Volume 23, Issue 11 p. 7152-7167
Research article
Open Access

Microbes mediating the sulfur cycle in the Atlantic Ocean and their link to chemolithoautotrophy

Daniele De Corte

Corresponding Author

Daniele De Corte

Institute for Chemistry and Biology of the Marine Environment, Carl Von Ossietzky University, Oldenburg, Germany

For correspondence. E-mail [email protected]. Telephone: (+49) 441-798-3704 Fax: (+49) 441-798-3358

Search for more papers by this author
Simone Muck

Simone Muck

Department of Functional and Evolutionary Ecology, University of Vienna, Vienna, Austria

Search for more papers by this author
Johanna Tiroch

Johanna Tiroch

Department of Functional and Evolutionary Ecology, University of Vienna, Vienna, Austria

Search for more papers by this author
Catalina Mena

Catalina Mena

Instituto Español de Oceanografía, Centro Oceanográfico de Baleares, Palma, Spain

Search for more papers by this author
Gerhard J. Herndl

Gerhard J. Herndl

Department of Functional and Evolutionary Ecology, University of Vienna, Vienna, Austria

NIOZ, Department of Marine Microbiology and Biogeochemistry, Royal Netherlands Institute for Sea Research, Utrecht University, Den Burg, The Netherlands

Search for more papers by this author
Eva Sintes

Eva Sintes

Department of Functional and Evolutionary Ecology, University of Vienna, Vienna, Austria

Instituto Español de Oceanografía, Centro Oceanográfico de Baleares, Palma, Spain

Search for more papers by this author
First published: 07 September 2021
Citations: 1


Only about 10%–30% of the organic matter produced in the epipelagic layers reaches the dark ocean. Under these limiting conditions, reduced inorganic substrates might be used as an energy source to fuel prokaryotic chemoautotrophic and/or mixotrophic activity. The aprA gene encodes the alpha subunit of the adenosine-5′-phosphosulfate (APS) reductase, present in sulfate-reducing (SRP) and sulfur-oxidizing prokaryotes (SOP). The sulfur-oxidizing pathway can be coupled to inorganic carbon fixation via the Calvin–Benson–Bassham cycle. The abundances of aprA and cbbM, encoding RuBisCO form II (the key CO2 fixing enzyme), were determined over the entire water column along a latitudinal transect in the Atlantic from 64°N to 50°S covering six oceanic provinces. The abundance of aprA and cbbM genes significantly increased with depth reaching the highest abundances in meso- and upper bathypelagic layers. The contribution of cells containing these genes also increased from mesotrophic towards oligotrophic provinces, suggesting that under nutrient limiting conditions alternative energy sources are advantageous. However, the aprA/cbbM ratios indicated that only a fraction of the SOP is associated with inorganic carbon fixation. The aprA harbouring prokaryotic community was dominated by Pelagibacterales in surface and mesopelagic waters, while Candidatus Thioglobus, Chromatiales and the Deltaproteobacterium_SCGC dominated the bathypelagic realm. Noticeably, the contribution of the SRP to the prokaryotic community harbouring aprA gene was low, suggesting a major utilization of inorganic sulfur compounds either as an energy source (occasionally coupled with inorganic carbon fixation) or in biosynthesis pathways.


Prokaryotes play a key role in the biochemical cycles of the global ocean (Azam et al., 1983). 10%–30% of the epipelagic primary production is exported as organic carbon into the meso- and bathypelagic layers (Buesseler et al., 2007; Buesseler and Boyd, 2009), limiting heterotrophic prokaryotes growth in the dark ocean (Herndl and Reinthaler, 2013). Recent studies indicate that dissolved inorganic carbon (DIC) fixation driven by chemical energy represents an alternative source of organic carbon (Walsh et al., 2009; Reinthaler et al., 2010; Swan et al., 2011; Sintes et al., 2013), together with non-sinking particulate organic matter for deep ocean prokaryotes (Baltar et al., 2010; Herndl and Reinthaler, 2013).

Several reduced inorganic compounds, such as ammonia, hydrogen and diverse sulfur compounds, can potentially be used as an energy source to fuel mixotrophic and/or chemoautotrophic processes in the dark ocean (Herndl et al., 2005; Swan et al., 2011; Anantharaman et al., 2013). The oxidation of reduced sulfur compounds is a widespread process in hydrothermal vent plumes and oxygen minimum zones (OMZs) (Sievert et al., 2007; Walsh et al., 2009; Ulloa et al., 2012). However, the extent to which reduced sulfur is used as an electron donor or energy source in the oxygenated oceanic water column remains poorly investigated. Reduced sulfur is typically low in concentration and under the detection limit of current methods outside of upwelling regions and the vicinity of hydrothermal vents (Radford-Knoery et al., 2001; Sievert et al., 2007). A potential source for reduced sulfur in the otherwise oxygenated water column are marine snow particles, where respiration can be fuelled by sulfate reduction and sulfide can be generated in oxygen-depleted or anoxic microenvironments (Shanks and Reeder, 1993; Bianchi et al., 2018). Organosulfur compounds primarily released by eukaryotes such as dimethylsulfoniopropionate (DMSP) and taurine are utilized by marine microbes (Clifford et al., 2019; Durham et al., 2019). Sulfate-reducing and sulfur-oxidizing prokaryotes (SOP) harbour the apr gene (Blazejak et al., 2006; Meyer and Kuever, 2007a). The apr gene encodes APS reductase, which forms a heterodimer with one alpha- and one beta-subunit (Lampreia et al., 1994; Hipp et al., 1997; Fritz et al., 2000) and is a key enzyme in the dissimilatory sulfate reduction pathway (Canfield et al., 2005; Meyer and Kuever, 2007a) of sulfate-reducing prokaryotes (SRP). However, this enzyme also operates in reverse direction in SOP, catalysing the transformation of sulfite to sulfate (Hipp et al., 1997; Kappler and Dahl, 2001; Meyer and Kuever, 2007a, 2007b). APS reductase has also been suggested to function as an intracellular detoxification mechanism of sulfite, a metabolite of organic sulfur degradation, in prokaryotes with an incomplete dissimilatory sulfur oxidation pathway (Meyer and Kuever, 2007b).

