ORIGINAL RESEARCH published: 15 July 2020 doi: 10.3389/fmicb.2020.01645 Sampling Device-Dependence of Prokaryotic Community Structure on Marine Particles: Higher Diversity Recovered by in situ Pumps Than by Oceanographic Bottles Viena Puigcorbé 1* , Clara Ruiz-González 2* , Pere Masqué 1,3,4,5 and Josep M. Gasol 1,2 1 School of Science, Centre for Marine Ecosystems Research, Edith Cowan University, Joondalup, WA, Australia, 2 Department of Marine Biology and Oceanography, Institut de Ciències del Mar (ICM-CSIC), Barcelona, Spain, 3 Institut de Ciència i Tecnologia Ambientals (ICTA), Bellaterra, Spain, 4 Department of Physics, Autonomous University of Barcelona, Barcelona, Spain, 5 International Atomic Energy Agency, Monaco City, Monaco Microbes associated with sinking marine particles play key roles in carbon sequestration in the ocean. The sampling of particle-attached microorganisms is often done with sediment traps or by filtration of water collected with oceanographic bottles, both Edited by: involving a certain time lapse between collection and processing of samples that may Jose M. Gonzalez, result in changes in particle-attached microbial communities. Conversely, in situ water University of La Laguna, Spain filtration through submersible pumps allows a faster storage of sampled particles, but Reviewed by: Phoebe J. Lam, it has rarely been used to study the associated microbial communities and has never University of California, Santa Cruz, been compared to other particle-sampling methods in terms of the recovery of particle United States microbial diversity. Here we compared the prokaryotic communities attached to small Angelina Lo Giudice, Institute of Polar Sciences (CNR), Italy (1–53 µm) and large (>53 µm) particles collected from the mesopelagic zone (100– *Correspondence: 300 m) of two Antarctic polynyas using in situ pumps (ISP) and oceanographic bottles Viena Puigcorbé (BTL). Each sampling method retrieved largely different particle-attached communities, viena.puigcorbe@outlook.com Clara Ruiz-González suggesting that they capture different kinds of particles. These device-driven differences clararg@icm.csic.es were greater for large particles than for small particles. Overall, the ISP recovered 1.5- to 3-fold more particle-attached bacterial taxa than the BTL, and different taxonomic Specialty section: This article was submitted to groups were preferentially recovered by each method. In particular, typical particle- Aquatic Microbiology, attached groups such as Planctomycetes and Deltaproteobacteria recovered with ISP a section of the journal were nearly absent from BTL samples. Our results suggest that the method used Frontiers in Microbiology to sample marine particles has a strong influence in our view of their associated Received: 29 March 2020 Accepted: 24 June 2020 microbial communities. Published: 15 July 2020 Keywords: prokaryotic communities, marine particles, size-fractionation, in situ pumps, oceanographic bottles, Citation: polynyas Puigcorbé V, Ruiz-González C, Masqué P and Gasol JM (2020) Sampling Device-Dependence of Prokaryotic Community Structure INTRODUCTION on Marine Particles: Higher Diversity Recovered by in situ Pumps Than by The biological carbon pump (Volk and Hoffert, 1985) is a key component of the global Oceanographic Bottles. carbon cycle, removing CO2 from the atmosphere by gravitational settling of particulate organic Front. Microbiol. 11:1645. carbon. Microbial decomposition of particles during sinking affects their properties, in general doi: 10.3389/fmicb.2020.01645 leading to the attenuation of the flux of carbon with depth (Herndl and Reinthaler, 2013; Frontiers in Microbiology | www.frontiersin.org 1 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes Legendre et al., 2015). In fact, only a small percentage (<5%) Several studies comparing the chemical nature of particles (Antia et al., 2001; Honjo et al., 2008) of the particulate organic (particularly organic carbon, nitrogen, and trace metals) collected carbon produced in surface waters reaches abyssal depths, and with different methods have reported chemically different types whether the decomposition processes occur in upper or in of particles recovered even when the type and pore size of deep ocean layers determines how long carbon is sequestered filters are the same (Gardner et al., 2003; Liu et al., 2005, 2009; in the ocean (Kwon et al., 2009). Since not all microbes affect Bishop et al., 2012; Puigcorbé et al., 2015; Twining et al., 2015). organic matter in the same manner (Kujawinski et al., 2016; Recent studies have attempted to compare whether the recovered Robinson et al., 2018), the structure of the microbial community microbial communities differ when sampling with oceanographic growing on the sinking particles will impact the efficiency of bottles or with in situ filtration systems (Torres-Beltrán et al., carbon sequestration. 2019; Wang et al., 2019) reporting device-driven taxonomic A number of studies have characterized the microbial differences in most cases, but these studies targeted mostly free- communities colonizing the material that potentially sinks to living communities. To the best of our knowledge, no study the deep ocean, unveiling that the prokaryotes attached to those has explicitly assessed the impact that the sampling device can particles differ from free-living assemblages (DeLong et al., 1993; have on the recovered structure of the prokaryotic communities Rieck et al., 2015) and between different particle types (Mestre associated with marine particles. et al., 2017; Duret et al., 2019), display compositional changes In this context, we explored whether sampling marine during sinking (Pelve et al., 2017), and might even source particles with oceanographic bottles (BTL) or in situ pumps (ISP) prokaryotes to the bathypelagic realm (Mestre et al., 2018). In has an impact on our view of the structure of the associated general, most studies on deep-sea marine particles have used microbial communities. Using Illumina sequencing