Overview of metagenomic data
Biomass from Ace Lake was collected from the upper oxic zone (U1, U2, U3), the interface (I), and the lower anoxic zone (L1, L2, L3) (Additional file 1: Table S1; also see the “Methods” section). The precise depths sampled varied marginally due in part to differences in the water level in the lake at the time of sampling; the sampling depth ranges were as follows: U1, surface; U2, 5 m; U3, 11.5–13 m; I, 12.7–14.5 m; L1, 14–16 m; L2, 18–19 m; L3, 23–24 m. A total of 120 metagenomes were analysed (Additional file 1: Table S1). Abundance calculations are described in the “Methods” section. Briefly, for operational taxonomic units (OTUs), ‘relative abundance’ describes the sum of coverages of the contigs assigned to an OTU in a metagenome relative to the sum of coverages of all contigs in the metagenome; ‘peak relative abundance’ describes the highest relative abundance of an OTU within a set of metagenomes being considered (e.g. all 120 metagenomes, a season, a lake depth); and ‘total abundance’ describes the sum of the coverages (contig read depth × contig length) of the contigs of an OTU. OTUs with relative abundance ≥ 1% in at least one metagenome were considered abundant OTUs. For KEGGs, the abundance of a pathway or an enzyme in a metagenome was calculated using the sum of the read depth of contigs corresponding to predicted genes assigned KO annotations, and normalized to all 120 metagenomes. The relative abundance of viruses represented in OTU analyses were calculated as for other OTUs. The read depth of contigs corresponding to viral clusters or singletons were summed to obtain their total read depth (e.g. for a time period), with those having a total read depth > 4000 being considered abundant.
Taxonomy and taxa abundance were determined from the protein-coding genes (~ 40 million) present on ~ 25 million assembled contigs (20 Gbp); choosing contigs rather than individual proteins reduced incorrect assignments arising from horizontally transferred genes. A total of 17,157 reference-based OTUs were identified of which 117 had relative abundance greater than 1% in at least one metagenome. A total of 3.5 billion (0.5 Tbp) reads from the metagenomes were aligned against the OTU bins to refine them, and Blast was used to align the filtered OTU bins against metagenome-assembled genomes (MAGs) that were generated using MetaBAT [12] (hereafter referred to as MetaBAT MAGs), resulting in 51 high-quality OTU bins (Additional file 1: Table S2). The relative abundances of the 51 OTUs in each of the 120 metagenomes were compiled (Additional file 1: Dataset S1). A large number of low abundance contigs (< 1% relative to the sum of coverages of all contigs in a metagenome) could not be confidently taxonomically assigned, including those with best matches to uncultured bacteria, archaea, eukarya and viruses (see Additional file 2).
Overview of seasonal and depth variation of taxa and function
The effects of a number of lake (depth and salinity) and external (daylength, sunlight hours, air temperature) parameters were used to assess the effects of environmental variables on the microbial community. The overall microbial population of the lake exhibited seasonal patterning (Fig. 2). A seasonal progression from summer through spring to winter was evident from a distance-based redundancy analysis (dbRDA) (Fig. 2, dbRDA2), in addition to separation by lake depth (Fig. 2, dbRDA1). The specific external factors, daylength (P = 0.001), sunlight hours (P = 0.002), and air temperature (P = 0.001) were significant explanatory variables for seasonality, with lake salinity (which increases with depth, P = 0.001) contributing to differences by lake depth (P = 0.001) (Fig. 2).
Alpha diversity was highest in the lower zone and lowest at the interface during summer (Fig. 3). The largest seasonal changes in alpha diversity occurred at the interface (Fig. 3) as a result of the Chlorobium population dominating in summer but reducing significantly in winter and spring (Figs. 4 and 5; Additional file 1: Fig. S2; Additional file 1: Dataset S1; also see below in the “The importance of sunlight”). Decreases in alpha diversity in the 3-0.8 μm fraction in U2 (spring) and U3 (summer) (Fig. 3) arose from increases in the Synechococcus population (Fig. 4). Synechococcus and Chlorobium were also the biggest contributors to similarity within summer samples, and dissimilarity between samples from different seasons (Additional file 1: Tables S3 and S4). The peak relative abundance (highest abundance in any metagenome) for Chlorobi (84%) and Cyanobacteria (63%), and their seasonally high contributions to peak relative abundance in summer and spring (Fig. 4; Additional file 1: Fig. S2 and Dataset S1), further illustrate the importance of these two phototrophic taxa to the ecosystem (see in the “The importance of sunlight” section).