The aim of this study was to determine the distribution patterns and the phylogenetic affiliation of the aprA-harbouring prokaryotic community throughout the water column along a latitudinal transect in the Atlantic Ocean covering six different oceanographic provinces (Fig. 1) (Longhurst, 2007). The potential link between chemoautotrophy and sulfur oxidation was also explored by relating the aprA to cbbM gene abundances and characterizing their biogeographical patterns. Quantitative-PCR (q-PCR) of two functional genes cbbM encoding the 1,5-bisphosphate carboxylase (RuBisCO), the key CO2 fixing enzyme of the Calvin–Benson–Bassham cycle, and aprA (encoding the APS reductase alpha subunit) was used to assess the biogeographic and depth-related distribution of prokaryotes harbouring these genes in the Atlantic Ocean. This quantification of aprA and cbbM was related to physicochemical and biological environmental factors. Similar patterns of increasing abundance towards the meso- and bathypelagic layers and towards low latitude regions were observed for both genes.

Details are in the caption following the image
Location of stations sampled during the GEOTRACES expeditions. Oceanographic regions are classified according to Longhurst (2007): ARCT, North Atlantic Arctic Province; NADR, North Atlantic Drift Province; NAG, North Atlantic Gyral Province; WTRA, Western Tropical Atlantic; SATL, South Atlantic Gyral Province; SANT, Subantarctic Province. Tag sequencing of the aprA gene was conducted in the red encircled stations.

Additionally, tag sequencing of the aprA gene was used to characterize the SRP and SOP community composition in the different biogeographic provinces and depth zones, as well as to identify prokaryotes with a potentially incomplete sulfur oxidation pathway. Our results suggest that sulfur-oxidizing bacteria are important players in the sulfur cycling of the meso- and bathypelagic Atlantic Ocean and that sulfur oxidation is only partially coupled with DIC fixation as revealed by low aprA/cbbM ratios.


Distribution of prokaryotes harbouring the aprA and cbbM genes in the Atlantic

The abundance of the aprA harbouring prokaryotic community throughout the Atlantic was highest in the meso- and upper bathypelagic layers (with a median of 4.2 × 102 and 3.4 × 102 genes ml−1 respectively). Generally, a higher abundance of aprA genes was found at low latitude regions (NAG, WTRA and SATL) with a median of 3.1 × 102 genes ml−1 (Table S1; Fig. 2A and B) than at high latitude regions (with a median of 1.5 × 102 genes ml−1). The aprA containing cells to total prokaryotic abundance (PA) ratio (i.e. abundance of aprA gene divided by the total PA) significantly increased with depth in all the oceanographic provinces peaking in the lower bathypelagic layer of the low latitude regions (with a median ratio ranging between 1.89 × 10−2 and 3.46 × 10−2) (Fig. S1a, b; Table S1).

Details are in the caption following the image
AprA ml−1 (A), cbbM ml−1 (C) and aprA/cbbM (E) throughout the Atlantic Ocean determined by q-PCR. Box plot showing aprA ml−1 (B), cbbM ml−1 (D) and aprA/cbbM (F) ratios in different regions (colour coded) and depth layers. The left and right of the box represent the first and the third quartiles, while the thick vertical line represents the median. Outliers (larger black points) and jitter values (smaller black points) are shown.

The cbbM gene abundance increased from the epipelagic waters (with a median of 3.5 × 100 genes ml−1) to the mesopelagic layer (with a median of 1.30 × 102 genes ml−1) and decreasing again in the bathypelagic waters (Table S1; Fig. 2C and D). The ratio of cbbM containing cells to total prokaryotic cell abundance (i.e. abundance of cbbM genes divided the total PA) showed similar depth and latitudinal patterns as the aprA to PA ratio, with the highest ratio found in the mesopelagic layer of the WTRA region (with a median of 3.94 × 10−2) (Fig. S1c, d; Table S1).

The aprA/cbbM ratio decreased from the epipelagic layers (with a median of 31 for the whole dataset) to the mesopelagic layers (median 1.73) (Table S1; Fig. 2E and F). The aprA/cbbM ratios were highest in the epipelagic communities of the NAG province (with a median of 179), whereas the lowest ratio was found in the mesopelagic prokaryotic community of the WTRA province (with a median of 0.25) (Table S1; Fig. 2E and F).

Distance-based redundancy analysis (dbRDA) was used to explain the variation of aprA and cbbM gene abundances and aprA/cbbM ratios along the transect in different depth layers. The percent variation explained by the selected environmental parameters ranged between 0% and 42% for the aprA gene abundance, between 48% and 84% for the cbbM gene abundance and between 0% and 44% for the aprA/cbbM ratio, depending on the pelagic zone (Table 1). The environmental parameters analysed did not explain most of the variation of aprA gene in the upper and lower bathypelagic layers, nor the variation of the aprA/cbbM ratio in the epipelagic and upper bathypelagic layers. Latitude and temperature were the main explanatory factors influencing the distribution of the sulfur-oxidizing/-reducing and chemolithoautotrophic genes in the different depth layers (Table 1).

