.scitotenv.2018.04.143. Rect: 235, 130, 252, 117 ActionDest: 603 Annotation: Subtype: Link Contents: Kumar, A., and A.Yadav. 2024. “Next Generation Sequencing in Metagenomics and Metatranscriptomics.” In MultiOmics Analysis of the Human Microbiome: From Technology to Clinical Applications, edited by I.Mani and V.Singh, 49 75. Springer Nature. https://doi.org/10.1007/9789819718443_3. Rect: 119, 61, 138, 48 ActionDest: 510 Annotation: Subtype: Link Contents: Couturier, M., G.TommiMorin, M.Sirois, A.Rao, C.Nozais, and G.Chaillou. 2017. “Nitrogen Transformations Along a Shallow Subterranean Estuary.” Biogeosciences14, no. 13: 3321 3336. https://doi.org/10.5194/bg1433212017. Rect: 472, 693, 490, 680 ActionDest: 465 Annotation: Subtype: Link Contents: Wong, W. W., A.Applegate, S. C.Poh, and P. L. M.Cook. 2020. “Biogeochemical Attenuation of Nitrate in a Sandy Subterranean Estuary: Insights From Two Stable Isotope Approaches.” Limnology and Oceanography65, no. 12: 3098 3113. https://doi.org/10.1002/lno.11576. Rect: 410, 682, 428, 669 ActionDest: 603 Annotation: Subtype: Link Contents: Spiteri, C., C. P.Slomp, M. A.Charette, K.Tuncay, and C.Meile. 2008. “Flow and Nutrient Dynamics in a Subterranean Estuary (Waquoit Bay, MA, USA): Field Data and Reactive Transport Modeling.” Geochimica et Cosmochimica Acta72, no. 14: 3398 3412. https://doi.org/10.1016/j.gca.2008.04.027. Rect: 330, 670, 348, 657 ActionDest: 603 Annotation: Subtype: Link Contents: Goyetche, T., L.Luquot, J.Carrera, L.MartínezPérez, and A.Folch. 2022. “Identification and Quantification of Chemical Reactions in a Coastal Aquifer to Assess Submarine Groundwater Discharge Composition.” Science of the Total Environment838: 155978. https://doi.org/10.1016/j.scitotenv.2022.155978. Rect: 326, 659, 344, 646 ActionDest: 510 Annotation: Subtype: Link Contents: MartínezPérez, L., L.Luquot, J.Carrera, etal. 2022. “A Multidisciplinary Approach to Characterizing Coastal Alluvial Aquifers to Improve Understanding of Seawater Intrusion and Submarine Groundwater Discharge.” Journal of Hydrology607: 127510. https://doi.org/10.1016/j.jhydrol.2022.127510. Rect: 447, 647, 465, 634 ActionDest: 510 Annotation: Subtype: Link Contents: Euler, S., L. C.Jeffrey, D. T.Maher, S. G.Johnston, R.Sugimoto, and D. R.Tait. 2024. “Microbiome Mediating Methane and Nitrogen Transformations in a Subterranean Estuary.” Environmental Microbiology26, no. 1: e16558. https://doi.org/10.1111/14622920.16558. Rect: 459, 624, 477, 611 ActionDest: 465 Annotation: Subtype: Link Contents: Li, D., Y.Zhao, and Z.Liu. 2025. “Nitrogen Transformation and Drivers in Response to Hydrological Variability in the Yangtze Subterranean Estuary.” Journal of Environmental Management391: 126474. https://doi.org/10.1016/j.jenvman.2025.126474. Rect: 513, 624, 531, 611 ActionDest: 510 Annotation: Subtype: Link Contents: Wu, J., Y.Hong, S. J.Wilson, and B.Song. 2021. “Microbial Nitrogen Loss by Coupled Nitrification to Denitrification and Anammox in a Permeable Subterranean Estuary at Gloucester Point, Virginia.” Marine Pollution Bulletin168: 112440. https://doi.org/10.1016/j.marpolbul.2021.112440. Rect: 327, 613, 344, 600 ActionDest: 603 Annotation: Subtype: Link Contents: Data S1: Supplementary Methods. Rect: 306, 532, 401, 519 ActionDest: 639 Annotation: Subtype: Link Contents: Please see "figure 7" for your more reference Rect: 496, 509, 500, 496 ActionDest: 420 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 534, 509, 547, 496 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 330, 498, 339, 485 ActionDest: 603 Annotation: Subtype: Link Contents: Please see "figure 7" for your more reference Rect: 337, 245, 341, 232 ActionDest: 420 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 373, 245, 387, 232 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 414, 245, 423, 232 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 462, 210, 471, 197 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 66, 513, 74, 502 ActionDest: 603 Annotation: Subtype: Link Contents: Figure S1: Temporal variations between July 2021 and September 2022 in temperature (a) and electrical conductivity (b) recorded at different depths within nest N1 (64m inland, Figure1a) with permanent installed sensors. Orange shaded bands denote the periods during which the two sampling campaigns were performed. Note the different scales in the Y axes of conductivity plots.Figure S2: Precipitation and air temperature variations in the Argentona sampling site during the period 2021 2022. Vertical yellow areas indicate the two sampling campaigns conducted, each spanning 3days. Meteorological data were obtained from the ‘Xarxa d'Estacions Meteorolòliques Autonàtiques’ (XEMA) through the ‘Portal de Dades Obertes de la Generalitat de Catalunya’ (https://analisi.transparenciacatalunya.cat/MediAmbient/DadesmeteorolgiquesdiriesdelaXEMA/7bvhjvq2/data_preview), using daily records from the Cabrils station, the closest to the sampling site.Figure S3: Distribution and