Visser et al. Molecular Neurodegeneration (2022) 17:27 https://doi.org/10.1186/s13024-022-00521-3 RESEARCH ARTICLE Open Access Cerebrospinal fluid tau levels are associated with abnormal neuronal plasticity markers in Alzheimer’s disease Pieter Jelle Visser1,2,3* , Lianne M. Reus1, Johan Gobom4,5, Iris Jansen6, Ellen Dicks1, Sven J. van der Lee1,7, Magda Tsolaki8, Frans R. J. Verhey2, Julius Popp9,10, Pablo Martinez‑Lage11, Rik Vandenberghe12,13, Alberto Lleó14, José Luís Molinuevo15,16, Sebastiaan Engelborghs17,18, Yvonne Freund‑Levi3,19, Lutz Froelich20, Kristel Sleegers21,22, Valerija Dobricic23, Simon Lovestone24, Johannes Streffer17,25, Stephanie J. B. Vos2, Isabelle Bos2, ADNI, August B. Smit26, Kaj Blennow4,5, Philip Scheltens1, Charlotte E. Teunissen27, Lars Bertram22,28, Henrik Zetterberg4,5,29,30 and Betty M. Tijms1 Abstract Background: Increased total tau (t-tau) in cerebrospinal fluid (CSF) is a key characteristic of Alzheimer’s disease (AD) and is considered to result from neurodegeneration. T-tau levels, however, can be increased in very early disease stages, when neurodegeneration is limited, and can be normal in advanced disease stages. This suggests that t-tau levels may be driven by other mechanisms as well. Because tau pathophysiology is emerging as treatment target for AD, we aimed to clarify molecular processes associated with CSF t-tau levels. Methods: We performed a proteomic, genomic, and imaging study in 1380 individuals with AD, in the preclinical, prodromal, and mild dementia stage, and 380 controls from the Alzheimer’s Disease Neuroimaging Initiative and EMIF-AD Multimodality Biomarker Discovery study. Results: We found that, relative to controls, AD individuals with increased t-tau had increased CSF concentrations of over 400 proteins enriched for neuronal plasticity processes. In contrast, AD individuals with normal t-tau had decreased levels of these plasticity proteins and showed increased concentrations of proteins indicative of blood– brain barrier and blood-CSF barrier dysfunction, relative to controls. The distinct proteomic profiles were already present in the preclinical AD stage and persisted in prodromal and dementia stages implying that they reflect disease traits rather than disease states. Dysregulated plasticity proteins were associated with SUZ12 and REST signaling, suggesting aberrant gene repression. GWAS analyses contrasting AD individuals with and without increased t-tau highlighted several genes involved in the regulation of gene expression. Targeted analyses of SNP rs9877502 in GMNC, associated with t-tau levels previously, correlated in individuals with AD with CSF concentrations of 591 plasticity associated proteins. The number of APOE-e4 alleles, however, was not associated with the concentration of plasticity related proteins. *Correspondence: pj.visser@amsterdamumc.nl; pj.visser@maastrichtuniversity.nl 1 Alzheimer Center Amsterdam, Department of Neurology, Amsterdam Neuroscience, Vrije Universiteit Amsterdam, PO Box 7057 1007, MB, Amsterdam, The Netherlands Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​ mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 2 of 16 Conclusions: CSF t-tau levels in AD are associated with altered levels of proteins involved in neuronal plasticity and blood–brain and blood-CSF barrier dysfunction. Future trials may need to stratify on CSF t-tau status, as AD individuals with increased t-tau and normal t-tau are likely to respond differently to treatment, given their opposite CSF prot‑ eomic profiles. Keywords: Alzheimer’s disease, Molecular mechanisms, Biomarker discovery, Heterogeneity, Neuronal plasticity, Cerebrospinal fluid proteomics Background collaboration under the supervision of Principle Inves- The amyloid cascade hypothesis poses that Alzheimer’s tigator Michael W. Weiner, MD. The primary goal of disease (AD) starts with amyloid beta (Aβ) aggregation ADNI is to study whether serial biological markers, and followed by tau pathology [1]. Increased total tau (t-tau) clinical and neuropsychological measures can be com- levels in cerebrospinal fluid (CSF) are supposed to be bined to measure progression of MCI and early AD and caused by axonal loss [2]. Still, 25% of the AD individuals has enrolled 2850 individuals, see www.​adni-​info.​org. with mild cognitive impairment (MCI) or dementia have The EMIF-AD MBD study aimed to identify markers for normal t-tau levels despite axonal loss [3, 4]. Moreover, diagnosis and prognosis of predementia AD. It combined CSF t-tau is abnormally increased in 50% of preclinical existing clinical data, samples and scans of 1218 individu- AD individuals, when neurodegeneration is limited [5]. als with normal cognition, MCI or mild dementia from Furthermore, our study in cognitively normal monozy- prospective cohort studies [12]. gotic twins discordant for amyloid aggregation, suggested that CSF t-tau levels may rise even before amyloid aggre- Group definition and staging gation reaches abnormal levels [6]. An alternative expla- We selected individuals with AD pathology defined as nation for the increase in t-tau in AD may be increased abnormally decreased CSF amyloid beta 1–42 (Aβ1-42), gene expression. A SILK study found that elevated CSF and we subdivided this group into those with abnor- t-tau levels in AD were associated with increased pro- mally increased total tau (t-tau) and those with normal duction rather than from spilling of intracellular tau by t-tau. Based on cognitive performance, AD individuals dying neurons [7]. A study with induced pluripotent stem were classified in 3 clinical stages as preclinical AD (nor- cells (iPSC) showed that AD individuals had increased mal cognition), prodromal AD (MCI) and mild AD-type tau expression in neuronal progenitor cells compared dementia according to study specific criteria [12, 13]. The to controls [8]. Increased t-tau levels may also result of control group consisted of individuals with normal cog- increased neuronal activity, [9] or increased neuronal nition, and normal CSF Aβ1-42 and t-tau levels. plasticity as t-tau has an high expression in brain regions with plastic potential [10]. As the drivers of increased Clinical assessment tau levels in AD remain largely unknown, and because Global cognition was assessed by the Mini-Mental State tau-related mechanisms are emerging as new treatment Examination (MMSE) [14] and ADAS-Cog 11-item ver- targets for AD, we aimed to clarify molecular processes sion (ADNI) [15]. As a measure for memory function we associated with CSF tau dysregulation. To this end, we used the delayed recall of the logical memory subscale performed a proteomic, genomic and imaging study II of the Wechsler Memory Scale (ADNI), [16] or center including 1380 individuals in the AD continuum, [11] specific verbal word learning tasks (EMIF-AD MBD) defined by abnormal CSF Aβ1-42, spanning the clinical [12]. We selected the Clinical Dementia Rating (CDR) spectrum from preclinical AD to mild dementia, and 380 scale sum of boxes score as a measure of disease severity controls with normal cognition and normal CSF Aβ1-42 [17]. and t-tau from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the European Medical Information CSF analysis Framework for Alzheimer’s disease (EMIF-AD) Multi- CSF samples in EMIF-AD MBD were collected according modality Biomarker Discovery (MBD) study [2]. to the BIOMARKAPD protocol, [12, 18] and in ADNI as described elsewhere [19]. Methods Study participants EMIF‑AD MBD We selected 961 individuals from ADNI (adni.loni. CSF Aβ1-42 and t-tau were measured locally with usc.edu) and 799 individuals from the EMIF-AD MBD INNOTEST ELISA or INNOBIA AlzBio3 (Fujirebio, study [12]. ADNI started in 2003 as a public–private Ghent, Belgium). Cut-offs for Aβ1-42 and t-tau were Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 3 of 16 cohort-specific in EMIF-AD MBD [12]. Aβ1-42 cut-offs MBD sample). For related proteins that had identical were determined for each cohort using Gaussian mix- values due to fragment a-specificity (e.g. ACTA1 and ture modelling [20]. We measured neurogranin, neuro- ACTA2) we randomly selected one protein for analysis. filament-light, YKL-40, Aβ38 and Aβ40 by ELISA [2]. We performed untargeted mass spectrometry using tan- Proteomic annotation dem mass tag (TMT) with 10 + 1 plexing as previously Enrichment analysis described, using high-pH reverse phase HPLC for peptide We included proteins for enrichment analyses that were prefractionation [20, 21]. Peptides were mapped to 2535 increased or decreased at p < 0.05 relative to the contrast proteins using the software ProteomeDiscoverer v.2.2. tested. (Thermo Scientific), using Mascot (MatrixScience) for protein identification (precursor Dm tolerance 15 ppm, Biological process enrichment analyses We performed fragment tolerance 0.05 Da, max missed cleavage sites pathway enrichment analyses for Gene Ontology biologi- 2), searching the human subset of the UniProtKB Swis- cal processes (GO-BP) using the online Panther applica- sProt database (www.​unipr​ot.​org). Percolator (MatrixSci- tion and SynGo [24–27]. ence) was used for scoring peptide specific matches, and a strict 1% false discovery rate (FDR) was set as threshold Transcription factor enrichment analyses We per- for identification. For reporter ion quantification the fol- formed transcription factor enrichment analysis using lowing settings were used: Integration tolerance 20 ppm; ChEA through Harmonizome through Enrichr [28–30]. Integration Method Most Confident Centroid; normalize on the reference protein average. The median (IQR) CV for these proteins was 5.6 (3.8, 8.0). Predominant cell‑type of protein production Based on the BRAIN RNASeq database (http://​www.​ ADNI brain​rnaseq.​org), [31] we labelled proteins as being pre- Aβ1-42 and t-tau were measured on the xMAP platform dominantly produced by a cell type when levels were (Luminex Corp, Austin, TX). Cut-offs for Aβ1-42 and t-tau higher than 40% of older individuals produced across cell were used as published [19]. Changes in t-tau levels over types. time were assessed in ADNI using longitudinal samples measured within the same batch. YKL-40, neurogranin, Other annotations APP beta, neurogranin, neurofilament light, alpha synu- Proteins associated with high choroid plexus expression clein, HBB, CFH, sTREM2, VILIP-1, and BACE1 were were defined based on the Allen Brain Atlas, [32] through measured by ELISA or related assays, 190 analytes were Harmonizome [29] and ABAEnrichment analysis [33]. analyzed by the Human DiscoveryMAP panel (MAP- Blood–Brain barrier (BBB) associated proteins were RBM) and 225 proteins were analyzed by targeted mass selected from reference [34–36]; BACE1 substrates were spectroscopy [22, 23]. We used the quality checked and selected from reference [37, 38]; Alpha secretase sub- finalised ‘Normalized Intensity’ data as described in strates were selected from reference [39, 40]; and gamma https://​ a dni.