Table 1. Distance-based redundancy analysis results of the environmental parameters best explaining the variation of the aprA ml−1, cbbM ml−1 and aprA/cbbM ratios in different depth layers (Cont = Contribution, F = F-test, p = p-value).
aprA ml−1
Epipelagic Mesopelagic Upper_Bathypelagic Lower_Bathypelagic
Variables Cont F p Cont F p Cont F p Cont F p
Abs_Latitude 0.294 29.72 0.001 0.183 11.45 0.001
Depth (m)
Temperature (°C) 0.167 6.18 0.023 0.118 11.95 0.003
Salinity 0.126 4.64 0.044
Oxygen (μM)
Silicate (μM)
Phosphate (μM)
Nitrate (μM)
Nitrite (μM)
DIC (μM) 0.126 4.67 0.059
Fe (nM)
Cd (nM) 0.07 4.38 0.037
Zn (nM)
Mn (nM)
Cumulative 0.42 0.41 0.25
cbbM ml−1
Surface Mesopelagic Upper_Bathypelagic Lower_Bathypelagic
Variables Cont F p Cont F p Cont F p Cont F p
Abs_Latitude 0.064 32.54 0.009 0.312 48.28 0.001 0.123 13.78 0.002 0.052 7.64 0.006
Depth (m)
Temperature (°C) 0.170 62.00 0.004 0.260 40.20 0.001 0.175 19.61 0.001 0.180 26.27 0.001
Salinity 0.084 30.61 0.024 0.041 6.03 0.03
Oxygen (μM) 0.204 22.83 0.002
Silicate (μM) 0.422 153.8 0.017
Phosphate (μM) 0.090 13.10 0.001
Nitrate (μM) 0.059 21.41 0.006 0.051 7.43 0.017
Nitrite (μM)
DIC (μM) 0.067 7.46 0.016
Fe (nM) 0.046 5.09 0.04 0.032 4.73 0.048
Cd (nM) 0.038 5.61 0.036
Zn (nM)
Mn (nM) 0.04 14.19 0.019 0.023 3.67 0.041
Cumulative 0.84 0.60 0.62 0.48
Surface Mesopelagic Upper_Bathypelagic Lower_Bathypelagic
Variables Cont F p Cont F p Cont F p Cont F p
Abs_Latitude       0.102 13.73 0.001
Depth (m) 0.048 5.83 0.018
Temperature (°C) 0.248 30.25 0.001 0.179 24.10 0.001
Oxygen (μM) 0.114 13.92 0.001
Silicate (μM) 0.067 9.08 0.006
Phosphate (μM) 0.037 4.98 0.035
Nitrate (μM)
Nitrite (μM)
DIC (μM)
Fe (nM) 0.082 5.56 0.024 0.051 6.97 0.01
Cd (nM)
Zn (nM)
Mn (nM) 0.035 4.37 0.055 0.099 6.71 0.012
Cumulative 0.44 0.18 0.44

Diversity and richness of the prokaryotic community harbouring the adenylyl-sulfate reductase (APS) subunit alpha (aprA) gene

Twenty-four samples were taken for determining the composition of the aprA harbouring prokaryotic community from four depth layers at six oceanic provinces (Fig. 1). In total, 787 447 paired reads were obtained, with 3.3 ± 1.3 × 10 4 reads per sample. After trimming and low-quality reads filtering, the total number of reads decreased to 403 613 paired reads with 1.7 ± 0.7 × 104 reads per sample. These total reads resulted in 2472 amplicon sequence variants (ASVs) after applying the DADA2 algorithm. The 2472 ASVs clustered in 775 different amino acid sequences at 100% identity. Therefore, the aprA ASVs might not always represent different protein sequences, and the nucleotide-based diversity might not always correspond to a similar functional diversity. However, in this study, we sequenced only a fraction of the aprA gene (~410 bp vs. ~1700 bp) and thus, we cannot explore the diversity of the entire protein. Taking this into consideration, we decided to use the ASVs to obtain more detailed information on the diversity of the aprA harbouring community.

The epi- and mesopelagic layers harboured 718 (29%) and 612 (25%) unique ASVs respectively (Fig. 3). In the lower and upper bathypelagic layers 339 (14%) and 476 (19%) unique ASVs respectively, were identified. Epi- and mesopelagic layers shared ~8% ASVs, whereas the upper and lower bathypelagic layers shared ~3% ASVs. Only eight (0.3%) ASVs were ubiquitous throughout the four pelagic zones (Fig. 3).

Details are in the caption following the image
Venn diagram showing the shared and unique aprA amplicon sequence variants (ASVs) and their relative contribution (in %) in four different depth layers throughout the Atlantic Ocean.

The cumulative read abundance showed clear differences between epi- and mesopelagic communities and bathypelagic aprA harbouring communities (Fig. 4). The cumulative abundance of the bathypelagic communities increased more steeply than the epipelagic communities with few very abundant ASVs (Fig. 4A). The lower bathypelagic layer exhibited the lower Chao richness and Pielou evenness indices of all the depth layers. Shannon diversity was significantly lower in the lower bathypelagic compared to the surface and mesopelagic communities (ANOVA on rank, Bonferroni corrected, p < 0.05). The phylogenetic diversity of the aprA harbouring community significantly differed between the mesopelagic and lower bathypelagic realms (ANOVA on rank, Bonferroni corrected, p < 0.05) (Fig. 4C). The epipelagic, mesopelagic and bathypelagic aprA harbouring communities were clearly separated, however, the upper and lower bathypelagic communities overlapped (Fig. 4B). The first and second coordinates of the principal coordinates analyses explained 59% and 13% of the community variation respectively.

Details are in the caption following the image
Cumulative abundance (A) and principal coordinate analysis (B) of the aprA harbouring prokaryotic community at the four depth layers studied. Box plot showing Chao richness, Shannon and Phylogenetic Diversity (PD) indexes and Pielou evenness (C) of the aprA containing prokaryotic community at different depths. The bottom and the top of the box represent the first and the third quartiles, while the thick horizontal line represents the median. Outliers (black dots) and jitter values (coloured dots) are shown.

Taxonomic assignment and community composition of aprA harbouring prokaryotes

Taxonomic assignment yielded 12 unclassified ASVs and two assigned only at the kingdom level, together accounting for ~0.2% of the total reads. Within the classified ASVs (at or above the phylum level), Alphaproteobacteria dominated the epipelagic aprA containing community (77 ± 24%), whereas Gammaproteobacteria dominated the mesopelagic (54 ± 18%) and bathypelagic (48 ± 9%) aprA harbouring community. Deltaproteobacteria also significantly contributed to the bathypelagic community accounting for 19 ± 9% of the aprA containing prokaryotic community (Table 2; Fig. 5). At the order level, Pelagibacterales dominated the epi- and the mesopelagic aprA containing communities, contributing 77 ± 23% and 33 ± 17% to the communities respectively. Pelagibacterales also significantly contributed (16 ± 13%) to bathypelagic aprA harbouring communities. Three taxa [Delta_proteobacterium_SCGC (19 ± 9%), Candidatus_Thioglobus (18 ± 9%) and Chromatiales (13 ± 3%)] dominated the upper and lower bathypelagic aprA containing prokaryotic communities.

