Metaproteomics reveals differential modes of metabolic coupling among ubiquitous oxygen minimum zone microbes

Water Column Chemistry and Molecular Sampling.

To evaluate changes in water column redox gradients associated with different stages of stratification and renewal, samples were collected from the Saanich Inlet water column from station S3 on April 9, 2008 (Apr08) and from multiple stations from the inlet mouth (S4) midpoint (S3) and end (S2) on September 1, 2009 (Sep09) (Fig. S1A). Water column chemistry profiles indicated four redox zones: upper oxycline (UO), lower oxycline (LO), sulfide nitrate transition zone (SNTZ), and sulfidic zone (SZ) (Fig. 1 and Fig. S1B), generally corresponding to dysoxic (20–90 µmol O₂), suboxic (1–20 µmol O₂), anoxic (<1 µmol O₂), and anoxic sulfidic conditions (2) (Fig. S1B). Water column redox zonation and associated microbial community structure was consistent with other OMZs (2, 24), making Saanich Inlet a tractable model ecosystem for studying microbial community responses to changing levels of water column oxygen deficiency.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Hierarchical clustering of metaproteome by NSAF (see Methods) for detected proteins from Sep09 S2, S3, and S4 indicating compartments of the water column, with adjacent sparklines for oxygen (O₂), nitrate (NO₃⁻), and hydrogen sulfide (H₂S) for each sample.

To explore changes in microbial community structure and function along water column redox gradients, we analyzed paired metagenomic and metaproteomic datasets from Apr08 and paired small subunit ribosomal RNA gene pyrosequencing and metaproteomics datasets from Sep09 (Fig. S1B). Sanger end sequencing of small insert clone libraries from the three Apr08 samples yielded a total of 54,701 ORFs, with an average of 18,234 ORFs per sample. Small subunit ribosomal RNA (SSU rRNA) gene pyrosequencing of the 12 Sep09 samples yielded 87,138 sequences that clustered into 3,385 nonsingleton operational taxonomic units at the 97% identity threshold. Tandem MS-coupled LC (LC/MS/MS) metaproteomic sequencing identified a total of 5,019 unique proteins, a number comparable to previous marine metaproteomic studies (25). A consistent number of proteins were identified across the Sep09 samples, averaging 695 unique proteins per sample (Table S1). Although variability in protein detection in the Apr08 samples was considerable, the high number of unique proteins detected in the Apr08 200-m sample (4,344) enabled identification of more complete metabolic pathways.

Patterns of Redox-Driven Niche Partitioning.

To determine patterns of redox-driven niche partitioning, we compared community composition with ORF counts and protein normalized spectral abundance factors (NSAF; SI Methods) between UO, LO, SNTZ, and SZ. Hierarchical clustering of NSAF values was consistent with redox zonation (Fig. 1). Clear trends in protein abundance were observed in relation to redox zonation not reflected in pyrotag and metagenomic datasets, consistent with alternative forms of coupling or regulated gene expression (Fig. S2). Ammonia oxidizing Thaumarchaeota, mediating the first step of nitrification, dominated UO and LO samples and decreased in abundance within the SNTZ and SZ. Similar trends were observed with respect to ORF counts and NSAF values (Tables S2–S4). The nitrite oxidizing bacterium Nitrospina gracilis (26), mediating the second step of nitrification, was abundant in UO and LO samples and decreased in abundance within the SNTZ and SZ. A second nitrite oxidizing bacterium Nitrospira defluvii (18), although absent from pyrotag datasets, exhibited high NSAF values with a similar distribution pattern as Thaumarchaeota and N. gracilis. Anammox bacteria affiliated with the Planctomycetes (Tables S2–S4) exhibited intermediate abundance (∼1%) in the UO and LO samples, decreasing in abundance within the SNTZ before increasing again in the SZ. Planctomycete ORF abundance increased along the redoxcline, whereas protein NSAF values were high in the UO and LO, decreasing to intermediate values within the SNTZ and SZ. These patterns of protein expression confirm previous reports of coupled nitrification and anammox observed in OMZs based on process rate and functional marker gene abundance (27, 28).