​ l oni. ​ u sc. ​ e du/ ​ w p- ​ conte ​ nt/ ​ u ploa ​ d s/ ​ 2 012/​ secretase substrates were selected from reference [41]. 01/​2011D ​e c28-​Bioma​rkers-​Conso​r tium-​Data-​Primer-​ FINAL1.​pdf Proteomic process scores (PPS) Proteins and protein fragments (ADNI) values were For illustration purposes we combined proteins from standardized according the mean and standard deviation selected GO-BPs, and proteins associated with BBB dys- values of the control group and expressed as z-scores, function and BACE1 substrates into a PPS by averaging with a score of 0 indicating the average concentration z-scores of individual proteins belonging to the process of the control group, z-scores > 0 higher concentrations, that differed between AD individuals with normal or and z-scores < 0 lower concentrations than controls. For increased t-tau in cross-sectional analysis in the total ADNI, we averaged peptides that mapped to the same group. protein into a composite protein score when they cor- related with r > 0.5, and included them as single peptides Neuroimaging analysis otherwise [20]. When the same protein was measured by We studied neuroimaging data only from ADNI, because different platforms in ADNI, values were averaged if they this study collected longitudinal scans. As a measure of correlated with r > 0.5 and otherwise we included them brain atrophy, we took cortical thickness data from 34 as separate proteins. From the EMIF-AD MBD TMT cortical areas from the longitudinal processing pipeline in mass spectrometry analysis proteins were included when Freesurfer version 4.3 for 1.5 T T1-weighted MRI scans, observed in at least 92 individuals (30% of EMIF-AD and v5.1 for 3 T T1-weighted MRI scans (http://​adni.​ Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 4 of 16 loni.​usc.​edu). As a measure of amyloid accumulation, we of the meta-analysis summary statistics. We considered used region-specific SUVR values for florbetapir bind- for further analysis gene scores based on at least 6 SNPs ing assessed by PET imaging in 34 cortical areas as par- and geneset scores with information on at least 6 genes. cellated by Freesurfer v4.5.0 [42]. As a measure of brain metabolism we analysed fluorodeoxyglucose (FDG)-PET scans, [43] and determined average glucose metabolism Polygenic and other genetic risk score analysis for brain areas standardized to the average uptake in the Polygenic risk scores (PGRS) for AD were computed for vermis and brain stem according to the Desikan-Killiany each subject using PRSice (V2.3) [48]. PGRS were cal- atlas (http://​adni.​loni.​usc.​edu/​metho​ds/​pet-​analy​sis-​ culated by adding the sum of each allele weighted by method/​pet-​analy​sis/). the strength of its association with AD risk as calculated previously [49]. AD cases were defined as patients clini- cally diagnosed with AD-type dementia or individuals Genomic assessment with a parental history of AD (i.e., AD-by-proxy) [49]. EMIF-AD MBD samples were genotyped using the Illu- Clumping was performed prior to calculating PGRS, to mina Global Screening array (Illumina, Incl) and 936 remove SNPs that are in linkage disequilibrium (r2 > 0.1) samples passed post-experiment QC [44]. ADNI samples within a slicing 1 M bp window. After clumping, PGRS were genotyped using the Illumina 2.5-M or Illumina were computed using various SNP inclusion thresholds. OmniQuad array [45]. In order to explore how specific genetic alterations were associated with CSF proteomic profiles, we calculated for Genotype imputation and quality control a selection of genesets based on MAGMA analysis (see Data processing and quality control was performed using results), gene-specific risk scores (GRS) including only GenomeStudio Software (v2.0.04, Illumina, Inc.), as SNPs located on the respective genes with clumping as described previously [44]. To identify ethnic outliers, a described above and computation of GRS using various principal component analysis (PCA) of ancestry was per- SNP inclusion thresholds. We used weights from one formed based on 1000 Genomes clustering, phase 3 using cohort to generate GRS in the other cohort. In order to PLINK [44]. Individuals of non-European descent and reduce the number of tests, GRS scores were selected for family relations up to second degree were excluded. After SNP inclusion threshold that best differentiated within filtering, the ADNI genotype data included 747 individu- AD between abnormally increased and normal t-tau. als and EMIF-AD MBD 931 individuals. SNPs were locally imputed using Minimac 3 to the Haplotype Reference Consortium reference panel [46]. Statistics To account for population structure, principal com- Our main outcomes are the biological process and tran- ponents (PC1-PC20) were computed on a subset of scription enrichment analyses for which we report FDR relatively uncorrelated (r2 < 0.2) single nucleotide poly- corrected p-values. For other analysis we report uncor- morphisms (SNPs). For ADNI subjects, Apolipopro- rected p-values. Test statistics can be found in the sup- tein E (APOE) ε4 genotype was assessed with two SNPs; plemental tables. Group comparisons between AD rs429358 the ‘ε4-allele’ and rs7412 the ‘ε2 allele’. APOE subgroups and controls were performed using ANOVA ε4 genotypes in EMIF-AD MBD were generated as correcting for age and gender. Change in cognition and described elsewhere [44]. imaging markers were assessed with linear mixed models including as main terms group, time, and the interaction GWAS and post‑GWAS analyses group by time, and correcting for age and gender, and Genome-wide association study (GWAS) analyses on additionally for level of education for cognitive markers. increased CSF t-tau in subjects with abnormal Aβ1-42 We investigated within the AD individuals with normal were performed separately in ADNI (abnormal Aβ1-42 t-tau and AD individuals with increased t-tau whether and increased t-tau n = 246, abnormal Aβ1-42 and nor- protein concentrations increased or decreased with mal t-tau n = 238) and EMIF-AD MBD (abnormal Aβ1-42 increasing disease severity ranging from preclinical AD, and increased t-tau n = 294, abnormal Aβ1-42 and nor- prodromal AD, to mild AD-dementia with linear models. mal t-tau n = 155) using PLINK software (V1.90). We A protein was considered to decrease or increase with used a logistic regression model including PC1-PC3 and disease severity if the linear trend was statistically signifi- sex. Genome-wide significance was defined as p ≤ 5e-08. cant or if there was a difference in concentration between Meta-analysis on ADNI and EMIF-AD MBD GWAS preclinical AD and prodromal AD, preclinical AD and summary statistics was performed using METAL [47]. mild AD-dementia, or between prodromal AD and MAGMA software was used to compute gene scores and mild AD-dementia and the trend analysis supported the geneset scores for biological pathways based on p-values change with severity. Associations between GRS scores Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 5 of 16 Fig. 1 Protein concentrations relative to control group in individuals with AD according to t-tau status. Protein concentrations of individuals with abnormal Aβ1-42 and increased t-tau (a + t + , red) and abnormal Aβ1-42 and normal t-tau (a + t-, blue) in ADNI A and EMIF-AD MBD B. Concentrations are expressed as z-score relative to the control group (CN) that had normal cognition, normal Aβ1-42 and normal t-tau. Proteins are sorted according to change relative to control group. Shown are proteins that differed between individuals with AD with increased t-tau or AD individuals with normal tau from controls. Venn diagram shows number of proteins that differed and CSF protein levels were tested with linear models decreased levels of 2 proteins in ADNI (1% of proteins adjusted for age and sex. measured) and of 43 proteins in EMIF-AD MBD (3% of proteins measured) (Fig. 1, Data S2). AD individuals Results with normal CSF t-tau, showed an opposite pattern with Of the individuals with AD, 788 (57%) had increased CSF increased levels of only 7 proteins in ADNI (3% of pro- t-tau levels and 592 (43%) normal t-tau levels. Abnormal teins measured) and of 60 proteins in EMIF-AD MBD and normal t-tau groups showed generally similar demo- (4% of proteins measured) and decreases in 107 proteins graphics and baseline cognitive performance (Data S1a). in ADNI (43% of proteins measured) and of 411 proteins We first determined stability of t-tau levels over time in EMIF-AD MBD (28% of proteins measured). Around in 499 ADNI participants who had repeated CSF sam- 50% of the proteins with increased levels in individuals pling up to 7 years. CSF t-tau increased in the prodromal with increased t-tau had decreased levels in individu- AD stage at similar rates for both t-tau groups, and this als with normal t-tau. This indicates that t-tau groups increase did not differ from the increase in t-tau in the show disruptions in the same molecular processes, but in control group (Figure S1a). At last follow-up, 20% of the opposite directions. AD individuals with normal t-tau at baseline developed Enrichment analyses with GO-BP for CSF proteins abnormally increased t-tau at follow-up and these indi- increased in the AD group with abnormal t-tau rela- viduals had t-tau levels just below the cut-point at base- tive to controls, showed in both ADNI and EMIF-AD line (Figure S1b). Taken together, these findings suggest MBD involvement of plasticity-related processes such that CSF t-tau levels reflect a trait rather than a stage. as nervous system development, axonogenesis, syn- We then studied CSF proteome profiles for t-tau sub- apse assembly, myelination, gliogenesis, angiogenesis, groups in 559 individuals with proteomic data (Data mitogen-activated protein