Introduction

Deep Lake is a unique, cold, marine-derived hypersaline lake in Antarctica. It is the most hypersaline system in the Vestfold Hills, a permanently ice-free region on the eastern shore of Prydz Bay, East Antarctica, where over 300 lakes and ponds exist (Gibson, 1999; Wilkins et al., 2013). The temperature close to the surface of Deep Lake ranges from −16 to 12 °C, with the temperature of the few meters at the lake surface exceeding 0 °C only for a few months of the year during the austral summer (Campbell, 1978; Wright and Burton, 1981; Ferris and Burton, 1988; Franzmann et al., 1988). Nevertheless, because of high salinity (270 g l−1), the water column remains ice-free even in winter when the lake temperature falls to −20 °C (Campbell, 1978; Wright and Burton, 1981; Ferris and Burton, 1988).

Typical of many hypersaline environments, the halotolerant unicellular green alga Dunaliella (Chlorophyta) is implicated as the major primary producer (Wright and Burton, 1981). However, productivity measured in the lake is extremely low (<10 g C m−2 per year) (Campbell, 1978; Gibson, 1999). In addition to hypersalinity and seasonal extremes in light availability, low productivity has been attributed to a deficiency in available nutrients and the depth of the lake (36 m) preserving very low annual temperatures (for example, compared with shallower hypersaline Antarctic lakes) (Campbell, 1978; Wright and Burton, 1981; Ferris and Burton, 1988).

Based on the analysis of Deep Lake metagenome data, it was recently shown that while the lake community is dominated by heterotrophic halophilic archaea (haloarchaea) of the family Halobacteriaceae, the community composition differs greatly to temperate and tropical hypersaline environments (DeMaere et al., 2013). The community was found to have low complexity and be hierarchically structured. Genome sequences were obtained for four isolates: Halorubrum lacusprofundi strain ACAM34 (Franzmann et al., 1988), strain tADL (Candidatus Halohasta litchfieldiae; Mou et al., 2012), strain DL31 and strain DL1 (Halobacterium sp.), which represent phylogenetically distinct (85% 16S rRNA gene similarity) lineages. Based on small subunit rRNA pyrotag sequence data, the distribution of these four haloarchaea was found to be reasonably uniform throughout the water column, and to account for a total 72% of all pyrotags. However, the relative proportion of the four haloarchaea varied considerably, with tADL being the most abundant (comprising 44% of pyrotags), DL31 18%, H. lacusprofundi 10% and DL1 only 0.3%.

A remarkable feature of the Deep Lake system was an observed high level of inter-genera gene exchange between tADL, DL31, H. lacusprofundi and DL1, including the sharing of long regions (up to 35 kb) of virtually identical (100%) DNA (DeMaere et al., 2013). Given the nature and extent of gene exchange, and a very low predicted rate of cell division of 6 rounds per year (and hence vertical inheritance), a notable feature of the lake system was the maintenance of sympatric speciation (that is, distinct Halobacteriaceae lineages). Genetic differences that describe ecotypes with a capacity to perform niche adaptation would promote sympatric speciation, and consistent with this, preliminary genomic traits suggestive of physiological distinction were noted between the four Deep Lake haloarchaea (DeMaere et al., 2013).

In order to advance our understanding of the potential for niche differentiation, in the present study, we performed an in-depth comparative genomic analysis of the four Deep Lake haloarchaea and assessed functional properties through substrate growth studies. A focus was placed on elucidating characteristics of tADL that promote its dominance. By inferring apparent physiological differences, and linking this to physicochemical characteristics of the lake, we were able to better define ecotype distinctions and overall lake ecology.

Materials and methods

Strains and genomic analyses

Deep Lake water samples from 5, 13, 24 and 36 m depths were sequentially size fractionated through a 20-μm prefilter onto 3.0, 0.8 and 0.1 μm pore size, 293 mm polyethersulfone membrane filters (DeMaere et al., 2013). Isolation of strains and DNA sequencing of genomes and metagenomes have been described (DeMaere et al., 2013). Pyrosequencing of PCR-amplified V6–V8 region of small subunit rRNA genes was used to generate microbial community profiles, with operational taxonomic units (OTUs) defined based on 97% sequence identity, as described previously (DeMaere et al., 2013). COG (Clusters of Orthologous Groups) scrambler and ALL scrambler were used to assess the statistical representation of COGs categories and individual COGs between the four Deep Lake haloarchaeal genomes, as described previously (Allen et al., 2009). For each comparison, the test genome was compared against the other three Deep Lake genomes, with subsample size of 10 000, confidence level 0.99 and 10 000 bootstrap replicates. Genomes were interrogated for the presence or absence of genes that encode proteins and metabolic pathways that may provide ecophysiological distinctions between the four Deep Lake haloarchaea. To assess the veracity of gene functional assignments, all annotations of genes involved in pathways described here were manually examined as described previously (Allen et al., 2009).

