Introduction

Most (∼75%) of the Earth's biosphere is cold (⩽5 °C), consisting of polar and alpine regions, the deep ocean, terrestrial and ocean subsurface, caves, the upper atmosphere, seasonally and artificially cold environments. Diverse cold-adapted (psychrophilic) microorganisms have evolved a capacity to proliferate in all these habitats. However, despite the enormous contribution cold environments make to the Earth's biosphere, relatively little is known about the genetic makeup and molecular mechanisms of adaptation of the resident microorganisms and how they drive the critical biogeochemical processes that help to maintain the planet in a habitable state (Cavicchioli, 2006; Murray and Grzymski, 2007).

The majority of microbiological studies of cold environments have focused on psychrophilic bacteria, and genomic analyses have provided valuable insight into unique characteristics of sea ice, sediment and planktonic species (Rabus et al., 2004; Medigue et al., 2005; Methé et al., 2005; Riley et al., 2008). In addition to bacteria, it is now clearly recognized that the archaea are numerically, phylogenetically and functionally important members of cold aquatic and terrestrial ecosystems (Cavicchioli, 2006; Murray and Grzymski, 2007). Their role in global biogeochemical cycles is diverse, including critical roles in the nitrification of soil and ocean waters mediated by ammonia oxidation (Konneke et al., 2005; Leininger et al., 2006; Cavicchioli et al., 2007), and the cycling of simple carbon compounds by methanogenesis and reverse methanogenesis (anaerobic oxidation of methane) (Kruger et al., 2005; Cavicchioli, 2006).

Cold anaerobic environments (for example, ocean and lake sediments, permafrost) harbor methanogens that have the capacity to contribute significantly to global carbon emissions through the production of methane as a green house gas. Although presently marine environments only contribute approximately 2% of the world's methane flux, this appears to be kept in check by the microbial communities that anaerobically oxidize methane (Moran et al., 2008). The methane cycle is still poorly understood, and gaining an understanding of how these cellular processes occur in the cold will require fundamental knowledge of the molecular mechanisms of cold adaptation of psychrophilic methane-producing and methane-oxidizing archaea (Cavicchioli, 2006). Psychrophilic archaea have generally proven to be difficult to isolate. However, methanogens have been isolated from Antarctic lakes, marine sediment in Alaska and the Baltic Sea, a freshwater lake in Switzerland and Arctic permafrost (Franzmann et al., 1992, 1997; Simankova et al., 2001; Chong et al., 2002; von Klein et al., 2002; Singh et al., 2005; Kendall et al., 2007; Kotsyurbenko et al., 2007; Morozova and Wagner, 2007).

Methanococcoides burtonii was the first formally characterized archaeal psychrophile and was isolated from cold (1–2 °C) methane-saturated, anaerobic bottom waters (25 m depth) of Ace Lake, Antarctica (Franzmann et al., 1992). M. burtonii is a methylotrophic methanogen utilizing methylamines and methanol, but not H2:CO2 or acetate for growth. It is a eurypsychrophile with a relatively broad growth temperature range (−2.5 to 29 °C), in contrast to Methanogenium frigidum that is a stenopsychrophile (0–18 °C) also isolated from Ace Lake (Franzmann et al., 1992, 1997; Reid et al., 2006). Unlike M. frigidum, which has proven to be very difficult to grow, M. burtonii is amenable to laboratory cultivation and has become a model psychrophilic archaeon for dissecting molecular mechanisms of cold adaptation (Cavicchioli, 2006). Studies on M. burtonii have examined enzyme structure/function, gene regulation, tRNA modification, membrane lipid composition and proteomics (reviewed in Cavicchioli, 2006), and intracellular solutes (Thomas et al., 2001; Costa et al., 2006). The draft genome sequences of M. burtonii and M. frigidum have also enabled comparative genomic analyses assessing the compositional and structural basis of thermal adaptation of proteins for archaea spanning growth temperatures ranging from 0 to 110 °C (Saunders et al., 2003).

The genome sequence of M. burtonii was recently completed paving the way for determining for the first time, the genomic basis for growth and survival of a psychrophilic archaeon. Due to the value of the genome sequence and importance of M. burtonii to the scientific community, the genome sequence was exhaustively manually annotated to generate a high-quality genome sequence that was then used to evaluate the evolution and biology of M. burtonii.

Materials and methods

Genome sequencing, assembly, automated annotation

DNA was isolated from M. burtonii DSM 6242 grown at 23 °C. The genome of M. burtonii was sequenced at the Joint Genome Institute (JGI) using a combination of 3, 8 and 40 kb DNA libraries. All general aspects of library construction and sequencing can be found at the JGI's Web site (http://www.jgi.doe.gov/). The Phred/Phrap/Consed software package (www.phrap.com) was used to assemble all three libraries and to assess quality. Possible misassemblies were corrected, and gaps between contigs were closed by editing in Consed, custom primer walks or PCR amplification. The error rate of the completed genome sequence of M. burtonii is less than 1 in 50 000. Putative coding regions were identified using Critica, Generation and Glimmer, with automated annotation of proteins according to search results from the databases TIGRFam, PRIAM, Pfam, Smart, COGs, Swiss-Prot/TrEMBL and KEGG, as described for other JGI genomes (Medigue et al., 2005; Chain et al., 2006; Klotz et al., 2006; Ivanova et al., 2007). Additional curation involved the calculation of homolog, paralog and ortholog gene relationships, and generation of genome statistics. A round of manual curation was performed on the predicted genes of M. burtonii by IMG staff in which 346 changes were made. This corresponds to 13.9% of predicted genes. The start sites of 136 genes were modified, 62 genes were deleted and 61 new genes were added. Also, 87 genes were identified as pseudogenes. This includes new pseudogenes that were added as well as previously existing genes that were converted to pseudogenes. IMG's curation staff also performed horizontal annotation involving manual investigation of proteins and biochemical pathways, with ‘IMG term’ annotations propagated across genomes according to strict homology criteria (http://imgweb.jgi-psf.org/w/doc/img_er_ann.pdf). In the M. burtonii genome, 635 proteins were assigned IMG terms by this process.

Manual annotation, evidence rating system and comparison genomes

The protein sequence for each gene was searched against the Swiss-Prot and Protein Data Bank (PDB) databases using BLAST to identify the most closely related, experimentally characterized homolog available in the literature. The curated part of Swiss-Prot database was selected because of the extent and quality of annotations associated with each protein sequence, and the PDB provides an archive of experimentally determined three-dimensional protein structures. BLAST matches were examined sequentially, searching for a match with direct experimental verification of the function. Matching proteins were not considered if they had no function listed, a function determined ‘by similarity’ or no reference cited. Experimental evidence that was considered acceptable to define function included papers detailing the expression and characterization of the protein, protein crystallography studies and mutation and complementation studies that defined function. Papers that only documented the nucleotide sequence, including genome sequences, were not considered to provide sufficient evidence of function, nor were unpublished protein crystal structures or the (published or unpublished) crystal structures of hypothetical proteins for which function was not known. In most cases papers were read in sufficient depth to establish if conclusions about function were justified. Choosing published experimental evidence as a prerequisite for defining function probably excluded some valid unpublished data; however, it was not possible to assess this data and it was excluded. Careful attention was also given to the recent literature to identify valid experimental data (for example, several methanogenesis-related proteins: Mbur_0808, Mbur_0811) that had not yet been updated in protein databases.

After the best experimentally characterized homolog had been determined, InterPro and Pfam domains were checked for their presence in the M. burtonii homolog to confirm that all identifiable functional domains were conserved. The graphical representation on the Swiss-Prot BLAST output was used for assessing the arrangement and extent of domain matches throughout the length of the query and matching proteins, thereby providing a rapid way to assess both the global (whole protein) and local (domain and motif) similarities between the M. burtonii and the functionally characterized proteins.