kinase (MAPK) signaling, S1b). This included targeted proteomics of 248 proteins cell-cycle, gene expression and glycolysis (Fig. 2 selec- in ADNI and targeted and untargeted proteomics of tion of representative processes; Figure S2 shows 1458 proteins in EMIF-AD MBD (Data S2). Compared synaptic processes; Data S3a, S3b show all enriched to controls, AD individuals with increased CSF t-tau had processes). Abnormality in each of these processes has increased levels of 130 proteins in ADNI (52% of ADNI been reported before in separate studies, [8, 50–55] proteins measured) and of 477 proteins in EMIF-AD and our findings now suggest that they are part of com- MBD (33% of EMIF-AD MBD proteins measured) and mon plasticity response. We also observed involvement Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 6 of 16 Fig. 2 Association of selected GO biological processes and transcription factors with CSF proteins, MAGMA geneset score, and cell type. A,B GO-BP and SUZ12 and REST transcription factors enriched for proteins in AD individuals with increased t-tau (a + t +) or normal t-tau (a + t-) with increased (red) or decreased (blue) concentrations relative to controls in the total sample (A) or in the preclinical AD stage (B). P-values of all GO-BPs are listed in Data S3a. C P-value of association of GO-BP with abnormal t-tau in GWAS-based MAGMA genset analysis. P-values of all GO-BPs are listed in Data S4d. D Proportion of proteins with cell-type specific expression. A: ADNI cohort, E: EMIF-AD MBD cohort; A + t + : abnormal Aβ1-42 and increased t-tau; A + t-: abnormal Aβ1-42 and normal t-tau; CN A + t + : abnormal Aβ1-42 and increased t-tau in preclinical AD stage; CN A + t-: abnormal Aβ1-42 and normal t-tau in preclinical AD stage. Oligo = oligodendrocyte; endothelial = endothelial cell. (protein enrichment) and Data S4d (MAGMA analysis) of cytokine signaling, leukocyte activation, oxidative through histone acetylation [58, 59]. Previous studies stress, mitochondrial dysfunction, and apoptosis. Amy- showed evidence of REST/SUZ12 de-repression in AD loid production was increased, reflected by increased brain tissue, iPSC neurons from individuals with sporadic levels of amyloid precursor protein (APP), Aβ1-40, and AD, and tangle bearing AD neurons [8, 60, 61]. Interest- Aβ1-38, and increased levels and activity of BACE1, and ingly, of the proteins with increased gene expression in increased concentrations in substrates of the main these previous studies that were measured in our CSF secretases involved in Aβ metabolism (i.e., BACE1, alpha study, the majority showed increased CSF concentrations and gamma secretase; annotated in Data S2, column in AD individuals with increased t-tau: 4 of 4 (CALB1, DA-DC). Increased Aβ production is known to set off the NRXN3, SCN3B, SNAP25) proteins from reference [60]; amyloid cascade in autosomal dominant AD, [56] and our 56 of 67 (84%) proteins from reference [8]; and 200 of 238 data suggest that increased amyloid production may also (84%) proteins from reference [61] (annotated in data S2, play a role in sporadic AD [57]. To identify potential driv- column DF-DG). ers of increased protein levels we performed CHeA tran- AD individuals with normal t-tau showed, relative to scription factor-binding site enrichment analysis, which the control group, decreased levels of proteins associ- indicated SUZ12 (p-FDR corrected = 1.62E-11) and ated with neuronal plasticity and regulation of MAPK REST (p-FDR corrected = 1.04E-9) as most significantly signaling, with concomitant lower levels of APP, Aβ1-40 enriched. SUZ12 and REST repress gene transcription and Aβ1-38, and secretase substrates (Fig. 2, Figure S2; Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 7 of 16 Data S3a, S3b for all enriched processes). Proteins with proteins were enriched for the same processes that dif- decreased concentrations in the AD group with normal fered for each group relative to controls (Figure S3, Data t-tau, also converged on SUZ12 and REST transcription S3a). Analysis of cell-type specific proteins revealed that factors, suggesting that increased gene repression activ- proteins of all major brain cell types differed in con- ity is driving this CSF proteomic profile. Furthermore, centration between AD individuals with increased and the AD group with normal t-tau showed 67 proteins with normal t-tau and that plasticity proteins were typically increased levels relative to controls. Twenty of these pro- neuron-specific (Fig. 2, Data S2). teins were associated with blood–brain barrier (BBB) We then investigated whether the opposite protein pat- dysfunction (annotated in Data S2, column DD) and 25 terns could already be detected in the preclinical stage proteins had a high expression in choroid plexus (ABAE- of AD. We found that relative to controls, preclinical nrichment minimum pFWER = 0.004), suggesting blood- AD individuals with increased t-tau had higher levels of CSF barrier (BCSFB) dysfunction (annotated in Data S2, plasticity proteins while preclinical AD individuals with column DE). Increased proteins were mainly seen in the normal t-tau showed lower plasticity protein levels. Pre- EMIF-AD MBD cohort, as ADNI proteomics was tar- clinical AD individuals with normal t-tau also showed geted for brain specific proteins. higher concentrations of proteins associated with bar- We continued analysis by comparing CSF protein con- rier dysfunction (Table 1, Fig. 2b, Fig. 3, Data S2 column centrations between AD individuals with increased t-tau BK-BN, Data S3a for all enriched GO-BP). These findings and normal t-tau directly. AD individuals with increased indicate that both aberrant neuronal plasticity processes t-tau showed, relative to those with normal t-tau, an and barrier dysfunction are very early events in AD. increase in the concentration of 167 proteins in ADNI We next investigated how protein concentrations (67% of ADNI proteins measured) and of 768 proteins in changed with increasing disease severity stages. In AD EMIF-AD MBD (53% of EMIF-AD MBD proteins meas- individuals with increased t-tau levels, the concentra- ured) and a decrease in the concentration of 2 proteins tion of 59 (ADNI) to 129 proteins (EMIF-AD MBD) in ADNI (1% of proteins measured) and of 136 proteins decreased with increasing disease severity and the con- in EMIF-AD MBD (9% of proteins measured). These centration of 2 (ADNI) to 34 proteins (EMIF-AD MBD) Table 1 Top 20 proteins that differed in preclinical AD from the control group according to CSF t-tau status Contrast Cohort Number of Top 20 proteins with largest effect size proteins that differ Preclinical AD with increased t-tau Proteins with increased concentration relative to CN EMIF-AD MBD 262 YWHAH, LAMP5, PCSK5, CHI3L1, SMOC1, ADCYAP1, SPP1, GDA, CRYM, TAGLN3, PLXNA2, PCDH8, HPRT1, CPD, GAP43, ENPP5, CAMK2A, PKM, MELTF, NAXE ADNI 93 NRGN, PKM, VSNL1, GOT1, ALDOA, SPP1, GOT2, HGF, NEO1, NCSTN, MOG, APLP2, SOD1, BACE1, APP, CHI3L1, FABP3, ENO2, NCAM2, SPON1 Proteins with decreased concentration relative to CN EMIF-AD MBD 19 S100A6, ADAMDEC1, GPLD1, ANXA5, KNG1, SERPINA4, HSPA1A, IGHV4-30–2, CPB2, IGHV4-30–4, FLNA, IGHV3-30, PRAP1, IGLV3-10, HSPA6, ABI3BP, IGHV2-70D, HSPA7, PON1 ADNI 2 GOLM1, LEP Preclinical AD with normal t-tau Proteins with increased concentration relative to CN EMIF-AD MBD 36 SLC39A12b, ­ADIPOQa, ANGPTL7, L­ GI1b, ­CD9b, KRT24, I­FI30a, TTR​b, ­FOLR1b, NCMAP, ­SLC5A5b, SELPLG, E­ NPP2b, ­F5b, CTSA, KRT12, KRT9, ­SIAEb, COL15A1, KRT10 ADNI 3 EDN1, APOE-e4, ­FGF4b Proteins with decreased concentration relative to CN EMIF-AD MBD 102 HS6ST1, CDH9, ASTN1, PLXNA1, APP, PMP2, STMN3, CDH8, PRKAR1A, GABARAPL2, GALNT6, PITHD1, GALNT1, CPM, MMP17, FAM19A2, CD99, FURIN, TAGLN, CDH12 ADNI 24 abeta1-40, abeta1-38, ADGRL1, CADM3, NEO1, NPTX1, MCAM, CHGB, PCSK1, NEGR1, L1CAM, UBB, PTPRN, CAC‑ NA2D1, TIMP1, PAM, BTD, VEGFA, DAG1, NBL1 T-tau Total tau, CN Control group with normal cognition, normal Aβ1-42 and normal t-tau a Protein associated with Blood Brain Barrier dysfunction b Proteins with high expression in choroid plexus. Full list of proteins is provided in Data S2, column BK-BN Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 8 of 16 Nervous system development Gene expression Cytokine-mediated signalling Leukocyte activation 4 4 4 4 3 3 3 3 2 2 2 2 1 1 1 1 PPS PPS PPS PPS 0 0 0 0 -1 -1 -1 -1 -2 -2 -2 -2 -3 -3 -3 -3 -4 -4 -4 -4 CN MCI Dem CN MCI Dem CN MCI Dem CN MCI Dem Angiogenesis Myelination Blood-brain barrier BACE1 substrates 4 4 4 4 3 3 3 3 2 2 2 2 1 1 1 1 PPS PPS PPS PPS 0 0 0 0 -1 -1 -1 -1 -2 -2 -2 -2 -3 -3 -3 -3 -4 -4 -4 -4 CN MCI Dem CN MCI Dem CN MCI Dem CN MCI Dem Fig. 3 Association of protein process score (PPS) with disease severity. Data are based on cross-sectional analysis and combines data from ADNI and EMIF-AD MBD. In grey PPS of control group with normal cognition and normal CSF Aβ1-42 and t-tau; In red PPS of AD individuals with increased t-tau; In blue PPS of AD individuals with normal t-tau. CN = cognitively normal; MCI = mild cognitive impairment; Dem = dementia increased. Proteins of which the concentration decreased Presence of the APOE ε4 allele, the major genetic AD with disease severity were enriched for neuronal plastic- risk factor, was more common in AD individuals with ity and proteins of which the concentration increased increased t-tau compared to AD individuals with nor- were enriched for mitochondrial outer membrane per- mal t-tau (66% vs 53% p < 0.001) and was in both AD meabilization and complement activation (Fig. 3, Data groups more common than in controls (17%, p < 0.001). S2, S3a (all enriched GO-BP), Figure S3). In AD individu- Compared to controls, AD individuals with increased als with normal t-tau, the concentration of 48 (ADNI) to t-tau and with normal t-tau had both higher AD PGRS 265 (EMIF-AD MBD) proteins decreased with increas- across SNP inclusion thresholds (Fig. 4a, Data S4a) ing disease severity and the concentration of 10 proteins [49]. PGRS did not differ between AD individuals with (ADNI) to 54 (EMIF-AD MBD) increased with disease increased t-tau and normal t-tau after stratification for severity (Fig. 3, Data S2, S3a (all enriched GO-BP), Fig- clinical stage and correction for APOE ε4 carriership ure S3). Proteins of which the concentration decreased and age (Data S4a), suggesting that these groups have with