Growth studies

The four Deep Lake haloarchaea, tADL, DL31, H. lacusprofundi and DL1 were tested for growth using potential carbon and nitrogen sources in batch cultures based on DBCM2 media (Burns and Dyall-Smith, 2006). DBCM2 contains low concentrations of peptone (0.025% w/v) and yeast extract (0.005% w/v), and growth of cultures did not occur unless these components were present in the media. Because essentially no change in turbidity occurred when strains were grown in unsupplemented DBCM2 media (data not shown), it is likely that the peptone and/or yeast extract provided essential nutrients, such as cofactors. Growth was assessed as maximum turbidity (optical density at 600 nm) achieved for cultures grown aerobically in liquid broth (20 ml volume, 150 r.p.m., 30 °C), as described previously (Franzmann et al., 1988). Growth was evaluated against a blank where each strain was grown in DBCM2 without the addition of the defined substrates. The following defined carbon and/or nitrogen sources were tested (all at 10 mM): glycerol, urea, sucrose, glucose, fructose, mannose, acetate, citrate, succinate or putrescine. These were tested alone or in combination with pyruvate (10 mM), except for succinate and citrate where pyruvate was not added. For growth assessments of carbon sources, ammonia was used as the defined nitrogen source. For growth with urea, urea was tested as a carbon source with ammonia, and as a nitrogen source with pyruvate. In addition, peptone was tested at 0.1%, either alone or with pyruvate. Cultures grown on urea were tested for urease activity using a rapid urease test (Marshall et al., 1987).

Analyses of lake ecology

Ammonia, nitrate, nitrite, total nitrogen, dissolved reactive phosphorus, total phosphorus, total organic carbon, total dissolved carbon, total sulfur and total dissolved sulfur for Deep Lake were determined by American Public Health Associations Standard Methods at Analytical Services, Tasmania, as described previously (Yau et al., 2013). Statistical analyses (Pearson Product Moment Correlation, ANOVA (analysis of variance) regression analysis for statistical significance) were performed to identify correlations between microbial community composition (OTUs by depth and filter size; Supplementary Table S1) and lake chemistry (Supplementary Table S2). The relative abundances of individual OTUs were compared with one another and also with physicochemical parameters measured at the same depths where the biomass was sampled.

Results

Genome analysis overview

A summary of basic genome characteristics including rRNA operons, origins of replication, prediction of proteins with transmembrane domains and secretion signals, is provided in Supplementary Information. Statistical evaluation of COGs between the four genomes was performed (Table 1; Supplementary Table S3), and the COG categories (Figure 1) and individual COGs (Table 1) with statistically significant over-representation or under-representation provided clues about the functional traits characteristic of each of the haloarchaea. This information guided an in-depth appraisal of gene composition, including the compilation of genes associated with ATP-binding cassette (ABC) transport systems (Supplementary Table S4) and other functions (Supplementary Table S5). Based on these analyses, the important functional characteristics of the Deep Lake haloarchaea were inferred (described below). A graphical representation of functional traits that are inferred to be most important to the four Deep Lake haloarchaea are portrayed in Figure 2.

Table 1 Selected cellular functions and individual COGs over- or under-represented in Deep Lake haloarchaeal genomes
Figure 1
figure 1

COGs category abundance in the genomes of Antarctic haloarchaea. (a) tADL, (b) DL31, (c) H. lacusprofundi and (d) DL1. Categories were determined to be significantly over-represented (black circle), significantly under-represented (white circle) or not statistically significant (gray circle) when compared with the other three Deep Lake genomes. White bars indicate 0.99 confidence interval. COGs category: [A] RNA processing and modification; [B] chromatin structure and dynamics; [C] energy production and conversion; [D] cell cycle control and mitosis; [E] amino acid transport and metabolism; [F] nucleotide metabolism and transport; [G] carbohydrate transport and metabolism; [H] coenzyme transport and metabolism; [I] lipid transport and metabolism; [J] translation; [K] transcription; [L] replication; recombination and repair; [M] cell wall/membrane/envelope biogenesis; [N] cell motility; [O] post-translational modification, protein turnover and chaperones; [P] inorganic ion transport and metabolism; [Q] secondary metabolites biosynthesis, transport and catabolism; [R] general function prediction only; [S] function unknown; [T] signal transduction mechanisms; [U] intracellular trafficking, secretion and vesicular transport; [V] defense mechanisms.

Figure 2
figure 2

An integrated view of the ecophysiologies of the Antarctic haloarchaea. (a) DL31, (b) tADL, (c) H. lacusprofundi and (d) DL1. Cellular traits inferred to be important for survival in Deep Lake are shown, with depictions highlighting energy production, nutrient uptake, motility and osmotic balance. ABC, ATP-binding cassette transporter; Amt, ammonia transporter; APC, acid-polyamine-organocation transporter; BR, bacteriorhodopsin; C–P lyase, phosphonate degradation complex; Cap, poly-gamma-glutamate capsule; GC, glyoxylate cycle; HR, halorhodopsin; MFS, Major Facilitator Superfamily transporter; Mrp, multi-component Na+/H+ antiporter; Nar, respiratory nitrate reductase; Nas, assimilatory nitrate reductase; Nha, Na+/H+ antiporter; Nir, assimilatory nitrite reductase; OADHC, 2-oxoacid dehydrogenase complex; PHA, polyhydroxyalkanoate; PTS, phosphotransferase system; SR, sensory rhodopsin; TRAP, tripartite ATP-independent periplasmic transporter; Trk, low-affinity K+ transporter.

Primary transport

ABC transporters drive the translocation of substrates across membranes by the hydrolysis of ATP. A complete ABC transporter for the uptake of substrates comprises ATP-binding and permease and substrate-binding lipoprotein components. Because haloarchaea typically do not use organic solutes to maintain osmotic balance (Oren, 1999), imported organic solutes such as simple sugars, glycerol and amino acids are presumably consumed as substrates. All genomes have genes for ABC transporters that target carbohydrates, including those with the highest sequence identities to ABC transporters for glycerol-3-phosphate, oligosaccharides or nucleosides. ABC transporters for glycerol-3-phosphate (Ugp) or oligosaccharides such as maltose share a close relationship with each other, and also with multiple sugar ABC transporters (Msm), and therefore cannot be distinguished based on sequence alone. Nevertheless, broad specificity (including functional exchangeability) has been documented for Ugp and maltose ABC transporters, at least in bacteria (Hekstra and Tomassen, 1993), from which haloarchaeal ABC transporters are thought to have been derived (Nelson-Sathi et al., 2012). Each genome has a single set of genes for a nucleoside ABC transporter for importing exogenous nucleotide sources, as an alternative to nucleotide biosynthesis.