In addition to these photoautotrophic bacteria, individual anaerobic Deltaproteobacteria showed variable abundance according to season, with certain OTUs showing decreased abundance in the winter correlated with the decline of Chlorobium, suggesting their growth is tightly coupled to Chlorobium (see the “Sulfur cycling: a tangled web” section). The Verrucomicrobia, represented by several abundant OTUs, were also more abundant in winter than summer, which may suggest that polymeric substrates preferred by this group were more available in winter (see the “Ace Lake nutrient cycles: carbon cycling” section).
The metabolic features inferred for each of the most abundant bacterial and archaeal OTUs were compiled (Additional file 1: Table S5; also see descriptions of nutrient cycles below). In general, OTUs from the upper zone were aerobic, whereas obligate anaerobes were restricted to the anoxic zone and the interface.
The importance of sunlight
The penetration of photosynthetically active radiation (PAR) into the lake varies greatly due to seasonal effects on the level of incident radiation; the opaqueness, depth, and age of the surface ice; the depth of snow on surface ice; and the microbial density, which is particularly high at the interface [5, 13,14,15]. In summer, daily maximum incident radiation has been recorded as high a 1225 μE m−2 S−1 [14], with light intensity during ice-free conditions penetrating to the interface at levels as high as 100 μE m−2 S−1 [15]. However, such high light levels would not occur often at the interface [15]. Seasonal variation in light levels is very high, with daily maximum incident radiation of 1.3 μE m−2 S−1 recorded in winter [14]. This translates to PAR within the lake at a depth of 5 m being over four orders of magnitude lower in winter (0.03 μE m−2 S−1; 1.2 m of surface ice) than in summer (245 μE m−2 S−1; ice-free) [14].
The Micromonas (picoeukaryote) OTU was abundant in summer and spring but absent in winter, consistent with a reliance on sunlight for proliferation (Fig. 4; Additional file 1: Dataset S1). Synechococcus relative abundance was generally highest in spring, higher in the water column of the upper zone (U2, 51%) compared to summer where abundance was highest, deeper in the upper zone (U3, 58%) (Fig. 5; Additional file 1: Dataset S1). The depth distribution may relate to grazing pressure and the response of Ace Lake Synechococcus to light intensity [5, 16]. Synechococcus is a major primary producer in Ace Lake, performing oxygenic photosynthesis via the Calvin-Benson-Bassham (CBB) cycle. It also has the potential for photoheterotrophic and photomixotrophic growth, as reported for other Synechococcus strains that utilize glucose or glycerol [17,18,19]. Consistent with other Synechococcus, Ace Lake Synechococcus encodes transporters for the uptake and catabolism of sugars and glycerol [20], and exogenous sugars or glycerol may also be utilized for the synthesis of compatible solutes such as glucosylglycerol, sucrose, or trehalose [12,13,14,15,16,17,18,19,20,21,22]. Genes for the uptake and utilization of organic nitrogen sources, and for the reductive assimilation of nitrate to ammonia were present in Ace Lake Synechococcus (Additional file 1: Table S5), although nitrogenase was not identified, which is consistent with the inability of Synechococcus strains cultivated from the lake to fix nitrogen [16].
Synechococcus also exhibited high, peak relative abundance at the interface (25%), in the lower zone (8%), and during winter (16%) (Fig. 5; Additional file 1: Dataset S1). In mid-winter, relative abundance had dropped to 6% (July 2014) but the population increased by August of the same year (16%) (Additional file 1: Dataset S1). As the upper waters would be devoid of any light in winter (Additional file 1: Fig. S3 and Table S9), heterotrophic growth in the dark would account for such an increase. In spring (Oct. 2014), a Synechococcus bloom appears to have occurred (51%) (Additional file 1: Dataset S1), similar to previous observations from September 1992 [16]. The October 2014 spring bloom appears to have extended into the anoxic interface where the Chlorobium abundance was at its lowest (see below). It is possible that fermentative abilities allow Synechococcus to survive in dark and anoxic conditions, as inferred for Synechococcus in the Black Sea [23]. Ace Lake Synechococcus encodes enzymes potentially involved in fermentation, including synthesis and mobilization of glycogen [23, 24], the latter coupled to hydrogen production using a Hox hydrogenase [25] or lactate production using D-lactate dehydrogenase [23].
