Subglacial lakes can form where water collects at hydrostatic stable points beneath an ice sheet, when geothermal heat, pressure-induced freezing point depression, and/or high salinity prevent complete freezing (Livingstone et al., 2022). These lakes are relatively isolated ecosystems that can host life despite low temperatures, low nutrient abundance, and the absence of sunlight as an energy source. Terrestrial subglacial lakes are considered accessible if imperfect, analogues to ice-covered environments in the past and present Solar system: the ice-covered oceans of “Snowball” Earth (Kirschvink et al., 2000), lakes beneath the south polar cap of Mars (Lauro et al., 2020) and oceans in some of the icy satellites of Jupiter and Saturn (Kivelson et al., 2000, 2002; Thomas et al., 2016). Studies of these systems complement research on other lakes with comparatively thin (metres) ice covers, e.g., the Dry Valleys lakes (Chinn, 1993) and epishelf lakes of Antarctica (Davies et al., 2017), where sufficient photosynthetic active radiation can reach the water column and support phototrophic communities (Howard-Williams et al., 1998).

There are more than 700 reported subglacial lakes in Antarctica (Livingstone et al., 2022), three in Iceland under the Vatnajökull glacier (Bjornsson, 2003), two in Greenland (Palmer et al., 2013), and one recently discovered in the Canadian Arctic (Rutishauser et al., 2018). The biological exploration of some of these lakes has begun, namely Lake Vostok (Karl et al., 1999; Priscu et al., 1999; Christner et al., 2001; Bulat, 2016; Gura and Rogers, 2020), Subglacial Lake Whillans (Tulaczyk et al., 2014), Subglacial Lake Mercer in Antarctica (Priscu et al., 2021), and all three Icelandic lakes (Gaidos et al., 2004, 2009; Marteinsson et al., 2013). In the cases of Antarctica and Iceland, the data unambiguously point to the presence of active assemblages of bacterial taxa which are distinct from the distribution in the overlying ice or surrounding glaciated terrain (Gaidos et al., 2004, 2009; Marteinsson et al., 2013; Achberger et al., 2016).

The three main Icelandic lakes, called Grímsvötn, Western, and Eastern Skaftárkatlar lakes, are distinguished by their location on active volcanoes and are maintained by geothermal melting at the base of the 250–300-m thick Vatnajökull ice cap. The chemistry of the approximately 100-m-deep water columns of these lakes is substantially influenced by volcanic gases and hydrothermal fluids, maintaining an anoxic and highly sulphidic environment (Agustsdottir and Brantley, 1994; Johannesson et al., 2007). Mixing of anoxic lake water with oxygenated glacial meltwater creates chemical disequilibrium that can serve as an energy source for chemolithotrophic microorganisms (Gaidos et al., 2004, 2009; Marteinsson et al., 2013). Previous investigations have used molecular tag-based methods to identify the dominant microbial taxa in the water column of the Skaftárkatlar lakes and link chemical reactions of the inorganic substrates thought present in the lakes with potential metabolisms of these taxa, e.g., acetogenesis, sulphide oxidation, sulphate reduction, iron reduction, and hydrogen oxidation by members of Acetobacterium, Geobacter, Sulfuricurvum, Sulfurospirillum, and Desulfosporosinus (Gaidos et al., 2009; Marteinsson et al., 2013). In Subglacial Lake Whillans, the chemoautotrophic microbial taxa, distinct from that taxa of the Icelandic subglacial lakes, use reduced nitrogen, iron or sulphur compounds as energy sources (Christner et al., 2014).

Here, we performed a metagenomic analysis on four water column samples collected from the Eastern Skaftárkatlar lake in 2007, as well as on enrichment cultures from the same samples (Marteinsson et al., 2013). Our study aimed to: (1) confirm the previously reported microbial community structure, (2) investigate the metabolic strategies of the microorganisms in the lake, and (3) analyse potential pathways within microbes grown in enrichment cultures under different conditions. Compared to the previous study done on the same water samples but only with a 16S rRNA sequencing approach, we adopted a metagenomic approach as a more informative, less biassed method to gain further understanding of the microbial diversity and community structure but also and most importantly of the functional capabilities both at the taxon and community level occurring in such a specific environment. Our analyses show unambiguously that the microbial communities of this Icelandic subglacial lake are dominated by seven taxonomic genera: Sulfuricurvum, Sulfurospirillum, Acetobacterium, Pelobacter/Geobacter, Saccharibacteria, Caldisericum, and an unclassified Prolixibacteraceae. Chemolithoautotrophy is the main metabolism of these communities and is adapted to their environment with oxidation/reduction of sulphur, nitrogen metabolism, carbon fixation via the reverse Krebs cycle, and acetogenesis.