tADL has the highest number of putative carbohydrate ABC transporters, but has no identifiable ABC transporters for oligopeptides/dipeptides. In contrast, DL31 has only one identifiable carbohydrate ABC transporter, but at least 15 oligopeptides/dipeptide ABC transporters, which is more than the other Deep Lake haloarchaea. Consistent with the abundance of oligopeptides/dipeptide ABC transporters, the genome of DL31 encodes the highest number of aminopeptidases that target di- and oligopeptides, including members of peptidase families S9 and M28 (13 and 5, respectively), and the highest number of secreted proteases of the peptidase family S8/S35 (5), which includes halolysin. DL31 and DL1 each have a prolyl iminopeptidase gene (family S33) that is not found in the other two Deep Lake haloarchaea. Many of these peptidase genes in DL31 are located close to the genes for oligopeptides/dipeptide ABC transporters. ABC transporters for amino acids are present in the genomes, particularly those with the highest identities for branched-chain amino acid (BCAA) ABC transporters. For DL1, BCAA ABC transporter genes were the largest class of ABC transporters represented in the genome.

Phosphate ABC transporters are present across all four genomes, often occurring in multiple copies. In addition, tADL contains genes for a phosphonate ABC transporter. In bacteria these high-affinity phosphate or phosphonate uptake systems are induced under conditions of phosphate limitation (Gebhard et al., 2006). The same is likely to be true for Deep Lake haloarchaea, given the presence of a regulatory protein gene adjacent to the ABC transporter genes. All four genomes have homologs of a corrinoid ABC transporter gene identified in Halobacterium salinarum NRC-1 (Woodson et al., 2005), although it is not clear whether these transporters in Deep Lake haloarchaea are specific for corrinoids or iron (III), as these types share a high level of sequence identity.

Secondary transport

Haloarchaea use a bioenergetically efficient ‘salt-in’ strategy of importing K+ and Cl ions (see Rhodopsins below), and expelling Na+ ions. The elevated intracellular K+ concentration is maintained by active import of K+ (Oren, 2002). All four genomes have genes that encode low-affinity K+ transporters (TrkAH) driven by membrane potential, but no genes are present for ATP-driven K+ import systems (KdpABC). For Na+ efflux, there are genes for multiple orthologs in DL1, H. lacusprofundi and DL31 for the NhaC Na+/H+ antiporter. All four have genes for the multi-component Mnh (=Mrp) Na+/H+ antiporter (Swartz et al., 2005).

Other transporters use electrochemical gradients to import organic solutes. Genes for amino acid transporters of the amino acid-polyamine-organocation (APC) superfamily are present in all four genomes, and most encode UspA domains and/or are adjacent to a UspA domain protein, as has been observed in other haloarchaeal genomes (Anderson et al., 2011). The UspA domains are inferred to be involved in transporter regulation, although the precise mechanism is unclear (Schweikhard et al., 2010). All four genomes possess genes that encode phosphate permeases (sodium/phosphate symporters). These belong to the same family (PiT) as a sulfate permease from Gram-positive bacteria, and it is therefore possible that some of these PiT transporter genes are involved in sulfate rather than phosphate uptake in haloarchaea. Sulfate uptake implies a role in assimilatory sulfate reduction, a pathway that is currently not elucidated in haloarchaea (see Sulfur and phosphorus metabolism, below).

Unlike ABC transporters, tripartite ATP-independent periplasmic (TRAP) transporters utilize a proton motive force rather than ATP hydrolysis to drive the translocation of substrates (usually carboxylic acids) into the cytoplasm (Forward et al., 1997). H. lacusprofundi has the highest number of TRAP transporter genes, and DL31 has none. H. lacusprofundi appears to lack symporters for the coupled uptake of sodium ions and di/tricarboxylic acids, although these are encoded in the genomes of the other three Deep Lake haloarchaea.

Carbohydrate metabolism

Glycerol may enter haloarchaeal cells by passive diffusion (Bolhuis et al., 2006), although a new family of glycerol transporters has been proposed to be present in haloarchaea (Anderson et al., 2011). Genes for the putative glycerol transporters are present in tADL and H. lacusprofundi, adjacent to genes for glycerol catabolism (Supplementary Table S5). Once in the cell, glycerol can be phosphorylated to glycerol-3-phosphate and then directed to glycolysis. A gene for glycerol kinase is present in H. lacusprofundi and DL31, and tADL has three orthologous glycerol kinase genes. All four genomes have genes for glycerol-3-phosphate dehydrogenase, which allows glycerol-3-phosphate to be oxidized to dihydroxyacetone phosphate (DHAP), and thereby enter glycolysis or gluconeogenesis. Genes for glycerol dehydrogenase and dihydroxyacetone kinase are present in tADL and H. lacusprofundi along with genes for a phosphoenolpyruvate (PEP)-dependent phosphotransferase system (PEP–PTS) that involves the cytosolic phosphorylation of dihydroxyacetone to DHAP by dihydroxyacetone kinase (Bolhuis et al., 2006). DHAP is also a precursor to glycerol-1-phosphate, for construction of the glycerophosphate backbone of membrane phospholipids, and a gene for glycerol-1-phosphate dehydrogenase was identified in all four genomes. There is no evidence in any of the four genomes of a dephosphorylating glycerol-3-phosphate phosphatase for the direct liberation of phosphate.