Chlorobium at the interface appears to be markedly affected by the changes in light conditions. While peak relative abundance was as high as 83% in summer, the population decreased considerably in winter (to ~ 6%), falling further through spring (1%; Oct. 2014) before blooming to high abundance 2 months later in summer (59%; Dec. 2014) (Additional file 1: Dataset S1). The biomass associated with Chlorobium at the interface is considerable with cell counts exceeding 108 cells mL−1 [10]. A video taken down through the water column into the interface illustrates the remarkable layer that forms and the intensity of the GSB within it (Additional file 3 including Movie S1). The Chlorobium biomass sinks as particulate matter in the lake [5, 10], explaining its abundance in the lower zone in summer (L1, 17%; L2, 14%; L3, 6%) (Fig. 5; Additional file 1: Dataset S1).
Chlorobium spp. have a genetic capacity for anoxygenic photoautotrophy via the reverse tricarboxylic acid (rTCA) cycle [26] and are able to efficiently harvest light in low-light surroundings using chlorosomes [27]. Chlorobium spp. may also contribute to organic cycling via photoassimilation of simple organic compounds (such as acetate and dicarboxylates) (Additional file 1: Table S5), which can boost photoautotrophic growth [28]. The closest match to the Ace Lake Chlorobium is Chlorobium phaeovibrioides which has 99% 16S rRNA gene identity, but only 85% average nucleotide identity (ANI). Unlike C. phaeovibrioides, Ace Lake Chlorobium is green rather than brown in colour (Additional file 3: Movie S1). As well as possessing the biosynthetic pathway for the carotenoid chlorobactene, found in green-coloured GSB, Ace Lake Chlorobium lacks the capacities to synthesize bacteriochlorophyll e or the carotenoid isorenieratene, both found in brown-coloured GSB [29, 30]. Because these enzymes are encoded in a bacteriochlorophyll e gene cluster that may be subject to horizontal gene transfer [31] and given the considerable depth of sequence coverage for the Ace Lake Chlorobium IMG MAGs, we infer these genes to be absent.
In addition to the dominant phototrophic bacteria in Ace Lake that utilize light to fuel carbon fixation (Synechococcus and Chlorobium), a diverse range of OTUs also harvest light to augment a heterotrophic metabolism. The potential for photoheterotrophic growth is attested by the prevalence of light-driven, proton-pumping rhodopsins (proteorhodopsins, actinorhodopsins, xanthorhodopsins) that derive from phylogenetically diverse bacteria in Ace Lake, including members of Bacteroidetes, Verrucomicrobia, Actinobacteria, Alphaproteobacteria, Gammaproteobacteria, and Betaproteobacteria (Fig. 6; Additional file 1: Table S5). Based on sequence analysis [32,33,34], the majority of rhodopsins are inferred to be green-absorbing, consistent with individual OTUs being found either predominantly or exclusively in the layers above the interface, with several OTUs detected only in surface waters. However, Ace Lake Hydrogenophaga had a blue-absorbing proteorhodopsin, and Methylophilaceae BACL14 had both green- and blue-absorbing rhodopsins, encoded by adjacent genes. Although blue-absorbing rhodopsins may indicate expression deeper in the water column [32], it can also indicate expression in ice since solar radiation is highly scattered in ice, and blue light predominates [35]. Certain Ace Lake OTUs (Yoonia, Hydrogenophaga) have genes for Type II reaction centres and bacteriochlorophyll synthesis, as well as for proton-pumping rhodopsins, but lack ribulose bisphosphate carboxylase/oxygenase (RubisCO), suggesting the capacity for bacteriochlorophyll-dependent photoheterotrophy. In these Ace Lake OTUs, it is possible that light is used for auxiliary energy production, such as under oligotrophic growth conditions [36].
