Organic matter processing by microbial communities throughout the Atlantic water column as revealed by metaproteomics

As a first step toward establishing a proteomic inventory of marine microbial communities, we performed protein mass spectrometry (Fig. S1) at 14 sample sites (Table S1) in the Atlantic Ocean (Fig. S2). Additionally to protein annotations retrieved by homology searches against genomic databases (Supporting Information and Fig. S3), functional classifications were inferred using (i) clusters of orthologous groups (COGs) of proteins and (ii) the Kyoto Encyclopedia of Genes and Genomes (KEGG) database for higher-order cellular processes (Table 1). A combined set of 1,002 nonredundant transport-related membrane proteins (29% of total proteins identified; Datasets S1 and S2), subsequently referred to as TMPs, was identified. Approximately 71% of the TMPs were identified in at least two depth layers, whereas only 7% (75 proteins) were detected in all metaproteomes (Fig. 1A). COG categories, defined for the transporter protein repertoire, resulted in a conserved set of 96 COG families (Dataset S3), subsequently referred to as transporter COGs. The comparison between the major water layers, namely the euphotic, the mesopelagic (300–850 m depth), and upper (1,500–2,000 m depth) and lower (2,750–4,050 m) bathypelagic zones, enabled us to identify 29 COG families (∼30%) common to all metaproteomes (i.e., involved in amino acid, inorganic ion, and carbohydrate transport; Fig. 1B). Among the 31 COG families present exclusively in the bathypelagic layer, we identified transporter COGs from the categories of “Carbohydrate transport and metabolism” (G; i.e., organic acids, xylose, maltose), “Inorganic ion transport and metabolism” (P), and “Poorly characterized” (R). The high number of unique transporter COGs recovered from deep ocean likely reflects the overall higher number of proteins recovered from these samples. Noteworthy, despite the greater diversity of differential transporter COGs, no additional substrate specificities were predicted from the protein-coding sequences. This observation points at the importance of suitable and “complete” genomic databases and to accurately annotate protein-coding sequences. These results highlight also the potentially broad substrate range encounter by marine microbes, which needs yet to be defined. To account for differences in protein recovery between the samples, we applied semiquantitative analyses of transporter COGs, introducing normalized area abundance factors (NAAFs) (Supporting Information). Resulting expression profiles of transporter COGs indicate that functionally similar transport processes are present throughout the water column yet differ in their relative abundances between sampling sites (Fig. 2).

Similar to previous metaproteomic studies (14, 21, 24, 25, 27, 28), ABC transporters (662 proteins, 20% of total identified proteins) comprised the most prevalent transporter system in our study. As reported from coastal surface waters off the Antarctic Peninsula (14), the SBPs were the most frequently encountered components of the ABC complex, whereas transmembrane domains were less often identified (4%; Dataset S1). Roughly 85% of ABC transporters were predicted periplasmic SBPs and 11% remained unassigned. TRAP transporters were among the 10 most expressed proteins in the metaproteomic repertoire (86 proteins; 9% of TMP and 2.5% of total identified proteins) and mostly comprised SBPs homologous to the DctP family. Only three SBPs were of the TRAP-associated extracytoplasmic immunity (TAXI) family (29, 30) (Dataset S1). Together with TRAP transporters (31), TTTs are considered as secondary active transporters found in Bacteria and Archaea that employ SBPs to capture their ligands (29, 30, 32, 33). TTTs accounted for the smallest fraction of prokaryotic transport systems (29 proteins; 3% of TMP and 1% of total identified proteins; Fig. 2), which might be explained by their narrow substrate range we know of until now (31, 34, 35), or due to the limited number of reference sequences available in public databases.

The third most abundant transporter system was the outer-membrane TBDT, which plays an important role in high–molecular-weight OM uptake (22). Metaproteomic (21), metagenomic (26), and metatranscriptomic (36) surveys indicate the ocean-wide (coastal to open ocean) presence of TBDTs in surface oceans and in deep-sea hydrothermal vents. In this study, we provide additional information on the vertical distribution and expression levels of TBDTs (107 proteins) in pelagic waters, accounting for 11% of TMP and 3% of total identified proteins.