The sugar PEP–PTS catalyzes the concomitant uptake and phosphorylation of sugars, in the first step in the breakdown of the imported sugar to pyruvate. tADL has two bacterial-type PEP–PTS gene clusters for sugar uptake, each of which encodes separate PEP-protein kinase (E1), phosphocarrier Hpr and enzyme II components (EIIA, EIIB and EIIC). In Haloferax volcanii, this bacterial-type PTS is used for the uptake and phosphorylation of fructose (Pickl et al., 2012). In tADL, each PTS gene cluster is adjacent to genes encoding enzymes involved in the modified Embden–Meyerhof pathway of glycolysis.

All four genomes have genes for the Emden–Meyerhof pathway for the breakdown of glucose to pyruvate, which is initiated by a broad-specificity hexokinase. tADL, DL31 and H. lacusprofundi also have genes for a modified (semiphosphorylated) Entner–Doudoroff pathway by which glucose is first oxidized to gluconate, and the gluconate is further converted to glyceraldehyde-3-phosphate and pyruvate via the intermediates 2-dehydro-3-deoxygluconate and 2-dehydro-3-deoxygluconate 6-phosphate (Danson et al., 2007; Falb et al., 2008). Genes responsible for the remaining steps in the glycolytic pathway (catabolism of glyceraldehyde-3-phosphate to pyruvate) are present in the four genomes. Under the growth conditions tested, only tADL and H. lacusprofundi showed abilities to grow on sugars as the sole defined carbon source (Supplementary Table S6), although both DL31 and DL1 have genes that encode putative carbohydrate uptake (ABC transporters) and metabolism genes (Tables S4 and S5). The presence of carbohydrate uptake and metabolism genes in haloarchaea that are reported not to grow on sugars has been observed previously (Ng et al., 2000), and may reflect potential saccharolytic abilities. Both tADL and H. lacusprofundi also have α-amylase genes, although all four Deep Lake genomes lack genes for enzymes (for example, xylanase and chitinase) that are required for the degradation of plant or fungal biomass. All four genomes also have genes for a complete gluconeogenesis pathway, allowing amino acids to be used as sole carbon sources (Ng et al., 2000; Danson et al., 2007).

There are no genes associated with glycogen biosynthesis. Polyhydroxyalkanoate is accumulated by certain haloarchaea as a carbon and energy source, and can be sourced from unrelated precursors, including carbohydrates (Lu et al., 2008). tADL is the only one of the four Deep Lake haloarchaea to have genes for the polyhydroxyalkanoate biosynthetic pathway.

Citric acid cycle (oxidative)

Consistent with the heterotrophic metabolism of all four Deep Lake haloarchaea, all genes in the oxidative citric acid cycle were present in each genome. Although all four genomes have genes for at least two orthologs of acetyl-CoA synthetase, genes for enzymes of the glyoxylate cycle, which allows acetate to be used as the sole carbon source, were identified only in DL1 (isocitrate lyase and archaeal-type malate synthase). H. lacusprofundi has a gene for isocitrate lyase, but not for malate synthase, as previously observed (Khomyakova et al., 2011). None of the four genomes has genes associated with the methylaspartate cycle, which is used for acetate assimilation in certain haloarchaea (Khomyakova et al., 2011).

Amino acids and 2-oxoacids

All four Deep Lake haloarchaea have the genomic capacity to biosynthesize the 20 amino acids necessary for protein biosynthesis. However, the presence of alternative pathways for amino acid biosynthesis varied between organisms. For example, whereas all four genomes have the potential to generate proline directly from ornithine, typical of haloarchaea, only tADL has a pathway predicted to derive proline from glutamate. The likely biosynthetic pathway for cysteine is clear in the DL1 and tADL genomes, deriving cysteine via activation of serine with acetyl-CoA, with acetylserine subsequently sulfhydrylated with sulfide. But in DL31 and H. lacusprofundi, the only available pathway is via a phosphoserine intermediate (generated in an earlier step of serine biosynthesis) that is sulfhydrylated using a cysteine synthase homolog, of which there are multiple copies in each genome. In all cases, the source of sulfide is unclear (see Sulfur and phosphorus metabolism below).

The capacity for degrading amino acids is encoded by all four genomes, but the ability to catabolize individual amino acids varies among the four Deep Lake haloarchaea. For example, DL1 is the only one of the four that has genes for the degradation of histidine (via urocanate) and tryptophan (to indole, or anthranilate via kynurenine). All Deep Lake genomes possess 2-oxoacid dehydrogenase complex (OADHC) gene clusters; DL1 has four, compared with one each in the other three genomes. Deamination of BCAAs generates branched-chain 2-oxoacids that can be further catabolized to citric acid cycle intermediates. However, none of the Deep Lake haloarchaea possess an identifiable arginine deiminase gene, indicating that none of them are capable of anaerobic growth by arginine fermentation (Dundas and Halvorson, 1966).

Nitrogen metabolism

Ammonia transporter genes are present in tADL, DL1 and H. lacusprofundi, but not in DL31. For ammonia assimilation, all genomes possess genes for glutamine synthetase, glutamate synthase and glutamate dehydrogenase. tADL also has two genes for the PII protein GlnK that are adjacent to each other and the ammonia transporter gene. In Haloferax mediterranei, both PII proteins have been shown to increase glutamine synthetase activity in the presence of 2-oxoglutarate (Pedro-Roig et al., 2013).