An evidence rating (ER) system was developed to enable the confidence of functional assignments to be clearly displayed for each gene (Supplementary Information: manual annotation). ER1 indicates that the protein from M. burtonii had been experimentally characterized (a self-match); ER2, the most closely related functionally characterized homolog is not from M. burtonii but the BLAST alignments share ⩾35% sequence identity along the entire length of the protein; ER3, the most closely related functionally characterized homolog shares <35% sequence identity along the length of the protein, but all required motifs/domains for function are present and complete; ER4, an experimentally characterized full-length homolog is not available but conserved protein motifs or domains can be identified; ER5 (hypothetical protein), no functionally characterized homolog can be found, and no characterized protein domains above the Pfam and InterProScan cutoff thresholds can be identified. The accession number for the M. burtonii genome is CP000300. The manually annotated genome is available through IMG/GEBA (img.jgi.doe.gov/geba) and will become available through the regular IMG system. Genome updates will also be made available through IMG. Comparative genomics (see methods below) was performed using local or external databases (for example, NCBI, IMG) with between 37 and 48 closed archaeal genome sequences. The draft sequence of M. frigidum was included in some comparisons (specified in the text).

COG scrambler and ALL scrambler

Two in-house programs, COG_scrambler.pl and ALL_scrambler.pl, were created for identifying significant differences in gene composition between samples. The COG_scrambler program compares all the COGs present in two data sets and calculates which COG categories are statistically over- or underrepresented in a data set by difference of medians analysis. The ALL_scrambler program performs an analogous process but reports which individual COGs are statistically over- or underrepresented. All programs were run with a confidence level setting of 0.99 (99% confidence that the difference is not due to chance) and 10 000 resampling replicates. The sample size was 6000 for the comparison of M. burtonii vs methanosarcinal genomes, and 10 000 for all other comparisons. The approach to determine the statistical significance was developed by FML and is similar to that described by Rodriguez-Brito et al. (2006).

Phylogenetic analyses

To visualize the phylogenetic relationships between proteins in the genome, we created an in-house PERL script ‘phylo_profiler.pl’ program. This program takes an input file (all M. burtonii proteins) and uses BLAST to create a matrix of BLAST scores against selected completed genome sequences. The best BLAST bit score for each M. burtonii protein was then normalized both within and across genomes as previously described (Enault et al., 2003). Pertinent data such as the COG ID, COG category, annotation and ER value were incorporated in the matrix to facilitate various searches and analyses. The matrix produced by phylo_profiler.pl was viewed using BioLayout Express 3D (Freeman et al., 2007) with typical settings as follows: Graph Size vs Corr. Threshold=95 (Nodes=630, Edges=1981, R2=0.7987), Start temperature=250, No. of iterations=60, K-value modifier=1.0, Iterations for Burst=10, Minimum component size=3. The protein sequences of transposases occurring multiple times (and those identified by proteomics to be expressed in the cell) were aligned using ClustalW (Thompson et al., 1994) and used to build phylogenetic trees by neighbor-joining (Saitou and Nei, 1987) using the Poisson-corrected distance implemented in MEGA4 (Tamura et al., 2007). The stability of the relationships was assessed by 1000 bootstrap replicates.

Identification of predicted horizontal gene transfer events

The program Alien Hunter (Vernikos and Parkhill, 2006) was used to identify genomic islands with altered nucleotide signatures. These islands were further divided in groups depending on their interpolated variable order motif (IVOM) scores and the open-reading frames (ORFs) contained within these regions were parsed using custom PERL scripts. The ORFs overlapping an island border were eliminated from further analyses. Each ORF was then phylogenetically assigned according to the best two BLAST hits with a minimum bit score of 50 against a customized version of the NCBI nonredundant (nr) database from which all M. burtonii proteins had been eliminated. The phylogenetic distribution was computed using MEGAN (Huson et al., 2007). Codon usage was analyzed with CodonW (Peden, 2000). The genes on the primary axis of inertia in the correspondence analysis with values above the mean value plus one standard deviation were considered efficiently expressed; those below the mean value minus one standard deviation were considered not efficiently expressed. Amino-acid percentage composition and principal component analysis were performed using customized PERL scripts and R (Ihaka and Gentleman, 1996) as described by Saunders et al. (2003). An artificial genome was generated by removing the coding sequences of efficiently expressed ORFs from the FASTA sequence of the genome using the in-house PERL script ‘reduce_genome.pl’.

Identification of hypothetical proteins unique to M. burtonii

The IMG Phylogenetic Profiler tool was used to search for genes in the M. burtonii genome, which did not contain a homolog in any of the 814 standard reference genomes (40 completed archaeal genomes, 5 draft archaeal genomes, 19 completed eukaryal genomes, 21 draft eukaryal genomes, 486 completed bacterial genomes and 243 draft bacterial genomes). The cutoff settings for a homolog was minimum amino-acid identity of 30% and maximum E-value of 1 × 10−5 using the present/absent algorithm. The list of proteins obtained was then curated further by removing any proteins that were not annotated as a hypothetical protein (ER5). BLAST searches against the NCBI nr database and the CAMERA Global Ocean Survey combined assembly ORF peptides database were performed (2 May 2008), and any proteins with matches of E-value smaller than 1 × 10−5 were also discarded to obtain the final list of 117 proteins that are unique to M. burtonii. Conserved gene context analysis (Saunders et al., 2005) was also used to infer possible functions for 28 proteins that were known from proteomics analysis to be expressed in the cell (Goodchild et al., 2004a, 2004b, 2005; Saunders et al., 2005). Some of the unique expressed hypothetical proteins were found to be located in pairs (for example, Mbur_2063 and Mbur_2064) or larger clusters with unique hypothetical proteins that (to date) have not been detected in expression studies (for example, Mbur_0640 to Mbur_0643 and Mbur_0275 to Mbur_0278).

Results and discussion

M. burtonii genome overview

The completed genome sequence of M. burtonii DSM 6242 contains 257 5032 bp in a single circular chromosome (Figure 1). Manual annotation (see Materials and methods section) resulted in changes to the functional annotation of 1079 of the 2494 genes that had been identified by the auto-annotation process, and functions were assigned to 364 genes that previously had no functional prediction. A complete description of the manual annotation approach is provided in Supplementary Information: manual annotation.

Figure 1
figure 1

Circular representation of the Methanococcoides burtonii genome. The circles show (outermost to innermost): (1) DNA coordinates (black); (2) genes on forward strand, color coded by COG categories; (3) genes on reverse strand, color coded by COG categories; (4) RNA genes, including tRNAs (green), rRNAs (red) and other RNAs (black); (5) genes involved in putative polysaccharide/capsule biosynthesis operons (red); (6) GC content; (7) GC skew.

Full size image

To identify characteristics of the M. burtonii genome that distinguishes it from other archaea, we compared the three completed Methanosarcinales genomes, all 14 methanogen genomes or 39 archaeal genomes to the M. burtonii genome using COG_scrambler and ALL_scrambler (Table 1). In all three sets of comparisons, the Signal transduction [T] and Replication, recombination and repair [L] COG categories were significantly overrepresented, and the General Function prediction only [R] COG category was underrepresented in the M. burtonii genome. Individual COGs that were overrepresented included signal transduction histidine kinases, RecA-superfamily ATPases and CheY-like response regulators [T], and numerous transposases [L]. In addition, Cell wall, membrane, envelope biogenesis [M] was overrepresented in all but the Methanosarcinales genomes.