The semiquantitative assessment of these transporter systems further confirmed the omnipresence and wealth of ABC transporters in the euphotic zone and the dark ocean (Fig. 2). Relative abundances of expressed ABC transporter proteins were on average six times higher than TRAP-Ts in the euphotic zone and 22 times higher than TBDTs in bathypelagic metaproteomes. TRAP-Ts and TTTs exhibited similar vertical expression patterns, with higher average abundances in the mesopelagic and bathypelagic zones, whereas TBDTs were expressed at higher values in the upper water column.

With the exception of TTT systems, transporter proteins mediating the uptake of both low– and high–molecular-weight OM were detected at all depths. However, relative expression levels of transporter systems varied with depths, suggesting depth-related quantitative changes in substrate uptake patterns while qualitative changes in the transporter systems were not apparent (Fig. S4). Importantly, the fraction of transporters increased from ∼23% in the euphotic layers, to 32% in the mesopelagic and 39% in bathypelagic waters (Table 1). The increase in the fraction of transporter proteins from the euphotic to the bathypelagic zone as observed in this study (Table 1) might be interpreted as an adaptation of the deep-sea microbial community to the changes in quality and quantity of the OM in the water column. Prokaryotic abundances and leucine incorporation rates, used as a proxy for heterotrophic prokaryotic production, typically decrease from the sunlit to the abyssopelagic waters by roughly three orders of magnitude (37). Cell-specific leucine incorporation rates, however, analyzed along the cruise track of Geotraces and Medea (Supporting Information) remained fairly constant in bathypelagic and abyssopelagic waters (1,000–5,500 m; Table S2), with average rates of 2.1 ± 2.2 × 10⁻⁵ fmol Leu⋅cell⁻¹⋅d⁻¹, indicating a fairly constant heterotrophic activity below the mesopelagic zone.

The taxonomic diversity derived from expressed transporter proteins, referred to as the substrate-active community, comprised mainly Bacteria (69%) and 2% Archaea (Fig. 3). Roughly 30% of TMPs (294 proteins) remained unclassified due to divergent assignments at the phylum level or insignificant hits (Dataset S1). Expression profiles of the substrate-active community revealed the following contribution of phyla: Proteobacteria (Alphaproteobacteria, 40% of NAAF values; Gammaproteobacteria, 10%; Deltaproteobacteria, 7%), Actinobacteria (5%), Bacteroidetes (5%), Thaumarchaeota (4%), Cyanobacteria (3%), and Euryarchaeota and Planctomycetes (each 1%). A depth-dependent stratification of substrate-responsive phyla, based on the expression of transporter-related proteins, was evident for Cyanobacteria, Actinobacteria, Bacteroidetes, Verrucomicrobia, Firmicutes, and the candidate phylum Marinimicrobia, accounting for higher relative abundances in euphotic and mesopelagic layers than in bathypelagic waters (Fig. 3). Conversely, transporters from Deferribacterales, Planctomycetes, Chloroflexi, and members of the Gammaproteobacteria and Deltaproteobacteria were relatively more abundant in the bathypelagic than in water masses above.

Concomitant metagenomes recovered between 1,809 and 65,228 predicted protein-coding genes (Table S3) and 32–438 near–full-length 16S rRNA gene sequences (Supporting Information and Table S4). The recovery of 16S rRNA genes from metagenomic libraries supports the findings on the diversity deduced from 16S rRNA gene PCR amplifications. Vertical distribution patterns based on variations in estimated 16S rRNA gene abundances revealed similar trends in the microbial community structure (Supporting Information and Fig. S5). Noteworthy, however, estimated 16S rRNA gene abundances indicated higher relative abundances of Thaumarchaeota throughout the water column than suggested by transporter protein abundances, contributing up to ∼20% of the microbial community in bathypelagic waters (Fig. S5). Thus, transporter proteins primarily selected as a proxy for transport and degradation of complex organic molecules, might underestimate the activity and abundance of marine Archaea and the relative contribution of chemolithoautotrophic metabolism versus heterotrophy. The genetic capabilities of Thaumarchaeota for the uptake or translocation of compounds are described in detail elsewhere (38). In our metaproteomes, predicted thaumarchaeal transporters included ammonium channel proteins (Amt), urea, and SSS family transporters revealing, despite the comparatively low expression levels, the potential of the organism to utilize inorganic and organic energy sources (39).

