Remote sensing and field surveys have identified ∼400 lakes beneath the Antarctic ice sheet (Smith et al., 2009; Wright and Siegert, 2012), provided evidence for the widespread occurrence of water-saturated sediments (e.g., Studinger et al., 2001), and demonstrated hydrologic connections between certain lakes (Fricker et al., 2007). The abundance of water and availability of metabolic energy sources beneath the ice sheet (Bell, 2008; Smith et al., 2009; Skidmore, 2011; Palmer et al., 2013) supports the possibility for microbial ecosystems and has motivated international exploration efforts in West and East Antarctica (Fricker et al., 2011; Lukin and Bulat, 2011; Ross et al., 2011). Although little information has been available to infer conditions in Antarctica’s vast and complex subglacial aquifer, microbial cells inhabiting this biome are estimated at ∼10²⁹ (Priscu et al., 2008), a value similar to global estimates for open ocean waters (Whitman et al., 1998). Fortunately, recent progress in subglacial exploration (Tulaczyk et al., 2014) has provided an opportunity to directly examine the nature, evolution, and biogeochemical contributions of microbial communities at the bed of Antarctica’s ice sheet.

Much of the data amassed on subglacial microbial communities has been derived through analyses of subglacial water outflows from alpine and polar glaciers (Skidmore et al., 2005; Cheng and Foght, 2007; Mikucki and Priscu, 2007; Boyd et al., 2011; Hamilton et al., 2013; Dieser et al., 2014) or basal materials retrieved from deep ice boreholes (e.g., Priscu et al., 1999; Lanoil et al., 2009; Christner et al., 2012). This approach has been valuable for providing baseline data and identifying common trends in disparate subglacial ecosystems. For example, there is geochemical and microbiological evidence that subglacial microbes derive their energy through the weathering of bedrock minerals and thus influence subglacial water chemistry (Skidmore, 2011; Mitchell et al., 2013). The prevalence of phylotypes related to species of Thiobacillus and Sideroxydans suggests primary production at the glacier bed may rely on reduced iron and sulfur compounds liberated through glacial comminution and microbiological processes occurring in the sediments or at the bedrock interface (Skidmore et al., 2010; Boyd et al., 2014). Evidence for the activity of methanogenic, methanotrophic, and ammonia oxidizing species has also been provided in several subglacial environments (Boyd et al., 2010, 2011; Dieser et al., 2014), implying these pathways could also play important roles in carbon and nitrogen cycling beneath larger ice masses. Although these pioneering efforts have provided valuable data to generate hypotheses on the structure and function of subglacial microbial ecosystems, their applicability to environments beneath ice sheets has remained uncertain.

Directly sampling sub-ice aquatic environments in a microbiologically clean manner is logistically challenging (Doran et al., 2008; Siegert et al., 2012), requiring strategies to reduce microbial cells associated with the drilling process and minimize exchange between the surface and subglacial environment (Priscu et al., 2013). During January 2013, the Whillans Ice Stream Subglacial Access Research Drilling (WISSARD) Project conducted the first successful sampling of an Antarctic subglacial lake (Christner et al., 2014; Tulaczyk et al., 2014). Christner et al. (2014) reported that planktonic bacteria and archaea in the aerobic water column were at an average concentration of 1.3 × 10⁵ cells mL^-1 and morphologically diverse. Molecular analysis of 16S rRNA gene sequences amplified from the water column and surficial sediments (0–2 cm) revealed a rich prokaryotic community consisting of several phylotypes similar to chemosynthetic species that have been observed in alpine and polar glacier environments (e.g., members of Thiobacillus, Sideroxydans, and Methylobacter (Lanoil et al., 2009; Boyd et al., 2014; Dieser et al., 2014). Furthermore, primary and heterotrophic production data revealed that Subglacial Lake Whillans (SLW) contained a metabolically functional microbial community that may be sustained by dark autotrophic activity (Christner et al., 2014).