Table 2. Phylogenetic affiliation and relative contribution (%, ±SD) of the aprA containing prokaryotic community in four different depth layers at high and low latitude oceanic provinces.
Surface Mesopelagic Upper bathypelagic Lower bathypelagic
High_latitude Low_latitude High_latitude Low_latitude High_latitude Low_latitude High_latitude Low_latitude
Average SD Average SD Average SD Average SD Average SD Average SD Average SD Average SD
Acidiferrobacterales 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.7 0.9 1.0 1.1
Alpha_environmental_samples 0.0 0.0 0.0 0.0 0.1 0.2 1.2 0.5 5.9 4.5 4.6 1.5 14.1 9.9 7.2 5.9
Candidatus_Thioglobus 4.6 7.0 0.1 0.1 5.8 2.2 10.8 4.5 10.7 6.9 17.7 4.0 19.4 13.2 24.7 7.1
Chromatiales 8.2 6.9 2.2 2.4 16.5 8.1 9.6 1.2 11.3 4.0 14.4 2.8 10.8 2.8 14.3 1.1
Delta_proteobacterium_SCGC 0.0 0.0 0.0 0.0 0.3 0.3 0.0 0.0 18.5 8.5 23.4 3.0 9.0 3.2 25.4 10.7
Desulfobacterales 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0
Environmental_marine_bacterium_NA 24.5 18.8 1.2 2.0 6.7 11.2 0.0 0.1 0.0 0.0 0.0 0.0 0.6 0.6 0.0 0.0
Gamma_environmental_samples 0.0 0.0 0.1 0.2 3.0 3.4 11.7 8.9 4.9 0.4 9.3 1.5 1.4 1.0 4.5 1.7
Gamma_proteobacterium_SCGC 0.1 0.1 0.0 0.0 7.7 10.3 1.9 2.5 6.2 2.8 1.8 1.0 6.3 4.7 0.8 0.8
Nitrosomonadales 0.0 0.0 0.1 0.2 0.0 0.0 0.7 1.2 1.5 0.5 1.1 1.4 0.0 0.0 0.0 0.0
Pelagibacterales 57.8 18.0 95.1 3.1 44.0 13.2 22.0 13.2 16.9 3.0 13.3 0.3 22.1 28.0 13.4 7.1
Proteobacteria_Alphaproteobacteria_NA 0.0 0.1 0.2 0.2 0.3 0.5 1.1 1.1 0.0 0.0 0.7 0.5 0.0 0.0 0.0 0.0
Proteobacteria_Deltaproteobacteria_NA 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Proteobacteria_Gammaproteobacteria_NA 3.1 4.6 0.2 0.3 6.2 1.5 33.3 1.5 11.0 2.2 6.7 1.0 6.1 4.8 5.3 1.8
Proteobacteria_NA_NA 0.9 1.4 0.0 0.0 8.1 9.3 7.5 3.7 9.9 4.6 7.1 1.4 9.2 8.0 3.2 2.1
SAR116_cluster 0.2 0.2 0.5 0.8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Sulfur_oxidizing_symbionts 0.8 0.9 0.2 0.2 1.4 2.1 0.1 0.1 3.2 5.6 0.0 0.0 0.0 0.1 0.0 0.0
Details are in the caption following the image
Normalized abundance (by the median sequencing depth) of abundant aprA harbouring taxa at the order level in different depth layers (A) and their relative contribution at different stations (B) along the Atlantic transect.

Significant differences were observed between aprA containing communities of high and low latitude regions. Specifically, Alphaproteobacteria contributed more to epipelagic communities at low latitudes than at high latitudes (96 ± 2% and 58 ± 18% respectively; T-test p < 0.05). Conversely, Alphaproteobacteria accounted for a substantially lower fraction of the community in meso- and bathypelagic communities of low latitude regions compared to the high latitude communities (Table 2; Figs 5B and S2). Gammaproteobacteria showed opposite trends compared to the Alphaproteobacteria in epipelagic zones, and no significant differences between high and low latitudes for bathypelagic communities (T-test p > 0.05). Two groups (Pelagibacterales and Chromatiales) were ubiquitously present and therefore significantly contributed to the epi-, meso- and bathypelagic communities. However, their contribution differed between different regions and depth layers.

Four ASVs, two closely related to the Delta_proteobacterium_SCGC and the other two putatively identified as closely related to Chromatiales, and unassigned Gammaproteobacteria composed the core bathypelagic bacterial community harbouring aprA genes. Only one ASV related to Pelagibacterales composed the core community of the mesopelagic layer. Epipelagic prokaryotic communities did not exhibit a core aprA gene harbouring community. The bathypelagic resident aprA containing community was dominated by ASVs with similar taxonomic affiliation as those found in the core community with the addition of Candidatus Thioglobus and Pelagibacterales. Eight ASVs (two closely related to Pelagibacterales, two related to Gamma environmental samples, and the other four putatively identified as closely related to Chromatiales, Candidatus Thioglobus, unassigned Gammaproteobacteria and unassigned Proteobacteria) composed the resident mesopelagic bacterial community harbouring aprA genes. The epipelagic aprA harbouring resident community was dominated by ASVs putatively assigned to Pelagibacterales.

The epi- and mesopelagic were clearly separated from bathypelagic samples as indicated by a constrained principal coordinates analysis (CAP) based on weighted UniFrac distance with the first and the second coordinates accounting for 58.3% and 9.4% of the community variation respectively (Fig. 6). Furthermore, the upper and lower bathypelagic communities partially overlapped, whereas epi- and mesopelagic communities were clearly separated. Pelagibacterales ASVs comprised most of the variation of epi- and mesopelagic communities, while Chromatiales and Deltaproteobacteria accounted for most of the variation in the bathypelagic communities (Fig. 6A). Spatial and physicochemical factors explained a significant portion of the beta-diversity of the aprA harbouring communities. All the selected parameters accounted together for 79% (p < 0.001) of the community variation. Specifically, depth (32%, p = 0.001), temperature (13%, p = 0.001) and latitude (11%, p = 0.001) accounted for most of the variation (summing up 56%), followed by oxygen, nitrite, DIC and silicate (Fig. 6).