Table 1 COG categories with a statistically significant difference in abundance
Full size table

Gene categories associated with cold adaptation

To identify functional gene categories characteristic of cold adaptation in archaea, we compared genomes from archaeal thermophiles and hyperthermophiles (4 methanogens, or 25 archaea in total) to M. burtonii using COG_scrambler and ALL_scrambler (Table 1). The psychrophilic genome of M. burtonii was overrepresented with Defense mechanisms [V] and Motility [N] and underrepresented with Translation [J] and Nucleotide metabolism [F] COG categories compared to both the methanogen and total archaeal genome sets.

In the Defense mechanisms category, the individual COGs overrepresented in M. burtonii were mainly from Type-I restriction modification (RM) systems (COG0286, COG0610, COG0732 and COG4096), and ABC transporters (see below). The RM COGs were present at four locations on the M. burtonii genome (Supplementary Figure S1). One set of restriction endonuclease, methyltransferase and specificity genes was arranged in a typical operon-like RM cluster (Mbur_1841 to Mbur_1843). Another cluster was spaced over 10 genes (Mbur_1213 to Mbur_1222) with numerous genes for hypothetical proteins (ER5) interspersed. Two clusters (Mbur_0506 to Mbur_0509 and Mbur_0480 to Mbur_0484) contain a divergent AAA domain protein (ER4), in addition to the specificity COG0732, methyltransferase COG0286 and a variant helicase as the restriction endonuclease (Mbur_0509 and Mbur_0484). Divergent AAA domain proteins have been implicated in a wide variety of roles, including functioning as transcriptional regulators. The relatively large number of diverse RM systems may be indicative of the exposure of M. burtonii to high levels of foreign DNA (see Cold adaptation is associated with specific signatures of genome evolution section).

The other overrepresented COGs from the Defense mechanisms category were COG0577 and COG1136. Proteins belonging to these two COGs were located in six short clusters in the genome, in association with proteins from COG1361 and COG4591 (Figure 2a). COGs 0577 and 4591 both contain proteins that annotated as DUF214-containing protein (ER4). These proteins contain four transmembrane domains each, and one member of the DUF214 family has been characterized as an ABC-transporter permease involved in lipoprotein export (Narita et al., 2003). Interestingly, although COG1361 is designated as S-layer domain proteins, the hypothetical proteins found in COG1361 form a cluster distinct from the true S-layer proteins, and have similarity to the ABC-transport substrate-binding proteins observed elsewhere in the M. burtonii genome (Figure 2b). These proteins each possess a signal peptide motif and one transmembrane domain, similar to ABC-transporter substrate-binding proteins. These clusters of genes may therefore represent novel ABC transporters.

Figure 2
figure 2

Putative ABC transporters involved in cell defense. (a) COG0577, ABC-type antimicrobial peptide transport system, permease component, all annotated DUF214 protein (ER4); COG4591, ABC-type transport system, involved in lipoprotein release, permease component, all annotated DUF214 protein (ER4); COG1136, ABC-type antimicrobial peptide transport system, ATPase component, all annotated ABC-transporter ATPase (ER2); COG1361, S-layer domain, all annotated hypothetical protein (ER5). (b) Unrooted phylogenetic tree of substrate-binding proteins from ABC transporters constructed using the neighbor-joining method from a multiple alignment of amino-acid sequences of COG1361 hypothetical proteins (aqua), S-layer proteins (pink) and ABC-transporter substrate-binding proteins (black) from M. burtonii. The sequence alignments are in given Supplementary Figure S2.

Full size image

The overrepresentation of the putative novel ABC transporters contrasts with the lack of identifiable ABC transporters for peptides in M. burtonii. This is one of the major differences between the genomes of M. burtonii and the Methanosarcina spp., with the latter containing between 6 and 19 substrate binding proteins for peptide ABC transporters. The lack of peptide transporters is consistent with the inability of M. burtonii to use peptides for growth. M. burtonii also lacks drug exporters from subfamilies 2 and 3 of the major facilitator superfamily (TC 2.A.1) and proteins from the ACR3 family (TC 2.A.59) and the ArsB family (TC 2.A.45) that are found in Methanosarcina spp. The susceptibility that M. burtonii may face toward antimicrobial compounds and toxic metalloids may be alleviated by the novel ABC transporters, some of which may also have a role in active transport leading to the formation of extracellular polymeric substances (EPS) that is characteristic of growth of M. burtonii in the cold (Reid et al., 2006; see A large genomic commitment to polysaccharide biosynthesis section). It is also noteworthy that, unlike the Methanosarcina spp., M. burtonii has a coenzyme F420-dependent sulfite reductase (Mbur_0619) similar to the one characterized in Methanocaldococcus jannaschii (Johnson and Mukhopadhyay, 2005) that allows it to tolerate sulfite in the environment. In the waters where M. burtonii was isolated, hydrogen sulfide concentrations reach very high levels (8 mM) (Rankin et al., 1999), indicative of high sulfite reductase activity (also see Linking the evolution of the M. burtonii genome to its ecological niche section).

Cold adaptation is associated with specific signatures of genome evolution

All members of the Methanosarcina genus (for which M. burtonii is closely related) have been shown to possess extremely dynamic genomes (Maeder et al., 2006), where genome rearrangements, acquisition of novel metabolic capabilities (Fournier and Gogarten, 2008) and capacity to express altered or foreign genes (Li et al., 2007b) contribute to their capacity to colonize a wide variety of ecological niches. To examine the importance of horizontal gene transfer (HGT) to the M. burtonii genome, we performed gene-independent analysis of potential HGT using IVOM scores calculated from the program Alien Hunter (Vernikos and Parkhill, 2006). An extremely high proportion (51%) of the M. burtonii genome (higher than for any other completed archaeal genome) was identified as having aberrant sequence composition compared to the average of the genome (Figure 3a). However, across the 48 archaeal genomes, the number of base pairs implicated in HGT did not correlate with growth temperature (Figure 3a). For example, the mesophilic Methanococcus maripaludis C5 had very little predicted HGT, whereas the hyperthermophilic Methanopyrus kandleri had almost 30% predicted HGT.

Figure 3
figure 3figure 3

Predictions of genome plasticity in M. burtonii. (a) Percentage of genome predicted to have undergone horizontal gene transfer by the Alien Hunter program. Bars are color coded according to the maximum growth temperature of the organism: purple, <30 °C; blue, 40–49 °C; green, 50–59 °C; yellow, 60–79 °C; orange, 80–99 °C; red, 100 °C and higher. (b) COG category classification (top pie charts), normalized expression (center bar graphs) and phylogenetic assignments (lower pie charts) of the open-reading frames (ORFs) contained in islands with different IVOM (interpolated variable order motif) scores. Norm (normal), IVOM⩽17.223; low, 17 223 < 26; medium, 26 ⩽IVOM < 40; high, 40⩽IVOM. (c) Correspondence analysis of codon usage of ORFs in regions with different IVOM scores. Norm (normal), IVOM ⩽17.223; low, 17.223 <26; medium, 26 ⩽IVOM <40; high, 40⩽IVOM. (d) 106 446 ORFs from archaeal genomes represented in Figure 3a, and ORFs from the draft sequence of M. frigidum were subjected to correspondence analysis of codon usage (Lynn et al., 2002). Growth temperatures: purple, <30 °C; blue, 40–49 °C; green, 50–59 °C; yellow, 60–79 °C; orange, 80–99 °C; red, 100 °C and higher.