In addition to known players in the nitrogen cycle, taxa involved in sulfur cycling or coupled nitrogen and sulfur cycling were also abundant and active in the water column. Multiple lineages affiliated with SAR11 within the Alphaproteobacteria mediating dimethylsulfoniopropionate (DMSP) oxidation (29) were abundant in the UO and LO samples, decreasing in abundance within the SNTZ and SZ (Fig. S2). Similar trends were observed with respect to ORF counts and NSAF values (Tables S2–S4). Multiple lineages affiliated with SUP05/ARCTIC96BD-19 and symbiont-related Gammaproteobacteria (Tables S2–S4) mediating oxidation of reduced sulfur compounds using O₂ (19) or NO₃⁻ (16), as TEAs were also abundant. The ORFs for ARCTIC96BD-19, SUP05, and symbionts exhibited reciprocal distribution patterns, with ARCTIC96BD-19 ORFs decreasing and SUP05 and symbiont ORFs increasing in abundance within the SNTZ and SZ. A similar pattern was observed with respect to NSAF values, with high SUP05 NSAF values in the LO, SNTZ, and SZ. These distribution patterns support previous reports of ARCTIC96BD-19 and SUP05 population structure (2, 16, 19, 30, 31).

Collectively, Thaumarchaeota, Nitrospina, Nitrospira, Planctomycetes, SAR11, SUP05/ARCTIC96BD-19, and symbiont-related Gammaproteobacteria comprised on average 48% of pyrotag, 41% of metagenomic, and 64% of metaproteomic datasets (Tables S2 and S3). Several taxonomic groups that were abundant based on pyrotags (>1%) including Marine Group II Euryarchaea, Crenarchaeota, Acidomicrobiales, Bacteroidetes, Chloroflexi, Flavobacteria, and Desulfobacteraceae, and candidate divisions OD1, OP11, Marine Group A, and SBR1093 were not well represented in metagenomic or metaproteomic datasets (Fig. S2). Lack of indigenous reference genomes likely caused many sequences originating from these groups to be classified as no hit or below cutoff (Methods). Consistent with this observation, BLAST queries against the Genomic Encyclopedia of Bacteria and Archaea Microbial Dark Matter (GEBA-MDM) single-cell genome collection (32) yielded only 23 additional protein sequences, which had otherwise been classified as below cutoff or no hit. Conversely, several taxonomic groups including N. defluvii and ARCTIC96BD-19 that were absent in pyrotag datasets exhibited intermediate ORF counts and NSAF values. This discrepancy was likely due to incomplete taxonomic resolution within the Greengenes database. Approximately 1% of pyrotag and 10% of metagenomic and metaproteomic datasets remained unaffiliated with any taxonomic group. Taken together, these results indicate that active nitrogen and sulfur cycling microorganisms are the primary contributors to both genetic potential and gene expression along the redoxcline in Saanich Inlet.

Differential Gene Expression Patterns.

To investigate patterns of gene expression driving carbon and energy metabolism along the redoxcline in Saanich Inlet, we identified nitrification, anammox, denitrification, sulfur oxidation, and inorganic carbon fixation pathway components in metagenomic and metaproteomic datasets using BLAST. By summing the NSAF values for each component, we observed differential patterns of gene expression and metabolic coupling along the redoxcline (Fig. 2). Expression of these pathways was remarkably stable under similar redox conditions in space (Sep09 S2-S4) and time (Apr08 to Sep09; Figs. S3–S5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Distribution and NSAF value of proteins involved in nitrogen and sulfur-based energy metabolism and inorganic carbon fixation for taxa abundant in the metaproteome. For metagenome (gray, Apr08 only) and metaproteome in upper oxycline (green), lower oxycline (teal), S/N transition zone (blue), and sulfidic zone (purple). See Table S4 for full list of protein names; Anx indicates anammox hydroxylamine oxidoreductase and hydrazine oxidoreductase proteins.