Details are in the caption following the image
Constrained principal coordinate analysis (CAP) based on weighted UniFrac distance of the aprA harbouring bacterial community showing the environmental parameters related to the bacterial variation. Variation of the most abundant orders (A) and samples (B).

Additionally, variation partition analysis revealed that 39% of the total variability in the aprA containing community was explained by the selected environmental parameters. Specifically, physicochemical parameters explained 27% of the variation followed by the spatial variables (depth and latitude) with 6% and temporal factors with 0.6%. The combination of the physicochemical and spatial variables explained an additional 5% of the aprA harbouring community variation (Fig. 7).

Details are in the caption following the image
Variation partitioning of physicochemical, spatial and time factors explaining the aprA-harbouring bacterial community composition.

Distribution of Candidatus Pelagibacter oligotypes

Oligotyping analysis at the amino acid level of the Candidatus Pelagibacter aprA containing community identified 35 oligotypes (Fig. 8A). The epi- and mesopelagic oligotypes clearly differed from the composition of bathypelagic oligotypes. The z-score distribution of aprA containing Candidatus Pelagibacter oligotypes exhibited four main clusters. The first cluster grouped communities from epipelagic, another cluster grouped high latitude communities of different depths, two other clusters grouped communities obtained from the upper and lower bathypelagic layers respectively.

Details are in the caption following the image
Distribution of Pelagibacter sp. oligotypes at six stations and four depths (A), different colours represent specific oligotypes. Heatmap of the z-score distribution of the Pelagibacter sp. oligotypes (B). The dendrogram clusters the samples according to Bray–Curtis similarity.


Factors driving the depth-related and biogeographic distribution of the prokaryotes with a role in sulfur cycle and chemolithoautotrophy

The distribution of prokaryotes involved in the sulfur cycle, such as those harbouring the aprA gene involved in sulfur oxidation or reduction, has been studied in marine sediments (Blazejak and Schippers, 2011; Watanabe et al., 2013; Aoki et al., 2015) and hydrothermal vents (Hügler et al., 2010; Frank et al., 2013). However, limited information is available on the distribution of aprA harbouring prokaryotes across different oceanic provinces and throughout the water column. Single amplified genomes (SAGs) and metagenomics studies revealed the importance of the potential for sulfur oxidation/reduction in the open ocean (Swan et al., 2011; Ulloa et al., 2012; Li et al., 2018), yet to which extent prokaryotes harbouring these genes contribute to the global sulfur cycle remains unknown.

The increase in the aprA abundance from the surface to the bathypelagic layers indicates an increasing contribution of genes involved in sulfur metabolism with depth coinciding with a decrease in bioavailable organic carbon (Hansell, 2013) potentially caused by dilution of dissolved organic matter (DOM) compounds (Arrieta et al., 2015; Wagner et al., 2020). This finding suggests that the limitation of bioavailable DOM in the deep ocean promotes the proliferation of alternative metabolic pathways, including chemoautotrophy or lithoheterotrophy, utilizing inorganic compounds as a supplementary source of energy. The increasing contribution of aprA containing prokaryotes to the communities in the deep ocean and in low latitude oligotrophic regions (reaching up to a median of 3.5% of the total prokaryotic community in the lower bathypelagic of the NAG, Table S1) supports the notion of an important role of inorganic sulfur compounds as a supplementary energy source in organic matter-limited environments. The contribution of aprA containing prokaryotes to the community determined in this study falls within the range (~3% to 20%) of values reported in a recent metagenomic study of the global deep ocean (Acinas et al., 2021). However, it is worth mentioning that our values might be underestimated compared to those of Acinas et al. (2021), since we calculated the aprA contribution to the community normalizing to total PA (bacteria and archaea), whereas Acinas et al. (2021) normalized to recA gene abundance (a housekeeping gene present only in bacteria). Moreover, some underestimation could also be present due to the DNA extraction efficiency when comparing gene abundance with PA determined by flow cytometry.

Chemolithoautotrophy has been demonstrated to occur in the meso- and bathypelagic realms at higher rates than previously assumed (Reinthaler et al., 2010). Chemoautotrophs such as ammonia-oxidizing archaea (AOA) show similar biogeographic and depth-related distribution patterns in the Atlantic (Sintes et al., 2016) as aprA containing prokaryotes (this study). Also, AOA increase in their contribution to the prokaryotic community towards oligotrophic regions and with depth (Sintes et al., 2016, Santoro et al., 2017). In contrast, relatively low contributions of AOA to total PA are found in highly productive regions such as the Arctic and Antarctic (Sintes et al., 2016), in agreement with the higher depth-integrated DIC fixation rates measured in the oligotrophic compared to mesotrophic North Atlantic Ocean regions (Reinthaler et al., 2010). Taken together, these findings suggest that the availability of organic compounds under mesotrophic conditions favours heterotrophic life strategies over slow-growing chemolithoautotrophic and mixotrophic prokaryotes.

Adenylyl sulfate reductase is present in both sulfate reducers and sulfur oxidizers playing different roles (Meyer and Kuever, 2007b). Therefore, to assess the link between potential chemoautotrophy and the aprA containing community, the distribution of the cbbM encoding form II ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) enzyme (Swan et al., 2011) responsible for DIC fixation was determined. The cbbM containing prokaryotes showed a similar distribution pattern as the aprA harbouring prokaryotes, exhibiting a higher contribution in oligotrophic regions and the meso- and bathypelagic prokaryotic communities compared to high latitudes and epipelagic communities. The ratio between aprA and cbbM gene abundance indicates that at least a fraction of the prokaryotic community carrying aprA genes oxidizes inorganic sulfur compounds as a source of energy coupled to CO2 fixation, in agreement with previous findings in the mesopelagic Atlantic (Swan et al., 2011).