Here we present a detailed description of SLW’s water column and sediment (to depths of 36 cm) communities based on analysis of amplified 16S rRNA genes (rDNA) and molecules (rRNA). This approach served the dual role of providing information on microbial community structures while also allowing an assessment of potentially metabolically active taxa and the biogeochemical reactions they are likely to catalyze. Our data provide an initial framework for discerning the diversity and ecology of Antarctic subglacial lake environments, and support the hypotheses that microbial transformations beneath ice masses are driven by chemosynthesis and have global biogeochemical significance (Wadham et al., 2012).

Antarctic subglacial lakes have been a scientific curiosity since their discovery (Oswald and Robin, 1973). The initial studies on accretion ice from Subglacial Lake Vostok concluded that viable microbes were present in the surface waters of the lake (Karl et al., 1999; Priscu et al., 1999), but due to concerns about sample contamination (e.g., Christner et al., 2005), evidence supporting the habitability of Antarctica’s subglacial environment was subject to criticism. More recently, studies have demonstrated the suitability of subglacial environments as ecosystems and shown that liquid water reservoirs at the base of glaciers harbor diverse assemblages of microorganisms (e.g., Skidmore et al., 2005; Boyd et al., 2011; Hamilton et al., 2013; Dieser et al., 2014). However, progress in understanding Antarctic subglacial lake microbiology has been hampered by logistical constraints that make sampling and direct observations challenging. The WISSARD project collected the first pristine water and sediment samples from an Antarctic subglacial lake (Christner et al., 2014), providing an unprecedented opportunity to examine microbial community structure and infer ecosystem processes in an aquatic habitat beneath the West Antarctic Ice Sheet (WAIS).

Conclusions on the composition of SLW microbial communities can be made with confidence given the significant differences they displayed from assemblages in the drilling water and procedural controls (Figure 1; Supplementary Results). With the exception of Janthinobacterium, Tumebacillus, and Herbaspirillum-related OTUs, OTUs identified in the drilling water were rare or absent in lake water samples. Microorganisms that were present in the borehole and drill water may be human-based contaminants, derived from the drill system plumbing, and/or have originated from the snow and glacial ice that sourced the water for the drill (Blythe et al., 2014; Rack et al., 2014). The latter is supported by the abundance of OTUs related to Tumebacillus and Janthinobacterium in drilling and borehole water, as these taxa have been frequently identified in polar environments (e.g., Steven et al., 2008; Kim et al., 2012). Alternatively, the presence of OTUs related to Herbaspirillum, a bacterial genus that frequently contaminates DNA extraction kits (Salter et al., 2014), may be an artifact of laboratory contamination. Determining the conclusive source of the Herbaspirillum phylotypes is complicated by the fact that limnological studies of surface Antarctic lakes have identified these taxa as bona fide members of the bacterioplankton communities (Pearce et al., 2003; Kuhn et al., 2014). Given these uncertainties, all of the OTUs discussed above were not considered to be members of the SLW community.

In the absence of light and photosynthetic activity, chemolithoautotrophic species may play a crucial role in carbon cycling through the generation of new organic carbon for the subglacial ecosystem. Christner et al. (2014) concluded that rates of CO₂ fixation in samples from the SLW water column were sufficient to support heterotrophic production. Given that the waters of SLW are predominantly derived from glacial melt, the bulk of nutrients needed to fuel primary production and support the ecosystem are likely sourced from subglacial sediments that contain relict marine organic matter deposits which are widespread beneath this portion of the WAIS (Wadham et al., 2012; Christner et al., 2014). Hence, the underlying geology and historical connection to the Ross Sea was expected to strongly influence the structure and metabolic function of microorganisms inhabiting the SLW ecosystem through nutrient input.

Similar to other subglacial environments (Christner et al., 2001; Skidmore et al., 2005; Lanoil et al., 2009; Hamilton et al., 2013; Dieser et al., 2014), the microbial species of SLW were largely composed of Proteobacteria, Actinobacteria, Bacteroidetes, and Firmicutes (Supplementary Figure S2), and also contained members of the Nitrospirae, Chloroflexi, Chlorobi, Thaumarchaeota, and Candidate Division JS1 phyla (Supplementary Figure S2). The most abundant taxa in SLW (Polaromonas, Sideroxydans, and Thiobacillus) were closely related to sequences characterized from sediment cores recovered from beneath the neighboring Kamb Ice Stream (Lanoil et al., 2009), while Methylobacter, Albidiferax, Candidatus Nitrotoga, and Thaumarchaeota species have been shown to be prevalent in subglacial outflows, basal ice, permafrost, and polar waters (Alawi et al., 2007; Cheng and Foght, 2007; Alonso-Sáez et al., 2011; Dieser et al., 2014; Doyle, 2015). The widespread distribution of these taxa in cold environments suggests an inherent tolerance of conditions in the cryosphere, and that they may serve important ecological functions in the subglacial habitat.