The influence of viruses
Viral sequences in metagenome data can derive from viruses integrated into cellular genomes (e.g. provirus), viruses associated with cells (e.g. infecting cells or remaining associated with cell surfaces during filtration), or from virions free in the water column. The size fractionation provided the capacity to infer the presence of virions if the viral signatures were in different size fractions to their hosts. This was also gauged from sequence coverage of contigs for viruses (e.g. lysogens having equivalent coverage with their host), correlation analyses, and by assembling closed viral genomes. Viral abundance varied mainly with depth, being most abundant in the upper, photic zone where peak relative abundance reached 46% (Fig. 7) (Additional file 1: Dataset S1). Consistent with this, analysis of complete, circular viral genomes identified 337 contigs representing 173 distinct viruses, most of which were phage (Caudovirales) that were most prevalent in the upper zone (U, 87; I, 21; L, 53) (Additional file 4: Table S1). In the upper zone, viral relative abundance in winter was similar to summer, and in fact was somewhat lower in spring indicating that a high viral presence was maintained effectively throughout the seasons (Fig. 7).
The 51 abundant Ace Lake OTUs included five algal viruses (Phycodnaviridae 1–5; Figs. 4 and 5), which represented 261 virus clusters and 109 singletons, many of which were taxonomically classified as nucleocytoplasmic large DNA viruses (NCLDVs). The viral clusters associated with Phycodnaviridae 3 differed to Phycodnaviridae 1, 2, 4, and 5 which shared clusters (Additional file 4: Table S2), and the relative abundance of Phycodnaviridae 1, 2, 4, and 5 positively correlated, but not with Phycodnaviridae 3 (Additional file 4: Table S3). While these associations were recognizable, co-occurrence was not observed between the most abundant green alga (Micromonas) and any of the Phycodnaviridae 1–5 (Additional file 4: Table S3), or any of the NCLDV clusters tested (7, 9, 20, 32, 35, and 66) which were chosen based on their abundance and representation in multiple metagenomes. It is important to define the seasonal contributions the algae and their viruses make to the ecosystem, and will require a dedicated effort, possibly involving single-cell and single-virus genome sequencing.
Using a cyanophage sequence assembled from the 2006 metagenome data, viral contigs from a number of cyanophage clusters and singletons were identified, all of which were exclusive to Ace Lake data as they did not cluster with any other IMG/VR sequences (Additional file 4: Table S4). Neither the Ace Lake Synechococcus OTU nor the Synechococcus IMG MAGs contain CRISPR-Cas genes, consistent with their reported absence in marine cyanobacteria [37], so viruses could not be mapped to the Synechococcus host through CRISPR spacer matches. In all the metagenome data, the cyanophage contigs were only assembled from the 0.1 μm fraction, and the Synechococcus contigs only from the 0.8 and 3 μm fractions (Fig. 5), consistent with size partitioning of the phage particles and the cells, and indicating the cyanophage sequences recovered were from virions released after active replication in the host. Despite this, no correlation was observed between the abundance of the cyanophage and Synechococcus, indicating the cyanophage were not primarily responsible for variations in Synechococcus abundance. PAR reaching Synechococcus is influenced not just by daylength (Additional file 1: Fig. S3), but also by cloud cover impacting sunlight reaching the lake surface (i.e. sunlight hours; Additional file 1: Fig. S4); snow cover on the lake, which in part is impacted by wind strength (Additional file 1: Fig. S5); and the presence and thickness of ice (particularly in summer when ice melts). We therefore speculate that the changeable nature of PAR availability influences Synechococcus growth to the extent that viral growth is sporadic and appears largely uncoupled from patterns of cellular growth.
Chlorobium spp. defend against viral predation using restriction enzymes and the CRISPR-Cas system [31, 38], and a subtype-2 Cas gene locus has been identified in the Ace Lake Chlorobium [10, 11]. Chlorobium spacers matched to two groups of Ace Lake contigs: one cluster (CL1024) and one singleton (SG14554). To confirm the specificity of these matches, these virus contigs were compared to the IMG/VR spacer database [39], which identified Chlorobi and Gammaproteobacteria as possible hosts (Additional file 4: Tables S5, S6 and S7). Contigs from the Ace Lake metagenomes that had high identity (≥ 98% identity to the two CL1024 contigs; ≥ 92% identity to SG14554) and a high alignment fraction (≥ 98%) were gathered to expand sequence representation. Additionally, two virus clusters (CL248 and CL400) were identified as being abundant at the interface (Additional file 4: Table S8). CRISPR spacers to CL248 were present in Chlorobium IMG MAGs assembled from lower zone metagenomes (as well as specific contigs for Gammaproteobacteria, including Klebsiella pneumoniae and Marinobacter antarcticus).