Multiple forms of RuBisCO occur in prokaryotes (Tabita et al., 2008), and some sulfur oxidizers might harbour other types of cbbM than cbbM form II. Consequently, the potential contribution of DIC fixation associated with sulfur oxidation in the present study is likely a conservative estimate. The ratios of aprA/cbbM in the mesopelagic and bathypelagic realm (with a median of 1.7 and 4.8 respectively; Fig. 2E and F) support the notion of a larger role of the prokaryotic community carrying aprA as chemoautotrophs than in epipelagic waters where the aprA/cbbM ratio had a median of 31 (Fig. 2E and F). This finding suggests that in the epipelagic communities most adenylyl sulfate reductase activity is not coupled to DIC fixation. The dominance of Pelagibacterales in the aprA harbouring community (Fig. 5) and the proposed role of APS reductase in detoxification of sulfite in this taxon (see below) corroborates this notion. High productivity regions such as the high latitude provinces and the equatorial area showed lower aprA/cbbM ratios than oligotrophic provinces (Fig. 2). The relatively high abundance of the cbbM gene in these provinces (ARCT, WTRA and SANT) suggests that the inorganic carbon fixation was coupled to alternative sources of energy (such as ammonia, iron, hydrogen, methane) rather than sulfur. Our findings are in agreement with a previous study conducted in the Atlantic Ocean (Swan et al., 2011) showing that not all the SAGs containing RuBisCO (cbbM and cbbL) also contain sulfur-oxidizing genes.

Adenylyl sulfate reductase can alternatively perform dissimilatory sulfate reduction as a form of anaerobic respiration using sulfate as an electron acceptor. This process is strictly anaerobic and therefore limited to OMZs and anaerobic microenvironments (Ulloa et al., 2012; Bianchi et al., 2018). Suspended and sinking marine snow can provide anoxic or microaerobic microenvironments, thus promoting the proliferation of sulfate-reducing bacteria (Ploug et al., 1997; Bianchi et al., 2018). Therefore, a fraction of the aprA harbouring community in the meso- and bathypelagic layers could inhabit these microniches. However, our results indicate that sulfate reducers constitute a minor fraction of the aprA containing prokaryotes in the open oxygenated ocean, in agreement with the patchiness of suitable marine snow in these waters (Silver et al., 1978).

Major aprA harbouring taxa in the oxygenated water column of the open Atlantic

The aprA harbouring prokaryotic community shares only a few ASVs and amino acid sequence variants (AASVs) between different depth layers throughout the Atlantic Ocean (Figs 3 and S2), suggesting a clear niche differentiation according to environmental gradients associated to depth layers. The phylogenetic characterization reveals that the aprA-containing community in the oxygenated water column is dominated by sulfur-oxidizing bacteria in the open Atlantic (Fig. 5). Neither Archaea nor bacterial sulfate reducers significantly contributed to the aprA harbouring community, suggesting a minor role of these groups in the sulfur cycle of the pelagic open ocean. However, the occurrence of anaerobic microenvironments in oxygenated waters, e.g. within marine snow, may sporadically promote the proliferation of sulfate reducers mediating sulfate reduction in the otherwise oxygenated waters (Bianchi et al., 2018).

SAR11, one of the most abundant and ubiquitous bacterial groups in marine ecosystems (Morris et al., 2002) and of paramount significance in the global ocean biogeochemical cycles, dominates the aprA-harbouring community in the epi- and mesopelagic layers (Fig. 3). Candidatus Pelagibacter apr belongs to the sulfur-oxidizing apr lineage-I (Meyer and Kuever, 2008), however, the lack of other enzymes from the pathway prevents the dissimilatory function in these organisms (Meyer and Kuever, 2007b). Consequently, it has been suggested that the primary role of APS reductase in SAR11 is intracellular detoxification of sulfite, a metabolite of organic sulfur degradation (Meyer and Kuever, 2007b). Indeed, SAR11 clade members are deficient in assimilatory sulfate reduction genes and thus, require organic sulfur compounds for growth (Tripp et al., 2008). DMSP, released by phytoplankton in the surface ocean, and taurine, released by zooplankton and other metazoans down to bathypelagic waters, potentially match the sulfur requirements of SAR11 in the ocean (Clifford et al., 2019; Durham et al., 2019). The reported decreased transcription of aprAB in SAR11 under sulfur limited conditions (Smith et al., 2016) supports the notion that APS reductase plays a role in detoxification.

Oligotyping provides evidence that there are different SAR11 ecotypes adapted to different environmental conditions. The oligotype composition of the aprA-harbouring SAR11 community at the amino-acid level is depth-stratified (Fig. 8), suggesting different adaptations to metabolize sulfur compounds of different origins, such as DMSP in the epipelagic layers or taurine in deeper pelagic zones. Another possibility would be that deep ocean ecotypes might be using inorganic sulfur oxidation as a source of additional energy (i.e. chemolithotrophy). Chemolithotrophy has not been found in Candidatus Pelagibacter ubique (Meyer and Kuever, 2007a, 2007b) as a consequence of lacking sulfate adenylyltransferase (sat). However, knowledge on the metabolic potential of deep ocean ecotypes is generally limited, and the capacity (or absence of it) for chemolithoheterotrophy based on sulfur oxidation has not been assessed yet.

In the bathypelagic realm, the aprA harbouring community was dominated by Chromatiales, Deltaproteobacteria SCGC and SUP05 clade members (Fig. 5). Members of the Chromatiales order are known as anoxygenic phototrophic bacteria that utilize sulfide as an electron donor, and subsequently oxidize it to sulfate (Imhoff, 2014). Nevertheless, a number of species are aerophilic sulfur oxidizers common in illuminated stratified coastal waters (Imhoff, 2014), near hydrothermal vents (Han et al., 2018; Pjevac et al., 2018) in deep-sea sponges (Kennedy et al., 2014) and in deep sediments (Li et al., 2013). Members of the Chromatiales order have been also reported throughout the oxygenated water column in the global ocean based on 16S rRNA gene sequencing, e.g. in the Gulf of Alaska (Muck et al., 2019), North Atlantic (Frank et al., 2016), or the Pacific (Pham et al., 2008) with increased relative abundances in meso- and bathypelagic waters. However, our knowledge on their metabolic strategies in the open oxygenated water column remains limited and further studies with isolates are needed to shed light on their metabolic adaptations.