taxonomic composition of clusterindicator ASVs. (a) Relative contribution of ASVs identified as indicators of each groundwater cluster (Soil, Fresh_S, Fresh_D, Saline and Seawater) across all clusters. Bars show the percentage of reads assigned to the indicator ASVs of each cluster. (b) Taxonomic composition of indicator ASVs for each cluster at the order level. Donut charts show the percentage of indicator ASVs assigned to each order, with percentage labels displayed only for taxa representing more than 5% of the indicator ASVs within each cluster. The 25 most abundant orders are shown individually, whereas the remaining orders are grouped as “Other orders”. Taxonomic groups are ordered clockwise from the most to the least abundant, with “Other orders” always displayed last. Numbers at the center of each donut indicate the total number of indicator ASVs (n) identified for each cluster.Figure S4: Predicted distributions of heterotrophic nitrogen assimilation genes. Predicted relative abundances (expressed as % of PICRUSt2retained reads) of the KOs predicted by PICRUSt2 in each cluster, colourcoded by sampling campaign. Above each panel, the corresponding nitrogen transformation process is shown, with the targeted genes placed at the respective step. Pie charts indicate the predicted taxonomic carriers of each gene, combining winter and summer campaigns. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively. Taxonomic assignments are shown at the phylum level, except for Pseudomonadota, Bacteroidota and Planctomycetota, where the main classes and even orders were distinguished. Pseudo, Pseudomonadota; Alph, Alphaproteobacteria; Gam, Gammaproteobacteria; Bact, Bacteroidota; Bacter, Bacteroidia; Planct, Planctomycetota.Figure S5: Abundance and taxonomic assignation of ASVs excluded from the PICRUSt2 analysis. Read abundance (a) and relative contribution (b) of ASVs discarded by PICRUSt2 due to NSTI >2. Percentage values are indicated above each bar. Samples are organized by cluster and campaign. Grey indicates ASVs retained by the analysis.Figure S6: Taxonomic composition of PICRUSt2predicted nitrogen cycling genes. Taxonomic composition of the predicted microbial contributors to the selected nitrogen cycling and nitrogen assimilation genes (nirK, nirS, nosZ, nifH, nasC/nasA, nirA, nirB, nirD, hao, amoA, gltB, gltS and glnA) across the three groundwater clusters (Fresh_S, Fresh_D and Saline). Taxonomic assignment was performed at the order level. The 30 most abundant orders across all genes and groundwater clusters are shown individually, whereas the remaining orders were grouped as “Other orders”. Percentage values are displayed only for orders representing more than 5% of the predicted gene abundance within each donut. Orders are arranged clockwise from the most to the least abundant, with “Other orders” always shown last. Numbers at the center of each donut indicate the total number of ASVs predicted to contribute to each gene within each groundwater cluster.Figure S7: Distribution of anammoxrelated genera. Relative contribution (% of total reads) of Scalindua (a) and Brocadia (b) genera (anammoxassociated genera) of identified ASVs, separated by cluster and colourcoded by sampling campaign. Fresh_S and Fresh_D indicate Fresh shallow and FreshDeep groundwater clusters, respectively.Figure S8: Richness and diversity of KEGG Orthologs (KOs) across different systems. (a) Boxplot showing KO richness in Soil, STE, and Seawater samples, based on a total of 10,106 KOs predicted by PICRUSt2. (b) KO diversity within nitrogen related KOs classified under the “Nitrogen metabolism” pathway in the KEGG database (46 KOs in total, of which 39 were detected in our dataset). (c) KO diversity and (d) number of ASVs (taxonomic richness) within 13 selected nitrogenrelated genes included in this study (nirK, nirS, nosZ, nifH, amoA, hao, nirB, nirD, glnA, gltB, gltS, nasA, and nirA). Diversity in panels b and c was calculated using the Shannon index. Opencoloured circles indicate sample values, while dots represent statistical outliers.Figure S9: Distribution of nitrogencycle gene abundances quantified by qPCR. Abundance of amoA, nirK and nirS qPCRbased abundances, grouped by cluster and colourcoded by campaign. Fresh S and Fresh D indicate Freshshallow and Freshdeep groundwater clusters, respectively. Statistical differences among clusters (blue) were assessed using pairwise Wilcoxon ranksum test for independent samples. Differences between sampling campaigns within each cluster (red) were evaluated using paired Wilcoxon signedrank test on samples collected from the same wells. Significance levels are indicated as p<0.05 (*) and p<0.01 (**).FigureS10: Relationship between groundwater physicochemistry and nitrogencycle gene abundances. (a) Heatmap of Spearma's correlation coefficients (rho) between gene abundances and measured variables, including nutrients, DOC and physicochemical parameters, considering all groundwater samples together. Positive