This study presents the first analysis of metagenomes originating from volcanic subglacial lake: one sample collected from the bottom of the lake (A₃), a second from four pooled water column samples collected from different depths in the lake (B_mix), and a third from combined enrichments, i.e., growth under conditions that attempted to mimic those in the subglacial lake (E_mix). The metagenomic analysis of the collected samples enabled a reconstruction of the potential metabolic pathways existing in the lake and a more robust description of the microbial community structure than before. Environmental metagenome analysis such as these have been shown to be more quantitative for microbial community studies, compared to amplicon-tag-based studies, due to the absence of Polymerase Chain Reaction (PCR) amplification bias (Parada et al., 2016; Brumfield et al., 2020).

This study not only confirms a previous study of the lake using 16S rRNA tag sequencing and Fluorescence In Situ Hybridisation (FISH) (Marteinsson et al., 2013) but also provides more robust evidence that only a few bacterial taxa dominate the microbial community in the lake water column (Sulfuricurvum, Acetobacterium, Geobacter/Pelobacter, and Sulfurospirillum). Metagenomic analyses in this study also confirm the presence of lower abundance taxa (e.g., Caldisericum and Desulfosporosinus) and reveal new additional lineages (Saccharibacteria and Pelosinus) which had not been observed by the previous amplicon-based studies. The relatively low microbial diversity (∼20 taxa) represented in the metagenome of this ecosystem resulted in high sampling depth and allowed assemblage of full-length 16S, 23S, and often also 5S sequences for most of the taxa. Full-length sequences permitted a more precise taxonomic assignment of the most abundant community members as well as 14 minor taxa with relative abundance ranging from 0.05 to 1%. Many of these taxa are closely related to psychrotolerant strains, including a member of Brevundimonas, a taxon detected in the Arctic (Trivedi et al., 2018) and Antarctic (González-Toril et al., 2009) and which is known to be resistant to cold temperatures (Dartnell et al., 2010) and the genus Pedobacter, also found in Arctic environments (Zhou et al., 2012; Peter and Sommaruga, 2016; Trivedi et al., 2018).

In a broader context, this study has confirmed previous results (Marteinsson et al., 2013) that the community in the East Skaftárkatlar lake is substantially different from that of other subglacial lakes such as Subglacial Lake Whillans in Antarctica. The former is dominated by Betaproteobacteria such as Polaromonas, Sideroxydans or Thiobacillus, Bacteroidetes, and Actinobacteria (Achberger et al., 2016) while the East Skaftárkatlar lake is dominated by Epsilonbacteria: Sulfuricurvum and Sulfurospirillum, and Firmicutes: Acetobacterium. East Skaftárkatlar could be significantly influenced by hydrothermal activity emanating from the underlying lake bed and thus can host, among others, sulphur oxidisers or reducers (Johannesson et al., 2007). Furthermore, the lake is a mix of glacial melt (containing oxygen) as well as sulphide from disproportionation of dissolved volcanic SO₂ that supports a sulphur cycle. Then, Sulfurospirillum deleyianum is known to be able to oxidise sulphide with nitrate, producing ammonium and intracellular elemental sulphur (Eisenmann et al., 1995). This ability is also known in Sulfuricurvum species, the most abundant member of the microbial community in the sample collected at the bottom of the lake 75% (A₃) compared to 30% in the water column samples (B_mix). The relative abundance of Sulfuricurvum correlates with the sulphate concentration, which was about five times higher in A₃ than in the other pooled sample B_mix (4.71 ppm in A₃ vs. from 0.29 to 1.43 ppm in B₁ to B₄, Supplementary Table 1). Compared to A₃, a higher diversity was observed in B_mix (Shannon’s α-diversity of 1.18 and 1.83, respectively, data not shown) including three additional taxa: Peptococcaceae, Bacteroidetes, and Caldisericum spp. which might indicate that growth and metabolism are faster than the mixing time of the water column across any vertical or lateral chemical gradients in the lake. The higher diversity could also be a result of the fact that B_mix is derived from a mixture of samples.