of the 16S size-fractionation to capture the microbes associated with both rRNA gene, we examined the method-driven differences on suspended and sinking material, collecting seawater with Niskin the prokaryotic communities associated with small (1–53 µm) oceanographic bottles and filtering it through different pore- and large (>53 µm) particles collected at mesopelagic depths size filters. Other less used methods include sediment traps, from two Antarctic polynyas. Large particles (>50 µm) have which recover mainly sinking particles but might miss those historically been considered to be the main vectors for carbon with low sinking rates (i.e., the so-called suspended particles; (and other elements) export from surface waters to depth, LeCleir et al., 2014; Fontanez et al., 2015; Boeuf et al., 2019), although there is a growing body of literature highlighting the or marine snow-catchers, which allow separating the suspended importance of smaller particles (∼1–50 µm) in the marine and sinking material present in a sample by their sinking biogeochemical cycles (Durkin et al., 2015; Puigcorbé et al., 2015; velocities (Duret et al., 2019). All these methods involve a Le Gland et al., 2019). Knowledge on the effect of the sampling certain time lapse between collection of particles and storage device on the recovery of microbial communities associated with of samples (ranging from several hours in the case of sampling particles of different sizes is essential for a deeper understanding with oceanographic bottles to several days, or even weeks, in of particle microbial ecology, and ultimately, their role in defining sediment traps) that might result in changes in particle-attached the efficiency of the transport of carbon and other elements of microbial communities. Even in those cases where a fixative is interest to depth. added to sediment traps (e.g., Gutierrez-Rodriguez et al., 2019), it is possible that not all microbial groups are well preserved if the trap is deployed at depth for a long time or that the MATERIALS AND METHODS particle-microbiome structure is not well preserved. In situ size-fractionated filtration with submersible pumps Study Area (ISP) is widely used in marine biogeochemical studies for Sampling was performed on board of the R/V Aurora Australis collecting suspended and sinking particles down to 5000 m (AA-V02 2016/17) between December 8, 2016, and January 21, (McDonnell et al., 2015, GEOTRACES program1 ). Besides 2017, in the Dalton and the Mertz polynyas (East Antarctica; reducing the time between particle collection and storage, the 67.2–66.8◦ S and 119.5–145.8◦ E, Table 1). Polynyas are highly in situ filtration keeps the sampled microbial communities at productive, with average phytoplankton concentrations being in situ conditions for most of the time, thus minimizing the more than twice of all the open waters around Antarctica (Arrigo growth of typical copiotrophic bacteria upon changes in pressure et al., 2015). Polynyas are areas where particle fluxes are expected and temperature. There is an increasing interest in developing to be significant, and they also harbor high densities of upper sampling systems based on in situ filtration (e.g., the Clio device, trophic level organisms such as birds and marine mammals Jakuba et al., 2018, or the ISMFF—in situ microbial filtration (Karnovsky et al., 2007). Further information regarding the and fixation, Wang et al., 2019) to facilitate large-scale microbial sampling area can be found in Moreau et al. (2019). Samples ecology studies (e.g., Saito et al., 2019). So far, however, ISP have were collected at two mesopelagic depths (100 and 300 m), using not been extensively used to sample particle-attached microbial either a 12L-Niskin bottle (BTL-samples) or an in situ pump communities (Lincoln et al., 2014), and thus, it is not known (ISP-samples) deployed at each depth. whether they successfully capture particle-attached microbial diversity in comparison to other more commonly used strategies. Particle Sampling For BTL samples, 7.5–10.5 L of water was collected in a carboy 1 www.geotraces.org and filtered onboard sequentially through a 47 mm diameter Frontiers in Microbiology | www.frontiersin.org 2 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes TABLE 1 | Station ID, sampling date, coordinates, depth and hydrochemical OTUs assigned to chloroplasts were removed, resulting in a total variables. of 6,868 OTUs. To enable comparisons between samples, the Polynya Station Date Lat. Lon. Depth Temp. Sal. OTU table was randomly subsampled to ensure an equal number (◦ N) (◦ E) (m) (◦ C) of sequences per sample (5,000 sequences), retaining 133,885 sequences clustered into 5,786 OTUs. The raw sequence data Dalton 2 31/December/16 −66.8 119.5 100 −1.94 34.26 have been deposited in the Figshare data repository, together 300 −1.85 34.32 with the non-rarefied OTU table, the taxonomy table and the Mertz EM03 10/January/17 −66.9 145.5 100 −1.57 34.45 environmental data used in this study, doi: 10.6084/m9.figshare. 300 −1.86 34.56 12333107. MG08 11/January/17 −67.2 145.9 100 −1.77 34.43 300 −1.88 34.53 Statistical Analyses The spatial differences between prokaryotic communities were 53 µm pore-size Nitex screen, followed by a 0.8 µm pore-size visualized using non-metric multidimensional scaling (NMDS, polycarbonate membrane filter (Millipore) (for simplification Vegan metaMDS function) based on Bray–Curtis distances. hereafter the 0.8–53 µm will be refered as 1–53 µm) using Significant differences in taxonomic composition between a Masterflex peristaltic pump. For ISP samples, the pumps depths, stations, sampling method, or size fraction were (McLane) had heads (trace metal clean 3-tier model) equipped tested using ANOSIM (Vegan anosim function). Device-driven with a 142 mm diameter 53 µm pore Nitex screen, followed differences in community structure for each size fraction were by a 1 µm pore quartz filter (QMA), and filtered, on average, further calculated at each station as Bray–Curtis dissimilarities 