Candidatus Thioglobus (SUP05 clade) together with the uncultivated Deltaproteobacterium SCGC (SAR324 cluster) increased in its contribution to the prokaryotic aprA containing community with depth. Deltaproteobacterium SCGC (SAR324), contributing 19 ± 9% to the bathypelagic aprA containing bacterial community, exhibits DIC fixation in conjunction with sulfur oxidation (Swan et al., 2011). It contributes significantly to the chemoautotrophy of the deep ocean based on sulfur oxidation. The isolate Candidatus Thioglobus singularis PS1 (member of the SUP05 clade) has the potential to use organic substrates and also to perform DIC fixation and sulfur oxidation (Spietz et al., 2019b). However, Candidatus Thioglobus NP1 isolated from the open ocean lacks the RuBisCO encoding gene in its genome, suggesting that DIC fixation is not performed by all SUP05 (Spietz et al., 2019a), in agreement with the high aprA/cbbM ratios (Fig. 2).

In conclusion, our study showed a clear niche differentiation of the prokaryotes carrying aprA genes in the open ocean. The aprA harbouring prokaryotic community was dominated by sulfur oxidizers, partly capable of DIC fixation (chemoautotrophy). Adenylyl sulfate reductase putatively assigned to Archaea and sulfate reducers was neglectable in this study. This suggests that in the ocean's oxygenated water column the adenylyl sulfate reductase is mainly used in the chemolithoautotrophic or chemolithoheterotrophic pathways as an energy source or in the biosynthesis pathways. This takes place especially in areas depleted in bioavailable organic carbon, such as oligotrophic regions and/or the deep ocean. Taken together, our results suggest that sulfur oxidizers may play an important role in the sulfur and carbon cycle of the oxygenated waters of the global ocean.

Material and methods

Study area and sampling

Water samples were collected during the GEOTRACES (legs 1–3) cruises on board of R/V Pelagia (in April and July 2010) and R/V James Cook (in February–March 2011) along a latitudinal transect in the Atlantic Ocean from 64°N to 50°S (Fig. 1). Sampling was performed using a rosette water sampler equipped with 24 25-L Niskin bottles, a CTD (conductivity-temperature-depth; SBE43 Seabird, Bellevue, WA, USA), with additional sensors for chlorophyll fluorescence, turbidity and oxygen concentration. Samples for inorganic nutrients, trace elements and microbial parameters were collected at 24 depth layers and 51 stations. Samples for quantitative PCR analyses (aprA and cbbM genes) were collected at 6–8 depth layers from epi- to bathypelagic waters. Three pelagic zones were distinguished throughout the depth profiles: epipelagic (50–70 m), mesopelagic (200–1000 m) and bathypelagic (1000–5000 m). The bathypelagic layer was subsequently divided into upper bathypelagic (1000–2000 m) and lower bathypelagic (2000–5000 m) based on the different physicochemical characteristics of these two depth layers, resulting in a total of four pelagic zones sampled. The upper bathypelagic (collected at ~1250 m depth) included samples from Iceland-Scotland Overflow Water, Labrador Sea Water, upper North Atlantic Deep Water (NADW) and Antarctic Intermediate Water masses. The lower bathypelagic layer included samples collected between 2500 and 4500 m depth and included the Antarctic Bottom Water, intermediate NADW and Denmark Strait Overflow Water, depending on the oceanographic region and depth.

Six oceanic provinces were discriminated based on the description of Longhurst (2007): the North Atlantic Arctic province (ARCT; 70°N–55°N), the North Atlantic Drift province (NADR; 55°N–40°N), the North Atlantic Gyre (NAG; 40°N–12°N), the Western Tropical Atlantic province (WTRA; 12°N–6°S), the South Atlantic Gyre (SATL; 6°S–40°S), and the Subantarctic province (SANT; 40°S–55°S) (Fig. 1).

Inorganic nutrients

The concentrations of dissolved inorganic nutrients (NO3, NO2, PO43−) were determined on 0.2 μm filtered water samples in a TRAACS 800 autoanalyser system immediately after collecting the samples following established protocols (Reinthaler et al., 2008).

Microbial counts

The PA was estimated by flow cytometry after nucleic acid staining with Sybr Green I (Giorgio et al., 1996; Brussaard, 2004).

DNA extraction and amplification

Two (epipelagic) to ten litres (mesopelagic to lower bathypelagic) of seawater were filtered through 0.22 μm GTTP polycarbonate filters (Millipore) for DNA analyses. The extraction was performed using the Ultraclean soil DNA isolation kit (Mobio) following the manufacturer's protocol.


q-PCR was used to evaluate the abundance of two cbbM genes encoding the RuBisCO form II, subsequently named cbbM65 and cbbM68 according to the primer sets used (Swan et al., 2011), and the aprA gene encoding the alpha subunit of APS. cbbM65 and cbbM68 cover the two main clusters of cbbM (rubisco form II) (Swan et al., 2011), and were specifically developed for q-PCR.

q-PCR was conducted in triplicate for all functional genes on a LightCycler 480 SW 1.5 (Roche). The reaction mixture (10 μl) contained 1× LightCycler 480 DNA SYBRGreen I Master (Roche), 0.5 μM of specific primers (Table S2), 1 μl of DNA extract made up to 10 μl with PCR-grade water (Roche). All reactions were performed in 96-well q-PCR plates (Roche) with optical tape. Accumulation of newly amplified double-stranded DNA products was followed by the increase of fluorescence due to the binding of the fluorescent dye SYBRGreen to the newly produced DNA. Specificity of the q-PCR reaction was tested on agarose gel electrophoresis and with a melting curve analysis (65–95°C). Each gene was amplified using specific primer sets and annealing temperatures as described in Table S2, and quantified using specific standards. The q-PCR thermocycling for cbbM was composed of an initial denaturation step at 95°C for 10 min and 50 amplification cycles consisting of denaturation at 95°C for 5 s, annealing at the specific primer temperature (Table S2) for 5 s, extension at 72°C for 15 s and plate reading at 76°C for 3 s. The average q-PCR efficiencies for the cbbM65 and cbbM68 genes were 73% and 91% respectively. The total cbbM abundance was calculated as the sum of cbbM65 and cbbM68 gene abundance. The q-PCR cycling for aprA consisted of an initial denaturation step at 95°C held for 10 min, followed by 50 amplification cycles consisting of denaturation at 95°C for 5 s, annealing at 60°C for 15 s, extension at 72°C for 30 s and plate read at 80°C for 3 s. The average efficiency of the q-PCR for aprA gene was 81%. The specificity of the q-PCR reaction for the different genes was checked with melting curve analyses (65–95°C).