disease severity were enriched for neuronal plas- a similar AD genetic risk architecture. Next, we per- ticity and proteins of which the concentration increased formed an exploratory GWAS within AD individuals to proteins included BBB related proteins (n = 32) and pro- identify other potential genetic markers associated with teins with increased expression in the choroid plexus increased t-tau. We found tentative associations with (n = 14). Because the concentration of plasticity pro- SNPs in the APBB2 gene (nominal p-value 1.63–2.92E- teins decreased in both AD individuals with increased 06, Data S4b). The top 3 genes associated with increased t-tau and normal t-tau, differences in plasticity proteins t-tau from gene-based analyses included, in addition to between the groups that were present in the preclinical APBB2, TBC1D10B (nominal p = 2.9E-04), a Rab GTPA AD stage persisted in the prodromal and dementia stage activating protein involved in MAPK signaling, and (Fig. 3). The decrease in concentration of neuronal plas- LRP3 (nominal p = 3.36E-04), a low-density lipoprotein ticity proteins with disease severity aligns with the obser- receptor protein (Data S4c). The top 3 GO-BP gene-sets vation of decreased gene expression of neuronal plasticity associated with increased t-tau from MAGMA analysis proteins with increasing disease severity in post-mortem were positive regulation of cellular process (self-con- AD studies [57, 62]. tained p = 1.07E-04), protein K29-linked ubiquitination Next, we investigated whether genetic factors were (self-contained p = 1.52E-04) and negative regulation related to increased t-tau in AD individuals in the com- of actin nucleation (self-contained p = 1.57E-04, Data bined ADNI and EMIF-AD MBD cohorts (n = 1067). S4d). Twenty-eight percent of GO-BPs enriched for Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 9 of 16 Fig. 4 Genetic analysis. A PGRS across SNP thresholds SNP weights were based on [49]. Full data (including data after correction for APOE ε4 genotype and demographics, and separate for cohort and clinical stage) are shown in Supplemental Data S4a. a + t + : abnormal Aβ1-42 and increased t-tau (red); a + t-: abnormal Aβ1-42 and normal t-tau (blue). B Number of proteins associated with genetic risk score (GRS) and processes associated with these proteins. Left site shows number of proteins with a positive correlation with CSF t-tau GWAS-based GRS (coloured outer ring) and right site indicates to which GO-biological processes (GO-BP) these proteins below (GO-BP CSF, grey outer ring). Shown are the 10 GRS with the highest number of positive correlations with CSF proteins. The size of the arrows visualizes the number of correlations between GRS and protein concentration. For example, the GO-BP gene expression GRS showed a positive correlation with 256 proteins and these proteins were predominantly associated with nervous system development, gene expression and cell adhesion. Arrows to CSF GO-BP are shown in case the GRS correlated with at least 8 proteins from the CSF GO-BP. Supporting data in Data S5a. ECM = Extracellular matrix CSF proteins that differed in concentration between lowest self-contained p-value. For each of the GRS, we AD t-tau groups, differed also in GO-BP gene-set score used weights from one cohort to generate a GRS in the between the t-tau groups (Fig. 2; Data S3a). This sug- other cohort. We then correlated the individual GRS of gests that interindividual differences in CSF t-tau each gene or GO-BP with individual protein concen- concentrations may have partly a genetic basis. To trations (Data S5a). A positive correlation means that investigate whether genetic markers influenced CSF a higher GRS is associated with a higher protein con- protein concentrations, we correlated subject level GRS centration. The largest number of positive correlations with subject level CSF protein concentrations in 188 of GRS with CSF protein concentrations were observed individuals with abnormal Aβ1-42. We used the GRS of for the GO-BP gene expression GRS (288 proteins), the top 3 genes, the GRS of the top 3 GO-BP and the the GO-BP regulation of MAPK cascade GRS (146 GRS of 9 GO-BPs that differed between abnormal and proteins), and the GO-BP histone H3 acetylation GRS normal t-tau groups in both CSF and MAGMA geneset (127 proteins, Fig. 4b, left coloured part, Data S5a). analysis in Fig. 2. Given the association of CSF profiles The proteins that showed a positive correlation with with REST and SUZ12 we also selected the GRS of 2 these GRS were typically associated with nervous sys- GO-BPs associated with histone acetylation with the tem development, gene expression, and cell adhesion Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 10 of 16 (Fig. 4b, right grey part). These findings support the Results from our molecular process and transcription notion that the increase in plasticity-related processes factor enrichment analysis suggested that the increased in CSF has in part a genetic background. Moreover, levels of neuroplasticity proteins in the AD group with we tested the association of SNP rs9877502 in GMNC, increased t-tau could result from an increase in mitogenic previously associated with CSF t-tau levels without MAPK signaling, or from a reduction in gene repression stratification for amyloid aggregation, with tau status by REST and SUZ12. The de-repression of gene expres- [44, 63, 64]. Consistent with previous data, the A-allele sion by REST/SUZ12 in AD has previously been reported frequency at this site was more common in AD indi- in post-mortem AD studies and in iPSC neurons from viduals with increased t-tau compared to those with individuals with sporadic AD [8, 60, 61]. Our finding normal t-tau (nominal p-value = 0.02). We then asso- that a subset of proteins coded by genes with increased ciated this SNP with CSF protein levels using additive expression in these previous studies had increased CSF models. 591 of 1705 proteins (35%) showed an increase concentrations in AD individuals with increased t-tau as in CSF level with increasing number of rs9877502 well, further supports the role of REST/SUZ12 de-repres- A-alleles (Data S5b). These proteins were enriched sion in AD. Both APP and tau are regulated by REST or for neuronal plasticity related processes (Data S5c), SUZ12, and this may explain why the increases in levels and SUZ12 (p-FDR corrected = 5.27e-24) and REST of Aβ production markers and tau occur in tandem [6]. (p-FDR corrected = 6.34e-18) signaling. 105 proteins The increased concentration of neuronal plasticity pro- (6%) showed a negative correlation with the number teins might reflect a compensatory mechanism triggered of rs9877502 A-alleles, including 36 proteins associ- by tau-induced synaptic dysfunction such as impairment ated with BBB dysfunction and 36 immunoglobulins. of presynaptic vesicle release, trafficking of glutamatergic In a smilar way, we associated the number of APOE-e4 receptors, and maturation of dendritic spines [65]. AD alleles with CSF protein levels in individuals with AD. neuropathological studies suggested that such a plastic- Forty-five proteins showed an increase in CSF levels ity response is likely pathological and results in aberrant with increasing number of APOE-e4 alleles (Data S5b) synaptic sprouting (both at axons and at dendrites), dis- and these proteins were enriched for GO-BP learning organised capillairies and an abnormal cortical myelin or memory (p-FDR corrected = 0.046). The concentra- architecture [51, 52, 66]. Increased expression of APP tion of 36 proteins showed a negative correlation with and tau may initiate a vicious cycle. APP and Aβ peptides the number of APOE-e4 alleles (Data S5b) and were not can activate MAPK signaling, [53, 67, 68] which may fur- enriched for any GO-BP. ther increase APP production[69, 70]. The APP intracel- Finally, we studied the effect of t-tau on disease pro- lular domain (AICD) stimulates gene transcription, [71] gression using longitudinal ADNI data. and may increase expression of APP and BACE1 [72]. Relative to controls, AD groups with increased t-tau Abnormal tau may increase gene expression through and normal t-tau showed both faster memory decline and depletion of H3K9me2, altered spatial open chromatin PET-amyloid accumulation in the preclinical stage and organization, or altered H3K9 acetylation [73, 74]. showed faster decline in all cognitive and imaging meas- The group of AD individuals with normal t-tau was ures in the prodromal and dementia stage. Compared to characterized by decreased concentrations of plasticity AD individuals with normal t-tau, those with increased related proteins and increased concentrations of BBB t-tau declined faster on cognitive tests and in glucose and BCSFB permeability related proteins. The increase metabolism in the prodromal AD stage and showed a in barrier permeability may be caused by accumula- faster decline in cortical thickness in the prodromal and tion of Aβ in vessel walls and toxic effects of Aβ on cells dementia stage (Fig. 5, Data S6). that constitute these barriers [75, 76]. This can loosen tight junctions and increase paracellular transport [75, Discussion 77–79]. Impaired BBB and BCSFB function may lead Our main finding is that AD individuals with to hypoplasticity through impairments in their physi- increased CSF t-tau levels showed increased concen- ological role in glucose transport, capillary perfusion, trations of proteins associated with neuronal plas- and neurogenesis, [79] and possibly also by increasing ticity, while AD individuals with normal t-tau levels REST signaling, which can be triggered by ischemia showed decreased concentrations of plasticity pro- [80]. The increase in concentrations of CSF proteins teins. These opposite CSF proteomic profiles were that are produced in the choroid plexus suggests already observed in the preclinical stage of AD, and BCSFB dysfunction. The choroid plexus is involved in persisted in prodromal and mild dementia stages, the clearance of Aβ, [81] and the increase in concentra- indicating that they reflect different disease traits tion of proteins expressed in the choroid plexus could rather than different disease states. be a response to Aβ induced inflammation [82–84]. Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 11 of 16 Fig. 5 Longitudinal cognitive and imaging changes according to diagnostic group and clinical stage Upper images: change in cognitive markers. A MMSE, B CDR sum of boxes, C Logical memory, and D ADAS-Cog. Change of controls are shown in grey, of AD individuals with increased t-tau in red (a + t +), and of AD individuals with normal t-tau (a + t-) in blue. 1. Slope p < 0.05 different from control; 2. Slope p < 0.05 different from 0; 3. Slope p < 0.05 different between AD individuals with increased t-tau (a + t +) and AD individuals with normal t-tau (a + t-). Lower images: Annual rate of change in imaging markers. E amyloid aggregation, F cortical thickness and G FDG-PET glucose metabolism. PET amyloid data are not shown for AD individuals with dementia with normal t-tau because of low number of participants (n = 6). Data are shown in Data S6. Data are from ADNI only. CN = normal cognition; MCI = mild cognitive impairment; a + t +  = abnormal Aβ1-42 and increased t-tau; a + t- = abnormal Aβ1-42 and normal t-tau AD individuals with normal t-tau may have cerebral Our GWAS and GRS analyses suggested that levels of amyloid angiopathy (CAA), a condition in which Aβ is tau and neuroplasticity proteins partly depend on genes deposited in cerebral and leptomeningeal blood vessels involved in gene expression, although no genome-wide [85]. CAA is common in individuals with AD and cor- significant signals were identified. APBB2 showed the relates with Aβ plaque deposition and neurofibrillary strongest association with CSF t-tau in these analyses. tangles [85]. CAA has been associated with normal lev- While markers in APBB2 do not show strong evidence els of t-tau levels and a decrease in Aβ40 levels, which for association with AD risk in GWAS comparing con- was also observed in our AD individuals with normal trols with individuals with AD-dementia [49], the gene t-tau [86]. Although t-tau levels were within normal represents an interesting functional candidate as APBB2 limits, this group showed a typical AD genetic pattern binds AICD. Alternative splicing of APBB2 increased and a progressive disease course with increases in cog- the Aβ42/ Aβ40 ratio [87], and overexpression increased nitive and AD biomarkers abnormalities, which sup- Aβ1-40, APP, and AICD levels and changed gene expres- ports the view that this is a subtype of AD, and not a sion [68, 87, 88]. In addition, we found associations of different disease entity. rs9877502 in GMNC (previously associated with t-tau Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 12 of 16 levels in this and other datasets) [44, 63, 64] with levels of study showed that increased CSF t-tau levels were associ- 696 CSF proteins in AD individuals. GMNC is involved ated with long-term depression (LTD) after transcranial in neuronal plasticity and regulation of gene expres- magnetic stimulation in the motor cortex [96]. As this sion [89, 90]. We also found that GRS associated with study was performed in demented individuals, it is possi- t-tau were involved in neuronal plasticity and that these ble that LTD reflects synaptic failure. Future studies that GRS showed an association with the concentration of combine electrophysiological measures with proteomic plasticity-related proteins. On the contrary, the number data in human beings are required to study this question of APOEe4 alleles was not associated with the concen- in more detail. tration of proteins involved in neuronal plasticity. The We further found that AD individuals with increased genetic pattern associated with t-tau levels in individuals t-tau also showed increased cytokine signaling and leu- with AD is thus different from the genetic processes asso- kocyte activation. This activation could result from sol- ciated with AD-type dementia, which include immunity, uble tau oligomers, [97] although it is also possible that lipid metabolism, and intracellular trafficking [49]. This both increased t-tau levels and microglia activation result indicates that within the AD genetic risk architecture from amyloid pathology. other genetic processes could influence AD phenotypes. A strength of the study is the large sample size available As we found a wide range of processes deregulated, for proteomic analysis and the use of two independent it is possible that this may reflect an aspecific response cohorts. Even though these cohorts performed proteom- to brain pathology. This seems, however, unlikely as the ics in a different way (targeted proteomics in ADNI, non- same set of proteins that was increased in AD individu- targeted proteomics in EMIF-AD MBD) and the overlap als with increased t-tau was decreased in AD individuals in proteins measured in both studies was limited, enrich- with normal t-tau suggesting that these processes have a ment analysis showed very similar results between the common denominator. In addition, we found that these cohorts (Fig. 2), which supports the robustness of our proteins were enriched for SUZ12 and REST transcrip- findings. The consistency across cohorts was also pre- tion factors, which are master regulators of neuronal sent on the protein level. Of the 167 proteins measured plasticity. Moreover, abnormalities in nervous system in both cohorts, only 7 proteins (5%) showed a difference development related processes, myelination, angiogen- between the cohorts on both major outcomes (AD with esis, MAPK signaling, cell-cycle, gene expression and normal t-tau vs control and AD with increased t-tau ver- glycolysis have all been reported in separate AD post- sus controls, Data S2). A limitation is that CSF proteom- mortem, animal or cell studies [8, 50–55]. Still, some ics was performed cross-sectionally and the differences dysregulated processes may only be indirectly involved between disease stages need to be confirmed in longitu- in plasticity. For example, increased levels of proteins dinal studies. In addition, the association between change associated with glycolysis could be a response to the high in CSF protein concentrations and biological processes energy demand resulting from increased neuronal activ- needs to be further addressed in functional studies. The ity [91]. It is unlikely that the increased levels of plastic- sample size for genetic analysis was small and larger sam- ity proteins in AD individuals with increased t-tau simply ples are needed to confirm these results. reflect cell death as levels of plasticity-associated proteins were highest in the preclinical AD stage, when neurode- Conclusions generation is limited, and decreased with increasing dis- CSF t-tau levels in AD are associated with altered levels ease severity. of proteins involved in neuronal plasticity and blood– The observation that increased t-tau levels are associ- brain and blood-CSF barrier dysfunction. The asso- ated with an increase in concentration of plasticity pro- ciation of increased CSF t-tau with neuronal plasticity teins, aligns with observations that tau is associated with protein levels provides support for the role of tau in increased neuronal excitability in neurophysiological neuronal plasticity and gene expression and implies that studies [92]. In an AD mice model it was shown that tau increased CSF t-tau levels may not simply reflect axonal could be an enabler of neural network dysfunction caused loss [8, 73, 74]. It will be of critical importance to stratify by amyloid pathology as tau reduction reduced amyloid- future trials on t-tau status, as individuals with increased related overexcitation [92, 93]. Tau may also have direct t-tau and normal t-tau are likely to respond differently effects on excitability as overexpression of human tau to treatment, given their opposite proteomic profiles. in mice models increased delta/theta power and reduc- The dosing of amyloid and tau antibodies may need to tion of tau expression counteracted these changes [94]. be tailored to t-tau levels, because of differences in BBB/ Another study showed that nuclear translocation of tau BCSFB integrity between those with increased and nor- led to increased expression of VgluT1, which is involved mal t-tau. Treatments with drugs that target amyloid in glutamatergic synaptic transmission [95]. A clinical production, hyperexcitation, or hyperplasticity, such Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 13 of 16 as bace inhibitors, histone modifiers and anti-epileptic REST transcription factors associated with proteins that differed between drugs, could only be effective in AD individuals with AD individuals with increased t-tau and normal t-tau and proteins that increased t-tau in the predementia stage, when the CSF changed with disease severity. concentration of amyloid production markers and plas- ticity related proteins are highest. Retrospective analysis Acknowledgements of previous trial data according to CSF t-tau status may We thank Drs. Andre Franke and Michael Wittig as well as Mrs. Tanja Wesse and Sanaz Sedghpour Sabet for their help with the genome-wide genotyping experi‑ lead to a better understanding why effects in these trials ments. Further, we thank Dr. Fabian Kilpert and Shengjun Hong for his help with were absent or minimal and whether AD individuals with the processing and management of the genome-wide genotyping data. increased t-tau and normal t-tau have a different treat- Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the ment response. Future studies on the mechanisms that investigators within the ADNI contributed to the design and implementation of lead to t-tau associated proteomic profiles, such as REST ADNI and/or provided data but did not participate in analysis or writing of this and SUZ12 signaling, will help to clarify AD pathophysi- report. A complete listing of ADNI investigators can be found at: http://​adni.​loni.​ usc.​edu/​wpcon​tent/​uploa​ds/​how_​to_​apply/​ADNI_​Ackno​wledg​ement_​List.