375 L of water during ∼2 h. To make the comparison of sampled between ISP and BTL assemblages. OTUs unique to a given volumes more even between the BTL and ISP, we took subsamples sampling method (i.e., OTUs present exclusively in BTL or in of the material retained in the ISP filters. The subsamples for ISP) or shared between both methods were identified considering the 1–53 µm faction were obtained by cutting 25 mm replicate each sample separately. Finally, in order to explore whether each punches from a 142 mm QMA filter. The >53 µm fraction method recovered preferentially different taxonomic groups, was subsampled by rinsing the screen with filtered (<0.2 µm) indicator species analysis was done using the indval function of seawater and dividing the obtained volume into various aliquots the R package LabDSV (Roberts, 2013), at a significance level of that were subsequently filtered through QMA filters. As a result, p ≤ 0.05, pooling all samples together. Statistical analyses and the 1–53 µm and >53 µm ISP samples were representative of data handling were done in R (R Core Team, 2017). about 10 and 35 L, respectively. Considering the volume filtered and the filter diameter for BTL and ISP, the ISP filtered on average about 4 times more volume per unit of area than the BTL (2 L RESULTS cm−2 vs. 0.5 L cm−2 ). The taxonomic composition of the studied particle-attached Duplicates or triplicates were obtained from the 1–53 µm communities was first examined based on potential differences fractions and for most of the >53 µm fractions (Supplementary driven by station, depth, particle size, and/or sampling device. Figure S1). After the recovery of the ISP, the filters were prepared We observed that whereas communities did not cluster based and stored at −80◦ C within the following hour. The processing on station or depth, particle size and sampling device resulted time for BTL samples was about 2–3 times longer, with a CTD in clear taxonomic differences between communities (ANOSIM deployment of about 2 h, followed by storage at 4◦ C for a few R = 0.54 and 0.43, respectively, p < 0.001, Figure 1A). The hours until filtration and final storage at −80◦ C. fact that the ISP replicates showed high similarities (Figure 1A) discards a bias due to subsampling of the ISP filter (see section Characterization of Prokaryotic “Materials and Methods”): Bray–Curtis dissimilarities between Communities replicates of a sample were significantly smaller (0.301 ± 0.030) The prokaryotic community structure was determined by high- than between different samples (0.718 ± 0.007; Supplementary throughput Illumina sequencing of the 16S rRNA genes. A total Figure S1). Notably, the device-driven differences in taxonomic of 27 samples were obtained, including the replicates, which were composition of the microbial communities for a given sample stored at −80◦ C until further analyses. At the home laboratory, were much more conspicuous for the large particles than for DNA was extracted using the PowerWater DNA Isolation kit the small particles (Figure 1B), suggesting that the effect of the following the manufacturer’s instructions (MOBIO). The samples sampling method affects differentially particles of different sizes. were sequenced using Illumina MiSeq 2×300 bp flow cells at Moreover, communities from the large size-fractions collected RTL Genomics (Texas, United States) using primers 515F-Y with ISP appeared more heterogeneous across sampling sites than and 926R (Parada et al., 2015) to amplify the V4-V5 region of BTL assemblages; whereas the former clustered into two groups, the 16S rRNA gene. The sequences were processed following the latter clustered closely together (Figure 1A). (Logares, 2017). Quality-checked sequences were clustered into At the two stations in which both, large and small particles OTUs (operational taxonomic units) at 99% similarity. Singletons were collected (EM03 and MG08), we explored the Bray–Curtis and chimeric OTUs were removed, and the remaining OTUs dissimilarity between small and large fractions sampled with each were taxonomically annotated using the SILVA v123 database. device within a given site. We found that the average values Frontiers in Microbiology | www.frontiersin.org 3 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes FIGURE 1 | (A) Non-multidimensional scaling analysis (NMDS) based on the Bray–Curtis dissimilarity between all the studied communities collected with either in situ pumps (ISP, asterisks) or bottles (BTL, dots), for two different particle sizes (1–53 µm and >53 µm, indicated by the size of the symbol). Each color represents a community. (B) Box plot showing the distribution of pairwise Bray–Curtis dissimilarities between sampling devices for each size fraction. were 0.51 ± 0.27 and 0.86 ± 0.11 for ISP and BTL, respectively sample). Remarkably, the majority of the OTUs (range 73–99%) (Table 2) indicating that the community structure in the two found in a given ISP community were never detected in its BTL fractions were more similar in the ISP than in the BTL samples. counterpart. Likewise, 60–97% of the OTUs in BTL communities In order to assess if the device-driven structural differences were not recovered by the ISP. The “ISP-only” OTUs were mostly involved changes in taxonomic richness and composition, for rare OTUs, representing in most cases a much lower proportion each particular community we classified the OTUs as only present of the total number of sequences (14–46%) compared to the in the BTL sample, only present in the ISP sample, or present “BTL-only” OTUs, which accounted for a larger fraction of BTL- in both types of samples (Figure 2A). Overall, we found that community sequences (11–62%) (Figure 2B). Only in one case ISP-communities contained 1.5- to >3-fold more OTUs than (Mertz polynya, station EM03, 300 m depth) we found almost no BTL-communities at a similar sequencing depth (5,000 reads per overlap between the communities captured by the two methods (Figure 2B). In general, however, OTUs present in both ISP and