Despite many different enrichment conditions, none of the dominant taxa found in the water column samples were enriched under the conditions of our incubations. The absence of culturable members of the main diversity is not unexpected as uncultivated phyla frequently dominate diverse environments (Lloyd et al., 2018). Interestingly, the dominant taxa found in E_mix, Microbacterium, Hydrogenophaga, and Sphingopyxis were only marginally detectable in the metagenome (<10 raw rDNA reads) of the two environmental samples A₃ and B_mix (Figure 1) and not detected in the rRNA tag sequences of Marteinsson et al. (2013).

Remarkably, no evidence of members of Archaea in the lake was detected in the previous 16S rRNA amplicon-tag sequencing study (Marteinsson et al., 2013). This includes all known chemolithotrophic methanogens, which could potentially compete with homoacetogens for H₂. In this metagenome sequencing-based study, which is not affected by PCR bias, we identified a significant number of archaea-like sequences in A₃ (data not shown). Nevertheless, a detailed analysis of those reads shows that they are not randomly distributed over a complete reference archaeal 16S rDNA gene (Supplementary Figure 1) and are instead localised to two overlapping regions. The absence of flanking sequence to these clusters and their exclusive location indicates that those reads do not originate from a complete gene. Therefore, we are not able to detect the presence of archaea in the lake with our metagenomic approach. The DNA extraction method used might also be a bias with an incomplete lysis of recalcitrant archaeal cells as it was noted in previous subsurface communities (Webster et al., 2003).

Since light is absent in the lake, chemoautotrophy must be responsible for the microbial growth as primary producers such as in subglacial lakes Vostok or Whillans (Rogers et al., 2013; Vick-Majors et al., 2016). The metabolic pathways detected in a combination of A₃ and B_mix for carbon fixation are shown in Figure 2. Acetobacterium sp. contains most of the genes involved in the Wood–Ljungdahl homoacetogenesis pathway. This pathway uses H₂ as an electron donor and CO₂ both as an electron acceptor and carbon source to generate acetyl-CoA. Acetogenesis seems to be an important energy source in this ecosystem and supports the previous findings that the lakes contain taxon with homoacetogens as the closest relatives (Gaidos et al., 2004, 2009; Marteinsson et al., 2013). The closest cultured relative to Acetobacterium is Acetobacterium woodii, a strain using the Wood–Ljungdahl (reverse acetyl-CoA) pathway and H₂ to fix CO₂ into acetate and maintain a sodium ion gradient for ATP synthesis (Poehlein et al., 2012). Prokaryotes living close to the thermodynamic limit, like methanogens and acetogens, use the reductive acetyl-coA pathway for both CO₂ fixation and energy conservation (Ragsdale and Pierce, 2008; Thauer et al., 2008). We propose that environmental factors such as low temperature and high H₂ concentration in this ecosystem favour acetogens over methanogens (Nozhevnikova et al., 1994). The absence of methanogens in this unique ecosystem could also potentially be explained by the presence of bacteria such as Sulphate Reducing Bacteria (SRB) that out-compete methanogens for hydrogen or produced acetate in the lake, with lower K_s values for H₂ and acetate (Kristjansson et al., 1982; Schönheit et al., 1982).