correlations are indicated in blue and negative in red, with color intensity proportional to rho values. Gene abundance data source (PICRUSt2 or qPCR) is indicated in the bottom axis. Asterisks indicate significance levels: ***p<Bonferroniadjusted threshold; **p<0.01; *p<0.05. (b) Principal coordinates analysis (PCoA) based on BrayCurtis dissimilarities of qPCRmeasured amoA, nirK and nirS copy numbers, with nutrient variables fitted using envfit. (c) PCoA based on BrayCurtis dissimilarities of PICRUSt2predicted read abundances for the 13 most abundant nitrogencycle genes, overlaid with envfitfitted nutrient vectors. In panels (b) and (c), points are colored by sample cluster and shape indicates sampling campaign. Nutrients marked with an asterisk (*) denote significant (p<0.05) envfit correlations.Figure S11: Ratio of predicted DNRA to denitrification genes. The ratio between predicted DNRA (nirB and nirD, encoding nitrite reductases that reduce NO2 to NH4+) and denitrification (nirS and nirK, encoding nitrite reductases that reduce NO2 to NO) gene abundance, inferred using PICRUSt2, across environmental clusters. Boxplots are coloured by sampling campaign. Fresh_S and Fresh_D indicates Freshshallow and FreshDeep groundwater clusters, respectively.Figure S12: Coverage and taxonomic analysis of primers targeting amoA, nirS, and nirK genes compared to predictions from PICRUSt2. (a) Bar plots showing the number of genomes predicted by PICRUSt2 encoding amoA, nirS, or nirK (left bars), compared to the number of those genomes detected by specific primers for each gene (right bars). (b) Stacked bar plots showing the percentage of ASV reads (coloured by phylum) captured by the qPCR primers over all the ASV reads predicted by PICRUSt2 to carry amoA, nirS, or nirK, across shallow fresh groundwater (Fresh_S), groundwater (Fresh_D), deep and fresh saline groundwater samples in winter and summer. Grey areas indicate predicted genecarrying ASVs not detected by the primers. (c) Boxplots showing the relative abundance (% of total community reads) of ASVs associated with amoA, nirS, or nirK genes that were captured by the primers, for each habitat (Fresh_S, Fresh_D, Saline) and season (winter, summer). Only ASVs identified by PICRUSt2 predictions were included to estimate total community reads.Figure S13: Observed versus predicted concentrations of dissolved nitrogen and carbon compounds in groundwater samples (n=38), based on the best multiple linear regression (MLR) models indicated in TableS5. Models included only environmental predictors (a), microbial predictors (b) or a combination of both (c). Concentrations of NH4+ were log10 transformed. The dashed line represents the 1:1 relationship. Colours indicate the hydrogeochemical sample clusters.TableS1: List of the 13 nitrogen related KEGG Orthologs (KOs) selected for the spatiotemporal analysis, including the genes, full names, functional pathways, reaction, and corresponding KO identifiers. Note that the KO K10944 is shared between ammonia monooxygenase (amo) and particulate methane monooxygenase (pmo). Consequently, PICRUSt2 cannot differentiate amoA from pmoA solely based on KO (but see Discussion).TableS2: Primers used to target nitrogencycle genes and the synthetic internal positive control (IPCA). Columns lists: Ncycle process (biogeochemical step targeted), gene, primer name, primer sequence (53), amplicon size (bp, expected product length), final qPCR reaction mix (20L), qPCR cycles and reference (original publication source). Primer sequence share reported in 53 orientation and employ standard IUPAC nucleotide ambiguity codes for degenerate positions. BSA refers to bovine serum albumin.TableS3: gBlock sequences used in this study. For each entry the table lists: Gene gBlock, sequence (53), size (bp) and clone reference (citation and accession number). Sequences are shown in 53orientation. Uppercase letters correspond exactly tot the clone sequence deposited in GenBank. Lowercase letters denotate synthetic nucleotides appended to the 5 and 3 ends of the gBlocks to facilitate primer binding; these appended bases are not part of the reference clone sequence.TableS4: Coverage of commonly used PCR/qPCR primer pairs targeting amoA, nirS, and nirK genes, analysed using PrimerEvalPy (VázquezGonzález etal. 2024). The table reports the performance of each primer pair (left column) in terms of gene coverage across predicted ASVs from PICRUSt2. The primerpairs used in this study are indicated in bold. The “Coverage by number of ASVs” (middle column) indicates the percentage of ASVs targeted by each primer pair relative to the total number of ASV predicted to carry the respective gene. The “Coverage by reads” (right column) reflects the proportion of total reads assigned to those ASVs captured by each primer pair, relative to the total reads from all the genecontaining ASVs retained by PICRUSt2 (for details on the analysis, see Supplementary Methods).TableS5: Best multiple linear