BTL communities comprised a relatively large fraction of the sequences from each assemblage (range 39–88%, Figure 2B). TABLE 2 | Taxonomic differences (as measured by Bray–Curtis dissimilarity) between small and large size-fractions collected with either BTL or ISP devices. In most cases, ISP-communities harbored a higher diversity of groups, including OTUs belonging to Planctomycetes, Station Depth Device Bray–Curtis dissimilarity Deltaproteobacteria, or Verrucomicrobia that were almost absent (large vs. small)* from BTL communities, which were mostly comprised by EM03 100 BTL 0.81 Alpha-, Gammaproteobacteria, or Flavobacteriia (Figure 3A and ISP 0.46 Supplementary Table S1). In order to explore whether there 300 BTL 0.98 were specific taxa that were preferentially captured by each ISP 0.26 sampling strategy, we identified BTL- and ISP-indicator OTUs MG08 100 BTL 0.73 pooling together all BTL and all ISP samples, respectively. ISP 0.80 Out of the total 5,786 OTUs, we detected 123 ISP-indicators 300 BTL 0.93 and 129 BTL-indicators (Figures 3B,C), which accounted for ISP 0.89 up to 49% of the ISP community sequences (Figure 3B) and up to 85% of the BTL community sequences (Figure 3C), *Bray–Curtis values represent single values if only one sample replicate of each size fraction was available, or average of the pairwise distances in those cases where respectively. These results suggest that there was preferential more than one replicate was collected. recovery depending on the sampling device, since BTL indicators Frontiers in Microbiology | www.frontiersin.org 4 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes FIGURE 2 | (A) Number of total OTUs detected in each sample, distinguishing between OTUs detected exclusively with in situ pumps (ISP, orange), with Niskin bottles (BTL, blue) or shared between both devices (black). (B) Contribution of these three groups of unique or shared OTUs to total community sequences. The depths and the stations sampled are indicated on the X-axis of panel (B). comprised almost exclusively Alpha- and Gammaproteobacterial methods yield a similar picture of particle-attached microbial OTUs (Figure 3C), while OTUs belonging to Flavobacteriia, assemblages, even though different sampling devices are known Sphingobacteriia, and several Planctomycetes groups were to collect chemically different particles (Liu et al., 2005, 2009; exclusively detected among ISP indicators (Figure 3B). Puigcorbé et al., 2015; Twining et al., 2015). Here we provide the first comparison of the impact of two types of sampling devices, oceanographic bottles and in situ pumps, on the particle- DISCUSSION attached microbial communities. Even though these two methods should capture essentially the same particles as they are not Since the advent of DNA sequencing technologies, there has intended to preferentially collect sinking or suspended material, been a rising number of studies on the ecology and diversity of our results show that the microbial communities recovered are prokaryotic communities associated to marine sinking particles largely different. (DeLong et al., 1993; Boeuf et al., 2019). However, there are Regardless of the method used, the taxonomic structure of still methodological considerations that limit our understanding the studied microbial communities differed strongly depending of the microbial ecology of particles. For example, there is on the particle size, in agreement with previous investigations no clear consensus on the filter pore size to use to effectively indicating that particles of different sizes are colonized by distinct separate particles of different kinds (see references in Mestre, microbial communities (e.g., Mestre et al., 2018). However, the 2017), known to be colonized by largely different microbial differences due to the sampling device were larger than those communities (Mestre et al., 2018; Duret et al., 2019). Although due to the different sampling stations and depths. ISP and BTL- both in situ pumps and filtering of water collected with Niskin derived communities differed in terms of overall taxonomic bottles have been used to sample particle-attached microbial structure, number of OTUs and identity of the taxa recovered, communities (e.g., Lincoln et al., 2014; Mestre et al., 2018) with ISP generally recovering more OTUs and a higher diversity there has been no attempt to compare whether these different of taxonomic groups than the BTL sampling. Hereafter we will Frontiers in Microbiology | www.frontiersin.org 5 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes FIGURE 3 | (A) Taxonomic composition of the communities sampled with either in situ pumps (ISP, asterisks) or Niskin bottles (BTL, dots). The size of the symbols indicates communities associated with small or large particles. The depths and the stations sampled are indicated on the X-axis of panel (A) (see Table 1). (B,C) Identification of ISP- or BTL-indicator OTUs (see section “Materials and Methods”). (B) Relative contribution (% of community sequences) and taxonomic composition of ISP-indicator OTUs. (C) Relative contribution (% of community sequences) and taxonomic composition of BTL-indicator OTUs. The classification was performed at the class level, and before each class, the phylum is indicated: Thaum, Thaumarchaeota; Actin, Actinobacteria; Bact, Bacteroidetes; Lent, Lentisphaerae; Planc, Planctomycetes; Prot, Proteobacteria; Ten, Tenericutes; Verr, Verrucomicrobia. discuss the potential causes underlying such differences, which vs. QMA 1 µm), which might have contributed to some of the include the use of different pore size filters, filter clogging, differences observed in the sampled communities. However, the microbial community turnover during sample processing, or same type of 53 µm screen was used for the large size-fractions other sampling issues. in both devices, and this size-fraction was precisely the one For practical reasons, size fractionation and sampled volumes showing the greatest device-driven differences. This suggests that