The abundance of the expanded CL1024 cluster (Pearson correlation coefficient r = 0.7, P = 2e− 11), expanded SG14554 singleton (r = 0.97, P = 0.02), and CL400 cluster (r = 0.9, P = 5e− 11) had significant positive correlations with Chlorobium abundance; the CL248 cluster also had a positive, but not significant, correlation (r = 0.5, P = 0.7) (Fig. 8). These positive correlations may be explained by the viruses growing cooperatively with the Chlorobium, but not being a cause of its immediate decline, which would be characterized by a negative correlation. Moreover, in October 2014 when Chlorobium abundance was at its lowest, CL1024, SG14554, and CL248 were not detected, indicating the Chlorobium decline is not associated with large-scale virion production by these viruses, which are thus likely not responsible for the Chlorobium decline. The marked reduction in the Chlorobium population through winter and spring is therefore likely caused by other factors, such as cellular decay due to insufficient light for maintenance energy production, and/or predation by eukaryotes.
CRISPR spacers are derived from invading DNA, and up to 60 spacers were identified in spacer arrays from Chlorobium IMG MAGs. Many spacers were present more than once across the 10-year sampling period. We speculate that the capacity to defend against diverse viruses and grow cooperatively with specific viruses is likely to contribute to the ability of the species to regenerate so well in summer, with the ability to reach such high cell density being the main strategy for Chlorobium to prevail through winter and spring (Fig. 4). The Ace Lake Chlorobium and the viruses it supports (e.g. CL1024, SG14554, CL400, and CL248; Additional file 4: Table S9) have only been identified in Ace Lake. Viral contigs from the three clusters, plus the singleton, are distinct from Chlorobi viral genomes recovered from Trout Bog Lake in Wisconsin (PRJNA611599, SAMN14323023 [40]) and seem to represent distinct genera based on vContact2 classification into distinct genus-level clusters (Additional file 4: Dataset S1). Adaptation to environmental conditions including polar light cycle, low temperature (Additional file 1: Fig. S3, S6 and S7) and specific nutrient conditions is likely to be a strong driver making the Ace Lake Chlorobium virus system endemic to the Vestfold Hills in Antarctica or possibly polar environments.
While viruses of phototrophs were the most abundant, it was notable that one complete, circular genome was for a ~ 528 kb ‘huge’ phage [41] belonging to cluster 24. Contigs constituting complete, circular genomes for the phage were present in the 0.1 μm fraction from November 2008 (L1, 0.9 % relative abundance) and October 2014 (L1, 2 %; L2, 1%) (Additional file 4: Table S1 and S10). Large CL24 contigs were also present at 1 or 2% relative abundance in 0.1 μm-fraction metagenomes from L1 or L2 depths (November 2008 and 2013, and December 2013), and from the 3 μm fraction in L3, November 2008 at 0.2% relative abundance (Additional file 4: Table S10). The size and depth partitioning of contigs for the virus likely reflect virion (0.1 μm) and intracellular (3 μm) forms, and the presence of the phage in multiple years attests to its persistence within the lower zone population. The phage has a putative type I-C CRISPR-Cas system, and CRISPR arrays with spacer sequences were also present in CL24 virus contigs, even though the spacer acquisition genes (cas1, cas2, and cas4) were absent. These are characteristics of some, recently reported huge phage [41]. The spacers did not match any virus contigs in the IMG/VR viral database, but were identical to spacers from some Ace Lake bacterial contigs (a Gammaproteobacteria and a Firmicutes). Since the CRISPR repeat sequences of the phage and these bacteria (presumed hosts) were identical, it is possible that this huge phage uses host cas genes to acquire spacers that target other viruses infecting its hosts [41].
