The genome of Salinibacter ruber: Convergence and gene exchange among hyperhalophilic bacteria and archaea

Results

Genome and Proteome Characteristics. The genome of strain M31^T DSM 13855, the type strain of S. ruber, comprises a 3,551,823-bp chromosome of high G+C content (66.29%) and a 35,505-bp plasmid (57.9% G+C), containing 2,934 and 33 ORFs, respectively. Table 1, which is published as supporting information on the PNAS web site, shows a comparison of general genomic features with those of Chlorobium tepidum TLS (Salinibacter's closest sequenced relative), Halobacterium sp. NRC-1 (an archaeal halophile), and Bacillus halodurans (a halotolerant bacterium). Previous analysis of halophilic archaeal genomes (such as Halobacterium sp. NRC-1, Haloarcula marismortui and Haloferax volcanii) has suggested a bipartite organization, with a high overall G+C content but smaller replicons and chromosmal islands of lower G+C (12, 13), the low G+C regions bearing most of the insertion sequences (IS) and phage-related elements (12). In Salinibacter, chromosomal low G+C islands are also enriched in such elements (Fig. 5, which is published as supporting information on the PNAS web site). The first of the low G+C islands is located 250 kb from the inferred origin of replication and contains 15 transposases and five prophage components (including three glycosyl-transferases). The second, spanning over 55 kb, contains 12 transposases, several prophage-related ORFs and, unexpectedly, a number of ORFs involved in O chain polysaccharide (or capsular polysaccharide) synthesis. The third low G+C island (39 kb) contains the only restriction-modification system found in this organism. The Salinibacter plasmid codes for 19 hypothetical and conserved hypothetical genes (57.6% of the plasmid ORFs) but also contains several ORFs involved in DNA metabolism, replication, and recombination, and a gene involved in UV protection (SRU_p0002). It also encodes an IS5 family transposon element not found in the chromosome.

One striking feature of Salinibacter biology is its high intracellular potassium concentration, an adaptation to living in hypersaline conditions previously observed only in haloarchaea. Consistent with this feature, Salinibacter proteins, like those of haloarchaea, have a high content of acidic residues, mainly aspartate, and a low content of basic residues, particularly lysine (Fig. 1). This observation agrees with the previous estimation of the amino acid composition from the bulk protein (4). The genome sequence allows a global analysis of the predicted isoelectric points (pI) of the proteome. A biomodal distribution of pI values differentiating cytoplasmic and integral membrane proteins, characteristic of most proteomes (14, 15), is shown by Bacteroides fragilis and C. tepidum but not by the haloarchaea (Fig. 1). Salinibacter is intermediate in character, with a median pI value (5.2) nearer those of haloarchaea (Halobacterium sp. NRC-1, 4.6; Haloarcula marismortui, 4.6) than those of B. fragilis (7.0) or C. tepidum (7.0). In some cases (see below), this increase in pI may reflect orthologous replacement of genes encoding the less acidic proteins of the Bacteroides/Chlorobi group by LGT from haloarchaea. But because the majority of Salinibacter's genes have their closest relative outside the haloarchaea (Fig. 2), residue-by-residue selection during adaptation of its ancestors to hypersaline conditions must be more often the explanation for such extensive convergence at the molecular level.

Phylogeny and Lateral Gene Transfer. Phylogenetic analysis of its 16S rRNA sequence places Salinibacter within the Bacteroides/Chlorobi group, with its closest relative R. marinus (2). The whole genome methods described by Gophna and colleagues (16) show C. tepidum and B. fragilis to be its closest relatives among bacteria with sequenced genomes. To assess the nature and number of potential LGTs into this genome, we used a “competitive matching” query (17). In this analysis, any ORF that has a match only to genomes outside the Bacteroides/Chlorobi group or a match at least 0.05 normalized blast score units better outside this group was counted as a potential LGT; 1,470 ORFs showed such discordant similarities by this blast-based method. The 10 most frequently matched genomes outside Bacteroides/Chlorobi are displayed in Fig. 2 (black bars).

blast analyses can mislead us for many reasons, and surely the number of within-group best matches will rise (apparent transfers fall) when more genome sequences appear from members of Bacteroides/Chlorobi. Phylogenetic reconstruction may be a more reliable method for identifying LGTs, although applicable only when adequate taxa are available and a rooting is possible. For each of these 10 genomes, we examined maximum likelihood trees generated by phyml (18) (with an assumed Bacteria/Archaea rooting) in which Salinibacter was the deepest branching member of a clade containing that genome. The total number of such candidate trees examined for specific support of transfer to or from Salinibacter and the indicated genome is shown by the gray bars in Fig. 2. In the pie charts, the fraction of these trees actually supporting LGT from the indicated genome to Salinibacter is shown in blue, those supporting LGT from Salinibacter to that genome in chartreuse, those supporting LGT of uncertain (either) direction in green, and those not supporting an LGT with this candidate genome in red. Phylogeny, like blast, identified haloarchaea as the most frequent donors or recipients in LGTs involving Salinibacter, although the total number of apparent transfers between Salinibacter and haloarchaea appears to be modest. (And aside from the haloarchaea and Rhodopirellula baltica, many LGTs indicated by blast are not supported by trees.) Where appropriate in the following sections of this report, we highlight individual instances in which LGT between these groups nevertheless has likely been a key factor in adaptation to the harsh saline environment.