COG and KEGG Orthology (KO) classifiers were used to resolve the compound specificity of transport proteins (Fig. 4), and for prevalent microbial groups, average NAAFs, referred to as “protein abundance” from here on, were summarized for the euphotic, the mesopelagic, and the upper and lower bathypelagic zones (Fig. 5).

Predicted substrate affinities of expressed transporter proteins suggest that amino acids, including branched-chain amino acids, proline/glycine betaine, and di- and oligopeptides, as well as carbohydrates and carboxylic acids, might represent essential components of the OM pool in the oligotrophic open ocean.

ABC transporters exhibited the highest expression values in every depth layer attributed to their involvement in the uptake of amino acids (245 proteins; Fig. 4). Predicted ligands of the respective SBPs included proteinogenic amino acids such as glutamate, arginine and derivatives, histidine, proline, and glycine or more general branched-chain amino acid (leucine, isoleucine, and valine) and polar- and l-amino acids (Fig. S6 and Dataset S1). Various bacterial phyla and Euryarchaeota Marine Group II (MG-II) expressed amino acid transporters, albeit the proportions of active cells varied among groups (Dataset S1). In agreement with actual measurements of amino acid uptake rates (40), Alphaproteobacteria and Gammaproteobacteria accounted for high relative abundances of amino acid transporters, together contributing ∼30% to the expression values in the lower euphotic, 83% in the mesopelagic, and 85% in the bathypelagic zone (Fig. 5). As previously described based on transcriptome data (41), members of the SAR11 clade invest substantially in the uptake of compatible solutes, for example, proline/glycine betaine (ProU operon), reportedly supporting cell growth (42). In sinking POM, amino acid-like material accounts for the largest component (40–50%) of particulate organic carbon (43). Interestingly, minimal changes in the chemical structure of the bulk POM composition were found throughout the water column (43). Together with amino acids and lipids, carbohydrates constitute a major group of biomolecules in the OM pool and comprise up to 30% (by mass) of sinking particles (43). We identified 109 proteins of ABC carbohydrate-binding transporters: (i) monosaccharide and (ii) disaccharide and oligosaccharide transporters, which are homologous to the family of oligopeptide and dipeptide transporters. Carbohydrate transporter abundances accounted for similarly high expression values as observed for peptides with only minor variations in relative abundances throughout the water column (Fig. 4). Predicted COG/KO functions indicate a broad range of potential target molecules (Fig. S6), with a remarkably high fraction of transporters involved in the uptake of glycerol-3-phosphate (G3P) (Dataset S4). Expression profiles of the carbohydrate-active community indicated a vertical stratification in the water column, with higher relative abundances of Actinobacteria, Firmicutes, and Gammaproteobacteria in the upper water column, whereas Deltaproteobacteria dominated in underlying waters (Fig. 5). SAR324 clade members showed high expression values of transporters specific to G3P, which might be utilized as a source of carbon or inorganic phosphate under nutrient-limiting conditions (44, 45). The expression of indicator genes for flagellar-related assembly (COG1344; Dataset S1) may indicate that at least some members of the SAR324 cluster possess a strong gliding or adhesion capability (46), suggesting the tendency to associate with particles (47).

Another omnipresent and abundantly expressed compound class was ABC peptide transporters, mediating the bulk uptake of dipeptides and oligopeptides (Dpp and Opp systems). We identified 115 transporter proteins involved in peptide utilization (Dataset S1). Independent of the sampling depth, relative abundances of dipeptide transporters were twice as high as those for oligopeptides (Fig. S6). While their primary function is the import of peptides as carbon and energy source, their role as messengers in virulence or signaling processes, for example, modulating chemotactic behavior or peptide export for biofilm formation, has to be taken into consideration. Thereby, peptide transporters exemplify that it is inevitable to perform biochemical analyses to verify substrate specificities and/or functions of the predicted transporter components. Various taxa demonstrated protein-processing capabilities, with the MG-II/III Euryarchaeota, Planctomycetes, and Gammaproteobacteria being the prime protein utilizers (Fig. 5). This is in agreement with high transcript levels of extracellular peptidases and carbohydrate-active enzymes (CAZymes) of MG-II and MG-III Euryarchaeota (48), suggesting the metabolic capacity of extracellular protein degradation and utilization of carbohydrates for heterotrophic growth (Table S5).