The relationship between abundances of sequenced 16S rRNA molecules and genes has been used as a means to assess the metabolic activity of individual taxa (e.g., DeAngelis et al., 2010; Jones and Lennon, 2010; Campbell et al., 2011; Dieser et al., 2014). Although cellular ribosome concentration correlates positively with growth rate, the interpretation of rRNA:rDNA ratios in natural communities is more complex because the relationship is not uniform for all species (Blazewicz et al., 2013) and may require a fundamentally different interpretation for oligotrophic versus copiotrophic lifestyles (Lankiewicz et al., 2015). In light of these uncertainties, high 16S rRNA abundance and rRNA:rDNA ratios were evaluated to specifically identify outliers in the SLW community data. The OTUs identified through this analysis are inferred to have the greatest potential for metabolic activity within the community and thus most likely to be influencing biogeochemical transformations in SLW at the time it was sampled. The positive correlation between the abundance of rRNA and rDNA in the SLW OTUs (Figure 3A) supported that taxa abundant in the rDNA data were ecologically important and unlikely to represent dead cell populations. Rank abundance of the taxa did not correlate with the rRNA:rDNA ratio (Figure 3B), suggesting that abundance per se was not a good predictor for potential metabolic activity. Although assumptions about metabolic function based on short 16S rRNA gene fragments are tenuous, many of the numerically abundant (Figure 2) and rare OTUs (Figure 3C) with large rRNA:rDNA ratios (>5.8; Figure 3C) were phylogenetically related to bacterial and archaeal species that use reduced nitrogen, iron, and sulfur compounds or C1 compounds as primary electron donors (Figure 4).

Geochemical analysis of the dissolved inorganic nitrogen pool in the SLW water column showed that it was comprised primarily of ammonium (∼2.4 μmol L^-1) and nitrate (0.8 μmol L^-1), with Δ¹⁷O-NO₃^- values (-0.1 to 0.2‰) indicating in situ nitrification (Christner et al., 2014). The potential for nitrification was also supported by the abundance of taxa classifying within the Thaumarchaeota (Supplementary Figure S2; Figure 3), a ubiquitous clade containing chemolithotrophic (Könneke et al., 2005) and mixotrophic (Qin et al., 2014) archaea that derive energy from ammonia oxidation. Water column OTUs related to the ammonia oxidizing archaea (AOA) were predominately found as cell populations that passed through 3.0 μm pores and were primarily retained on the 0.8 and 0.2 μm filters (Figure 2), which agreed well with individual cell sizes reported for this group (<1 μm, Könneke et al., 2005). The AOA were 1.6- and 14-times (rRNA and rDNA, respectively) more abundant than known ammonia oxidizing bacterial (AOB) genera in the water column; however, the abundance of AOA quickly diminished with sediment depth, while the AOB increased (55-fold higher in 0–2 cm rRNA, Figures 2 and 4). The most abundant AOB-related OTU in the sediments had 99% sequence identity with Nitrosospira multiformis (Figure 2). AOA have been found to outcompete AOB in environments with low ammonium concentrations (<1 μM), presumably due to the higher affinity of their ammonia monooxygenase for substrate (Martens-Habbena et al., 2009). This fact may explain the distribution of AOA and AOB in SLW, as ammonium was ∼40-times more abundant in sediment pore waters than in the lake (Vick-Majors, 2016). Further, SLW contained abundant and rare OTUs with large rRNA:rDNA ratios that were related to the anaerobic ammonia oxidizing bacteria (Figure 4) Candidatus Jettenia asiatica (OTU000042; Figure 2; Hu et al., 2012) and Brocadia fulgida (OTU000583; Figure 3C; Kartal et al., 2008), respectively. These microorganisms may serve an important role in nitrogen removal within the sediments during conditions of water column hypoxia.