regression (MLR) models predicting the concentrations of dissolved inorganic and organic nitrogen and carbon within the STE (n=38 groundwater samples). Models included only environmental predictors (top table), microbial predictors (middle table) and a combination of both (bottom table). Models were selected using forward stepwise selection based on the Akaike Information Criterion (AIC). The sign in parentheses denotes the sign of the coefficient in the MLR model. Prok. Abund., prokaryotic abundance; %HNA, % of HNA cells. Variables in bold indicate significant predictors; however, some nonsignificant predictors were also retained in certain cases when they improved the AIC. Rect: 115, 513, 127, 502 ActionDest: 603 Page: Sequence: 16 Annotations: Annotation: Subtype: Link Contents: Louca, S., A. K.Hawley, S.Katsev, etal. 2016. “Integrating Biogeochemistry With Multiomic Sequence Information in a Model Oxygen Minimum Zone.” Proceedings of the National Academy of Sciences113, no. 40: E5925 E5933. https://doi.org/10.1073/pnas.1602897113. Rect: 246, 579, 263, 566 ActionDest: 510 Annotation: Subtype: Link Contents: Reed, D. C., C. K.Algar, J. A.Huber, and G. J.Dick. 2014. “GeneCentric Approach to Integrating Environmental Genomics and Biogeochemical Models.” Proceedings of the National Academy of Sciences of the United States of America111, no. 5: 1879 1884. https://doi.org/10.1073/pnas.1313713111. Rect: 67, 567, 84, 555 ActionDest: 559 Annotation: Subtype: Link Contents: Buchwald, C., K.Grabb, C. M.Hansel, and S. D.Wankel. 2016. “Constraining the Role of Iron in Environmental Nitrogen Transformations: Dual Stable Isotope Systematics of Abiotic NO2− Reduction by Fe(II) and Its Production of N2O.” Geochimica et Cosmochimica Acta186: 1 12. https://doi.org/10.1016/j.gca.2016.04.041. Rect: 203, 521, 221, 509 ActionDest: 465 Annotation: Subtype: Link Contents: Erler, D. V., I. R.Santos, Y.Zhang, etal. 2014. “Nitrogen Transformations Within a Tropical Subterranean Estuary.” Marine Chemistry164: 38 47. https://doi.org/10.1016/j.marchem.2014.05.008. Rect: 109, 464, 126, 451 ActionDest: 465 Annotation: Subtype: Link Contents: Otwell, A. E., A.López García de Lomana, S. M.Gibbons, M. V.Orellana, and N. S.Baliga. 2018. “Systems Biology Approaches Towards Predictive Microbial Ecology.” Environmental Microbiology20, no. 12: 4197 4209. https://doi.org/10.1111/14622920.14378. Rect: 179, 464, 196, 451 ActionDest: 559 Annotation: Subtype: Link Contents: DiegoFeliu, M., V.Rodellas, A.AlordaKleinglass, M.Saaltink, A.Folch, and J.GarciaOrellana. 2022. “Extreme Precipitation Events Induce High Fluxes of Groundwater and Associated Nutrients to Coastal Ocean.” Hydrology and Earth System Sciences26, no. 18: 4619 4635. https://doi.org/10.5194/hess2646192022. Rect: 265, 452, 283, 440 ActionDest: 465 Annotation: Subtype: Link Contents: DiegoFeliu, M., V.Rodellas, A.AlordaKleinglass, M.Saaltink, A.Folch, and J.GarciaOrellana. 2022. “Extreme Precipitation Events Induce High Fluxes of Groundwater and Associated Nutrients to Coastal Ocean.” Hydrology and Earth System Sciences26, no. 18: 4619 4635. https://doi.org/10.5194/hess2646192022. Rect: 245, 395, 263, 382 ActionDest: 465 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.13039/501100011033 for your more information Rect: 497, 303, 549, 291 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.13039/501100011033 for your more information Rect: 306, 293, 333, 281 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.13039/501100011033 for your more information Rect: 388, 233, 520, 221 Page: Sequence: 17 Annotations: Annotation: Subtype: Link Contents: TableS6: Environmental, microbial and categorical variables used for most analyses in RomanoGude etal.TableS7: ASV count table used for analyses in RomanoGude etal.TableS8: Taxonomic categorization of the obtained ASVs in RomanoGude etal. Rect: 69, 706, 90, 695 ActionDest: 639 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2024.1501781 for your more information Rect: 167, 630, 289, 619 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2024.1501781 for your more information Rect: 45, 620, 72, 609 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2023JG007660 for your more information Rect: 103, 565, 234, 554 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2024.170940 for your more information Rect: 153, 510, 289, 499 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2024.170940 for your more information Rect: 45, 500, 69, 489 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1371/journal.pone.0165825 for your more information Rect: 263, 465, 289, 454 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1371/journal.pone.0165825 for your more information Rect: 45, 455, 175, 444 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fenvs.2021.653568 for your more information Rect: 169, 420, 289, 409 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fenvs.2021.653568 for your more information Rect: 45, 410, 69, 399 Annotation: Subtype: Link