Complete pathways for assimilatory sulphur reduction into sulphide were detected and mainly assigned to Sulfurospirillum sp. whereas the dissimilatory reduction of sulphur in sulphide was assigned mainly to Acetobacterium sp. (Figure 3A). This difference might be the result of Sulfurospirillum only performing assimilatory sulphate reduction (i.e., to produce organosulphur compounds) and Acetobacterium the dissimilatory one to conserve energy. Sulfuricurvum was described as a sulphur oxidiser (Han et al., 2012) and not as a sulphate reducer which can explain why assimilatory genes were mainly assigned to Sulfurospirillum. Moreover, the reason why Sulfurospirillum has phsA gene and not standard dissimilatory genes might be explained by the low sulphate concentration in the lake making sulphate reduction difficult. Sulphide production might be energetically unfavourable against the background of high sulphide in the lake (around 1 mM). Thus sulphur-cycling species might take sulphur only to sulphite and then cycle it back to sulphate or sulphur under microaerobic conditions. Sulfuricurvum sp. and Sulfurospirillum sp. also play a role in activating sulphur with a polysulphide reductase. These results emphasise the central role of sulphur species as sources of electrons (S^2–, SO₄^2–, and S₂O₃^2–) for chemolithoautotrophy in the lake and, as expected, in the metabolism of amino acids. Respiratory sulphate reduction is a common process in environments with a high sulphate concentration whereas in sulphate-depleted anoxic environments, acetogenesis is favoured over sulphate reduction (Laanbroek et al., 1982; Muyzer and Stams, 2008; Stams and Plugge, 2009).

No complete nitrogen metabolic pathways were found in the co-assembly metagenome except for nitrogen fixation. The genes involved in the nitrogen fixation pathway were detected and belong mainly to Sulfuricurvum sp. The genus Sulfuricurvum sp. has not been reported as being diazotrophic but its sequenced representative Sulfuricurvum kujiense YK-1^T also possesses the necessary genes (Kodama and Watanabe, 2004; Han et al., 2012). The dissimilatory nitrate reduction to ammonium, which is an important reaction of the reductive branch of the nitrogen cycle (Simon and Klotz, 2013), is mainly taxonomically affiliated with Geobacter/Pedobacter genus. Whereas nitrification is an important chemoautotrophic pathway of new organic carbon production in Subglacial Lake Whillans (Christner et al., 2014), none of the genes involved in this pathway were found in our metagenomes. The presence of nir (NADH-dependent) but absence of nrf (periplasmic) genes might imply that assimilatory reduction but not respiratory reduction (for energy conservation) might be occurring in this oligotrophic environment.

The study of microbial communities inhabiting subglacial lakes is of importance as such ecosystems remain underexplored. Only four subglacial lakes were sampled for microbial analyses at the time this paper was written (Gaidos et al., 2004, 2009; Christner et al., 2014; Priscu et al., 2021). The results in our research support the former hypothesis that was based on a taxonomy study, but it also gives insight into the metabolic pathways. The ecosystem in this subglacial lake appears to have a chemolithoautotrophic foundation with sulphur and carbon cycling that is fuelled by H₂, CO₂, and sulphur species originating from the geothermal activity and by O₂ from melted ice. By identifying the potential metabolic pathways to specific taxa in the microbial assemblage, we have gained insight into the ecology of microbiomes in the water column of such an extreme environment. However, we should assume that other ecological niches may exist, e.g., on the bottom of the lake, in the lake sediments, or at the ice/lake interface, and these remain undiscovered. Moreover, nearly half of the sequence reads in our metagenomic library were not assigned and are a source for future advances. Also, new targeted sampling with larger volume is needed e.g., samples collected close to potential geothermal vents at the bottom of the lake might reveal thermophiles belonging to both the Bacteria and Archaea by sequencing and cultivation strategies. Even if analysed samples were collected in 2007, the results of this study might still correlate with the current situation of the lake as the latter has experienced similar conditions and has undergone continuous filling and draining (called jökulhlaup) since records have been kept.

Finally, due to recent Arctic warming, ice caps are thinning, notably in Iceland (Gudmundsson et al., 2008). Despite their isolation from the surface, these lakes are influenced by the dynamics and melting of glaciers, which are changing with climate change (Livingstone et al., 2022). The existence of subglacial lakes is thus threatened in the long run. The complete exploration of the microbial diversity of this ecosystem needs to happen quickly to monitor such future changes.

The data presented in this study are deposited in NCBI Sequences Read Archive repository under the accession number: SRP011365 (https://www.ncbi.nlm.nih.gov/sra/?term=SRP011365).

PV, GF, and VM conceived and designed the study and analysed the data. AK performed the DNA extractions. GF performed the bioinformatic analyses. PV, GF, EG, and VM wrote the main manuscript text. All authors reviewed the manuscript.