between both sampling devices could not be identical. The filters the difference in the pore size and filter type in small size fraction used for collecting the small size-fractions were slightly different was not the main reason explaining the device-driven changes in between the BTL and the ISP samples (Polycarbonate 0.8 µm microbial composition. Frontiers in Microbiology | www.frontiersin.org 6 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes Taxonomic differences in particle-attached prokaryotes have small particles, as assemblages from large particles showed larger sometimes been related to differences in the filtered volume. differences in taxonomic structure depending on the device used. For example, taxonomic richness in prokaryotic free-living When sampling with BTL, fast sinking particles, which differ communities (0.22–1.6 µm) was found to decrease with in community composition from suspended particles (Duret increasing filtered volume (0.1–5 L), whereas it generally et al., 2019) could accumulate at the bottom of the Niskin even increased in the attached (>1.6 µm) fraction (Padilla et al., during transit of the rosette to the surface and be undersampled, 2015). The authors claimed that this was due to progressive especially when not all content of the bottle is filtered (Suter filter clogging (i.e., small cells being increasingly retained in the et al., 2017). In addition, we observed that the large differences large fraction). Our ISP samples filtered on average four times indicated by the NMDS between the ISP large size-fractions more volume per filter area than BTL (see section “Materials were not reflected by their BTL counterparts, which clustered and Methods”), which might have resulted in a more intense more closely together in a single group. Similarly, the NMDS clogging of ISP than BTL filters. Based on the results of Padilla ordination patterns of the small size-fraction communities from et al. (2015), clogging of ISP filters should have caused higher the ISP and BTL communities did not resemble each other. This richness in ISP than in BTL samples in the large-size fractions, but suggests that different sampling methods will affect not only the the opposite pattern in the small size-fractions. As we recovered taxonomic composition and richness, but also the beta-diversity much higher taxonomic diversity of taxa in both size fractions patterns (i.e., the observed differences between sites or stations) in ISP samples, clogging might have not be the main reason and hence, their links with environmental or spatio-temporal explaining the differences between devices. Besides, we did not gradients. These discrepancies could be due to the previously observe visual signs of clogging in any of the two fractions. As mentioned growth of copiotrophs in oceanographic bottles, an alternative explanation, also suggested by Padilla et al. (2015), explaining the observed homogenization of BTL communities, small volumes might fail to capture a fraction of the particles due or could be caused by fragmentation of large aggregates and to microscale patchiness in particle distribution (Long and Azam, loss of diversity during ISP filtration (Bishop et al., 2012; Maiti 2001). Hence, the larger amount of water filtered by the ISP et al., 2012). The sampling device could also have impacted could be much more efficient at recovering certain types of rarer the observed taxonomic differences between small and large particles and their associated communities. Should this be the size-fractions. It could be hypothesized that large particles case, ISP would clearly be advantageous for the characterization can be destroyed when collected with the BTL and converted of particle-attached communities with respect to Niskin bottles. into smaller particles because of the physical filtration with a Although it could be tested whether filtering larger volumes peristaltic pump. This would imply smaller differences between of water collected with Niskin bottles would lead to a more large and small particles collected with the BTL than with the ISP. representative characterization of the particle-attached bacterial However, direct comparison of the Bray–Curtis dissimilarities diversity, there are generally inherent practical difficulties in (Table 2) indicate that in BTL samples there were larger or doing so given the limitations in water budgets and filtration similar taxonomic differences between both particle sizes as times during oceanographic cruises. compared to ISP samples, thus discarding this hypothesis. In Another potential bias could be due to microbial population any case, these contrasting patterns have important implications turnover within the sample that may occur between particle for our capacity to understand the ecological drivers of particle- collection and sample processing, as ISP samples are stored more attached communities. quickly and are less manipulated than BTL samples; the ISP Interestingly, despite the higher taxonomic richness recovered samples were collected during an average of 2 h pumping time, with ISP, we also detected a significant number of OTUs exclusive and in less than 1 h the filters were cut or rinsed and stored of BTL-samples that were not detected with ISP. We thus believe at −80◦ C. On the contrary, the BTL samples were collected that the most likely explanation for the observed differences is during a CTD deployment (∼2 h), they then remained at low that the two devices are capturing different types of particles. temperature (4◦ C) until filtration a few hours later, which took Indeed, other studies have unveiled differences in the chemical place at room temperature (∼20◦ C). Such a larger processing composition of particles collected by either BTL or ISP (Liu time between sample collection and final storage of the filters et al., 2005, 2009; Twining et al., 2015), which were attributed at −80◦ C could have allowed the growth of some opportunistic to sampling designs, undersampling of certain particle types, prokaryotic groups. The