Contents: Please visit https://github.com/adriaaula/dada2_guidelines for your more information Rect: 263, 395, 289, 384 Annotation: Subtype: Link Contents: Please visit https://github.com/adriaaula/dada2_guidelines for your more information Rect: 45, 385, 183, 374 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1002/ecs2.3184 for your more information Rect: 208, 350, 289, 339 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1002/ecs2.3184 for your more information Rect: 45, 340, 78, 329 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2025.1532193 for your more information Rect: 45, 295, 194, 284 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.gca.2016.04.041 for your more information Rect: 168, 240, 289, 229 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.gca.2016.04.041 for your more information Rect: 45, 230, 68, 219 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1890/1540-9295(2007)5[89:HWOTRO]2.0.CO;2 for your more information Rect: 95, 185, 289, 174 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1890/1540-9295(2007)5[89:HWOTRO]2.0.CO;2 for your more information Rect: 45, 175, 75, 164 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1373/clinchem.2008.112797 for your more information Rect: 227, 140, 289, 129 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1373/clinchem.2008.112797 for your more information Rect: 45, 130, 140, 119 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1890/08-1823.1 for your more information Rect: 97, 95, 210, 84 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/nmeth.3869 for your more information Rect: 306, 741, 427, 730 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1002/lol2.70026 for your more information Rect: 426, 696, 542, 684 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.16160 for your more information Rect: 326, 650, 463, 639 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2024GL109621 for your more information Rect: 497, 615, 549, 604 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2024GL109621 for your more information Rect: 306, 605, 386, 594 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/0304-4203(95)00062-3 for your more information Rect: 420, 570, 549, 558 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/0304-4203(95)00062-3 for your more information Rect: 306, 560, 333, 548 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.5194/bg-14-3321-2017 for your more information Rect: 512, 524, 549, 513 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.5194/bg-14-3321-2017 for your more information Rect: 306, 514, 403, 503 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.tim.2016.05.004 for your more information Rect: 433, 479, 549, 468 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.tim.2016.05.004 for your more information Rect: 306, 469, 329, 458 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.5194/hess-26-4619-2022 for your more information Rect: 326, 414, 471, 403 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jhydrol.2021.126043 for your more information Rect: 306, 369, 459, 357 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41587-020-0548-6 for your more information Rect: 339, 333, 485, 322 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2019WR026554 for your more information Rect: 488, 298, 549, 287 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2019WR026554 for your more information Rect: 306, 288, 378, 277 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.marchem.2014.05.008 for your more information Rect: 306, 253, 466, 241 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.16558 for your more information Rect: 410, 207, 547, 196 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jhydrol.2020.125050 for your more information Rect: 306, 162, 460, 151 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1098/rstb.2013.0120 for your more information Rect: 390, 117, 521, 105 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41598-024-67735-x for your more information Rect: 389, 71, 538, 60 Page: Sequence: 18 Annotations: Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2022.155978 for your more information Rect: 251, 721, 289, 710 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2022.155978 for your more information Rect: 45, 711, 167, 700 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jhazmat.2023.130876 for your more information Rect: 73, 666, 230, 655 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.mimet.2004.07.002 for your more information Rect: 204, 621, 289, 609 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.mimet.2004.07.002 for your more information Rect: 45, 611, 111, 599 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2018.03343 for your more information Rect: 227, 566, 289, 554 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2018.03343 for your more information Rect: 45, 556, 124, 544 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.4319/lo.2005.50.5.1393 for your more information