Full size image

Parametric methods used for the detection of HGT events (such as Alien Hunter program) are prone to false positive predictions (Azad and Lawrence, 2007). To assess the reason for the high level of atypical nucleotide composition and better assess HGT, we performed a range of further analyses. Putative HGT islands were grouped together based on their IVOM scores and their phylogeny determined. For M. burtonii 1042 ORFs were found in 125 genomic islands with altered IVOM scores, 1097 were found outside of these regions and 164 were on the border and discarded from further analysis. The islands were 1 313 583 nucleotides in length and represented 51.01% of the genome, with an average island length of 10 509 nucleotides. For the mesophilic Methanosarcina genomes, the islands represented 40.89%, 40.79% and 35.36% for M. mazei, M. acetivorans and M. barkeri, respectively (Supplementary Table S1). Genes with the highest IVOM scores included a high proportion of unclassified genes and genes from unknown organisms (Figure 3b), and may have arisen through HGT events. The weakly atypical islands contained a high number of genes involved in translation, ribosomal structure and biogenesis (COG category J) that were phylogenetically congruent with the Methanosarcina genomes.

There are several lines of evidence that suggest that the ORFs within the weakly atypical islands have not arisen by HGT: (1) Genes encoding ribosomal proteins have high barriers to HGT (Sorek et al., 2007). (2) The codon usage of the ORFs within the islands is very similar to that of the rest of the genome (Figure 3c). (3) As evidenced by the proportion of genes most closely related to the Methanosarcinales genomes (Figure 3b), the phylogenetic distribution of the ORFs is indicative of vertical rather than horizontal transmission. In addition, the presence of a high number of ribosomal proteins, which often display atypical nucleotide composition (Tsirigos and Rigoutsos, 2005), suggests alternative reasons for the altered IVOM scores. Initially we hypothesized that the high incidence of predicted HGT genomic islands in M. burtonii was caused by variations in codon usage. However, codon usage of the ORFs within the islands was similar to that of the whole genome (Figure 3c) and to the Methanosarcinales genomes (Figure 3d). Moreover, although the usage of specific synonymous codons has been correlated with the ability to grow at high temperatures (Lynn et al., 2002), the codon usage of the psychrophilic archaea, M. burtonii and M. frigidum, was indistinguishable from that of the mesophilic Methanosarcina genomes (Figure 3d).

We subsequently reasoned that the alteration in IVOM scores might be caused by a high incidence of efficiently expressed genes whose protein products have been strongly selected for specific amino-acid usage. Principal component analysis of amino-acid composition of M. burtonii and the three Methanosarcina genomes shows a strong correlation (Spearman's correlation P<0.01) between the PC1 loadings of efficiently expressed genes and the PC2 loadings previously observed (Saunders et al., 2003) for whole genomes that was associated with temperature adaptation (Supplementary Figure S3). This correlation was not observed when the same analysis was performed with genes that are not efficiently expressed. We therefore conclude that (1) efficiently expressed genes in M. burtonii tend to have a stronger ‘psychrophilic’ component than those that are not efficiently expressed; (2) M. burtonii has unique amino-acid usage due to its psychrophilic lifestyle (for example, in comparison to the mesophilic Methanosarcina spp.).

As a further test, we generated an artificial M. burtonii genome where the coding regions of efficiently expressed ORFs were removed, reducing the genome size by 322 822 nucleotides (12.54%) and the number of genomic islands from 125 to 105. Of these only 63 were above the original IVOM score and encompass 782 651 nucleotides (34.75%) of the artificial genome. That is, the removal of the islands with efficiently expressed ORFs reduces the incidence of regions with atypical IVOM scores by twice as much as the number of nucleotides removed.

Our data highlight the need to carefully assess inferences about HGT when parametric methods are used. For M. burtonii, the exceptional amount of genome plasticity (that is, 51.01% altered IVOM scores) is primarily due to a strong selection for specific amino acids associated with cold adaptation present in efficiently expressed genes. This same coding bias is not seen in other genes. This may indicate that it is a recent evolutionary event and selection is yet to be manifested in other genes, or that low-temperature selection only has a significant effect on genes that need to be efficiently expressed in the cell (for example, high-abundance ribosomal proteins and other proteins critical to growth in the environment) (also see Linking the evolution of the M. burtonii genome to its ecological niche section). In comparison to hyper/thermophilic archaea that are subject to selection of synonymous codon usage (Lynn et al., 2002), M. burtonii has evolved cold adaptation through a genomic capacity to accommodate highly skewed amino-acid content; an ability it has obtained while retaining codon usage in common with related mesophilic Methanosarcina spp.

Another feature of the codon usage plot (Figure 3d) is that the pattern of codon usage for archaea with growth temperatures ⩽60 °C is similar and distinct from archaea with higher growth temperatures. A similar distinction at 60 °C was previously noted for the relationship between G+C content of tRNA and growth temperature of archaea (Saunders et al., 2003). In the case of tRNA in M. burtonii, high levels of dihydrouridine incorporation provide a mechanism for enhancing flexibility beyond that achievable through G+C content (Noon et al., 2003; Saunders et al., 2003). By analogy for M. burtonii proteins, amino-acid usage provides a mechanism for enhancing protein psychrophilicity beyond any requirements for specific codon usage.

To further assess possible HGT events, we analyzed the 500 ORFs that could not be reliably assigned to the domain Archaea (combined ‘other’ and ‘NA’ categories from Figure 3b). Compared to the whole genome these ORFs were overrepresented in COG categories M (Cell wall, membrane, envelope biogenesis) and T (Signal transduction mechanisms) and underrepresented in S (Function unknown), F (Nucleotide transport and metabolism) and J (Translation ribosomal structure and biogenesis) (Table 1). We note that the majority (44 of 70; 63%) of these ORFs were contained in islands with nonsignificant IVOM scores, illustrating that IVOM scores per se do not provide an unambiguous means of identifying HGT events in M. burtonii.

The underrepresentation of F and J is consistent with their central role in cell growth and survival and the inherent barriers for HGT of these genes. However, it was striking that category M was associated with HGT as this COG category is statistically overrepresented in the M. burtonii genome (Table 1) and the cell wall and lipid membrane have specifically been linked to cold adaptation (see A large genomic commitment to polysaccharide biosynthesis and Lipid biosynthesis sections). The 27 category M ORFs are mainly annotated as different types of glycosyltransferases with varying ERs (ER2–ER4), and a number of them appear to have been inherited from the ɛ-Proteobacteria. In category T, 39 ORFs had atypical phylogenetic assignments. These ORFs mainly belong to COG0642 (Signal transduction histidine kinase) and have ER3 or ER4 confidence ratings. Three that could be reliably assigned to a phylogenetic group (Mbur_1264, Mbur_2201 and Mbur_2108) clustered within the δ-Proteobacteria. This suggests that similar to category M, the overrepresentation of this Signal transduction category in M. burtonii (see M. burtonii genome overview section) is likely to have involved HGT (also see Signal transduction and adaptive potential section). In addition to these genes in categories M and T, HGT appears to have impacted on specific genes in central metabolism. Phylogenetic analysis of fructose-bisphosphate aldolase (Mbur_1969), serine O-acetyltransferase (Mbur_0414) and aconitase (Mbur_0316) shows that they are most closely related to counterparts from δ-Proteobacteria, and interestingly, M. burtonii is the first sequenced archaeon to have the B form of aconitase. Aspartate-semialdehyde dehydrogenase (Mbur_0379) does not cluster with archaeal/eukaryal sequences, but does branch within bacterial enzymes.

The presence of multiple genes in M. burtonii originating from ɛ- and δ-Proteobacteria suggests that a close, possibly syntrophic, relationship has existed between M. burtonii and members of these proteobacterial groups. In this respect, it is noteworthy that several abundant phylotypes similar to known sulfate-reducing bacteria of the δ-Proteobacteria, and less abundant signatures of ɛ-Proteobacteria have been detected in sediment and anaerobic water samples of Ace Lake (Bowman et al., 2000) (see Linking the evolution of the M. burtonii genome to its ecological niche section). Our analysis shows that HGT with ɛ- and δ-Proteobacteria has a specific role for acquiring genomic capabilities important for environmental adaptation.