lower taxonomic richness observed in or loss of labile compounds. Similarly, the two studies that BTL communities actually points to the growth and dominance compared free-living assemblages (>0.22 or 0.44 µm) sampled of a few taxa, and we observed higher relevance of some typical with bottles or in situ filtration devices reported taxonomic copiotrophic groups in the BTL samples, where Alteromonadales differences (Torres-Beltrán et al., 2019; Wang et al., 2019) and and Vibrionales were highly enriched (Supplementary Table S1). suggested that in situ sampling could overcome biases linked However, and as part of the larger study in which these results to bottle sampling and provide a more accurate representation were obtained, we also sampled other stations and water depths of the actual microbial structure. Our results agree with this with the ISP in which up to 80% of the sequences belonged to interpretation and suggest that ISP generally capture a higher these bacterial groups (details not shown), which would suggest diversity of particle-attached microbes. For example, groups like that their dominance is not solely the result of Niskin enclosure. Bacteroidetes, Planctomycetes, and Deltaproteobacteria, which Interestingly, the sampling method seemed to be more were recovered preferentially with ISP, have shown a consistent relevant for communities associated to the large than to the preference for a particle-attached lifestyle across the global Frontiers in Microbiology | www.frontiersin.org 7 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes mesopelagic and bathypelagic ocean (Salazar et al., 2015; Mestre AUTHOR CONTRIBUTIONS et al., 2018). While it is true that we cannot discard that some of the differences are caused by the differences in the amount VP, JG, and PM participated in the design of the sampling of volume filtered, the higher diversity recovered by the ISP scheme. VP performed the sampling and sample processing. actually reflects a higher efficiency of this sampling device to CR-G, VP, and JG conducted the data analysis. VP and capture the complexity of prokaryotic communities associated to CR-G compiled the needed data, analyzed the data, and wrote particles in the ocean. It should be noted, however, that ISP also this manuscript. All authors made relevant comments and have potential biases, such as the rupture of fragile aggregates editions to the text. due to cross filter pressure, or the loss of some of the large particles collected with a mesh, due to washout when recovering the pumps (Bishop et al., 2012). In particular, the type of filter FUNDING holders used in our study had been shown to lose large biogenic particles (Bishop et al., 2012), so it is possible that the differences VP received funding from Edith Cowan University (G1003456) in taxonomic richness between ISP and BTL large size fractions and from the School of Science at Edith Cowan University might be even larger than what we show here. We tried to (G1003362) to support this work. CR-G and JG were minimize these issues by filtering at a flow rate of maximum 3.8 L supported by the grants CTM2015-70340-R and RTI2018- min−1 [the limit to preserve sample integrity is thought to be 6 L 101025-B-I00 of the Spanish Ministry of Science, Innovation min−1 (Maiti et al., 2012)], cared for a smooth device recovery, and Universities and by the Generalitat de Catalunya and drained the excess of water once the pumps were on deck. In Consolidated Research Group 2017SGR/1568. PM acknowledges any case, there is still a clear need to explore in detail the reasons the support of the Generalitat de Catalunya (MERS 2017 behind the observed differences before we can unequivocally SGR – 1588). determine which sampling method is the most suitable for the characterization of particle-attached prokaryotic assemblages. ACKNOWLEDGMENTS CONCLUSION We are grateful to the crew of RSV Aurora Australis Our results suggest that sampling marine particles with in situ for their hospitality and help during the 2016/2017 AA- pumps or oceanographic bottles provides a significantly different v2 voyage. Thanks also to the Australian Antarctic Division view of the particle-attached microbial communities recovered, technical support crew for always finding ways to make calling for caution when studies using different methodologies things work and to all the scientific party for their tireless are to be compared. We show that sampling with in situ pumps curiosity and motivation to help. We are particularly grateful recovers more complex, diverse and heterogeneous particle- to the biogeochemistry group, led by Delphine Lannuzel, attached prokaryotic communities than sampling with Niskin Sébastien Moreau, Matthew Corkill, Julie Janssens, Lavenia bottles, and that some taxa are exclusively captured with one Ratnarajah, Mar Arroyo, and Cristina Genovese. We warmly device or the other. To our knowledge, this is the first evidence thank Montserrat Roca-Martí for her invaluable help with of compositional differences in particle-attached assemblages sample collection and preparation and Vanessa Balagué for caused by the sampling methods. Yet given that BTL and ISP are DNA extraction and initial data processing. We want to likely sampling different particle pools, combining devices may thank two reviewers for their constructive comments. This offer an opportunity for a more thorough understanding of the work contributes to the ICTA Unit of Excellence (MinECo, particle microbial ecology and its implications in the assessment MDM2015-0552). The IAEA is grateful for the support provided of the carbon sequestration in the oceans. to its Environment Laboratories by the Government of the Principality of Monaco. DATA AVAILABILITY STATEMENT SUPPLEMENTARY MATERIAL The raw sequence data have been deposited in the Figshare data repository, together with the non-rarefied OTU table, the The Supplementary Material for this article can be found taxonomy table and the environmental data used in this study, online at: https://www.frontiersin.org/articles/10.3389/fmicb. doi: 10.6084/m9.figshare.12333107. 