Another ubiquitous source of labile carbon and nitrogen in the ocean constitute polyamines like putrescine and spermidine. The occurrence of polyamine transporters has been studied in marine bacterial genomes (49), revealing the wide distribution of SBPs among the phyla of Actinobacteria, Chlamydiae/Verrucomicrobia, Cyanobacteria, Firmicutes, and Proteobacteria. Surveys of marine metagenomes revealed that up to 32% of surface ocean bacterioplankton potentially encode homologs implicated in the transport or degradation of polyamines, suggesting the important role of polyamines in carbon and nitrogen cycling (50). In our study, the active polyamine-transforming community comprised Alphaproteobacteria (25 proteins; including Rhodobacterales, Rhizobiales, and Pelagibacterales), Gammaproteobacteria (14 proteins), and Deltaproteobacteria (7 proteins), with SAR11 dominating in the euphotic zone (Fig. 5). Overall, relative abundances of polyamine transporter systems (PotD, PotF; 56 proteins) suggest a higher expression in the euphotic zones than underlying water masses (Fig. 4).

ABC transporters expressed at considerably lower abundances had predicted substrate specificities to urea, phosphate/phosphonate, or taurine (2-aminoethane sulfonate), sources of organic carbon, nitrogen, and/or sulfur moieties (Fig. 4). Taurine, in particular, constitutes a valuable food source for heterotrophic microbes due to its carbon, nitrogen, and sulfur moieties and is found at nanomolar concentrations in the marine environment (51). Key enzymes required to utilize these nutritional elements (52, 53) include the taurine-pyruvate aminotransferase (Tpa), alanine dehydrogenase (Ald), and sulfoacetaldehyde acetyltransferase (Xsc), facilitating the utilization of taurine as C source (52, 53). Thirteen proteins encoding the SBP (TauA) of the taurine transport system (TauABC) were detected in our metaproteomes at higher average abundances in the euphotic zone than in deeper layers (Fig. 4; NAAFs ranging between 0.003 and 0.06). Proteins encoding the Xsc (Dataset S1) were recovered at 500 m and 1,475- to 2,808-m depth, suggesting that taurine might serve as a favorable C source also in mesopelagic and bathypelagic waters. Function predictions of several proteobacterial single amplified genomes revealed genes or homologs for the TauD and/or Tpa, suggesting the genetic potential of C and S utilization. Our metagenomes clearly indicated an overall increase of TauD gene abundances below the euphotic zone, pointing at the importance of taurine as S source in the dark ocean (Dataset S5). Within the Alphaproteobacteria, SAR11 cluster members appeared as the prime utilizers of taurine in bathypelagic waters (Fig. 5). This is in agreement with a previous study, experimentally demonstrating the growth of SAR11 on taurine (54) and the ubiquity of SAR11 taurine transporters in metaproteomic and metatranscriptomic surveys in coastal surface waters (14, 28, 41).

Additionally to ABC transporters, diverse members of the microbial community expressed TRAP transporters at relatively high abundances, particularly in bathypelagic layers (Fig. 5). TRAP transporters are ATP independent, and thus less energy consuming than ABC transporters and, consequently, more advantageous under oligotrophic, deep-sea conditions. Known substrates recognized by the SBPs include a wide variety of compounds, all characterized by the presence of a carboxylate group, that is, organic acids (amino acids and acid sugars) and other carboxylate-containing small metabolites (29, 30, 32, 33, 55). Besides nutritional benefits, TRAP transporters support the thermostabilization of microbial cells by translocating osmoprotectants such as ectoine and 5-hydroxyectone (56). Predicted substrate affinities of TRAP SBPs identified in this study included mannitol/chloroaromatic compounds and aromatic acids (35) (COG4663, FcbT1; 45 proteins) and C4-dicarboxylates such as malate, fumarate, and succinate used as carbon and energy sources (COG1638, DctP; 36 proteins). Collectively, the phylogenetically diverse expression of these transporters and their ubiquity may indicate their ecological importance in dark ocean substrate turnover.

High proportional abundances of outer membrane TBDTs (107 proteins) were observed in euphotic and mesopelagic layers (NAAF euphotic, 0.0747 ± 0.047; meso, 0.1778 ± 0.117; bathy, 0.03 ± 0.03), similar to environmental metatranscriptomic (57) and metaproteomic (21) studies. TonB transporters were originally identified in the context of iron transport but are now reported to be involved in the uptake of a variety of compounds (58).