Expressed pathways for nitrogen-based energy metabolism progressed from ammonia oxidation and nitrification in the UO and LO to denitrification in the SNTZ and SZ (Fig. 2 and Fig. S3). Proteins catalyzing the first step of nitrification, ammonia monooxygenase subunits B and C (Amo), from Thaumarchaeota, were detected in the UO and LO, decreasing along the redoxcline. Proteins catalyzing the second step of nitrification, nitrite oxidase (NXR), from N. graclilis and N. defluvii followed the same pattern of expression as Amo. Moreover, the detection of both Amo and NXR from nitrifying taxa, albeit at lower NSAF values in the SNTZ and SZ, supports recent observations of NO₂⁻ oxidation in the Namibian OMZ with implications for NO₃⁻ supply for denitrification (28, 33). Nitrite oxidase from Planctomycetes (Fig. 2) (14) had the highest NSAF values of any protein in the UO and LO and exhibited a similar expression profile to Amo and NXR originating from N. gracilis and N. defluvii (Fig. 2 and Fig. S3). Conversely, proteins catalyzing anammox, including hydrazine and hydroxylamine oxidoreductases (Anx) from Planctomycetes (Fig. 2 and Fig. S3), exhibited opposing expression patterns, with low NSAF values in the UO and LO that increased in the SNTZ and SZ. Contrasting patterns of NXR and Anx expression from Planctomycetes could reflect a metabolic response to O₂, resulting in a shift between maintenance energy production in the UO and LO to anammox for growth under more favorable redox conditions in the SNTZ and SZ. Alternatively, close sequence similarity between Planctomycetes, N. gracilis, and N. defluvii NXR could confound BLAST-based taxonomic assignment.

Proteins mediating the partial denitrification pathway from SUP05 including dissimilatory nitrate reductase subunits G and H (Nar), periplasmic nitrate reductase subunits A and B (Nap), and nitrite reductase (NirK) were detected in the UO increasing in abundance along the redoxcline (Fig. 2 and Fig. S3). Protein NSAF values for SUP05 Nar increased relative to Planctomycetes NXR in the SNTZ and SZ. Additional proteins for SUP05 nitric oxide reductase subunits B and C (Nor) were detected with similar NSAF values in the LO, SNTZ, and SZ. Although denitrification pathway components from other taxonomic groups were detected in the water column, SUP05 was the only group to express consecutive proteins in the denitrification pathway, making up 50% of total NSAF values for all denitrification proteins (SI Text). These observations point to SUP05 as the dominant player in nitrogen-based energy metabolism in the SNTZ and SZ. The detection of SUP05 Nap and NirK in the UO and LO where O₂ concentrations approached 120 µM was unexpected given that 20 µM O₂ is a commonly accepted threshold for denitrification (34) and may have implications for the O₂ threshold for nitrogen loss processes in other OMZs. Additionally, detection of SUP05 Nor and absence of nitrous oxide reductase (NosZ) in the LO and low abundance in SNTZ and SZ point to SUP05 as a source of N₂O. Recent observations of enrichment of genes which encode Nor and NosZ on particles within OMZs suggest a distributed denitrification pathway across particle and nonparticle niches (35) and may account for low NSAF values observed for NorCB and NosZ.