2020.01645/full#supplementary-material REFERENCES Arrigo, K. R., van Dijken, G. L., and Strong, A. L. (2015). Environmental controls of marine productivity hot spots around Antarctica. J. Geophys. Res. Ocean 120, Antia, A., Koeve, W., Fischer, G., Blanz, T., Schulz-Bull, D., Scholten, J., et al. 5545–5565. doi: 10.1002/2015JC010888 (2001). Basin-wide particulate carbon flux in the Atlantic Ocean: regional Bishop, J. K. B., Lam, P. J., and Wood, T. J. (2012). Getting good particles: accurate export patterns and potential for atmospheric CO2 sequestration. Glob. sampling of particles by large volume in-situ filtration. Limnol. Oceanogr. Biogeochem. Cycles 15, 845–862. doi: 10.1029/2000GB001376 Methods 10, 681–710. doi: 10.4319/lom.2012.10.681 Frontiers in Microbiology | www.frontiersin.org 8 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes Boeuf, D., Edwards, B. R., Eppley, J. M., Hu, S. K., Poff, K. E., Romano, A. E., et al. sometimes differ from those using in-situ pumps? Deep Sea Res. Part I Oceanogr. (2019). Biological composition and microbial dynamics of sinking particulate Res. Pap. 52, 1324–1344. doi: 10.1016/j.dsr.2005.02.005 organic matter at abyssal depths in the oligotrophic open ocean. Proc. Natl. Logares, R. (2017). Workflow for Analysing MiSeq Amplicons Based on Uparse v1.5. Acad. Sci. U.S.A. 116, 11824–11832. doi: 10.1073/pnas.1903080116 Available online at: https://github.com/ramalok/amplicon_processing DeLong, E. F., Franks, D. G., and Alldredge, A. L. (1993). Phylogenetic diversity Long, R. A., and Azam, F. (2001). Microscale patchiness of bacterioplankton of aggregate−attached vs. free−living marine bacterial assemblages. Limnol. assemblage richness in seawater. Aquat. Microb. Ecol. 26, 103–113. doi: 10.3354/ Oceanogr. 38, 924–934. doi: 10.4319/lo.1993.38.5.0924 ame026103 Duret, M. T., Lampitt, R. S., and Lam, P. (2019). Prokaryotic niche partitioning Maiti, K., Buesseler, K. O., Pike, S. M., Benitez-Nelson, C., Cai, P., Chen, W., et al. between suspended and sinking marine particles. Environ. Microbiol. Rep. 11, (2012). Intercalibration studies of short-lived Thorium-234 in the water column 386–400. doi: 10.1111/1758-2229.12692 and marine particles. Limnol. Ocean. Methods 10, 631–644. doi: 10.4319/lom. Durkin, C. A., Estapa, M. L., and Buesseler, K. O. (2015). Observations of carbon 2012.10.631 export by small sinking particles in the upper mesopelagic. Mar. Chem. 175, McDonnell, A. M. P., Lam, P. J., Lamborg, C. H., Buesseler, K. O., Sanders, R., Riley, 72–81. doi: 10.1016/j.marchem.2015.02.011 J. S., et al. (2015). The oceanographic toolbox for the collection of sinking and Fontanez, K. M., Eppley, J. M., Samo, T. J., Karl, D. M., and DeLong, E. F. suspended marine particles. Prog. Oceanogr. 133, 17–31. doi: 10.1016/j.pocean. (2015). Microbial community structure and function on sinking particles in 2015.01.007 the North Pacific Subtropical Gyre. Front. Microbiol. 6:469. doi: 10.3389/fmicb. Mestre, M. (2017). Spatial and Temporal Patterns of Marine Prokaryotic Diversity 2015.00469 Along the Particulate Matter Continuum. Available online at: http://hdl.handle. Gardner, W. D., Richardson, M. J., Carlson, C. A., Hansell, D., and Mishonov, net/10261/156339 (accessed November 10, 2019). A. V. (2003). Determining true particulate organic carbon: bottles, pumps Mestre, M., Borrull, E., Sala, M. M., and Gasol, J. M. (2017). Patterns of bacterial and methodologies. Deep Sea Res. Part II Top. Stud. Oceanogr. 50, 655–674. diversity in the marine planktonic particulate matter continuum. Isme J. 11:999. doi: 10.1016/S0967-0645(02)00589-1 doi: 10.1038/ismej.2016.166 Gutierrez-Rodriguez, A., Stukel, M. R., Lopes dos Santos, A., Biard, T., Scharek, R., Mestre, M., Ruiz-González, C., Logares, R., Duarte, C. M., Gasol, J. M., and Sala, Vaulot, D., et al. (2019). High contribution of Rhizaria (Radiolaria) to vertical M. M. (2018). Sinking particles promote vertical connectivity in the ocean export in the California current ecosystem revealed by DNA metabarcoding. microbiome. Proc. Natl. Acad. Sci. U.S.A. 115, E6799–E6807. doi: 10.1073/pnas. ISME J. 13, 964–976. doi: 10.1038/s41396-018-0322-7 1802470115 Herndl, G. J., and Reinthaler, T. (2013). Microbial control of the dark end of the Moreau, S., Lannuzel, D., Janssens, J., Arroyo, M. C., Corkill, M., Cougnon, E., biological pump. Nat. Geosci. 6, 718–724. doi: 10.1038/NGEO1921 et al. (2019). Sea ice meltwater and circumpolar deep water drive contrasting Honjo, S., Manganini, S. J., Krishfield, R. A., and Francois, R. (2008). Particulate productivity in three antarctic polynyas. J. Geophys. Res. Ocean. 124, 2943–2968. organic carbon fluxes to the ocean interior and factors controlling the biological doi: 10.1029/2019JC015071 pump: a synthesis of global sediment trap programs since 1983. Prog. Oceanogr. Padilla, C. C., Ganesh, S., Gantt, S., Huhman, A., Parris, D. J., Sarode, N., 76, 217–285. doi: 10.1016/j.pocean.2007.11.003 et al. (2015). Standard filtration practices may significantly distort planktonic Jakuba, M. V., Breier, J. A., Gómez-Ibáñez, D., Tradd, K., and Saito, M. A. (2018). microbial diversity estimates. Front. Microbiol. 6:547. doi: 10.3389/fmicb.2015. “Clio: an autonomous vertical sampling vehicle for global ocean biogeochemical 00547 mapping,” in Proceedings of the 2018 IEEE/OES Autonomous Underwater Parada, A. E., Needham, D. M., and Fuhrman, J. A. (2015). Every base matters: Vehicle Workshop (AUV) (Porto: IEEE), 1–8. doi: 10.1109/AUV.2018.8729797 assessing small subunit rRNA primers for marine microbiomes with mock Karnovsky, N., Ainley, D. G., and Lee, P. (2007). “the impact and importance communities, time series and global field samples. Environ. Microbiol. 18, of production in polynyas to top-trophic predators: three case histories,” in 1403–1414. doi: 10.1111/1462-2920.13023 Polynyas: Windows to the World Elsevier Oceanography Series, eds W. O. Smith, Pelve, E. A., Fontanez, K. M., and DeLong, E. F. (2017). Bacterial succession on and D. G. Barber (Amsterdam: Elsevier), 391–410. doi: 10.1016/S0422-9894(06) sinking particles in the Ocean’s interior. Front. Microbiol. 