Sequencing and bioinformatics analyses of the bacterial aprA genes

Samples were collected at four depth layers covering the epipelagic to deep bathypelagic realm at six stations corresponding to different geographic provinces to determine the composition of the prokaryotic aprA harbouring community. The aprA gene was amplified using aprA-1F and aprA-5R primers (Meyer and Kuever, 2007b) (Table S2). The thermocycling PCR conditions were as follows: initial denaturation at 94°C for 4 min; amplification: 30 cycles, at 94°C for 1 min, 55°C for 1 min, and extension at 72°C for 1 min, followed by a final extension step at 72°C for 7 min and holding at 4°C. The PCR products were purified using PCRExtract MiniKit (5-PRIME) and quality checked in a 2% agarose gel. The purified PCR products were subsequently sequenced with Illumina Miseq (IMGM Laboratories GmbH) with V2 chemistry. The data are publicly available in the NCBI database under the accession number PRJNA641589.

The raw reads were quality-filtered and processed using dada2 version 2.1 (Callahan et al., 2016b). The obtained ASVs were taxonomically assigned using a naïve Bayesian classifier method with a minimum bootstrap threshold of 40 (implemented in dada2) against a reference aprA database. The database was built using aprA gene sequences obtained from NCBI database closely related to the ASVs according to blast results and of known taxa (Supplementary Table S3). The obtained data were subsequently analysed with phyloseq R package (McMurdie and Holmes, 2013; Callahan et al., 2016a). AprA gene sequences were further classified using custom phylogenetic trees. The ASVs were translated to amino acid sequences, and subsequently, CD-HIT (Huang et al., 2010) was used to cluster the amino acid sequences at 100% sequence identity. We named the resulting amino acid cluster sequences AASVs. The AASVs were aligned together with amino acid sequences obtained from NCBI database using MEGA-X (Kumar et al., 2018). The phylogenetic tree of the aprA AASVs was built using the Neighbour-Joining method (Saitou and Nei, 1987) with 100 bootstraps in MEGA-X. The resulting tree was subsequently drawn using iTOL (Letunic and Bork, 2019) (Fig. S2).

Shared and unique aprA ASVs in different depth layers were determined with the ‘Venn’ package in R software using presence/absence data. Richness and diversity indexes were calculated with phyloseq R package. Weighted UniFrac distances and principal coordinate analysis (Lozupone et al., 2011) were calculated with ampvis2 R package and used to compare the aprA community composition between samples (Andersen et al., 2018). The cumulative read abundance was also estimated using ampvis2 R package. ASVs reported in more than 80% of the samples from a specific depth layer were considered core phylotypes of that depth layer, ASVs reported in 50%–79% of the samples were considered resident phylotypes.

The composition of the aprA containing prokaryotic community and its relationship with the environmental variables was analysed by CAP implemented in phyloseq R package using weighted UniFrac distance data. To reveal the major environmental factors affecting the community composition, variance partitioning was performed on the Hellinger-transformed dataset using the varpart function of vegan (Oksanen et al., 2018). The variance partitioning results were visualized with eulerr R package (Larsson, 2020).

Oligotyping analysis of aprA assigned to the Candidatus Pelagibacter family at the amino acid level was conducted following the pipeline available at http://oligotyping.org (Eren et al., 2013).

Statistical analyses

Relationships between gene abundances (aprA, cbbM) and the aprA/cbbM ratio with environmental variables were analysed using a dbRDA with the ‘Vegan’ package (Oksanen et al., 2018) implemented in R software v.3.4. Environmental variables included latitude (absolute), depth, temperature, salinity, fluorescence, dissolved oxygen, DIC, macronutrients (phosphate, silicate, nitrite and nitrate) and trace elements (Cd, Fe, Mn, Zn). Micronutrients, such as trace metals, play an important role for the functioning of proteins and enzymes. Adenylyl-sulfate reductase consists of two subunits (alpha and beta), the beta subunit contains Fe. Other metals serve as cofactors during the sulfur oxidation processes. Moreover, during sulfate reduction and production of sulfide, Cd precipitates as CdS. Therefore, we decided to test whether the aprA gene abundance was influenced by the availability of specific micronutrients (such as trace metals) as they play an important role in the sulfur oxidation and sulfate reduction processes.

The response variables were normalized and Euclidean distance similarities matrices were calculated. The F-test and the p-value were obtained applying 999 permutations. The dataset of environmental parameters and the corresponding methods used for the measurements are available at the GEOTRACES website (http://www.geotraces.org) (Mawji et al., 2015).

Analysis of variance (ANOVA test) was performed to test possible variations within different depth layers and/or oceanic regions.


We thank the captain and crew of the R/V Pelagia for their support and splendid atmosphere on board. Shiptime was provided by the Geotraces program. Laboratory work was supported by the Austrian Science Fund (FWF) projects: I486-B09, Z194-B17, P28781-B21 and by the European Research Council under the European Community's Seventh Framework Program (FP7/2007-2013)/ERC grant agreement No. 268595 (MEDEA project) to G.J.H. D.D.C. was supported as an overseas researcher under the Postdoctoral Fellowship of Japan Society for Promotion of Science (P16085). E.S. was supported by the Austrian Science Fund (FWF) project P27696-B22. The valuable comments of the reviewers helped to improve the quality of the paper.

    Data Availability Statement

    The data are publicly available in the NCBI database under the accession number PRJNA641589.