With only one known exception (i.e., van Kessel et al., 2015), the second step of nitrification is catalyzed by a physiologically distinct group of microorganisms that oxidize nitrite to nitrate. Based on the SLW water column data, we conclude that this reaction is probably carried out by a group of Betaproteobacteria closely related to Candidatus Nitrotoga arctica, a nitrite-oxidizing isolate from Siberian permafrost (Alawi et al., 2007; Figure 4). OTUs related to Candidatus Nitrotoga were highly abundant in the water column (∼10% rRNA) and surficial sediment (0–2 cm; 3.8% rRNA; Figure 4). Also prevalent within the surficial sediments (0–6 cm) were OTUs closely related to Thiobacillus denitrificans (99% sequence identity; Figures 2 and 4), a microorganism with the capacity to use inorganic sulfur compounds or Fe²⁺ as an electron donor via aerobic respiration or denitrification (Beller et al., 2006).

Studies of subglacial sediments from alpine glaciers (e.g., Skidmore et al., 2005; Boyd et al., 2014; Harrold et al., 2016) and the WAIS (Lanoil et al., 2009) have provided evidence that sulfide and iron oxidation are microbially mediated weathering processes that occur at the bed. Boyd et al. (2014) inferred microbial pyrite oxidation beneath the Robertson Glacier based on geochemical and functional gene analysis, concluding that it may contribute to subglacial primary production. Within SLW’s pyrite containing sediments (Michaud et al., 2016), species of Sideroxydans, Ferriphaselus, and Albidiferax (Figures 2–4; Kato et al., 2014) could participate in iron cycling. A. ferrireducens is a heterotrophic, facultative anaerobe capable of coupling the reduction of Fe⁺³ to the oxidation of organic compounds such as acetate (Finneran et al., 2003), which was ∼1 μM in the SLW water column (Christner et al., 2014; Figure 4). Albidiferax-related OTUs were found within the surficial sediments, but in the water column, the majority were observed on the 10 and 3.0 μm filter size fractions, consistent with the size reported for the type strain (3–5 μm filaments; Finneran et al., 2003) and/or attachment to water column suspensoids. Although iron reduction is thermodynamically unfavorable in the oxic water column, members of Albidiferax may be associated with anaerobic microhabitats within large suspended sediment particles as has been found in marine snow aggregates (e.g., Stief et al., 2016). It is notable that members of this genera have also been found abundant in other glacier ecosystems, including basal ice facies from the Matanuska Glacier, Alaska and periglacial streams near Thule, Greenland (Doyle, 2015).

Sequences related to the adenosine-5′-phosphosulfate reductase (aprA) gene, a key enzyme involved in sulfur redox chemistry, of Sideroxydans and Thiobacillus species were detected in SLW sediments (Purcell et al., 2014). Purcell et al. (2014) used quantitative PCR and estimated 0.084 (28–34 cm) to 9.2 × 10⁵ (0–4 cm) aprA gene copies per gram of wet SLW sediment, which represented from 0.9% (28–34 cm) to 15% (0–4 cm) of the 16S rRNA gene copies that were also detected. This is consistent with our rDNA analysis which showed that OTUs related to the sulfur oxidizing species Sideroxydans and Thiobacillus accounted for 18% of rDNA sequences in the 0–2 cm sediment horizon (17% rRNA), decreasing to 0.01% at a depth of 34–36 cm (0% rRNA; Figure 2). Members of these genera were also inferred to be highly abundant in the SLW water column (OTUs000038 and 000064; Figure 2) and had large rRNA:rDNA ratios (OTU000380; Figures 3C and 4). The activity of sulfide oxidizing microorganisms within SLW sediments is further supported by the excesses of crustally derived sulfate observed in the upper 15 cm of the sediments (Michaud et al., 2016). Purcell et al. (2014) also measured low rates of sulfate reduction in laboratory experiments with SLW sediments. Although known sulfate reducing bacteria were rare in the SLW sediment community, OTUs related to the genera Desulfatiglans and Desulfatibacillum had high rRNA:rDNA ratios in the 4–6 and 16–18 cm sediment horizons, consistent with their potential to conduct sulfate reduction in the SLW sediments if redox conditions within the lake become favorable (Figure 4). Nonetheless, sulfate profiles in the sediment porewaters suggested sulfate reduction was not an active process in the upper 36 cm of the sediments (Michaud et al., 2016). Consequently, the in situ function of these organisms is unknown.