Transposons are involved in genome evolution

Phylogenetic profiling was used to look for patterns of gene evolution. Each gene in the M. burtonii genome was compared using BLAST against a set of completed genomes to create a profile of the gene's phylogenetic relationships. Two comparisons were made: all completed archaeal genomes, and 431 completed bacterial and archaeal genomes. Genes were clustered according to their phylogenetic occurrence and visualized using BioLayout (Supplementary Figure S4).

Five of the most tightly clustered phylogenetic groups were composed of transposases, with each group having specific phylogenetic occurrence (Table 2; Figure 4). The DDE superfamily of transposases (group 5) appeared to be restricted to members of the Methanosarcinales. Two of the 117 unique hypothetical proteins (Mbur_0556 and Mbur_0576; full list given in Supplementary Table S2) are located in a region adjacent to three cassettes each encoding genes for a DNA-directed DNA polymerase B domain protein and at least two conserved hypothetical proteins. The duplication evident in this region of the genome may also have arisen through transposition, as three transposons are also located nearby. The three cassettes are shared only between M. burtonii and M. mazei. Several other hypothetical genes appear to have been duplicated, possibly by transposition. Mbur_0169 and Mbur_0074 share 93% sequence identity and both are located adjacent to a Radical SAM family protein and an α/β-hydrolase domain protein, indicative of a duplicated cassette.

Table 2 Clusters formed from phlogenetic profiling using genes from M. burtonii
Full size table
Figure 4
figure 4

Phylogeny of transposases in M. burtonii. Sixty-three transposases grouped in six major clusters: Group 1 is found in environmental genomic fragments from deep-sea sediments (GZfos19A5) and has multiple hits to M. acetivorans (3) and M. mazei (5). Group 2 is found in environmental genomic fragments from deep-sea sediments (GZfos27B6, GZfos9C4, GZfos26D6) and to M. acetivorans (1). Group 3 has multiple hits to M. acetivorans (11), M. mazei (1) and M. barkeri (5) and is probably an IS1-like element. Group 4 has hits to M. mazei (2) and is likely to belong to the mutator type. Group 5 contains the DDE superfamily IS4 transposases and has hits to M. acetivorans (1), M. barkeri (2) and M. mazei (1). Group 6 was not detected by phylogenomic profiling and was found in M. acetivorans (2) and in genomic fragments from deep-sea sediments (GZfos34A6). Triangles indicate transposases identified during proteomic analysis to be expressed in the cell. Best matches to M. burtonii transposases were found by performing BLAST against a customized version of the NCBI nonredundant (nr) database from which all M. burtonii proteins had been eliminated. The sequence alignments are in given Supplementary Figure S5.

Full size image

Due to approaches used for phylogenetic profiling, more distantly related phylogenetic groupings of transposons were detected (Table 2) compared to using BLAST against NCBI nr database to identify best matches (Figure 4). For example, group 2 included matches to Thermoplasma species, and groups 3 and 4 to Sulfolobales (Table 2). However, sequences from other methanogens always produced the best BLAST matches (Figure 4), indicating that M. burtonii transposases did not appear to have been transferred across distant phylogenetic boundaries. Additional signatures of genes shared between M. burtonii and other methanogens include an entire M. hungatei CRISPR region, and a cluster of RNA-directed DNA polymerase genes (for example, possibly viral genes) common to the Methanosarcinales and M. hungatei.

Seven transposases have been found to be expressed in the cell during growth (Figure 4, triangles), indicating that transposases are active and not just markers of past genomic changes (Goodchild et al., 2004b). The three largest groups (1, 2 and 3) were all represented in the expressed list. A further indicator of the activity of transposons is the presence of seven ORFs interrupted by the coding region of a transposase, with five of the ORFs known to be expressed (detected by proteomics) (Table 3). Fragments of five transposases are also present in the genome and are likely to be remnants of transposition events that are gradually being purged from the genome; Mbur_0774 and Mbur_0265 are fragments of a group 2 transposase; Mbur_0771 is a fragment of a group 1 transposase; Mbur_2251 is related to either Mbur_800 or Mbur_2016; and Mbur_0442 is a fragment of a group 6 transposase. The coding region of two of the five transposase fragments have themselves been interrupted by transposable elements; Mbur_0771 possibly by Mbur_0774 (also truncated), and Mbur_2251 by Mbur_2252.

Table 3 Genes interrupted by transposases
Full size table

Transposases [L] were one of the main COG categories distinguishing the psychrophilic M. burtonii from other archaea (see M. burtonii genome overview section). In association with the above genomic and proteomic data, it appears that transposases have an important role in evolving the genome of M. burtonii. Transposase inactivation has previously been linked to cold sensitivity in the deep sea, cold-adapted bacterium Photobacterium profundum SS9 (Lauro et al., 2008), and transposases have been found to be one of the most overrepresented COG categories in metagenomic data of cold, deep (4000 m) ocean water samples (DeLong et al., 2006).

Functional capacity is associated with distinct phylogenetic clusters of genes

The largest phylogenetic cluster (106 proteins) was common to all archaeal species and included ribosomal proteins, chaperonins, translation machinery and 89 hypothetical proteins (Table 2). Other clusters with phylogenetic distribution across many archaeal species include groups of signal transduction proteins, chemotaxis proteins, serine protease inhibitors, O-phosphoseryl-tRNA (Cys) synthetase-linked proteins and DNA-gyrase-linked proteins. The latter two groups are particularly interesting as they link a well-characterized protein function (ER2) with a protein of unknown function in close phylogenetic association, even though the genes are physically separated on the genome. From this, it can be inferred that the DUF39-domain containing protein Mbur_0311 may be involved in production of Cys-tRNACys by phosphoserine, and based on the child–parent relationship of the DUF1119 and peptidase A22 InterPro domains the DUF1119 protein Mbur_1910 may have peptidase function involved in assisting the action of DNA gyrase. In six cases, poorly characterized genes (ER4 and ER5) clustered with well-characterized methanogenesis genes (only present in methanogenic archaea). The phylogenetic profile suggests that all the genes in the clusters are involved in methanogen-specific metabolic processes. A similar approach has been used to characterize hypothetical genes known to be expressed in M. burtonii (Saunders et al., 2005).

General metabolism and transport

Central metabolism appears overall to be similar to other methanogens. Complete pathways for glycolysis (through a modified Embden–Meyerhof pathway) and gluconeogenesis are present. Although no glycogen phosphorylase could be found, other enzymes of starch metabolism are present suggesting that M. burtonii can store carbon as starch or glycogen. Glycogen serves as a polysaccharide storage reserve in many methanogens. The only pentose synthesis pathway present is the reverse ribulose monophosphate pathway, similar to other archaea.

M. burtonii possesses carbon monoxide dehydrogenase/acetyl-coenzyme A synthase (CODH/ACS; see Methanogenesis section) to produce acetyl-CoA from methyl-tetrahydrosarcinapterin and endogenously generated carbon dioxide. Although in a previous report M. burtonii was suggested to also be capable of carbon fixation by RubisCO as it possesses a type III ribulose-1,5-bisphosphate carboxylase/oxygenase (Mbur_2322) (Goodchild et al., 2004b), no identifiable gene for phosphoribulokinase has been found in the completed genome. Further, M. burtonii does not grow autotrophically (Franzmann et al., 1992), and there is no evidence for operation of either the reductive acetyl-CoA pathway or Calvin–Benson–Bassham cycle for carbon fixation. Archaeal (type III) RubisCO was recently demonstrated to be involved in AMP metabolism in the archaeon Thermococcus kodakaraensis (Sato et al., 2007). In this pathway, AMP phosphorylase (Mbur_0255) and ribose-1,5-bisphosphate isomerase (Mbur_1938) supply RubisCO substrate (ribulose-1,5-bisphosphate, RuBP) from AMP and phosphate. RuBP can then be converted by RubisCO to two 3-phosphoglycerate (3-PGA) molecules, which can then enter central carbon metabolism. As homologous genes for all three enzymes are present, we suggest the M. burtonii RubisCO functions in the AMP phosphorylase pathway similar to T. kodakaraensis.