Rect: 85, 511, 225, 499 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2021.661201 for your more information Rect: 263, 475, 289, 464 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2021.661201 for your more information Rect: 45, 465, 166, 454 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1021/acs.est.3c05513 for your more information Rect: 92, 420, 225, 409 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.ibiod.2018.10.002 for your more information Rect: 204, 375, 289, 364 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.ibiod.2018.10.002 for your more information Rect: 45, 365, 105, 354 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/aem.63.12.4679-4685.1997 for your more information Rect: 154, 320, 289, 309 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/aem.63.12.4679-4685.1997 for your more information Rect: 45, 310, 79, 299 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/aem.49.3.599-607.1985 for your more information Rect: 98, 265, 256, 254 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/978-981-97-1844-3_3 for your more information Rect: 227, 220, 289, 209 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/978-981-97-1844-3_3 for your more information Rect: 45, 210, 135, 199 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1021/acs.est.5b00909 for your more information Rect: 110, 165, 244, 154 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3390/hydrology5040060 for your more information Rect: 65, 130, 210, 118 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2017.02360 for your more information Rect: 263, 95, 289, 83 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2017.02360 for your more information Rect: 45, 85, 160, 73 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jenvman.2025.126474 for your more information Rect: 524, 751, 549, 740 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jenvman.2025.126474 for your more information Rect: 306, 741, 440, 730 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s11157-025-09719-5 for your more information Rect: 512, 696, 549, 684 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s11157-025-09719-5 for your more information Rect: 306, 686, 417, 674 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.1602897113 for your more information Rect: 430, 640, 549, 629 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.1602897113 for your more information Rect: 306, 630, 326, 619 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41559-018-0519-1 for your more information Rect: 346, 595, 490, 583 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1186/s13568-021-01321-6 for your more information Rect: 354, 549, 502, 538 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.14806/ej.17.1.200 for your more information Rect: 524, 524, 549, 513 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.14806/ej.17.1.200 for your more information Rect: 306, 514, 401, 503 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jhydrol.2022.127510 for your more information Rect: 465, 469, 549, 457 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.jhydrol.2022.127510 for your more information Rect: 306, 459, 375, 447 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2023.1189281 for your more information Rect: 488, 423, 549, 412 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2023.1189281 for your more information Rect: 306, 413, 393, 402 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.5281/zenodo.4587955 for your more information Rect: 306, 378, 441, 366 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s12237-018-0485-5 for your more information Rect: 371, 332, 515, 321 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s10533-019-00587-0 for your more information Rect: 429, 287, 549, 276 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s10533-019-00587-0 for your more information Rect: 306, 277, 333, 266 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/S0304-4203(99)00014-6 for your more information Rect: 306, 241, 467, 230 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/feart.2021.600710 for your more information Rect: 306, 196, 448, 185 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3354/ame046141 for your more information Rect: 306, 151, 425, 139 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.earscirev.2021.103816 for your more information Rect: 524, 115, 549, 104 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.earscirev.2021.103816 for your more information Rect: 306, 105, 440, 94 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41396-022-01300-0 for your more information Rect: 512, 70, 549, 58 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41396-022-01300-0 for your more information Rect: 306, 60, 418, 48 Page: Sequence: 19 Annotations: Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.14378 for your more information Rect: 85, 696, 222, 684 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41561-022-00968-5 for your