Ace Lake nutrient cycles: carbon cycling
The abundance of enzymes or pathways involved in energy conservation and metabolism (KEGG analysis; Fig. 9), the metabolic traits of abundant OTUs (Additional file 1: Table S5), the glycoconjugate degradation enzymes of abundant OTUs (Additional file 1: Table S6), the hydrogenases of abundant OTUs (Additional file 1: Table S7), and the relative abundance of the major OTUs (Figs. 4 and 5; Additional file 1: Dataset S1) were integrated to generate seasonal nutrient cycles (Fig. 10). Three carbon fixation processes predominate in Ace Lake: CBB cycle, the rTCA cycle, and the Wood-Ljungdahl (WL) pathway (reductive acetyl-CoA pathway). The CBB cycle is linked to Synechococcus, which uses it for oxygenic photosynthesis in the oxic upper layers (U2 and U3). Chlorobium, which dominates the interface, employs a light-driven rTCA cycle with sulfide as the electron donor; thus, this carbon fixation pathway was dominant at the interface, where it showed a marked seasonal variation based on Chlorobium abundance. Chlorobium sinking into the lower zone contributed to the persistence of genes associated with the rTCA cycle. However, we infer that other bacteria that encode the rTCA cycle (e.g. Cloacimonetes JGIOTU-2) may have an active cycle that is fuelled by hydrogen oxidation (see the “Hydrogen cycling: pivotal to multiple nutrient cycles” section). The WL pathway was found exclusively in the lower zone, with its abundance linked to sulfate-reducing Deltaproteobacteria and methanogenic archaea. The latter includes hydrogenotrophic and acetoclastic methanogens, although in acetoclastic methanogens, the WL pathway is not used for carbon fixation [43].
Organic nutrients can be supplied by the primary producing algae and bacteria via exudation or cell lysis. Other nutrients are generated by Ace Lake zooplankton within the water column (i.e. the copepod P. antarctica – also see Additional file 3: Movie S1) and a more diverse range of organisms (algae, zooplankton, diatoms, bacteria) colonizing benthic mats, or are of exogenous origin, generated by microbial mats, birds, and seals [5]. While exogenous input would be at its highest in summer when the surface ice has melted and snow-melt drains into the lake, any seasonal changes in the availability of organic material have little impact on catabolic pathways and enzymes (as judged by KEGG analysis).
Recalcitrant polymeric algal material and particulate matter can be remineralized by bacteria throughout the Ace Lake water column. The genomic potential to catalyse the depolymerization of polysaccharides (e.g. chitin, starch, laminarin, cellulose, pectin, fucoidan, carrageenan) was associated with Bacteroidetes, Verrucomicrobia, Planctomycetes, Gammaproteobacteria, and Cloacimonetes OTUs. Bacteroidetes, Verrucomicrobia, and Planctomycetes OTUs also encode capacities for the degradation of fucoidan (fucosidases, sulfatases, fucose catabolic enzymes), which has the potential to be used as both a carbon and sulfur source. Of these groups, it is notable that the most abundant Verrucomicrobia OTUs fell to below 1% relative abundance in summer, suggesting that their higher abundance in winter (Additional file 1: Dataset S1) may be linked to recalcitrant organic matter that is more abundant in winter, especially polymeric substrates that derive from decaying algae.
Algae (including brown algae, green algae, diatoms) and zooplankton (including copepods) release organic solutes such as glycolate, simple sugars, amino acids, peptides, methanol, urea, organosulfonates, and phosphonates [44,45,46,47,48,49,50] (Additional file 1: Supplementary text). Although present in Synechococcus to detoxify endogenous glycolate (a photorespiratory by-product), the abundance of glycolate oxidase genes in Ace Lake is largely attributable to the breakdown of glycolate by heterotrophic bacteria. Amino acid and peptide primary transporters were found across the examined Ace Lake OTUs, although some differential abundances were apparent, with general amino acid ABC transporters predominating in the upper, oxic zone, and ABC transporters for branched-chain amino acids (BCAA) and peptides more prevalent in the lower, anoxic zone. The latter accords with peptide and amino acid oxidation (especially BCAA) by phylogenetically diverse anaerobes, with these substrates potentially used as carbon, nitrogen, and energy sources [51,52,53]. Methanol has the potential to be utilized as a carbon and energy source by Ace Lake methylotrophs. Urea transporters and urea catabolism genes were detected throughout the water column, although they were more prevalent in the upper zone due to their presence in Synechococcus and many aerobic Proteobacteria and Verrucomicrobia. Taurine uptake and catabolism genes were also more prevalent in the oxic layers, and taurine was predicted to be used as a nitrogen source by many OTUs, with some also using it as a sulfur source. Another organosulfonate, dimethylsulfoniopropionate (DMSP), is a potential source of carbon and sulfur and is a preferred source of reduced sulfur for roseobacters (Additional file 1: Supplementary text) [54] and Pelagibacter [55]; accordingly, DMSP catabolism genes were limited to the upper, oxic zone of the lake. Based on catabolic pathways encoded in Ace Lake OTUs, glycine betaine can be utilized as a source of glycine by both aerobic and anaerobic bacteria, although for certain sulfate-reducing Deltaproteobacteria, the methyl group of glycine betaine provides a source of reductant for anaerobic respiration [56]. The capacity to degrade phosphonates is exhibited by several Ace Lake OTUs that encode enzymes to degrade methylphosphonate and 2-aminoethylphosphonate as sources of phosphorus and, in the case of the latter, nitrogen [47, 57]. See Additional file 1: Supplementary text for more description about carbon cycling.