M. burtonii also has ADP-dependent (AMP-forming) sugar kinases that are used by many archaea in glycolysis. Given that 3-PGA can be used for ATP generation, these archaea may use the above pathway under anaerobic conditions to utilize AMP when energy levels are low and/or intracellular AMP levels are high (Sato et al., 2007). Alternatively, AMP can be produced from 5-phosphoribosyl-1-pyrophosphate by adenine phosphoribosyltransferase (Mbur_1435), which also recycles the adenine generated by AMP phosphorylase.

Pyruvate synthase (Mbur_2155 –to Mbur_2158) and pyruvate carboxylase (Mbur_2425 to Mbur_2426) provide a pathway for synthesis of oxaloacetate from acetyl-CoA. Synthesis of 2-oxoglutarate from oxaloacetate and acetyl-CoA is likely to take place through a truncated oxidative tricarboxylic acid (TCA) cycle. The genes involved in the reductive steps of the TCA cycle from oxaloacetate to 2-oxoglutarate are almost completely missing, with only a heterodimeric fumarase (Mbur_0250 to Mbur_0251) present. In the oxidative direction, citrate synthase was initially thought to be missing on the basis of the automated genome annotation. However, a citrate synthase with alternate stereospecificity (known as Re-citrate synthase) has been characterized in a Clostridium species, and an ortholog in M. burtonii was identified during the manual annotation process (Mbur_1075) (Li et al., 2007a). Aconitase (Mbur_0316) is present (see Cold adaptation is associated with specific signatures of genome evolution section), and Mbur_1073 (isocitrate/isopropylmalate dehydrogenase) is a likely candidate to catalyze the remaining step from isocitrate to 2-oxoglutarate.

M. burtonii has biosynthetic pathways for pyrrolysine in addition to the 20 standard amino acids. Akin to Methanosarcina spp., M. burtonii has two pathways for cysteine synthesis: the tRNA-dependent pathway (Sauerwald et al., 2005) and the O-acetylserine pathway. Cysteinyl-tRNA synthetase is absent, so the tRNA-dependent pathway is probably the only way to make cysteine for incorporation into proteins. Pyrrolysyl-tRNA synthetase (Mbur_2086) is located adjacent to the putative gene cassette for pyrrolysine synthesis (Mbur_2083 to Mbur_2085) (Longstaff et al., 2007); thus pyrrolysine is synthesized before being attached to its tRNA. Evidence for the incorporation of pyrrolysine in methyltransferase enzymes has been provided from proteomic studies (Goodchild et al., 2004a).

The way in which M. burtonii fixes nitrogen is unclear from the genome sequence. One nifH and one nifD homolog are present, but they cluster phylogenetically with group IV nitrogenase-related genes that are not expected to be involved in nitrogen fixation (Raymond et al., 2004). To test the ability of M. burtonii to fix nitrogen, we grew cells in defined media with N2 gas as the sole source of nitrogen. Growth was observed after several generations and repeated passaging, although rates of growth and growth yield were lower than in media supplemented with an organic or inorganic nitrogen source (not shown). Preliminary analysis of nifH mRNA levels indicates that gene expression occurs even when cells are grown in complex media (not shown). These data indicate that M. burtonii fixes nitrogen through an as yet, undefined pathway.

Nitrate reductase and nitrate transporters cannot be identified in the genome sequence. The only member of the ammonium transporter family in M. burtonii (Mbur_0933) is situated adjacent to the gene for the regulatory PII protein (GlnB) (Mbur_0934). Mbur_0933 has 64% sequence identity to the characterized transporter Amt-3 from Archaeoglobus fulgidus. However, it is truncated at both the N terminus and the C terminus by approximately 88 amino acids so its functional status is uncertain. Given that ammonia concentrations increase with depth in Ace Lake (Butler et al., 1988; Rankin et al., 1999), it is possible that the ammonia requirements of M. burtonii can be met by inefficient facilitated diffusion of ammonium ion, and/or by the passive diffusion of ammonia. After uptake, M. burtonii may assimilate ammonia using, (1) the two-step glutamine synthetase (Mbur_1975) and glutamate synthase (Mbur_0092) pathway, and/or (2) glutamate dehydrogenase (Mbur_1973). 2-Oxoglutarate serves as the carbon skeleton for both pathways, and can be provided by a partial oxidative TCA cycle (see above). The glutamate synthase of M. burtonii appears to be a ferredoxin-dependent species. All three proteins are expressed in M. burtonii under laboratory growth conditions (Goodchild et al., 2004a, 2004b, 2005).

Methanogenesis

As an obligately methylotrophic methanogen, M. burtonii obtains energy from the oxidation of methyl groups to carbon dioxide and reduction to methane (Figure 5). Methyltransferases required to initiate metabolism by demethylation of methanol, monomethylamine, dimethylamine and trimethylamine and subsequent methylation of their respective corrinoid-binding polypeptides were detected. Multiple copies of the each substrate-specific methyltransferase and corrinoid protein were present. However, the second gene encoding dimethylamine methyltransferase (Mbur_2291) is disrupted by a putative transposon and is likely to be nonfunctional. Three copies of the methyl CoM methyltransferase for transfer of the methyl group from the methanol-specific corrinoid, and two copies of the common methyl CoM methyltransferase for methyl transfer from the methylated amine-specific corrinoid proteins were identified. This is similar to observations in the closely related methansarcinal genomes, which also have multiple copies of initiating methyltransferases and cognate corrinoid proteins (Galagan et al., 2002; Deppenmeier et al., 2002; Maeder et al., 2006). However, in the methanosarcinal genomes only single copies of each CoM methyltransferase are present. Methyltransferases specific for methylthiols, which are found in some methanosarcinal species, were not present in M. burtonii (Tallant and Krzycki, 1997; Tallant et al., 2001). During methylotrophic growth, methyl CoM is disproportionated 3:1 by oxidation to carbon dioxide and reduction to methane, respectively. All of the genes encoding for enzymes required for methyl reduction to methane and oxidation of methyl CoM through the reversal of the carbon dioxide reduction pathway are present in the M. burtonii genome (Figure 5).

Figure 5
figure 5

Methanogenesis and biomass production in M. burtonii. Color coding indicates proteins involved in methanogenesis from methanol (red), monomethylamine (MMA; green), dimethylamine (DMA; blue) and trimethylamine (TMA; purple). Abbreviations are as follows: CH3OH, methanol; (CH3)3N, TMA; (CH3)2NH, DMA; CH3NH2, MMA; CH3-CP, methyl-corrinoid protein; CoM, coenzyme M; CoB, coenzyme B; MePh, oxidized methanophenazine; MePhH2, reduced methanophenazine; F420, coenzyme F420; F420H2, reduced coenzyme F420; CH3-H4SPT, methyl-tetrahydrosarcinapterin; CH2=H4SPT, methylene-tetrahydrosarcinapterin; CH≡H4SPT, methenyl-tetrahydrosarcinapterin; CHO-H4SPT, formyl-tetrahydrosarcinapterin; CHO-MFR, formyl-methanofuran; Fd, ferredoxin; MtaB and MtaC, methanol methyltransferase and corrinoid protein; MttB and MttC, TMA methyltransferase and corrinoid protein; MtbB and MtbC, DMA methyltransferase and corrinoid protein; MtmB and MtmC, MMA methyltransferase and corrinoid protein; Mcr, methyl-CoM reductase; MtaA, methanol:CoM methylase; MtbA, methylamine:CoM methylase; Mcr, methyl-CoM reductase; Hdr, heterodisulfide reductase; Fpo, F420H2 dehydrogenase; Mtr, methyl-H4SPT:CoM methyltransferase; Mer, methylene-H4SPT reductase; Mtd, methylene-H4SPT dehydrogenase; Mch, methenyl-H4SPT cyclohydrolase; Ftr, formyl-methanofuran:H4SPT formyltransferase; Fmd or Fwd, formyl-methanofuran dehydrogenase; CODH/ACS, carbon monoxide dehydrogenase/acetyl-CoA synthase.