​pdf ology and may eventually lead to novel drug targets. Authors’ contributions Conceptualization: PJV, BMT. Methodology proteomics EMIF-AD: JG, HZ. Abbreviations Formal analyses: BMT, PJV, LR, JG, VD, SH, ED. Resources: JG, LR, IJ, ED, SH, SL, AD: Alzheimer’s disease; ADNI: Alzheimer’s Disease Neuroimaging Initiative; VD, MT, FRJV, JP, PMLA, RV, AL, JLM, SE, YFL, LF, KS, LB, SL, JS, SV, IB, DP, GS, KB, PS, AICD: APP intracellular domain; APOE: Apolipoprotein E; APP: Amyloid precur‑ CET, HZ, PJV. Data curation: SV, IB. Writing—Original Draft: BMT, PJV. Writing— sor protein; Aβ: Amyloid beta; BBB: Blood–Brain barrier; BCSFB: Blood-CSF Review & Editing: all authors. Supervision: PJV, BMT. Funding acquisition: PJV, barrier; CAA​: Cerebral amyloid angiopathy; CDR: Clinical Dementia Rating; CN: SL, HZ, JB, KB, LB, KS. All authors read and approved the final manuscript. Cognitively normal; CSF: Cerebrospinal fluid; ECM: Extracellular matrix; EMIF- AD: European Medical Information Framework for Alzheimers disease; FDG: Funding Fluorodeoxyglucose; FDR: False discovery rate; GO-BP: Gene Ontology biologi‑ This work has been supported by ZonMW Memorabel grant programme cal processes; GRS: Gene-specific risk scores; GWAS: Genome-wide associa‑ #73305056 (BMT), #733050512 (SL) and #733050824 (BMT and PJV), the tion study; iPSC: Induced pluripotent stem cells; LD: Linkage disequilibrium; Swedish Research Council (#2018–02532, HZ), the European Research LTD: Long-term depression; MAPK: Mitogen-activated protein kinase; MBD: Council (#681712, HZ) and Swedish State Support for Clinical Research Multimodality Biomarker Discovery; MCI: Mild cognitive impairment; MMSE: (#ALFGBG-720931, HZ), the Alzheimerfonden (Grant no. AF-930934) and Mini-Mental State Examination; PC1: Principal component; PCA: Principal Dtiftelsen Gamla tjänarinnor (JG), and the Innovative Medicines Initiative Joint component analysis; PET: Positron Emission Tomography; PGRS: Polygenic risk Undertaking under EMIF grant agreement #115372 (PJV, HZ, SV, IB). HZ is a scores; PPS: Proteomic process scores; SNP: Single nucleotide polymorphisms; Wallenberg Academy Scholar. The Leuven cohort was funded by Stichting SUVR: Standardized uptake value; T-tau: Total tau; TMT: Tandem mass tag. Alzheimer Onderzoek (#11020, #15005, #13007) and the Vlaamse Impulsfi‑ nanciering voor Netwerken voor Dementie-onderzoek (IWT #135043). The Lausanne cohort was supported by the Swiss National Research Foundation Supplementary Information SNF (#320030_141179). Data was used for this project of which collection and The online version contains supplementary material available at https://​doi.​ sharing was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) org/​10.​1186/​s13024-​022-​00521-3. (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Depart‑ ment of Defense award number W81XWH-12–2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging Additional file 1: Data S1. Participant characteristics. S1a: Characteristics and Bioengineering, and through generous contributions from the following: of individuals with CSF Aβ1-42 and tau measurements available; S1b: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Characteristics of individuals with CSF proteomic data. Data S2. Protein Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; Cer‑ annotation and statistics of group comparisons of protein levels. Data eSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; S3a. Full list of GO biological processes associated with proteins that EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, differ according to group and clinical stage. Data S3b. SynGO enriched Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy synaptic cellular components and biological processes that differ accord‑ Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research ing to group. Data S4a. Estimated marginal means of AD GWAS-based & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale polygenic risk scores in controls, AD individuals with increased t-tau and Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis AD individuals with normal t-tau. Data S4b. Top 1000 SNPS from GWAS Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Phar‑ on AD individuals with increased t-tau and normal t-tau in pooled ADNI maceutical Company; and Transition Therapeutics. The Canadian Institutes of and EMIF-AD MBD cohorts. Data S4c. Difference in MAGMA gene score Health Research is providing funds to support ADNI clinical sites in Canada. between AD individuals with increased t-tau and normal t-tau based Private sector contributions are facilitated by the Foundation for the National on t-tau GWAS in pooled ADNI and EMIF-AD MBD cohorts. Data S4d. Institutes of Health (www.​fnih.​org). The grantee organization is the Northern Difference in GO biological process MAGMA geneset score between AD California Institute for Research and Education, and the study is coordinated individuals with increased t-tau and normal t-tau based on t-tau GWAS in by the Alzheimer’s Therapeutic Research Institute at the University of Southern pooled ADNI and EMIF-AD MBD cohorts. Data S5a. Correlation between California. ADNI data are disseminated by the Laboratory for Neuro Imaging at genetic risk score and CSF protein level in individuals with abnormal Aβ1- the University of Southern California. 42. Data S5b. Association of the number of GMNC rs9877502-A risk alleles and number of APOE-e4 alleles with CSF protein concentrations in a linear Availability of data and materials model in individuals with AD. Data S5c. GO-BP processes enriched for ADNI data can be downloaded from adni.loni.usc.edu. Raw proteomic data proteins that have a positive or negative association with the number of from EMIF-AD MBD has been deposited in the ProteomeXchange Consortium rs9877502-A risk alleles in an additive model. Data S6. Annual change in via the PRIDE partner repository with the dataset identifier https://​doi.​org/​ imaging measures. 10.​6019/​PXD01​9910. Access to other EMIF-AD MBD data can be requested Additional file 2: Figure S1. Longitudinal change in CSF total tau. Figure from the authors. Datasharing may be restricted by consent given by research S2. Enrichment of synaptic processes in individuals with AD according to participants within each contributing cohort and by European GDPR regula‑ t-tau status. Figure S3. Enriched GO biological processes and SUZ12 and tions, which currently exclude data sharing with a number of non-European countries. Statistical data can be found in the supplementary information files. Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 14 of 16 Declarations 3. Vos SJ, Verhey F, Frolich L, Kornhuber J, Wiltfang J, Maier W, Peters O, Ruther E, Nobili F, Morbelli S, et al. Prevalence and prognosis of Alzhei‑ Ethics approval and consent to participate mer’s disease at the mild cognitive impairment stage. Brain. 2015;138(Pt Local institutional review boards approved the procedures for this study and 5):1327–38. written informed consent was obtained. 4. Ferreira D, Shams S, Cavallin L, Viitanen M, Martola J, Granberg T, Shams M, Aspelin P, Kristoffersen-Wiberg M, Nordberg A, et al. The contribu‑ Consent for publication tion of small vessel disease to subtypes of Alzheimer’s disease: a study Not applicable. on cerebrospinal fluid and imaging biomarkers. Neurobiol Aging. 2018;70:18–29. Competing interests 5. Vos SJ, Xiong C, Visser PJ, Jasielec MS, Hassenstab J, Grant EA, Cairns NJ, The authors declare that they have no competing interests. Morris JC, Holtzman DM, Fagan AM. Preclinical Alzheimer’s disease and its outcome: a longitudinal cohort study. Lancet Neurol. 2013;12(10):957–65. Author details 6. Konijnenberg E, Tomassen J, den Braber A, Ten Kate M, Yaqub M, Mulder 1 Alzheimer Center Amsterdam, Department of Neurology, Amsterdam SD, Nivard MG, Vanderstichele H, Lammertsma AA, Teunissen CE, et al. Neuroscience, Vrije Universiteit Amsterdam, PO Box 7057 1007, MB, Amster‑ Onset of Preclinical Alzheimer Disease in Monozygotic Twins. Ann Neurol. dam, The Netherlands. 2 Alzheimer Center Limburg, School for Mental Health 2021;89:987–1000. and Neuroscience, Maastricht University, PO Box 616, 6200, MD, Maastricht, 7. Sato C, Barthelemy NR, Mawuenyega KG, Patterson BW, Gordon BA, The Netherlands. 3 Department of Neurobiology, Care Sciences and Society, Jockel-Balsarotti J, Sullivan M, Crisp MJ, Kasten T, Kirmess KM, et al. Tau Division of Neurogeriatrics, Karolinska Institutet, Stockholm, Sweden. 4 Clini‑ kinetics in neurons and the human central nervous system. Neuron. cal Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, 2018;97(6):1284-1298 e1287. Sweden. 5 Department of Psychiatry and Neurochemistry, Institute of Neu‑ 8. Meyer K, Feldman HM, Lu T, Drake D, Lim ET, Ling KH, Bishop NA, Pan Y, roscience and Physiology, Sahlgrenska Academy at the University of Goth‑ Seo J, Lin YT, et al. REST and neural gene network dysregulation in iPSC enburg, Mölndal, Sweden. 6 Department of Complex Trait Genetics, Center models of Alzheimer’s disease. Cell Rep. 2019;26(5):1112-1127 e1119. for Neurogenomics and Cognitive Research, Amsterdam Neuroscience, VU 9. Yamada K, Holth JK, Liao F, Stewart FR, Mahan TE, Jiang H, Cirrito JR, Patel University, Amsterdam, the Netherlands. 7 Section Genomics of Neurodegen‑ TK, Hochgrafe K, Mandelkow EM, et al. Neuronal activity regulates extra‑ erative Diseases and Aging, Department of Clinical Genetics, Vrije Universiteit cellular tau in vivo. J Exp Med. 2014;211(3):387–93. Amsterdam, Amsterdam UMC, Amsterdam, the Netherlands. 8 1St Department 10. Grothe MJ, Sepulcre J, Gonzalez-Escamilla G, Jelistratova I, Scholl M, Hans‑ of Neurology, AHEPA University Hospital, Thessaloniki, Makedonia, Greece. 9 Old son O, Teipel SJ. Alzheimer’s disease neuroimaging i: molecular properties Age Psychiatry, University Hospital Lausanne, Lausanne, Switzerland. 10 Depart‑ underlying regional vulnerability to Alzheimer’s disease pathology. Brain. ment of Geriatric Psychiatry, University Hospital of Psychiatry and University 2018;141(9):2755–71. of Zürich, Zürich, Switzerland. 11 Fundación CITA-Alzhéimer Fundazioa, San 11. Jack CR Jr, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, Sebastian, Spain. 12 Neurology Service, University Hospitals Leuven, Leuven, Holtzman DM, Jagust W, Jessen F, Karlawish J, et al. NIA-AA research Belgium. 13 Laboratory for Cognitive Neurology, Department of Neurosciences, framework: toward a biological definition of Alzheimer’s disease. Alzhei‑ KU Leuven, Leuven, Belgium. 14 IIB-Sant Pau, Hospital de La Santa Creu I Sant mers Dement. 2018;14(4):535–62. Pau, Universitat Autonoma de Barcelona, Barcelona, Spain. 15 Barcelonaβeta 12. Bos I, Vos S, Vandenberghe R, Scheltens P, Engelborghs S, Frisoni G, Brain Research Center (BBRC), Barcelona, Spain. 