There was no genomic evidence that any of the four Deep Lake haloarchaea are capable of fixing dinitrogen. tADL and H. lacusprofundi each have a single gene for a nitrate/nitrite transporter of the Major Facilitator Superfamily family, whereas DL1 has four. Genes that encode proteins involved in assimilatory nitrate reduction were located only in H. lacusprofundi and tADL genomes, including ferredoxin-dependent assimilatory nitrate and nitrite reductases (Nas and Nir, respectively).

There is evidence for a denitrification capacity among the four Deep Lake haloarchaeal genomes. H. lacusprofundi has a gene cluster encoding a putative respiratory nitrate reductase (narGHI) and an enzyme chaperone (narJ), suggesting that H. lacusprofundi might have the capacity to reduce nitrate to nitrite. Multicopper oxidase permease genes are present in the genomes of H. lacusprofundi, DL1 and tADL, and might represent nitrite reductase (nirK). Putative nitric oxide reductase (norB) genes are present in H. lacusprofundi and tADL, and putative nitrous oxide reductase (nosZ) genes are present in H. lacusprofundi, DL31 and DL1.

The tADL genome is the only one possessing a gene cluster for a urea ABC transporter. Near the urea ABC transporter gene cluster is a urease gene cluster that encodes a cytoplasmic urease complex that hydrolyzes urea to ammonia and carbon dioxide. The DL1 genome also has a urease gene cluster, but no identifiable urea transporter genes. Genes for urea amidolyase were absent from all four Deep Lake genomes.

Sulfur and phosphorus metabolism

The biosynthesis of reduced sulfur (hydrogen sulfide) has not been experimentally characterized in haloarchaea. A proposed pathway involves the direct reduction of sulfate to sulfite by sulfite oxidoreductase, followed by reduction of sulfite to sulfide by sulfite reductase (Feng et al., 2012). Genes encoding both of these enzymes were identified in the genomes of tADL, H. lacusprofundi and DL1, and a gene for sulfite oxidoreductase, but not sulfite reductase, was identified in DL31. All four genomes have genes for rhodanese (thiosulfate sulfurtransferase) for the conversion of thiosulfate to sulfite. It is possible that one of the rhodanese proteins encoded in the DL31 genome actually functions as a sulfite reductase. Alternatively, DL31 may be dependent on exogenous reduced sulfur (for example, from oligopeptides) for biosynthesis. Dimethylsulfoniopropionate transport or catabolism genes were not detected in any of the genomes, and Dunaliella from nearby Organic Lake has been reported not to produce dimethylsulfoniopropionate when grown in culture (Franzmann et al., 1987). Therefore, dimethylsulfoniopropionate does not appear to be a potential source of reduced sulfur for the Deep Lake haloarchea.

Only tADL possesses the genomic capacity for phosphonate degradation. The essential genes for the ATP-dependent degradation of phosphonate to cyclic ribose–phosphate (Kamat et al., 2011) are located within a gene cluster that also includes phosphonate ABC transporter genes (Supplementary Table S4).

Rhodopsins

Archaeal rhodopsins are light-responsive transmembrane proteins that are used as light-driven ion pumps or as signal receptors for phototaxis. Bacteriorhodopsin (BR) is a proton pump that uses light energy to create an electrochemical gradient for ATP production; halorhodopsin uses light energy to transport Cl ions into the cytoplasm, assisting the cell in maintaining osmotic balance; sensory rhodopsins (SRs) mediate attraction to orange light (SRI) or avoidance of UV/blue light (SRII) (Spudich, 2006; Sharma et al., 2007). Based on sequence identities (and conserved functional residues), only tADL possesses BR, whereas only tADL and H. lacusprofundi possess halorhodopsin and SRII genes. For the latter SR, genes for the cognate transducers (HtrII) were also present.

Motility: flagella and gas vesicles

tADL, DL1 and H. lacusprofundi all contain gene clusters for flagellum biosynthesis (=archaellum; Jarrell and Albers, 2012). In these cases, the flagellar gene cluster is adjacent to one or more genes for chemotaxis proteins. However, only tADL possesses genes for gas vesicle biosynthesis. Gas vesicles are hollow, gas-filled structures made of protein that confer buoyancy to planktonic cells, allowing more efficient vertical migrations than swimming by flagella, particularly for slow-growing organisms (Walsby, 1994), and a function in combatting osmotic stress, or orientating cells positioned at the water surface, has also been proposed (reviewed in Pfeifer, 2012). Multiple GvpA orthologs, which function in the construction of the gas vesicle wall, are encoded, but GvpC, which strengthens the gas vesicle wall, or the regulatory proteins GvpD and GvpE (Walsby, 1994; Oren et al., 2005) were not identified in the tADL genome, as is the case for many other haloarchaea (Pfeifer, 2012). Having multiple copies of gvpA and lacking gvpC and gvpDE is characteristic of certain methanogenic archaea that produce gas vesicles (Pfeifer, 2012). Halobacterium salinarum ΔgvpC deletion mutants are still capable of producing functional gas vesicles, albeit these are more fragile and irregularly shaped compared with wild-type cells (Offner et al., 1996), indicating that tADL may produce an unusual form of gas vesicles.

Cell envelope

Haloarchaeal surface layer glycoproteins can undergo both N- and O-glycosylation (reviewed in Eichler, 2003). The presence of a gene for the oligosaccharyltransferase AglB, plus multiple genes for glycosyltransferases, indicates that N-glycosylation pathways function in all four Deep Lake haloarchaea. The glycosyltransferase genes vary between the four genomes, indicating that the profiles of N-linked glycans would differ in these haloarchaea as has been observed between other species of haloarchaea (Kaminski et al., 2013). Genes for a bacterial-type CapBCA biosynthesis complex are encoded in tADL and H. lacusprofundi, indicating that they may be able to synthesize a poly-gamma-glutamate capsule (Ashiuchi and Misono, 2002). tADL and DL31 also have gene clusters associated with the construction of capsular polysaccharide using nucleotide sugars. In addition, the tADL genome encodes a halomucin homolog (Bolhuis et al., 2006). The presence of neuA and neuB genes in tADL and DL31 indicates that they can probably synthesize N-acetylneuraminic acid (sialic acid).