The basal sediments beneath ice sheets may harbor globally significant reservoirs of organic matter and be methane sources to the atmosphere (Wadham et al., 2008). Wadham et al. (2012) estimated a methane reservoir of ∼10 Pg C beneath the WAIS, where thick organic rich sediments combined with low temperature and high pressure would promote methane hydrate storage. Methanogenesis has been reported in a number of subglacial environments (Boyd et al., 2010; Stibal et al., 2012a,b), including the bed of the Greenland Ice Sheet (Christner et al., 2012; Dieser et al., 2014). In the sediment core analyzed for this study, OTUs related to methanogenic archaea were rare (0.1%) and only detected in the lowest sediment depth analyzed (34–36 cm). This coupled with measurements of methane in sediment pore water suggests that methanogenic populations and activity may have been more prevalent in deeper portions of SLW’s sediments (Michaud, 2016; Figure 4). An OTU (000112) closely related to the type I methanotroph, Methylobacter tundripaludum (99% identity), was ∼10-fold lower in the water column than in the 0–2 cm sediment rRNA where it represented ∼15% of the sequences (Figure 2). Dieser et al. (2014) identified a similar phylotype of Methylobacter that was inferred to be responsible for methane oxidation (320 nM CH₄ d^-1) in subglacial water outflows at the western margin of the Greenland Ice Sheet. The presence of active methanotrophy in SLW was supported not only by the high abundances of methanotroph-related OTUs and their large rRNA:rDNA ratios (Figure 3C), but also by geochemical and isotopic data which showed a significant decrease in methane concentration between the surficial sediments (0–2 cm) and the water column concurrent with a positive shift in the δ¹³C-CH⁴ (Michaud, 2016). Together, these results imply that microbial methane oxidation may serve a role as a substantial methane sink beneath the WAIS. The effect of subglacial methane release to the atmosphere during ice sheet wastage is estimated to be significant (e.g., Wadham et al., 2012), but such efforts have not considered aerobic methane oxidation as an aspect of carbon cycling beneath ice masses. Our data and those of Dieser et al. (2014) imply that bacterial methane consumption could be a significant methane sink and pathway for primary production in portions of the subglacial hydrological system where methane and oxygen coexist.

There is a lack of fundamental information on the microbial biomes at the base of polar ice sheets because accessing these subglacial environments in a clean, environmentally conscious way is technically and logistically challenging. The WISSARD project at SLW provided the first opportunity to examine the limnology and ecology of an Antarctic subglacial lake, revealing a chemosynthesis-based community comprised of bacteria and archaea, with no conclusive evidence for the presence of eukaryotic species (Christner et al., 2014; Achberger, 2016). The members of SLW’s microbial community appear to derive their nitrogen, iron, sulfur, and carbon compounds from mineral weathering and relict organic matter in the sediments (Figure 4; Michaud et al., 2016), but various nutrients may be actively cycled in SLW and the interconnected WIS subglacial hydrologic system. At present it is unclear how representative the SLW microbial ecosystem is of other lakes in this region or to the many hundreds of others beneath the Antarctic ice sheet. However, taxa phylogenetically related to many of the dominant OTUs present in SLW have been previously observed in other icy subsurface environments, raising possibilities for microbial biogeographical studies of specific taxa endemic to the polar regions. Most species inhabiting the SLW ecosystem were distinct from those entrapped in the overlying glacial ice (i.e., the water melted for drilling), implying that microbial inocula for the lake is derived from another source. Hence, microorganisms may enter into SLW as plankton or attached to particles that are transported by subglacial water source from more internal regions of the Antarctica continent.

All authors contributed to the study design, acquisition of samples, and writing of the manuscript. The generation and analysis of molecular data was conducted by AA.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.