16 Alzheimer’s Disease Unit Molinuevo JL, Wallin A, Lleo A, Popp J, et al. The EMIF-AD multimodal and Other Cognitive Disorders Unit, Hospital Clinic de Barcelona, Barcelona, biomarker discovery study: design, methods and cohort characteristics. Spain. 17 Reference Center for Biological Markers of Dementia (BIODEM), Alzheimers Res Ther. 2018;10(1):64. Institute Born‑Bunge, University of Antwerp, Antwerp, Belgium. 18 Depart‑ 13. Bertens D, Knol DL, Scheltens P, Visser PJ. Alzheimer’s disease neuroimag‑ ment of Neurology, UZ Brussel and Center for Neurosciences, Vrije Universiteit ing i: temporal evolution of biomarkers and cognitive markers in the Brussel, Brussels, Belgium. 19 Department of Psychiatry at School of Medical Sci‑ asymptomatic, MCI, and dementia stage of Alzheimer’s disease. Alzhei‑ ences, Örebro University, Örebro, Sweden. 20 Department of Geriatric Psychiatry, mers Dement. 2015;11(5):511–22. Zentralinstitut Für Seelische Gesundheit, University of Heidelberg, Mannheim, 14. Folstein MF, Folstein SE, McHugh PR. “Mini-Mental State”, a practical Germany. 21 Complex Genetics Group, VIB Center for Molecular Neurology, method for grading the cognitive state of patients for the clinician. J VIB, Antwerp, Belgium. 22 Department of Biomedical Sciences, University Psychiatr Res. 1975;12:189–98. of Antwerp, Antwerp, Belgium. 23 Lübeck Interdisciplinary Platform for Genome 15. Rosen WG, Mohs RC, Davis KL. A new rating scale for Alzheimer’s disease. Analytics, Institutes of Neurogenetics and Cardiogenetics, University of Lübeck, Am J Psychiatry. 1984;141(11):1356–64. Lübeck, Germany. 24 Jansen UK, High Wycombe, UK. 25 AC Immune SA, 16. Wechsler D. Wechsler Memory Scale-. IIIrd. San Antonio: The Psychologi‑ Lausanne, Switzerland. 26 Department of Molecular and Cellular Neurobiology, cal Corporation Administration and scoring manual; 1997. Center for Neurogenomics and Cognitive Research, Amsterdam Neuroscience, 17. Morris J. The Clinical Dementia Rating (CDR): current version and scoring Vrije Universiteit Amsterdam, Amsterdam, the Netherlands. 27 Neurochemistry rules. Neurology. 1993;43:2412–4. Laboratory, Department of Clinical Chemistry, Amsterdam University Medical 18. del Campo M, Mollenhauer B, Bertolotto A, Engelborghs S, Hampel H, Centers (AUMC), Amsterdam Neuroscience, Netherlands. 28 Center for Lifespan Simonsen AH, Kapaki E, Kruse N, Le Bastard N, Lehmann S, et al. Recom‑ Changes in Brain and Cognition, Dept. of Psychology, University of Oslo, Oslo, mendations to standardize preanalytical confounding factors in Alzhei‑ Norway. 29 Department of Neurodegenerative Disease, UCL Institute of Neurol‑ mer’s and Parkinson’s disease cerebrospinal fluid biomarkers: an update. ogy, London, UK. 30 Dementia Research Institute at UCL, London, UK. Biomark Med. 2012;6(4):419–30. 19. Shaw LM, Vanderstichele H, Knapik-Czajka M, Clark CM, Aisen PS, Petersen Received: 16 November 2021 Accepted: 13 February 2022 RC, Blennow K, Soares H, Simon A, Lewczuk P, et al. Cerebrospinal fluid biomarker signature in Alzheimer’s disease neuroimaging initiative subjects. Ann Neurol. 2009;65(4):403–13. 20. Tijms BM, Gobom J, Reus L, Jansen I, Hong S, Dobricic V, Kilpert F, Ten Kate M, Barkhof F, Tsolaki M, et al. Pathophysiological subtypes of References Alzheimer’s disease based on cerebrospinal fluid proteomics. Brain. 1. Selkoe DJ, Hardy J. The amyloid hypothesis of Alzheimer’s disease at 25 2020;143(12):3776–92. years. EMBO Mol Med. 2016;8(6):595–608. 21. Batth TS, Francavilla C, Olsen JV. Off-line high-pH reversed-phase 2. Bos I, Vos S, Verhey F, Scheltens P, Teunissen C, Engelborghs S, Sleegers fractionation for in-depth phosphoproteomics. J Proteome Res. K, Frisoni G, Blin O, Richardson JC, et al. Cerebrospinal fluid biomark‑ 2014;13(12):6176–86. ers of neurodegeneration, synaptic integrity, and astroglial activation 22. Spellman DS, Wildsmith KR, Honigberg LA, Tuefferd M, Baker D, Raghavan across the clinical Alzheimer’s disease spectrum. Alzheimers Dement. N, Nairn AC, Croteau P, Schirm M, Allard R, et al. Development and evalu‑ 2019;15(5):644–54. ation of a multiplexed mass spectrometry based assay for measuring Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 15 of 16 candidate peptide biomarkers in Alzheimer’s Disease Neuroimaging 42. Landau SM, Fero A, Baker SL, Koeppe R, Mintun M, Chen K, Reiman EM, Initiative (ADNI) CSF. Proteomics Clin Appl. 2015;9(7–8):715–31. Jagust WJ. Measurement of longitudinal beta-amyloid change with 23. Leung YY, Toledo JB, Nefedov A, Polikar R, Raghavan N, Xie SX, Farnum M, 18F-florbetapir PET and standardized uptake value ratios. J Nucl Med. Schultz T, Baek Y, Deerlin VV, et al. Identifying amyloid pathology-related 2015;56(4):567–74. cerebrospinal fluid biomarkers for Alzheimer’s disease in a multicohort 43. Landau SM, Harvey D, Madison CM, Koeppe RA, Reiman EM, Foster NL, study. Alzheimers Dement (Amst). 2015;1(3):339–48. Weiner MW, Jagust WJ. Alzheimer’s disease neuroimaging i: Associations 24. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis between cognitive, functional, and FDG-PET measures of decline in AD AP, Dolinski K, Dwight SS, Eppig JT, et al. Gene ontology: tool for the and MCI. Neurobiol Aging. 2011;32(7):1207–18. unification of biology. The Gene Ontology Consortium Nat Genet. 44. Hong S, Prokopenko D, Dobricic V, Kilpert F, Bos I, Vos SJB, Tijms BM, 2000;25(1):25–9. Andreasson U, Blennow K, Vandenberghe R, et al. Genome-wide associa‑ 25. Gene Ontology C. The Gene Ontology resource: enriching a GOld mine. tion study of Alzheimer’s disease CSF biomarkers in the EMIF-AD Multi‑ Nucleic Acids Res. 2021;49(D1):D325–34. modal Biomarker Discovery dataset. Transl Psychiatry. 2020;10(1):403. 26. Mi H, Muruganujan A, Ebert D, Huang X, Thomas PD. PANTHER version 14: 45. Saykin AJ, Shen L, Foroud TM, Potkin SG, Swaminathan S, Kim S, Risacher more genomes, a new PANTHER GO-slim and improvements in enrich‑ SL, Nho K, Huentelman MJ, Craig DW, et al. Alzheimer’s disease neuro‑ ment analysis tools. Nucleic Acids Res. 2019;47(D1):D419–26. imaging initiative biomarkers as quantitative phenotypes: Genetics core 27. Koopmans F, van Nierop P, Andres-Alonso M, Byrnes A, Cijsouw T, Coba aims, progress, and plans. Alzheimers Dement. 2010;6(3):265–73. MP, Cornelisse LN, Farrell RJ, Goldschmidt HL, Howrigan DP, et al. SynGO: 46. McCarthy S, Das S, Kretzschmar W, Delaneau O, Wood AR, Teumer A, Kang An Evidence-Based, Expert-Curated Knowledge Base for the Synapse. HM, Fuchsberger C, Danecek P, Sharp K, et al. A reference panel of 64,976 Neuron. 2019;103:217–34. haplotypes for genotype imputation. Nat Genet. 2016;48(10):1279–83. 28. Lachmann A, Xu H, Krishnan J, Berger SI, Mazloom AR, Ma’ayan A. ChEA: 47. Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of transcription factor regulation inferred from integrating genome-wide genomewide association scans. Bioinformatics. 2010;26(17):2190–1. ChIP-X experiments. Bioinformatics. 2010;26(19):2438–44. 48 Choi SW, O’Reilly PF. PRSice-2: Polygenic Risk Score software for biobank- 29 Rouillard AD, Gundersen GW, Fernandez NF, Wang Z, Monteiro CD, scale data. Gigascience. 2019;8(7):giz082. McDermott MG, Ma’ayan A. The harmonizome: a collection of processed 49. Jansen IE, Savage JE, Watanabe K, Bryois J, Williams DM, Steinberg S, Seal‑ datasets gathered to serve and mine knowledge about genes and pro‑ ock J, Karlsson IK, Hagg S, Athanasiu L, et al. Genome-wide meta-analysis teins. Database (Oxford). 2016;2016:baw100. identifies new loci and functional pathways influencing Alzheimer’s 30. Kuleshov MV, Jones MR, Rouillard AD, Fernandez NF, Duan Q, Wang Z, disease risk. Nat Genet. 2019;51(3):404–13. Koplev S, Jenkins SL, Jagodnik KM, Lachmann A, et al. Enrichr: a compre‑ 50. Ihara Y. Massive somatodendritic sprouting of cortical neurons in Alzhei‑ hensive gene set enrichment analysis web server 2016 update. Nucleic mer’s disease. Brain Res. 1988;459(1):138–44. Acids Res. 2016;44(W1):W90-97. 51. Bulk M, Abdelmoula WM, Nabuurs RJA, van der Graaf LM, Mulders CWH, 31. Zhang Y, Chen K, Sloan SA, Bennett ML, Scholze AR, O’Keeffe S, Phatnani Mulder AA, Jost CR, Koster AJ, van Buchem MA, Natte R, et al. Postmor‑ HP, Guarnieri P, Caneda C, Ruderisch N, et al. An RNA-sequencing tran‑ tem MRI and histology demonstrate differential iron accumulation and scriptome and splicing database of glia, neurons, and vascular cells of the cortical myelin organization in early- and late-onset Alzheimer’s disease. cerebral cortex. J Neurosci. 2014;34(36):11929–47. Neurobiol Aging. 2018;62:231–42. 32. Hawrylycz MJ, Lein ES, Guillozet-Bongaarts AL, Shen EH, Ng L, Miller JA, 52. Biron KE, Dickstein DL, Gopaul R, Jefferies WA. Amyloid triggers extensive van de Lagemaat LN, Smith KA, Ebbert A, Riley ZL, et al. An anatomically cerebral angiogenesis causing blood brain barrier permeability and comprehensive atlas of the adult human brain transcriptome. Nature. hypervascularity in Alzheimer’s disease. PLoS One. 2011;6(8):e23789. 2012;489(7416):391–9. 53. Greenberg SM, Koo EH, Selkoe DJ, Qiu WQ, Kosik KS. Secreted beta- 33. Grote S, Prufer K, Kelso J, Dannemann M. ABAEnrichment: an R package amyloid precursor protein stimulates mitogen-activated protein to test for gene set expression enrichment in the adult and developing kinase and enhances tau phosphorylation. Proc Natl Acad Sci U S A. human brain. Bioinformatics. 2016;32(20):3201–3. 1994;91(15):7104–8. 34. Dayon L, Cominetti O, Wojcik J, Galindo AN, Oikonomidi A, Henry H, 54. Seward ME, Swanson E, Norambuena A, Reimann A, Cochran JN, Li R, Migliavacca E, Kussmann M, Bowman GL, Popp J. Proteomes of Paired Roberson ED, Bloom GS. Amyloid-beta signals through tau to drive Human Cerebrospinal Fluid and Plasma: Relation to Blood-Brain Barrier ectopic neuronal cell cycle re-entry in Alzheimer’s disease. J Cell Sci. Permeability in Older Adults. J Proteome Res. 2019;18(3):1162–74. 2013;126(Pt 5):1278–86. 35. Rapoport SI, Pettigrew KD. A heterogenous, pore-vesicle membrane 55. Vlassenko AG, Vaishnavi SN, Couture L, Sacco D, Shannon BJ, Mach RH, model for protein transfer from blood to cerebrospinal fluid at the cho‑ Morris JC, Raichle ME, Mintun MA. Spatial correlation between brain roid plexus. Microvasc Res. 1979;18(1):105–19. aerobic glycolysis and amyloid-beta (Abeta ) deposition. Proc Natl Acad 36. Neumeier M, Weigert J, Buettner R, Wanninger J, Schaffler A, Muller AM, Sci U S A. 2010;107(41):17763–7. Killian S, Sauerbruch S, Schlachetzki F, Steinbrecher A, et al. Detection of 56. Israel MA, Yuan SH, Bardy C, Reyna SM, Mu Y, Herrera C, Hefferan MP, adiponectin in cerebrospinal fluid in humans. Am J Physiol Endocrinol Van Gorp S, Nazor KL, Boscolo FS, et al. Probing sporadic and familial Metab. 