Expressed pathways for sulfur-based energy metabolism were detected in the UO and increased in NSAF value along the redoxcline (Fig. 2 and Fig. S4A). Proteins catalyzing sulfide oxidation predominantly originated from SUP05/ARCTIC96BD-19 and symbiont-related Gammaproteobacteria. With the exception of ARCTIC96BD-19 adenylylsulfate reductase (Apr), the vast majority of proteins originated from SUP05 and symbionts. With respect to SUP05, flavocytochrome C (Fcc), sulfide oxidation proteins (Sox), dissimilatory sulfate reductase (Dsr), and Apr were detected in the UO and increased in NSAF value along the redoxcline. In addition, SUP05 ATP sulfurylase (Sat) and sulfide:quinone oxidoreductase (Sqr) were detected in the LO, SNTZ, and SZ, and SNTZ and SZ, respectively. These results are consistent with recent SUP05 protein expression profiles observed in hydrothermal plume and overlying waters (25). With the exception of Sox, symbiont proteins catalyzing sulfide oxidation followed the same expression pattern as SUP05. The expression of sulfur oxidation pathway components from SUP05/ARCTIC96BD-19 in the UO and LO is consistent with a cryptic sulfur cycle. However, no proteins from defined sulfate (SO₄²⁻)-reducing bacteria were identified in the metaproteome (13). This observation could reflect a bias against particle-associated microorganisms capable of SO₄²⁻ reduction during sample processing or the use of alternative electron donors including DMSP, elemental sulfur, thiosulfate, or polysulfide in the UO and LO. Additionally, proteins with BLAST hits to hydrogenase subunit HupL originating from Guaymas Basin SUP05 metagenomes (36) were detected in the SNTZ, with NSAF values comparable to SUP05 NapA (Fig. S4B), expanding the range of potential substrates for SUP05 energy metabolism in the Saanich Inlet water column.

Expressed proteins for three inorganic carbon fixation pathways including the 3-hydroxypropionate/4-hydroxybutyrate (3HP-4HB) from Thaumarchaeota, reductive acetyl-CoA (rACoA) from Planctomycetes, and Calvin Benson Basham (CBB) cycle from SUP05 were differentially expressed along the redoxcline (Fig. 2 and Fig. S5). Unlike proteins mediating nitrogen and sulfur-based energy metabolism, ORFs encoding carbon fixation pathway components were found in higher relative abundance in the metagenome (Fig. 2).

Proteins catalyzing the 3HP-4HB pathway were detected predominantly in the UO including 4-hydroxybutytyl-CoA dehydratase, acetyl-CoA carboxylase, and propionyl CoA carboxylase (15, 17, 37). Similar expression patterns were observed for Amo and other ammonia oxidation pathway components (SI Text), providing evidence of inorganic carbon fixation coupled to ammonia oxidation by Thaumarchaeota in the UO. Consistent with previous reports, proteins catalyzing a putative Planctomycete rACoA pathway were detected in the SZ in Apr08 along with Anx proteins providing evidence for inorganic carbon fixation coupled to anammox under sulfidic conditions (2.1 µM) (38). Protein NSAF values for SUP05 CBB pathway components increased relative to other bacteria in the SNTZ and SZ, providing compelling evidence for inorganic carbon fixation coupled to sulfide-oxidation and partial denitrification by SUP05. Indeed, CBB pathway components had the highest ORF counts and protein NSAF values of all carbon fixation pathways, comprising 47% of all carbon fixation proteins within the SNTZ and SZ. In addition to inorganic carbon fixation pathways, the abundance of SAR11 DNA and protein in the UO and LO (Fig. S2 and Tables S2 and S3) suggest that heterotrophic remineralization of dissolved organic matter (DOM) is an active process in the UO and LO. Specifically, ABC transporter proteins for uptake of glycine betaine, spermidine/putrescine, and taurine (sources of carbon, nitrogen, and sulfur, respectively) were detected with moderate NSAF values within the UO and LO. In addition to consuming molecular oxygen, remineralization of DOM by SAR11 and other heterotrophic microbes in the UO and LO could act as a source of NH₄⁺, SO₄²⁻, and CO₂.

Regulated Gene Expression.