Evaluation of growth properties

To assess specific metabolic properties inferred from the genome sequences, growth studies were performed on the Deep Lake haloarchaea (Supplementary Table S6). All grew using pyruvate as the defined carbon and energy source, which is a common feature of haloarchaea. tADL and H. lacusprofundi were able to use simple sugars as defined carbon and energy sources, but DL31 and DL1 could not. Thus, only tADL and H. lacusprofundi can be described as being saccharolytic. The strongest growth of tADL was using a combination of pyruvate and sucrose or pyruvate and glycerol. However, tADL growth on sucrose as the defined carbon and energy source was weak. Initially, no growth was detectable on glycerol as the defined carbon and energy source, but when tADL grown on medium containing both pyruvate and glycerol was then transferred to medium containing just glycerol, growth was observed.

For DL31 and DL1, pyruvate was the only substrate that served as a defined carbon and energy source. Growth of both strains was enhanced in media supplemented with 0.1% peptone. Overall, H. lacusprofundi had the ability to grow on the widest range of substrates.

Urea was used as a nitrogen source by tADL and H. lacusprofundi. Urea-grown tADL cultures were strongly urease positive, whereas urea-grown H. lacusprofundi cultures were urease negative.

Ecology inferred from physicochemical data

Although Deep Lake is clearly dominated by haloarchaea, OTUs for bacteria were present, including those for Bacteroidetes, Gammaproteobacteria, Betaproteobacteria and Firmicutes (Supplementary Table S1). Within Gammaproteobacteria, a subset of OTUs matched Ectothiorhodospiraceae (purple sulfur bacteria), which suggests the presence of a potentially photoautotrophic bacterium in Deep Lake, although its low abundance (0.6%) indicates that it would make a negligible contribution to primary productivity. The relative abundances of tADL and DL31 were negatively correlated (Pearson correlation coefficient r=−0.82; P<0.01), and DL31 was negatively correlated with Dunaliella (r=−0.65; P<0.05). The relative abundance of DL1 was positively correlated with OTUs for the anaerobe Halanaerobium (0.9% abundance; r=0.65; P<0.01) and total Firmicutes (1.2% abundance; r=0.68; P<0.01). The relative abundance of DL31 also positively correlated with total organic carbon concentrations (r=1.00;, P<0.01) and with OTUs for the low abundance (0.8%) gammaproteobacterium Halomonas subglaciescola (r=0.95; P=0.05). The relative abundance for Hm. subglaciescola positively correlated with nitrate concentration (r=0.96; P<0.05). Dunaliella relative abundance positively correlated with the depth of the water column (r=1.0; P<0.01, for both chloroplast and 18S rRNA genes), and Dunaliella chloroplast and 18S rRNA gene relative abundances were also positively correlated (r=1.0; P=0.01).

Discussion

Integration of genomic characteristics, growth properties, analysis of haloarchaeal distribution by filter size and depth, and assessment of lake chemistry enabled inferences to be drawn about the metabolic capacities and specific ecophysiological traits of the four Deep Lake haloarchaea. The distinctive features provided insights into substrate preferences occurring in the lake, how the four Deep Lake haloarchaea co-exist in the lake and why tADL is dominant. Below we discuss key aspects of their inferred ecophysiologies.

tADL

We identified a number of reasons that may explain the numerical dominance of tADL in Deep Lake. Of the four Deep Lake haloarchaea, only tADL possesses genes for BR and gas vesicles. The use of the latter as flotation devices facilitates vertical migration and enables access to solar irradiation at the lake surface by membrane-bound BR. Although the distribution of tADL within Deep Lake is generally homogeneous, it is less abundant at lower depths (DeMare et al., 2013; Supplementary Table S1). Gas vesicle formation in tADL may therefore be linked to photoheterotrophic growth. This may particularly boost tADL numbers in summer when illumination and algal primary production are highest.

The halotolerant microalga Dunaliella accumulates molar concentrations of glycerol as a compatible solute for osmotic stabilization (Borowitzka, 1981). Although the cytoplasmic membrane of Dunaliella is far less permeable to glycerol than that of other algae, leakage of glycerol does occur (Bardavid et al., 2008), and glycerol will also be released because of cell death and lysis. Glycerol is therefore regarded as a major carbon and energy source for haloarchaea in hypersaline habitats (Bardavid et al., 2008; Sherwood et al., 2009), including Deep Lake (Franzmann et al., 1988). Of the four Deep Lake genomes, only DL1 lacks genes for enzymes required for glycerol catabolism. Overall, tADL and H. lacusprofundi appear to be best equipped to utilize glycerol as a carbon and energy source in Deep Lake: both have dual pathways for glycerol degradation to DHAP (initiated by glycerol kinase or glycerol dehydrogenase), and tADL has genes for multiple orthologs of glycerol kinase and a putative membrane protein implicated in glycerol uptake (Anderson et al., 2011). Growth studies confirmed the ability of both tADL and H. lacusprofundi to utilize glycerol, with especially strong growth observed for tADL when both glycerol and pyruvate were provided in the medium (Supplementary Table S6).