Sulfur cycling: a tangled web
Chlorobium and the anaerobic, sulfate-reducing bacteria (SRB) constitute a sulfur cycle, whereby Chlorobium oxidizes sulfide to sulfate and the SRB reduce sulfate to sulfide [10, 11]. The Deltaproteobacteria Desulfobacterium, Desulfobacterales S5133MH16, and Desulfatiglanales NaphS2 were the most abundant SRB in Ace Lake, and encode the ability to respire by dissimilatory sulfate reduction (DSR) to sulfide. Certain other anaerobic Deltaproteobacteria present in the lake lack the ability for DSR, including the abundant OTUs Desulfocapsa and Syntrophales UBA2210. Desulfocapsa can disproportionate sulfur (and thiosulfate) to sulfide and sulfate, thereby providing potential substrates for both Chlorobium (sulfide) and SRB (sulfate) [58,59,60]. Syntrophales UBA2210 is inferred to reduce sulfur to sulfide as part of a fermentative metabolism. Chlorobium is a potential major source of sulfur: extracellular sulfur globules (generated and mobilized via water-soluble polysulfides) are an intermediate in sulfide oxidation [61]. Additionally, Desulfocapsa is likely to be dependent on the ability of Chlorobium to scavenge sulfide, since sulfur disproportionation can only proceed in environments where sulfide levels are kept low [60, 62].
Desulfobacterium and Desulfocapsa exhibited seasonal variation between summer and winter, which may be linked to the marked decline of Chlorobium (see the “The importance of sunlight” section). Both these Deltaproteobacteria OTUs fell to below 1% relative abundance in winter (Additional file 1: Dataset S1). Further, Chlorobium relative abundance was positively correlated with the relative abundance of Desulfobacterium (r = 0.94, P = 0.002, 3 μm size fraction) and Desulfocapsa (r = 0.81, P = 0.026, 3 μm size fraction; r = 0.8, P = 0.031, 0.8 μm size fraction). By contrast, the relative abundances of Desulfatiglanales NaphS2, Desulfobacterales S5133MH16, and Syntrophales UBA2210 were each negatively correlated with Chlorobium relative abundance, although the relationship was not significant.
The Ace Lake Chlorobium lacks the Sox system for thiosulfate utilization to sulfate [11]. The considerable sequence coverage for the Chlorobium OTU, and knowledge that the sox gene cluster is encoded on a mobile genetic element in some Chlorobiaceae [63] indicates sox genes are absent in the Ace Lake Chlorobium population. However, Sox systems are present in diverse Alpha- and Betaproteobacteria consistent with the presence of KEGG hits in the upper, oxic zone. These proteobacteria also potentially provide a source of sulfate to the lake that augments the major contribution made by Chlorobium.
DSR is dominant in the anoxic, lower zone due to SRB belonging to Deltaproteobacteria. The detection of DSR-related genes in the upper, oxic zone is attributed mainly to the presence of adenylylsulfate reductase genes (aprAB) in Pelagibacter; however, rather than being employed for DSR, it has been proposed that this enzyme is employed solely for the detoxification of sulfite (such as generated by taurine catabolism) to adenosine 5′-phosphosulfate [8].
Typical of GSB, Chlorobium in Ace Lake lacks any genomic evidence of assimilatory sulfate reduction (ASR) and therefore requires exogenous sulfide for biosynthetic purposes [64]. Accordingly, based on the abundance of ASR-related genes within the water column, this process was lowest at the interface at the times when Chlorobium was most abundant. ASR-related genes were prevalent throughout the water column, with numerous bacteria capable of using inorganic sulfate as a source of biosynthetic sulfur, including Synechococcus, Verrucomicrobia, various Proteobacteria (but not Deltaproteobacteria), and certain Bacteroidetes.