2007;293(4):E965-969. Alzheimer’s disease using induced pluripotent stem cells. Nature. 37. Dislich B, Wohlrab F, Bachhuber T, Muller SA, Kuhn PH, Hogl S, Meyer-Lue‑ 2012;482(7384):216–20. hmann M, Lichtenthaler SF. Label-free quantitative proteomics of mouse 57. Bossers K, Wirz KT, Meerhoff GF, Essing AH, van Dongen JW, Houba P, cerebrospinal fluid detects beta-site APP cleaving enzyme (BACE1) Kruse CG, Verhaagen J, Swaab DF. Concerted changes in transcripts in the protease substrates in vivo. Mol Cell Proteomics. 2015;14(10):2550–63. prefrontal cortex precede neuropathology in Alzheimer’s disease. Brain. 38. Kuhn PH, Koroniak K, Hogl S, Colombo A, Zeitschel U, Willem M, Volbracht 2010;133(Pt 12):3699–723. C, Schepers U, Imhof A, Hoffmeister A, et al. Secretome protein enrich‑ 58. Kouznetsova VL, Tchekanov A, Li X, Yan X, Tsigelny IF. Polycomb ment identifies physiological BACE1 protease substrates in neurons. repressive 2 complex-Molecular mechanisms of function. Protein Sci. EMBO J. 2012;31(14):3157–68. 2019;28(8):1387–99. 39. Kuhn PH, Colombo AV, Schusser B, Dreymueller D, Wetzel S, Schepers 59. Hwang JY, Zukin RS. REST, a master transcriptional regulator in neurode‑ U, Herber J, Ludwig A, Kremmer E, Montag D, et al. Systematic substrate generative disease. Curr Opin Neurobiol. 2018;48:193–200. identification indicates a central role for the metalloprotease ADAM10 in 60. Lu T, Aron L, Zullo J, Pan Y, Kim H, Chen Y, Yang TH, Kim HM, Drake D, Liu axon targeting and synapse function. Elife. 2016;5:e12748. XS, et al. REST and stress resistance in ageing and Alzheimer’s disease. 40. Brummer T, Muller SA, Pan-Montojo F, Yoshida F, Fellgiebel A, Tomita Nature. 2014;507(7493):448–54. T, Endres K, Lichtenthaler SF. NrCAM is a marker for substrate-selec‑ 61. Otero-Garcia M, Xue Y-Q, Tamara Shakouri T, Yongning Deng Y, Samuel tive activation of ADAM10 in Alzheimer’s disease. EMBO Mol Med. Morabito S, Thomas Allison T, Lowry WE, Kawaguchi R, Swarup V, Cobos I. 2019;11(4):e9695. Single-soma transcriptomics of tangle-bearing neurons in Alzheimer’s dis‑ 41. Haapasalo A, Kovacs DM. The many substrates of presenilin/gamma- ease reveals the signatures of tau-associated synaptic dysfunction. bioRxiv. secretase. J Alzheimers Dis. 2011;25(1):3–28. 2020:2020.2005.2011.088591. https://​doi.​org/​10.​1101/​2020.​05.​11.​088591. Visser et al. Molecular Neurodegeneration (2022) 17:27 Page 16 of 16 62. Hondius DC, van Nierop P, Li KW, Hoozemans JJ, van der Schors RC, van 84. Prineas JW, Parratt JD, Kirwan PD. Fibrosis of the choroid plexus filtration Haastert ES, van der Vies SM, Rozemuller AJ, Smit AB. Profiling the human membrane. J Neuropathol Exp Neurol. 2016;75(9):855–67. hippocampal proteome at all pathologic stages of Alzheimer’s disease. 85. Thal DR, Ronisz A, Tousseyn T, Rijal Upadhaya A, Balakrishnan K, Van‑ Alzheimers Dement. 2016;12(6):654–68. denberghe R, Vandenbulcke M, von Arnim CAF, Otto M, Beach TG, et al. 63. Cruchaga C, Kauwe JS, Harari O, Jin SC, Cai Y, Karch CM, Benitez BA, Jeng Different aspects of Alzheimer’s disease-related amyloid beta-peptide AT, Skorupa T, Carrell D, et al. GWAS of cerebrospinal fluid tau levels identi‑ pathology and their relationship to amyloid positron emission tomogra‑ fies risk variants for Alzheimer’s disease. Neuron. 2013;78(2):256–68. phy imaging and dementia. Acta Neuropathol Commun. 2019;7(1):178. 64. Deming Y, Li Z, Kapoor M, Harari O, Del-Aguila JL, Black K, Carrell D, Cai Y, 86. Banerjee G, Ambler G, Keshavan A, Paterson RW, Foiani MS, Toombs Fernandez MV, Budde J, et al. Genome-wide association study identifies J, Heslegrave A, Dickson JC, Fraioli F, Groves AM, et al. Cerebrospinal four novel loci associated with Alzheimer’s endophenotypes and disease Fluid Biomarkers in Cerebral Amyloid Angiopathy. J Alzheimers Dis. modifiers. Acta Neuropathol. 2017;133(5):839–56. 2020;74(4):1189–201. 65. Wu M, Zhang M, Yin X, Chen K, Hu Z, Zhou Q, Cao X, Chen Z, Liu D. The 87. Penna I, Vassallo I, Nizzari M, Russo D, Costa D, Menichini P, Poggi A, Russo role of pathological tau in synaptic dysfunction in Alzheimer’s diseases. C, Dieci G, Florio T, et al. A novel snRNA-like transcript affects amyloi‑ Transl Neurodegener. 2021;10(1):45. dogenesis and cell cycle progression through perturbation of Fe65L1 66. Arendt T. Alzheimer’s disease as a disorder of mechanisms underlying (APBB2) alternative splicing. Biochim Biophys Acta. 2013;1833(6):1511–26. structural brain self-organization. Neuroscience. 2001;102(4):723–65. 88. Bruni P, Minopoli G, Brancaccio T, Napolitano M, Faraonio R, Zambrano N, 67. Young KF, Pasternak SH, Rylett RJ. Oligomeric aggregates of amyloid beta Hansen U, Russo T. Fe65, a ligand of the Alzheimer’s beta-amyloid precur‑ peptide 1–42 activate ERK/MAPK in SH-SY5Y cells via the alpha7 nicotinic sor protein, blocks cell cycle progression by down-regulating thymidylate receptor. Neurochem Int. 2009;55(8):796–801. synthase expression. J Biol Chem. 2002;277(38):35481–8. 68. Chang Y, Tesco G, Jeong WJ, Lindsley L, Eckman EA, Eckman CB, Tanzi RE, 89. Kyrousi C, Arbi M, Pilz GA, Pefani DE, Lalioti ME, Ninkovic J, Gotz M, Guenette SY. Generation of the beta-amyloid peptide and the amyloid Lygerou Z, Taraviras S. Mcidas and GemC1 are key regulators for the gen‑ precursor protein C-terminal fragment gamma are potentiated by eration of multiciliated ependymal cells in the adult neurogenic niche. FE65L1. J Biol Chem. 2003;278(51):51100–7. Development. 2015;142(21):3661–74. 69. Zhu X, Lee HG, Raina AK, Perry G, Smith MA. The role of mitogen- 90. Balestrini A, Cosentino C, Errico A, Garner E, Costanzo V. GEMC1 is a activated protein kinase pathways in Alzheimer’s disease. Neurosignals. TopBP1-interacting protein required for chromosomal DNA replication. 2002;11(5):270–81. Nat Cell Biol. 2010;12(5):484–91. 70. Schnoder L, Gasparoni G, Nordstrom K, Schottek A, Tomic I, Christmann A, 91 Diaz-Garcia CM, Mongeon R, Lahmann C, Koveal D, Zucker H, Yellen Schafer KH, Menger MD, Walter J, Fassbender K, et al. Neuronal deficiency G. Neuronal Stimulation Triggers Neuronal Glycolysis and Not Lactate of p38alpha-MAPK ameliorates symptoms and pathology of APP or Tau- Uptake. Cell Metab. 2017;26(2):361-374 e364. transgenic Alzheimer’s mouse models. FASEB J. 2020;34(7):9628–49. 92 Chang CW, Shao E, Mucke L. Tau: Enabler of diverse brain disor‑ 71. Cao X, Sudhof TC. Dissection of amyloid-beta precursor protein-depend‑ ders and target of rapidly evolving therapeutic strategies. Science. ent transcriptional transactivation. J Biol Chem. 2004;279(23):24601–11. 2021;371(6532):eabb8255. 72. von Rotz RC, Kohli BM, Bosset J, Meier M, Suzuki T, Nitsch RM, Konietzko 93. Roberson ED, Scearce-Levie K, Palop JJ, Yan F, Cheng IH, Wu T, Gerstein U. The APP intracellular domain forms nuclear multiprotein complexes H, Yu GQ, Mucke L. Reducing endogenous tau ameliorates amyloid and regulates the transcription of its own precursor. J Cell Sci. 2004;117(Pt beta-induced deficits in an Alzheimer’s disease mouse model. Science. 19):4435–48. 2007;316(5825):750–4. 73. Frost B, Hemberg M, Lewis J, Feany MB. Tau promotes neurodegeneration 94. Das M, Maeda S, Hu B, Yu GQ, Guo W, Lopez I, Yu X, Tai C, Wang X, Mucke through global chromatin relaxation. Nat Neurosci. 2014;17(3):357–66. L. Neuronal levels and sequence of tau modulate the power of brain 74. Klein HU, McCabe C, Gjoneska E, Sullivan SE, Kaskow BJ, Tang A, Smith RV, rhythms. Neurobiol Dis. 2018;117:181–8. Xu J, Pfenning AR, Bernstein BE, et al. Epigenome-wide study uncovers 95. Siano G, Varisco M, Caiazza MC, Quercioli V, Mainardi M, Ippolito C, large-scale changes in histone acetylation driven by tau pathology in Cattaneo A, Di Primio C. Tau Modulates VGluT1 Expression. J Mol Biol. aging and Alzheimer’s human brains. Nat Neurosci. 2019;22(1):37–46. 2019;431(4):873–84. 75. Bergen AA, Kaing S, ten Brink JB, Netherlands Brain B, Gorgels TG, Janssen 96. Koch G, Di Lorenzo F, Del Olmo MF, Bonni S, Ponzo V, Caltagirone C, Boz‑ SF. Gene expression and functional annotation of human choroid plexus zali M, Martorana A. Reversal of LTP-Like Cortical Plasticity in Alzheimer’s epithelium failure in Alzheimer’s disease. BMC Genomics. 2015;16:956. Disease Patients with Tau-Related Faster Clinical Progression. J Alzheimers 76. Sweeney MD, Zhao Z, Montagne A, Nelson AR, Zlokovic BV. Blood- Dis. 2016;50(2):605–16. Brain Barrier: From Physiology to Disease and Back. Physiol Rev. 97. Sun XY, Li LJ, Dong QX, Zhu J, Huang YR, Hou SJ, Yu XL, Liu RT. Rutin 2019;99(1):21–78. prevents tau pathology and neuroinflammation in a mouse model of 77. Hartz AM, Bauer B, Soldner EL, Wolf A, Boy S, Backhaus R, Mihaljevic I, Alzheimer’s disease. J Neuroinflammation. 2021;18(1):131. Bogdahn U, Klunemann HH, Schuierer G, et al. Amyloid-beta contributes to blood-brain barrier leakage in transgenic human amyloid precursor protein mice and in humans with cerebral amyloid angiopathy. Stroke. Publisher’s Note 2012;43(2):514–23. Springer Nature remains neutral with regard to jurisdictional claims in pub‑ 78. Wolburg H, Paulus W. Choroid plexus: biology and pathology. Acta Neu‑ lished maps and institutional affiliations. ropathol. 2010;119(1):75–88. 79. Marques F, Sousa JC, Sousa N, Palha JA. Blood-brain-barriers in aging and in Alzheimer’s disease. Mol Neurodegener. 2013;8:38. 80. Calderone A, Jover T, Noh KM, Tanaka H, Yokota H, Lin Y, Grooms SY, Regis R, Bennett MV, Zukin RS. Ischemic insults derepress the gene silencer REST in neurons destined to die. J Neurosci. 2003;23(6):2112–21. 81. Solar P, Zamani A, Kubickova L, Dubovy P, Joukal M. Choroid plexus and the blood-cerebrospinal fluid barrier in disease. Fluids Barriers CNS. 2020;17(1):35. 82. Marques F, Falcao AM, Sousa JC, Coppola G, Geschwind D, Sousa N, Correia-Neves M, Palha JA. Altered iron metabolism is part of the choroid plexus response to peripheral inflammation. Endocrinology. 2009;150(6):2822–8. 83. Brkic M, Balusu S, Van Wonterghem E, Gorle N, Benilova I, Kremer A, Van Hove I, Moons L, De Strooper B, Kanazir S, et al. Amyloid beta oligomers disrupt blood-CSF barrier integrity by activating matrix metalloprotein‑ ases. J Neurosci. 2015;35(37):12766–78.