8:2269. doi: 10.3389/ 74012-0 fmicb.2017.02269 Kujawinski, E. B., Longnecker, K., Barott, K. L., Weber, R. J. M., and Kido Soule, Puigcorbé, V., Benitez-Nelson, C. R., Masqué, P., Verdeny, E., White, A. E., M. C. (2016). Microbial community structure affects marine dissolved organic Popp, B. N., et al. (2015). Small phytoplankton drive high summertime matter composition. Front. Mar. Sci. 3:45. doi: 10.3389/fmars.2016.00045 carbon and nutrient export in the Gulf of California and Eastern Tropical Kwon, E. Y., Primeau, F., and Sarmiento, J. L. (2009). The impact of North Pacific. Glob. Biogeochem. Cycles 29, 1309–1332. doi: 10.1002/2015GB0 remineralization depth on the air–sea carbon balance. Nat. Geosci. 2:630. doi: 05134 10.1038/ngeo612 R Core Team (2017). R: A Language and Environment for Statistical Computing. Le Gland, G., Aumont, O., and Mémery, L. (2019). An estimate of thorium 234 Vienna: R Project for Statistical Computing. partition coefficients through global inverse modeling. J. Geophys. Res. Ocean. Rieck, A., Herlemann, D. P. R., Jürgens, K., and Grossart, H.-P. (2015). Particle- 124, 3575–3606. doi: 10.1029/2018JC014668 associated differ from free-living bacteria in surface waters of the Baltic sea. LeCleir, G. R., DeBruyn, J. M., Maas, E. W., Boyd, P. W., and Wilhelm, S. W. Front. Microbiol. 6:1297. doi: 10.3389/fmicb.2015.01297 (2014). Temporal changes in particle-associated microbial communities after Roberts, D. W. (2013). Labdsv: Ordination and Multivariate Analysis for Ecology. interception by nonlethal sediment traps. FEMS Microbiol. Ecol. 87, 153–163. R package. Available online at: https://cran.r-project.org/web/packages/labdsv/ doi: 10.1111/1574-6941.12213 index.html Legendre, L., Rivkin, R. B., Weinbauer, M. G., Guidi, L., and Uitz, J. (2015). The Robinson, C., Wallace, D., Hyun, J.-H., Polimene, L., Benner, R., Jiao, N., et al. microbial carbon pump concept: potential biogeochemical significance in the (2018). An implementation strategy to quantify the marine microbial carbon globally changing ocean. Prog. Oceanogr. 134, 432–450. doi: 10.1016/j.pocean. pump and its sensitivity to global change. Natl. Sci. Rev. 5, 474–480. doi: 10. 2015.01.008 1093/nsr/nwy070 Lincoln, S. A., Wai, B., Eppley, J. M., Church, M. J., Summons, R. E., and DeLong, Saito, M. A., Bertrand, E. M., Duffy, M. E., Gaylord, D. A., Held, N. A., Hervey, E. F. (2014). Planktonic Euryarchaeota are a significant source of archaeal W. J., et al. (2019). Progress and challenges in ocean metaproteomics and tetraether lipids in the ocean. Proc. Natl. Acad. Sci. 111, 9858–9863. doi: 10. proposed best practices for data sharing. J. Proteome Res. 18, 1461–1476. doi: 1073/pnas.1409439111 10.1021/acs.jproteome.8b00761 Liu, Z., Cochran, J. K., Lee, C., Gasser, B., Miquel, J. C., and Wakeham, S. G. (2009). Salazar, G., Cornejo−Castillo, F. M., Borrull, E., Díez−Vives, C., Lara, E., Vaqué, Further investigations on why POC concentrations differ in samples collected D., et al. (2015). Particle−association lifestyle is a phylogenetically conserved by Niskin bottle and in situ pump. Deep Sea Res. Part II Top. Stud. Oceanogr. trait in bathypelagic prokaryotes. Mol. Ecol. 24, 5692–5706. doi: 10.1111/mec. 56, 1558–1567. doi: 10.1016/j.dsr2.2008.12.019 13419 Liu, Z., Stewart, G., Cochran, J. K., Lee, C., Armstrong, R. A., Hirschberg, D. J., et al. Suter, E. A., Scranton, M. I., Chow, S., Stinton, D., Medina Faull, L., and Taylor, (2005). Why do POC concentrations measured using Niskin bottle collections G. T. (2017). Niskin bottle sample collection aliases microbial community Frontiers in Microbiology | www.frontiersin.org 9 July 2020 | Volume 11 | Article 1645 Puigcorbé et al. Effect of Sampling Method on Particle-Attached Prokaryotes composition and biogeochemical interpretation. Limnol. Oceanogr. 62, 606– DC: American Geophysical Union), 99–110. doi: 10.1029/GM032 617. doi: 10.1002/lno.10447 p0099 Torres-Beltrán, M., Mueller, A., Scofield, M., Pachiadaki, M. G., Taylor, C., Wang, Y., Gao, Z.-M., Li, J., He, L.-S., Cui, G.-J., Li, W.-L., et al. (2019). Hadal water Tyshchenko, K., et al. (2019). Sampling and processing methods impact sampling by in situ microbial filtration and fixation (ISMIFF) apparatus. Deep microbial community structure and potential activity in a seasonally anoxic Sea Res. Part I Oceanogr. Res. Pap. 144, 132–137. doi: 10.1016/j.dsr.2019.01.009 fjord: Saanich Inlet, British Columbia. Front. Mar. Sci. 6:132. doi: 10.3389/ fmars.2019.00132 Conflict of Interest: The authors declare that the research was conducted in the Twining, B. S., Rauschenberg, S., Morton, P. L., Ohnemus, D. C., absence of any commercial or financial relationships that could be construed as a and Lam, P. J. (2015). Comparison of particulate trace element potential conflict of interest. concentrations in the North Atlantic Ocean as determined with discrete bottle sampling and in situ pumping. Deep Sea Res. Part Copyright © 2020 Puigcorbé, Ruiz-González, Masqué and Gasol. This is an open- II Top. Stud. Oceanogr. 116, 273–282. doi: 10.1016/j.dsr2.2014. access article distributed under the terms of the Creative Commons Attribution 11.005 License (CC BY). The use, distribution or reproduction in other forums is permitted, Volk, T., and Hoffert, M. I. (1985). “Ocean carbon pumps: analysis of relative provided the original author(s) and the copyright owner(s) are credited and that the strengths and efficiencies in ocean-driven atmospheric CO2 changes,” original publication in this journal is cited, in accordance with accepted academic in The Carbon Cycle and Atmospheric CO2: Natural Variations Archean practice. No use, distribution or reproduction is permitted which does not comply to Present, eds E. T. Sundquist, and W. S. Broecker (Washington, with these terms. Frontiers in Microbiology | www.frontiersin.org 10 July 2020 | Volume 11 | Article 1645