more information Rect: 183, 660, 289, 649 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s41561-022-00968-5 for your more information Rect: 45, 650, 88, 639 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.5194/hess-24-2121-2020 for your more information Rect: 208, 605, 289, 593 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.5194/hess-24-2121-2020 for your more information Rect: 45, 595, 108, 583 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.13023 for your more information Rect: 251, 549, 289, 538 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.13023 for your more information Rect: 45, 539, 145, 528 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2024JG008346 for your more information Rect: 45, 484, 177, 473 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/fwb.14152 for your more information Rect: 193, 439, 289, 427 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/fwb.14152 for your more information Rect: 45, 429, 65, 417 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2022.1064112 for your more information Rect: 135, 383, 284, 372 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.watres.2026.125629 for your more information Rect: 45, 338, 196, 326 Annotation: Subtype: Link Contents: Please visit https://www.R-project.org/ for your more information Rect: 123, 302, 216, 291 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.1313713111 for your more information Rect: 190, 257, 289, 246 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.1313713111 for your more information Rect: 45, 247, 84, 236 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.biortech.2022.127896 for your more information Rect: 129, 201, 287, 190 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s43247-025-02594-6 for your more information Rect: 45, 156, 195, 145 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.advwatres.2017.10.041 for your more information Rect: 125, 101, 287, 89 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.1419049112 for your more information Rect: 469, 751, 549, 740 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.1419049112 for your more information Rect: 306, 741, 364, 730 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2018.06.095 for your more information Rect: 433, 696, 549, 684 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2018.06.095 for your more information Rect: 306, 686, 346, 674 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/AEM.02056-09 for your more information Rect: 524, 650, 549, 639 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/AEM.02056-09 for your more information Rect: 306, 640, 414, 629 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/mec.70201 for your more information Rect: 469, 595, 549, 584 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/mec.70201 for your more information Rect: 306, 585, 342, 574 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2022.861557 for your more information Rect: 341, 530, 486, 518 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2024WR038087 for your more information Rect: 488, 494, 549, 483 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1029/2024WR038087 for your more information Rect: 306, 484, 379, 473 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1088/2515-7620/ab1695 for your more information Rect: 512, 449, 549, 438 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1088/2515-7620/ab1695 for your more information Rect: 306, 439, 410, 428 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/msystems.01125-25 for your more information Rect: 524, 394, 549, 383 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/msystems.01125-25 for your more information Rect: 306, 384, 428, 373 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1093/femsre/fuab010 for your more information Rect: 345, 339, 479, 327 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/mec.16695 for your more information Rect: 370, 293, 487, 282 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.soilbio.2025.109858 for your more information Rect: 334, 248, 485, 237 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3390/w11040804 for your more information Rect: 368, 203, 488, 191 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s10040-009-0526-z for your more information Rect: 524, 177, 549, 166 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s10040-009-0526-z for your more information Rect: 306, 167, 427, 156 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/AEM.72.3.2102-2109.2006 for your more information Rect: 345, 122, 516, 111 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/j.1462-2920.2007.01547.x for your more information Rect: 488, 77, 549, 65 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/j.1462-2920.2007.01547.x for your more information Rect: 306, 67, 410, 55 Page: Sequence: 20 Annotations: Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.gca.2008.11.029 for your more information Rect: 237, 721, 289, 710 