Nitrogen cycle: all about ammonia
Nitrogen appears to be the limiting nutrient in the upper zone due to its rapid depletion at the onset of primary productivity in spring [5]. Bacteria in the lower zone contribute to replenishment of ammonium and other reduced nitrogen in the upper zone through degradation of amino acids [5, 10]. The Ace Lake metagenome data is dominated by genes for nitrogen fixation and ammonia assimilation, with no evidence for nitrification and a low capacity for denitrification. Chlorobium and Desulfocapsa encode nitrogenases; however, Chlorobium abundance is highly dependent on season, and diazotrophy would only be necessary when ammonia levels are low [10, 11]. Several OTUs exhibit the genomic potential for incomplete denitrification: respiratory nitrate reductase (NarG) (Nisaea), periplasmic nitrate reductase (NapA) (Verrucomicrobia BACL24; Pseudomonas E; Desulfocapsa) for nitrate dissimilation, nitrite reductase (nitric oxide forming) (NirK) (Balneolaceae UBA2664, Nisaea, Pelagibacter, Bacteroidales UBA4459), and nitrous oxide reductase (NosZ) (Balneolaceae UBA2664, Bacteroidales UBA4459). There is also the potential for dissimilatory nitrate reduction to ammonia via cytochrome c nitrite reductase (NrfA) (Desulfocapsa). The Ace Lake metagenome data is dominated by genes for nitrogen fixation and ammonia assimilation, with no evidence for nitrification and a low capacity for denitrification.
Hydrogen cycling: pivotal to multiple nutrient cycles
Multiple types of hydrogenases are represented among Ace Lake bacteria and archaea, including respiratory uptake hydrogenases, and hydrogenases linked to fermentation and/or redox balance (Additional file 1: Table S7). Hydrogenases were more abundant in the lower zone than the upper zone (Fig. 9), where they are associated with the use of hydrogen as an energy source and/or production of hydrogen as part of anaerobic respiration or fermentation. Hydrogen-evolving hydrogenases can be used by aerobes that adopt a fermentative metabolism in response to oxygen limitation, and are employed for survival rather than growth (e.g. Gimesia) [65]. Synechococcus has the genomic potential for fermentative growth under prolonged anoxic conditions, including hydrogen evolution via a Hox hydrogenase (see the “The importance of sunlight” section). Thus, Synechococcus is a potential source of hydrogen in Ace Lake. Ace Lake Chlorobium, which lacks any evidence of hydrogenases, is potentially also a major source of hydrogen [10], derived from a membrane-bound nitrogenase that generates hydrogen as a by-product of nitrogen fixation [66]. Thus, hydrogen production by Chlorobium would be seasonal, and occur only when ammonia is limiting [10].
Anaerobes such as sulfate-reducing Deltaproteobacteria (Desulfobacterium and Desulfatiglanales NaphS2) and hydrogenotrophic methanogens (Methanomicrobiaceae 1) have hydrogen-uptake hydrogenases that exploit hydrogen as an energy source for carbon fixation via the WL pathway, coupled to the reduction of sulfate to sulfide or the reduction of CO2 to methane, respectively. Ace Lake Desulfocapsa encodes both a nitrogenase and a membrane-bound hydrogenase (Additional file 1: Table S5); this suggests that hydrogen released as a by-product of nitrogen fixation can be re-oxidized to protons, thus recovering the electrons for energy conservation. Hydrogen oxidation is also inferred for other OTUs using heterodisulfide-linked hydrogenases that could be employed for biosynthetic purposes (Bacteroidales UBA4459, Cloacimonetes JGIOTU-2) [67]. Certain anaerobic bacteria (Atribacteria 34-128, Cloacimonetes JGIOTU-2) were inferred to have respiratory systems that include a membrane-bound hydrogenase, which utilizes protons both as a substrate for hydrogen production and for ion translocation, with reductant provided by oxidation of organic substrates [68]. Cytosolic hydrogenases serve as hydrogen-evolving hydrogenases to re-oxidize redox carriers during fermentative growth by obligate anaerobes (Bacteroidales UBA4459, Desulfocapsa, Desulfatiglanales NaphS2, Atribacteria 34-128, Izimaplasma) [69]. These data for hydrogen cycling in Ace Lake are consistent with the proposal that microbial hydrogen metabolism is highly important to ecosystem function in both oxic and anoxic environments [70].