TonB-dependent transporter proteins encoded for cobalamine or vitamin B12 receptors (COG4206, BtuB), Ton box of ferric citrate (COG4772, FecA), outer membrane Fe transporters (COG1629, CirA), and outer membrane receptors for ferrienterochelin and colicins (COG4771, FepA). Also, we identified eight proteins encoding the SusC/RagA clade of TonB-linked outer membrane proteins, presumably utilizing large protein fragments (i.e., RagA) or carbohydrates (i.e., SusC) as organic compounds (Dataset S1). TonB-related proteins were also identified as among the most abundant transcripts assigned to DOM-responsive Idiomarinaceae and Alteromonadacea in a DOM-enriched marine microcosm (22). Both FepA and FecA facilitate the uptake of iron complexed to OM in diverse marine bacteria (59) in addition to other transport capacities. Fe is a critical trace element for bacterioplankton and impacts carbon and nitrogen fixation over broad regions of the ocean (60). However, to our knowledge, Fe(II) oxidation has not been observed in the open water column. In agreement with previous reports (26, 61), phylogenetic analyses of TBDTs strongly indicated that Gammaproteobacteria (34 proteins) and Bacteroidetes (including Cytophaga, Flavobacteria, and Sphingobacteria; 19 proteins) degrade polymeric matter throughout the water column. The taxonomic classification and relative abundance of TBDTs throughout the water column reveal a clear stratification of Flavobacteria, Alteromonadales, SAR86 cluster bacteria, and Marinimicrobia, dominating in the upper ocean. Conversely, TBDTs of Deferribacteres and Gemmatimonadetes were expressed at higher abundances in the dark ocean (Dataset S1). Taken together, these observations further highlight the significance of transporter processes throughout the water column and likely illustrate depth-related partitioning of marine bacteria according to environmental conditions.

Sampling was performed during the research cruises Geotraces III (March 2010), Moca (October 2010), and Medea I and II (October 2011, June/July 2012), spanning a latitudinal range in the Atlantic Ocean from 67°N to 49°S (Fig. S2). To investigate the vertical structure and activity of microbial communities, water samples (100–600 L) for metagenomic (100–5,002 m) and metaproteomic (100–4,050 m) analyses were collected from eight distinct water masses and 14 depth layers (Table S1).

Preparation of genomic DNA was performed using a standard phenol extraction protocol (Supporting Information). Contigs were assembled using paired-end Illumina HiSeq reads, and the assembly and prediction of ORFs were performed using the software tools MetaVelvet and Prokka (Fig. S1). 16S rRNA gene sequences were assembled using EMIRGE (66) and identified by BLASTn searching against the Silva SSU 128 database (Fig. S5).

Protein extraction and spin filter-aided in-solution digestion (SF-ISD) was performed according to León et al. (67) with modifications to optimize protein recovery (Supporting Information). Equal peptide aliquots generated from each sample were analyzed by ultrahigh-pressure nanoLC coupled to an LTQ-Orbitrap Velos Mass Spectrometer (Thermo Scientific). Peptide separation was performed with a nanoAcquity UPLC system (Waters) fitted with a 2-cm, 180-µm ID Symmetry C18 trap column (Waters) and a 25-cm, 75-µm ID, BEH C18 (1.7-µm particles) analytical column (Waters). Peptides were trapped for 2 min at 10 µL/min with 0.1% TFA and separated at 350 nL/min using a gradient of 5–35% acetonitrile with 0.1% formic acid for 75 min.

Detailed information on the experimental procedures, functional assignments, and quantification methods can be found in SI Materials and Methods.

Genomic sequence data that support the findings of this study have been deposited in GenBank with accession codes SRP081826 and SRP081823 and Pangaea (https://doi.org/10.1594/PANGAEA.883794). Accession codes of assembled 16S rRNA sequences are as follows: KY241481–KY241660, KY194331–KY194691, KY193976–KY194214, KY081807–KY081876, KX426906–KX426937, KX427579–KX428016, KX426472–KX426524, and KX426391–KX426464. Peptide sequences generated and analyzed during this study are included in Supporting Information.