Given the numerical abundance of SUP05, we were able to resolve changes in protein expression originating from a metabolic island integrating nitrogen- and sulfur-based energy metabolism with inorganic carbon fixation (16) (Fig. 3). Specifically, NSAF values for the SUP05 Sqr, NarH, and NarG subunits appeared to vary as a function of O₂ concentration, whereas the FccAB and NapAB subunits remained relatively constant in the LO, SNTZ, and SZ (Figs. 2 and 3 and Figs. S3 and S4A). Close proximity and similar expression profiles for napAB and fccAB are consistent with regulated gene expression along the redoxcline. Indeed, two ORFs encoding Crp/Fnr transcriptional regulators implicated in redox sensing (39) are located on either side of the nap/fcc gene cluster, with the potential to modulate gene expression, and Crp/Fnr proteins (SUP05_0428) were detected in the Apr08 SZ (Fig. 3). Protein NSAF values for CbbM, a RuBisCO subunit located in proximity to the nar gene cluster, increased between the UO and LO and remained relatively constant in the SNTZ and SZ. These results provide functional evidence in support of previous genomic observations positing a highly integrated and redox-sensitive energy metabolism in SUP05 with direct implications for energy supply to inorganic carbon fixation. In addition to coordinated Fcc, Sqr, Nar, Nap, and CbbM expression, an ORF encoding a hydroxylamine-oxidoreductase homolog (HAO-like) located in the nap/fcc gene cluster was among the most abundant SUP05 proteins detected in the SZ (Figs. 2 and 3 and Fig. S3). SUP05 hao is closely related to genes found in sulfur-oxidizing endosymbionts, as well as the anammox bacterium Candidatus Kuenenia stuttgartiensis. All four HAO homologs contain eight CxxCH multiheme motifs similar to those found in NrfA, a nitrite reductase catalyzing dissimilatory nitrate reduction to ammonia (DNRA) (40).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Relative abundance of SUP05 genes and proteins in two overlapping SUP05 fosmid sequences (GQ351266 and GQ351267) (16). Metagenome (gray, Apr08 only) and metaproteome for the upper oxycline (green), lower oxycline (teal), S/N transition zone (blue), and sulfidic zone (purple). Selected SUP05 genes involved in denitrification (dark gray shading), sulfide oxidation (black shading), and putative hydroxylamine oxidoreductase (diagonal lines) are indicated. Protein abundance shown as summed NSAF values for all detected ORFs with top hit to a given SUP05 protein. Metagenome abundances shown as percentage of ORFs with top hit to a given SUP05 gene with sparklines for oxygen (O₂), nitrate (NO₃⁻), and hydrogen sulfide (H₂S) for each sample.

Metabolic Coupling Model.

Although bulk inorganic carbon fixation rates within OMZs have been measured (20⇓⇓–23), few studies have directly linked inorganic carbon fixation with energy metabolism of defined OMZ microbes (41). With this linkage in mind, we construct a metabolic model describing taxonomic and metabolic networks coupling pathways of nitrogen- and sulfur-based energy metabolism and inorganic carbon fixation along the redoxcline in Saanich Inlet based on metaproteomic datasets (Fig. 4). In this model, heterotrophic remineralization of DOM releases CO₂, SO₄²⁻, and NH₄⁺ within the UO and LO. Thaumarchaeota couple oxidation of NH₄⁺ to inorganic carbon fixation via the 3HP-4HB pathway within the UO, producing NO₂⁻, a process that has been demonstrated both in culture (42) and in situ (43). Nitrous oxide is also produced as a byproduct of ammonia oxidation (43, 44). Nitrite produced via NH₄⁺ oxidation is oxidized in turn by N. defluvii, N. gracilis, and Planctomycetes in the UO (14, 28). The extent to which this process is coupled to inorganic carbon fixation within these groups remains to be determined. Ammonia oxidation attenuates as O₂ levels decline in the LO, SNTZ, and SZ, accompanied by a transition to partial denitrification and anammox. In the LO, SUP05 begins to couple oxidation of reduced sulfur compounds, and possibly hydrogen, with NO₃⁻ reduction to N₂O to fix inorganic carbon via the CBB pathway, a trend that increases in the SNTZ and SZ (16, 36, 45, 46). In parallel, Planctomycetes couple anammox to fix inorganic carbon via the rACoA pathway (14), although the broader water column occurrence of this process remains to be determined. Competition between SUP05 and Planctomycetes for oxidants could help explain variations in spatial and temporal dynamics of nitrogen loss processes observed in different OMZs. Alternatively, potential DNRA by SUP05 could supply NH₄⁺ for anammox resulting in a cometabolic linkage. Overall, the interactions described in the model are dynamic and reflect patterns of redox-driven niche partitioning regulating nitrogen loss processes and carbon flux through ubiquitous OMZ microbes.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Proposed metabolic model based on metaproteomic observations for heterotrophic remineralization (brown) and energetic coupling (yellow dashed lines) of nitrogen (green), sulfur (red), and hydrogen (orange) based chemolithotrophic energy metabolism with carbon fixation (yellow star) for taxa abundant in the metaproteome. Line weight and arrow size indicate magnitude of metabolic activity. Gray lines, activity not occurring under given conditions; light gray taxa, reduced abundance and metabolic activity.