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.gca.2008.11.029 for your more information Rect: 45, 711, 132, 700 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s43017-021-00152-0 for your more information Rect: 208, 676, 289, 665 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1038/s43017-021-00152-0 for your more information Rect: 45, 666, 113, 655 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2019.00760 for your more information Rect: 251, 631, 289, 620 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmicb.2019.00760 for your more information Rect: 45, 621, 148, 610 Annotation: Subtype: Link Contents: Please visit https://www.semanticscholar.org/paper/A-simple%2C-economical-method-for-measuring-bacterial-Smith-Azam/de4d1a5b9a27b6ed7f982976c60c85175a4e08b7 for your more information Rect: 80, 586, 289, 575 Annotation: Subtype: Link Contents: Please visit https://www.semanticscholar.org/paper/A-simple%2C-economical-method-for-measuring-bacterial-Smith-Azam/de4d1a5b9a27b6ed7f982976c60c85175a4e08b7 for your more information Rect: 45, 576, 289, 565 Annotation: Subtype: Link Contents: Please visit https://www.semanticscholar.org/paper/A-simple%2C-economical-method-for-measuring-bacterial-Smith-Azam/de4d1a5b9a27b6ed7f982976c60c85175a4e08b7 for your more information Rect: 45, 566, 143, 555 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.gca.2008.04.027 for your more information Rect: 204, 521, 289, 510 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.gca.2008.04.027 for your more information Rect: 45, 511, 100, 500 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2024.171473 for your more information Rect: 263, 466, 289, 455 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2024.171473 for your more information Rect: 45, 456, 179, 445 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.3389/fmars.2022.822939 for your more information Rect: 73, 411, 219, 400 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.femsec.2004.04.011 for your more information Rect: 135, 366, 287, 355 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/j.1462-2920.2007.01563.x for your more information Rect: 85, 321, 251, 310 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1186/s12859-024-05805-7 for your more information Rect: 166, 276, 289, 265 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1186/s12859-024-05805-7 for your more information Rect: 45, 266, 71, 255 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1128/AEM.00062-07 for your more information Rect: 108, 221, 241, 210 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1002/lol2.10390 for your more information Rect: 237, 186, 289, 175 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1002/lol2.10390 for your more information Rect: 45, 176, 109, 165 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.16504 for your more information Rect: 171, 141, 289, 130 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1111/1462-2920.16504 for your more information Rect: 45, 131, 65, 120 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1002/lno.11576 for your more information Rect: 172, 86, 286, 75 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1093/bioinformatics/btaf269 for your more information Rect: 469, 751, 549, 740 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1093/bioinformatics/btaf269 for your more information Rect: 306, 741, 385, 730 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.marpolbul.2021.112440 for your more information Rect: 427, 696, 549, 685 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.marpolbul.2021.112440 for your more information Rect: 306, 686, 349, 675 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1007/s13131-023-2257-8 for your more information Rect: 321, 641, 464, 630 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2018.04.143 for your more information Rect: 512, 606, 549, 595 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1016/j.scitotenv.2018.04.143 for your more information Rect: 306, 596, 425, 585 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.96.4.1463 for your more information Rect: 512, 561, 549, 550 Annotation: Subtype: Link Contents: Please visit https://doi.org/10.1073/pnas.96.4.1463 for your more information Rect: 306, 551, 399, 540 Annotation: Subtype: Link Contents: Please see "figure 1a" for your more reference Rect: 522, 481, 529, 470 ActionDest: 220 Page: Sequence: 21 Annotations: Annotation: Subtype: Link Contents: VázquezGonzález, L., A.RegueiraIglesias, C.BalsaCastro, N.VilaBlanco, I.Tomás, and M. J.Carreira. 2024. “PrimerEvalPy: A Tool for InSilico Evaluation of Primers for Targeting the Microbiome.” BMC Bioinformatics25, no. 1: 189. https://doi.org/10.1186/s12859024058057. Rect: 325, 381, 341, 370 ActionDest: 603 Checksum: e68b58f6 Type: CRC32 Checksum: e5ca2112a65a874dfb394a0e8cc3f35d Type: MD5 Checksum: 6e1118ca73f471d0efae5b4dfaa3e63e80bfd7bc Type: SHA-1 Checksum: 6e17de77be6e8cf23ce4fc0e336b756003f446ceae141ed8359bb8e15c0f6fb0 Type: SHA-256