Full size image

Methanosarcina spp. are capable of growth by all four known methanogenic pathways: hydrogen reduction of carbon dioxide, hydrogen reduction of methylated compounds, fermentation of acetate and dismutaton of methylotrophic compounds. The genome of M. burtonii provides the first direct insight into the minimal genes required for obligate methylotrophy by a methanogen. Growth with hydrogen requires three hydrogenases: Ech, which catalyzes ferredoxin reduction by hydrogen for subsequent reduction of carbon dioxide to the formyl level; Frh/Fre, which reduce coenzyme F420 for reduction of methenyl and methylene groups; Vho, which catalyzes the reduction of the coenzyme CoM reducing electron carrier methanphenazine (Meuer et al., 2002). Except for one protein with similarity to a coenzyme F420-reducing hydrogenase β-subunit (Mbur_2261), genes encoding these enzymes were not detected in M. burtonii; consistent with the inability of this species to grow with hydrogen (Franzmann et al., 1992). During growth, F420 dehydrogenase (Fpo) could reduce methanphenazine from reduced F420 generated during the oxidation steps of methyl and methylene groups in the methanogenic pathway. Reduced ferredoxin generated by oxidation of formylmethanofuran to carbon dioxide could be providing reducing potential for biosynthesis. M. acetivorans, which grows with acetate or methylotrophic substrates but does not grow with hydrogen, also lacks the Ech hydrogenase and although it has frh and vho genes they are not expressed (Guss et al., 2005). This latter observation combined with the lack of hydrogenase genes in M. burtonii confirms that hydrogenases are not required for the methylotrophic pathway.

Hydrogen reduction of methylated compounds requires the specific methyltransferases and the reductive methanogenic pathway after methanopterin. However, this pathway also requires electrons from oxidation of hydrogenase to reduce methyl groups, which are not present in M. burtonii. Aceticlastic methanogenesis requires CODH/ACS to catalyze the methylation of methanopterin by dismutation of acetyl CoA. In contrast to two gene copies in aceticlastic Methanosarcina spp., M. burtonii genome has only one copy of the CODH/ACS operon and lacks genes encoding acetate kinase and phophotransacetylase for generating acetyl CoA from acetate. This is consistent with other nonaceticlastic methanogens that likely use CODH/ACS for carbon assimilation from carbon dioxide. Overall, the limited catabolic capabilities of M. burtonii are consistent with the small genome size relative to Methanosarcina spp.; a characteristic shared with the genomes of other methanogens with a single catabolic pathway.

In one respect the ability of Methanosarcina spp. to grow with all known methanogenic substrates provides them with the ability to adapt their catabolism to changes in substrate availability. In contrast M. burtonii is obligately methylotrophic and restricted to methanol and methylamines as exogenous carbon sources, which raises the question of how this is an advantage for the survival of this species as a specialist. One phenomenon observed for Methanosarcina spp. is increased hydrogen production on acetate and methylotrophic substrates when grown in the presence of a sulfate-reducing species (Phelps et al., 1985). This also results in a reduction in methane yield per mole of substrate as more substrate is oxidized and reducing equivalents are diverted to hydrogen. This shift in the pathway would effectively reduce the energy yield for the methanogen and enable sulfate-reducing bacteria to utilize energy from the noncompetitive methylotrophic substrate through the reverse methanogenic pathway of the methanogen. The lack of hydrogenases in M. burtonii allows this specialist to utilize methylotrophic substrates without diversion of electrons to sulfate reducers and prevent the indirect utilization of these substrates by the latter. This may have provided a growth advantage to obligate methylotrophs such as M. burtonii in Ace Lake when the lake was (in the past) a sulfate-rich environment (see Linking the evolution of the M. burtonii genome to its ecological niche section).

Signal transduction and adaptive potential

The signal transduction system of M. burtonii is similar to that of its Methanosarcina spp. relatives, despite M. burtonii having a much smaller genome size. M. burtonii is motile by means of a single flagellum (Franzmann et al., 1992) and encodes a chemotaxis system that includes a single methyl-accepting chemotaxis protein (Mbur_0356), its methylase and demethylase (Mbur_0360, Mbur_0399), chemotaxis histidine kinase CheA (Mbur_0361) and chemotaxis response regulator CheY (Mbur_0359). In addition to CheA, M. burtonii encodes 29 other two-component histidine kinases, of which 10 contain a (predicted) extracytoplasmic sensor domain and are probably involved in environmental sensing (Galperin, 2006). The remaining histidine kinases are intracellular; most of them contain PAS domains and can be involved in sensing of the cellular level of oxygen, carbon monoxide, NO and other molecules. This could be particularly important for M. burtonii because this strict anaerobe might have to cope with the higher solubility of oxygen at low temperatures. Four of the M. burtonii histidine kinases have an N-terminal receiver (REC) domain similar to the proteins described recently in the haloarchaea (Galperin, 2006), and are likely to participate in complex phosphorelay signal transduction cascades.

Of the 14 response regulators encoded in the M. burtonii genome, eight consist of a single stand-alone REC domain and two more have an REC-PAS domain architecture. All these response regulators are likely involved in protein–protein interactions. Similar to other archaea, M. burtonii encodes no response regulators of the OmpR, NarL, NtrC, PrrA or LytR families. However, it encodes a response regulator (Mbur_0695) with a DNA-binding output domain of the GlpR type, which forms a typical operon-like structure with an environmental sensor histidine kinase Mbur_0694. This regulator is more abundant in cells grown at low temperature, and may form a temperature responsive regulatory system with the histidine kinase (Goodchild et al., 2004a). Two more response regulators (Mbur_0878 and Mbur_2185) contain previously uncharacterized output domains, one unique for M. burtonii and the other found only in Methanosarcina spp. In addition, M. burtonii encodes an adenylate cyclase (Mbur_1935) and five predicted Ser/Thr protein kinases, one of the ABC1/AarF family, two of the RIO family, one unusual protein kinase and one Kae1-associated serine threonine kinase. Similar to other archaea, the role of cAMP, presumably synthesized by the adenylate cyclase, remains unknown, as are the cellular targets of the (predicted) protein kinases.

The large number of two-component regulatory systems in M. burtonii with limited similarity to known functional proteins may reflect a requirement for complex internal regulation (Ashby, 2006), and M. burtonii has a high ‘IQ’ (a measure of the adaptive potential of an organism) compared to many methanogens (Galperin, 2005). The majority of signal transduction genes (35 of 45 proteins) have a functional evidence rating of ER4 (indicating a lack of experimentally characterized full-length homologs). In addition, five ORFs listed as pseudogenes contain histidine kinase domains, and at least two of these have been interrupted by transposons and appear to be nonfunctional. Given the general lack of experimentation on two component regulatory systems in archaea and the specific importance of signal transduction systems to M. burtonii (see Cold adaptation is associated with specific signatures of genome evolution section), there is a compelling reason to experimentally characterize the sensing and response mechanisms of these phosphorelay networks.