The genomic evidence indicates that tADL has a ‘sweet tooth’, with a highly saccharolytic metabolism served by abundant ABC transporters and regulation by PTS. Although both tADL and H. lacusprofundi are saccharolytic, only the tADL genome is statistically over-represented for the COGs category associated with carbohydrate transport and metabolism (Figure 1), and for individual COGs associated with carbohydrate uptake and metabolism (Table 1, Supplementary Table S3). PTS indicates that catabolite repression occurs, allowing a rapid response to preferred carbon and energy sources. Nevertheless, as with glycerol, our growth assays indicate that growth on sucrose and mannose is enhanced in the presence of pyruvate. Surplus carbon could be stored by tADL as polyhydroxyalkanoate and mobilized under carbon-depleted conditions. A preference for high-energy substrates could also permit the bioenergetic investment required to degrade phosphonates. Phosphonates are abundant in nature (for example, as components of lipids), but the ability to use phosphonates as a substrate requires cleavage of the strong C–P bond.

In addition to the above, the tADL genome is the only one with genes for a nitrogen regulatory system (involving PII), thereby integrating responses to nitrogen and carbon availability (Pedro-Roig et al., 2013). tADL also appears to be able to source ammonia directly from the environment using specific transporters, as does H. lacusprofundi and DL1. Ammonia levels do not appear to be limiting in Deep Lake. Nevertheless, tADL and H. lacusprofundi have genomic evidence for nitrate uptake and assimilatory reduction. Nitrate was detected in Deep Lake, and possibly derives from bird guano (for example, penguins and skuas). tADL also potentially derives nitrogen from a diverse array of organic sources, including urea, free amino acids and polyamines. However, in contrast to DL31 (see DL31, below), tADL appears not to utilize oligopeptides (Table 1, Supplementary Table S3). Although urea can cross the cell membrane by passive diffusion, the genome of tADL encodes a urea ABC transporter system. The major source of urea in the Vestfold Hills is likely from pinniped excreta; by contrast, guano is dominated by uric acid and does not contain urea (Lindeboom, 1984). The ability of tADL to grow using urea as a nitrogen source is consistent with the presence of urea transport and urease genes and detectable urease activity.

DL31

All data point to DL31 having a heterotrophic strategy very different to tADL. DL31 is the only one that is lacking genes for swimming motility and/or gas vesicles. The genome (including COGs analyses) indicates a highly proteolytic metabolism, with abundant peptide transporters and proteolytic enzymes (Table 1, Supplementary Tables S3 and S5). The negative correlation between the distributions of tADL and DL31 across filter sizes at different depths (Supplementary Table S1) is likely to reflect divergent ecological strategies within the water column. One hypothesis is that DL31 adheres to particulate matter and targets proteinaceous substrates in detritus. This is consistent with the strong positive correlation between the relative abundance of DL31 and total organic carbon concentrations.

However, the data do not support an affinity for Dunaliella cells per se, as the relative abundance of DL31 was in fact inversely correlated with Dunaliella OTUs. Similarly, there was no correlation between OTUs for Halomonas and Dunaliella, even though a syntrophic relationship between the two taxa has been proposed, with Halomonas growing on dissolved organic carbon exuded by primary producing Dunaliella (Keshtacher-Liebson et al., 1995). The positive correlation observed between the abundances of DL31 and Halomonas might therefore suggest a similar substrate preference. Hm. subglaciescola is a non-saccharolytic halophile originally isolated from nearby Organic Lake, and this bacterium has been proposed to grow on proteinaceous compounds released from decaying penguin feathers molted during the summer (Franzmann et al., 1987). These data support a preference for detritus rather than living algal cells.

Unlike the other three Deep Lake haloarchaea, DL31 lacks genes for the transport of ammonia or nitrate/nitrite. The increased genomic capacity of DL31 for uptake of organic nitrogen compounds likely obviates the need for either ammonia uptake from the environment or assimilatory nitrate reduction. This physiological characteristic may also distinguish the ecological niches that DL31 is capable of colonizing.

DL31 has the genomic capacity to degrade glycerol (similar to tADL and H. lacusprofundi), although the initial step in DL31 can only be catalyzed by glycerol kinase. In the absence of any identifiable glycerol transporters in the genome, DL31 would be reliant upon passive diffusion for glycerol uptake. Generally, cell membranes are permeable to glycerol (Bardavid et al., 2008), so this could be sufficient for accessing this substrate, especially if glycerol is obtained by DL31 attachment to Dunaliella-derived cell debris.

H. lacusprofundi

Similar to tADL, H. lacusprofundi is capable of using a range of organic acids and simple sugars for growth (Franzmann et al., 1988). By inference of metabolic capacity and substrate preference, H. lacusprofundi is the least specialized of the four Deep Lake haloarchaea. This is consistent with it having the lowest number of over- or under-represented metabolism-related COGs (Figure 1). The lack of specialization may explain why it is by far the most represented species arising from cultivation studies of Deep Lake water (Franzmann et al., 1988).

Beyond this general trait, a number of specific characteristics of H. lacusprofundi are worth noting. H. lacusprofundi has the highest number of TRAP transporters, indicating that it may prefer organic acids (Table 1; Supplementary Table S3). H. lacusprofundi is capable of growth on acetate (Franzmann et al., 1988), possibly by using a glyoxylate cycle that includes a novel malate synthase (Lynch et al., 2012). Similarly, H. lacusprofundi grew with urea as a sole nitrogen source, despite the genome lacking identifiable genes for urea transport or catabolism (urease and urea amidolyase) and cultures testing urease negative (see above; Franzmann et al., 1988).