Energy for inorganic carbon fixation within the LO, SNTZ, and SZ is derived in large part from denitrification and anammox nitrogen loss processes, with the balance between these processes impacting energy flow to either SUP05 or Planctomycetes, with concomitant feedback on growth rates. The numerical dominance of SUP05 DNA and protein in the LO, SNTZ, and SZ relative to Planctomycetes suggests that partial denitrification outcompetes anammox from a bioenergetic perspective. Indeed, the difference in free energy yield between the two processes (denitrification coupled to sulfide oxidation yields ∼3.5 times the Gibbs free energy as anammox under standard conditions) is consistent with lower cell abundance and biomass for anammox bacteria, even though anammox is observed more frequently than denitrification in many OMZs (34). As OMZs expand, the contribution of SUP05 to inorganic carbon fixation may have significant impact on global ocean carbon cycling if sufficient energetic substrates are available. However, the fate of carbon fixed in OMZ waters is largely unknown, as the balance between carbon transport and heterotrophic remineralization processes remains to be constrained.

Future Implications.

This study represents the first metaproteome of an O₂-deficient water column encompassing the range of redox conditions, from dysoxic to anoxic sulfidic, found in OMZs globally. Although a recent numerical model by Reed and colleagues attempted to integrate geochemical processes and functional gene markers in the Arabian Sea OMZ (47), our conceptual model uses protein expression to describe differential metabolic coupling among ubiquitous OMZ microbes. The Reed model implicitly assumes reaction rates scale linearly with gene abundance. Thus, the model does not account for biological information flow from DNA to RNA to protein, a regulated process resulting in assembly of pathways driving real world process rates. Incorporation of protein expression information into the model could be used to convert gene abundance into protein abundance or protein production rates, resulting in more accurate predictions. Based on the evidence provided here, SUP05 will likely play an important role in such a pathway-centric model. Indeed, we identified SUP05 as the dominant contributor to inorganic carbon fixation within OMZs, providing insight into ocean carbon and nutrient cycling.

In the future, global climate models predict expansion and intensification of OMZs, with concomitant shoaling and stabilization of sulfidic zones (38, 48). Such a scenario would provide an increased habitat for SUP05, supporting inorganic carbon fixation via direct oxidation of reduced sulfur compounds and cryptic sulfur cycling in oxygen-deficient waters, resulting in increased primary production and potentially increased carbon sedimentation (49). Given an estimate of 4.61 × 10¹⁸ L of O₂-deficient marine waters (50) and the range of observed dark carbon fixation rates from various OMZs of 0.2–2.5 µM/L/d (20⇓⇓–23), we estimate 0.4–5 Pg carbon/y fixed in OMZs globally. This number represents up to 10% of surface primary production [using 48.5 Pg C/y (51)] and will continue to increase with OMZ expansion. With 47% of observed carbon fixation proteins originating from SUP05, we suggest that SUP05 is responsible for 0.2–2.4 Pg carbon/y, representing up to 5% of surface primary productivity. Although OMZ expansion is a predicted consequence of global warming, negative feedback loops may ultimately lead to increased drawdown of atmospheric CO₂ driven in large part by blooming SUP05 populations.