Lipid biosynthesis

An important pathway associated with cold adaptation in M. burtonii is isoprenoid lipid biosynthesis (Nichols et al., 2004). Synthesis of unsaturated lipids increases during growth at low temperature and is thought to maintain the fluidity, and hence functionality of the membrane in the cold. The digeranylgeranylglyceryl phosphate (DGGGP) synthase that was not present in the draft genome was identified as Mbur_1679 in the closed genome (Figure 6). Phosphomevalonate kinase was also not previously identified, and to date the gene has only been identified in the Sulfolobales (Boucher, 2007). Recently M. jannaschii was hypothesized to use a modified mevalonate pathway, with the two steps required for conversion of mevalonate phosphate to isopentenyl diphosphate catalyzed by a novel (although not yet experimentally verified) mevalonate phosphate decarboxylase, followed by isopentenyl phosphate kinase (Grochowski et al., 2006). The corresponding enzymes in M. burtonii are Mbur_2394 and Mbur_2396, respectively. This mevalonate pathway differs from the steps previously proposed for M. burtonii, which requires diphosphomevalonate kinase and diphosphomevalonate decarboxylase (Nichols et al., 2004). In the M. burtonii genome, Mbur_2394 and Mbur_2396 lie within a putative gene cluster associated with lipid biosynthesis, including mevalonate kinase (Mbur_2395), isopentenyl-diphosphate δ-isomerase (Mbur_2397), and a bifunctional enzyme comprising farnesyl pyrophosphate synthetase and geranyltranstransferase (Mbur_2399). Other important members of the lipid biosythesis pathway include an experimentally characterized geranylgeranyl reductase responsible for conversion of DGGGP to archaetidic acid (Mbur_1077) (Murakami et al., 2007). This enzyme is thought to have an important role in facilitating the regulation of unsaturation levels by performing selective saturation (similar to plants) (Nichols et al., 2004). Other genes characterized on the basis of COG groupings involved in lipid formation include acetyl-CoA synthetases (3 proteins), acyl-CoA synthetase, pyruvate decarboxylase (α- and β-subunits located adjacent to each other in the genome), phosphatidyltransferases and cytidyltransferases.

Figure 6
figure 6

Lipid biosynthesis in M. burtonii. Figure based on analysis of the closed genome sequence (including new gene numbers), incorporating information from previous studies (Nichols et al., 2004; Grochowski et al., 2006; Boucher, 2007). Abbreviations are as follows: CoA, coenzyme A; HMG-CoA, 3-hydroxy-3-methylglutaryl-CoA; P, phosphate; IPP, isopentenyl diphosphate; DMAPP, dimethylallyl diphosphate; GPP, geranyl diphosphate; FPP, farnesyl diphosphate; GGPP, geranylgeranyl diphosphate; GGGP, geranylgeranylglyceryl phosphate; DGGGP, digeranylgeranylglyceryl phosphate; CDP, cytidine diphosphoglycerol; G-1-P, glycerol-1-phosphate; DHAP, dihydroxyacetone phosphate.

Full size image

A large genomic commitment to polysaccharide biosynthesis

Saccharide or polysaccharide moieties are important modifiers of the isoprenoid lipid membrane (Jahn et al., 2004) and S-layer (Karcher et al., 1993) of archaea, and can be secreted as EPS (Parolis et al., 1996; Paramonov et al., 1998). In M. burtonii, genes involved in polysaccharide biosynthesis comprise at least 3.3% of protein coding genes in the genome. In contrast to the 81 M. burtonii genes, the mesophilic M. acetivorans only contains approximately 30 polysaccharide biosynthesis genes representing 0.6% of its genome (Galagan et al., 2002). Four operon-like clusters of polysaccharide biosynthesis genes (containing 39, 16, 11 and 10 genes) and five single glycosyl-transferase genes are distributed around the M. burtonii genome (Figure 1). In addition, five unique hypothetical proteins are encoded in proximity to the polysaccharide biosynthesis gene clusters. Five homologs (Mbur_0724, Mbur_0725, Mbur_0726, Mbur_1581 and Mbur_2023) of a glycosyl transferase (COG0438), and four homologs (Mbur_0727, Mbur_1593, Mbur_2028 and Mbur_2225) of a membrane protein involved in the export of O-antigen and teichoic acid (COG2244) are present in four main regions of the genome where they are adjacent to a number of other genes involved in polysaccharide biosynthesis, such as sugar- and N-acetylglucosamine epimerases and sugar dehydratases (Figure 1). The proteins represented by COG0438 and COG2244 are homologs to P. profundum SS9 proteins, PBPRA2678 and PBPRA2684, respectively. Mutation of the P. profundum SS9 genes produces a cold-sensitive phenotype (Lauro et al., 2008; Ferguson et al. personal communication). EPS, including polysaccharides appear to have a role in cell aggregation and biofilm formation of archaea at low temperatures (Reid et al., 2006). Higher levels of EPS are produced by M. burtonii growing at low (compared to high) temperatures (Reid et al., 2006), and approximately half of the M. burtonii polysaccharide biosynthesis genes are known to be expressed, having been shown to be abundant proteins in proteomic analyses (Goodchild et al., 2004a, 2004b, 2005; Saunders et al., 2005). Collectively these data strongly suggest that polysaccharide biosynthesis has an important role in the cold adaptation of M. burtonii.

Linking the evolution of the M. burtonii genome to its ecological niche

Ace Lake is located on Long Peninsula in the Vestfold Hills, East Antarctica (Rankin et al., 1999). During the early Holocene (13 000–9400 years ago), Ace Lake was an aerobic freshwater system, becoming a seasonally isolated marine basin (9400–9000 years ago) and subsequently open marine basin (9400–5700 years ago) (Rankin et al., 1999; Coolen et al., 2004; Cromer et al., 2005). Approximately 5100 years ago Ace Lake became a permanently isolated saline lake, and developed meromixis, with an active methane cycle in existence for last ∼3000 years. The microbiota in the lake is clearly marine derived. However, isotopic data indicate that all the water now present in the lake is of meteoric origin and that the lake has been mixed for considerable periods before its present stable meromixis. Nutrient input presently into the lake is very limited and it is a cold (average ∼0 °C), oligotrophic system, although inorganic carbon levels are sufficient to lead to carbon dioxide efflux (Rankin et al., 1999). The concentrations of most trace metals are higher than the ocean and are not limiting. Gradients of nutrients occur throughout the lake and concentrations can vary significantly; for example, manganese concentration varies from 78 to 1460 nM within 5 m in the anaerobic zone and this concentration is orders of magnitudes higher than the 5.1 nM of the ocean. The anoxic waters (12–25 m depth) support stable increasing gradients of salt (up to 4.3%), methane (saturated below 20 m, ∼5 mM) and H2S (up to 8 mM), and decreasing gradients of sulfate (essentially depleted below 19 m) (Rankin et al., 1999).

Close relatives of M. burtonii have been identified in several cold ocean water locations in south and north polar marine waters and the deep sea, including Methanococcoides alaskense that has 99.8% 16S rRNA identity (Li et al., 1999; Purdy et al., 2003; Singh et al., 2005; Cavicchioli, 2006). Cold, deep-sea environments in the Atlantic Ocean have been found rich in extracellular DNA, and may promote opportunities for DNA exchange (Dell’Anno and Danovaro, 2005). The adaptation of M. burtonii to the cold will have occurred over millennia in the cold marine environment. However, the large environmental changes that have taken place in Ace Lake since the Holocene are likely to have provided strong selection pressure for ecotypes with genomic variation better suited to the new environment. In this regard it would be valuable to sequence the genome of M. alaskense to assess this. The specific capacity of M. burtonii to evolve through genome plasticity (including nucleotide skew, HGT and transposase activity) appears to have placed it in a strong position to not only adapt to the cold (for example, polysaccharide synthesis, lipid composition, amino-acid composition), but to the particular biotic and abiotic conditions that have changed in the lake throughout its recent several thousand year history (for example, central metabolism, novel ABC transporters, coenzyme F420-dependent sulfite reductase).