A complete denitrification pathway was identified in the H. lacusprofundi genome, despite the original description of H. lacusprofundi describing its inability to perform nitrate respiration (Franzmann et al., 1988). Specific growth conditions may be required to promote nitrate respiration and other denitrifying activities. In this regard, it is noteworthy that despite overall low OTU abundance in the lake (Supplementary Table S1), the two most abundant bacteria, Halospina denitrificans and Hm. subglaciescola, are both facultatively anaerobic Gammaproteobacteria that can reduce nitrate to nitrite and N2O (Franzmann et al., 1987; Sorokin et al., 2006). Nitrite and N2O intermediates of denitrification may serve as terminal electron acceptors for anaerobic growth by certain Deep Lake haloarchaea (possibly H. lacusprofundi), and dissimilatory nitrate reduction by halophilic bacteria and/or haloarchaea may contribute to the nitrite detected in Deep Lake (Supplementary Table S2).

DL1

Genomic evidence suggests that amino acids are preferred substrates for DL1, and that glycerol cannot be used. However, in common with the three other Deep Lake haloarchaea, DL1 has the genomic potential for the utilization of glycerol-3-phosphate (a degradation product of bacterial and eucaryal phospholipids). Amino acids are likely obtained by DL1 directly from the environment. However, although both DL1 and DL31 genomes were statistically over-represented for the COGs category related to amino acid metabolism (Figure 1), there was no genomic evidence that DL1 is specialized for the degradation of peptides. Unlike DL31, DL1 encodes a gene for ammonia permease. Degradation of amino acids by DL1 explains the presence of urease genes without genes for identifiable urea transporters, because urea is generated endogenously by degradation of arginine (Baliga et al., 2004). In addition, the DL1 genome was statistically over-represented for COGs associated with BCAA uptake and catabolism, including OADHC components (Table 1, Supplementary Table S3). In archaea, certain OADHCs can catalyze the decarboxylation of branched-chain 2-oxoacids that result from BCAA catabolism, as well as pyruvate (Heath et al., 2004, 2007; Sisignano et al., 2010). Thus, these COGs associated with general amino acid metabolism, BCAA uptake and catabolism, and OADHC components, likely describe a common substrate preference by DL1.

DL1 is by far the least abundant of the four Deep Lake haloarchaea. This may relate to its metabolic preference for amino acids as carbon and energy sources combined with the inability to catabolize glycerol. Free amino acids may not be readily available in Deep Lake; the abilities of all four Deep Lake haloarchaea to biosynthesize the 20 proteinogenic amino acids are consistent with a nutrient-depleted environment and the extremely low productivity of Deep Lake (Campbell, 1978). The high number of nitrate/nitrite transporter and multicopper oxidase genes in DL1 might also indicate a capacity, and possibly a preference, for anaerobic growth via nitrite respiration; this property would be unlikely to promote competitiveness throughout the oxic water column, but may provide niche adaptation.

Capsular polysaccharide biosynthesis

tADL, DL31 and H. lacusprofundi have a genomic capacity for capsular polysaccharide biosynthesis. The production of extracellular polymeric substances (EPS) and biofilm formation has been observed for certain strains of tADL and H. lacusprofundi, with N-acetylglucosamine and/or sialic acid identified as constituent glycoconjugates (Fröls et al., 2012). EPS promote cell aggregation and biofilm formation of archaea at low temperatures (Reid et al., 2006; Allen et al., 2009; Williams et al., 2011). As well as facilitating adhesion to solid surfaces, EPS is also inferred to serve in protection against desiccation, osmotic stress and phage attack (Sutherland, 2001). For the Deep Lake haloarchaea, the EPS may protect cells against dehydration, facilitate membrane function in the cold and also modulate viral infection; viruses have been detected and previously associated with gene transfer events between the Deep Lake haloarchaea (DeMaere et al., 2013). For non-motile DL31 cells, the EPS may have a particular role in particle attachment.

The presence of distinct glycosyltransferase genes and the specific complement of cell envelope genes in the four genomes indicate that the type of EPS produced is likely to vary between the Deep Lake haloarchaea. Futhermore, in tADL, the neuA and neuB genes involved in the synthesis of sialic acid reside in a gene cluster that has previously been suggested as a marker of genomic variation indicative of ecotypes; the region was found to have low fragment recruitment coverage of metagenome data, indicating that sub-populations with specific phenotypic differences occur in the lake (DeMaere et al., 2013).

Conclusions

Based on genomic comparisons, the four Antarctic haloarchaea examined showed distinct substrate preferences. tADL showed evidence of a ‘high-energy’ saccharolytic metabolism that complements the ability to utilize biogenergetically expensive substrates (for example, urea and phosphonate) and to construct gas vesicles to optimize light harvesting by BR at the lake surface. The extensive illumination of the lake surface during the Antarctic summer is likely to promote photoheterotrophic growth. H. lacusprofundi is also saccharolytic, but the genome suggests a more versatile heterotrophic metabolism than tADL and a higher reliance on organic acids as carbon sources. In contrast, DL31 and DL1 appear to be metabolic specialists that target peptides and free amino acids, respectively, with the former possibly associating with particulate organic matter. The data are also consistent with the hypothesis that a capacity to catabolize glycerol contributes toward the relative abundance of the four species, insofar as Deep Lake is a hypersaline environment where Dunaliella is the major primary producer. In general, the OTU and genomic data provide evidence for niche partitioning among Deep Lake haloarchaea. Coexistence of specialist and versatile haloarchaea has been proposed for other hypersaline environments (Javor, 1984). Further work on ecological interactions of the Deep Lake haloarchaeal community is currently being investigated using metaproteomics in order to determine whether the substrate preferences hypothesized for these four haloarchaea (tADL, DL31, H. lacusprofundi and DL1) are corroborated by protein abundance levels in Deep Lake.