Physiology. Salinibacter has a full complement of fermentative components and genes involved in transport and degradation of organic compounds. A chitinase, two cellulases, one amylase, a pectinase, and several proteases and lipases were among the polymer-degrading enzymes encoded on the chromosome. Additionally, the plasmid encodes a 1,4-β-cellobiosidase homolog. Genes related to the transport and metabolism of glycerol and glycine betaine (the most common compatible solutes found in more moderate halophiles and halotolerant species) were also present. The broad degradative abilities fit with the expected complexity of the organic pool present in the crystallizer as a result of the lysis of biomass produced at lower salinities. Contrary to a previous suggestion (19), glycolysis appears to take place through an Embden-Meyerhoff pathway and not a modified Entner-Doudoroff pathway.

Like the haloarchaea, Salinibacter has all of the genes for a complete tricarboxylic acid cycle and a cytochrome c-containing respiratory chain. It possesses a succinate dehydrogenase gene cluster (SRU_0484-0487) very similar to Actinobacteria and a second succinate dehydrogenase flavoprotein subunit, sdhA (SRU_2444), most similar to sdhA2 from H. marismortui. Similarly, comparison of respiratory chain proteins in Salinibacter indicates there are two clusters of cytochrome c oxidase subunit I and II genes, the first with coxA1 and coxB1 (SRU_2099 and SRU_2100) and the second with coxA2 and coxB2 (SRU_0314 and 0313). Phylogenetic analyses demonstrate coxA1 and coxB1 genes, which are also next to the genes for subunits III (SRU_2098) and IV (SRU_2097), are related to those of R. marinus, whereas coxB2 and coxA2 are clearly most closely related to their haloarchaeal homologs (Fig. 6, which is published as supporting information on the PNAS web site). This second gene cluster may have been imported from the haloarchaea, a shared adaptation to the microoxic conditions often associated with hypersalinity, perhaps functioning in a novel respiratory electron transfer pathway. In support of this notion, a nosZDF gene cluster and an oxygen-sensitive Cpr/Fnr-type transcription regulator (20) are found near to coxA2 and coxB2 in Salinibacter. The nos cluster encodes a nitrous oxide reductase similar to that found in a variety of proteobacteria (21, 22) and in H. marismortui. Phylogenetic analyses of NosZ (SRU_0308) and NosD (SRU_0310) demonstrate high similarity between Salinibacter and H. marismortui; however, the exact relationship remains unresolved (Fig. 7, which is published as supporting information on the PNAS web site).

Salinibacter may have other unique characteristics with respect to microoxic adaptation. Unlike haloarchaea but like R. marinus (23), it possesses components of a cbb3-type cytochrome oxidase (ccoO and ccoN) (SRU_0323 and SRU_0322). Cytochrome c oxidases of the cbb3 type have a very high affinity for O₂ and allow respiration to continue under low-O₂ levels (24). Both ccoO and ccoN were acquired likely by LGT from a β-proteobacterium. Strangely, Salinibacter does not possess a ccoP gene, thought to be an essential subunit of cbb3-type cytochrome oxidase (25). However, next to ccoON is a gene encoding a cytochrome c protein (SRU_0324), which also appears to have been transferred from a β-proteobacterium and may substitute for CcoP function in Salinibacter.

Haloadaptation. A cluster of 19 genes (Fig. 3) includes K⁺ uptake/efflux systems and cationic amino acid transporters of crucial importance to a hyperhalophilic lifestyle. This “hypersalinity island” is mosaic in nature, apparently pieced together from a variety of bacterial and archaeal sources. More than one-half of the genes are most similar to their haloarchaeal homologs, including the cationic amino acid transporters, one trkH gene, and three trkA homologs. The Trk system is responsible for the uptake of K⁺, where TrkH is the membrane bound translocating subunit and TrkA is a cytoplasmic membrane surface protein that binds NAD⁺ (26). The existence of multiple trkA genes in Salinibacter suggests complex regulation of the Trk system, a feature likely shared with haloarchaea. This complexity is increased by the presence of yet an additional TrkAH system in the hypersalinity island that is most similar to that found in Firmicutes (Fig. 3).

The recruitment of several nonhaloarchaeal hypersalinity genes may have involved an island-encoded IS1 transposase. In Escherichia coli, IS1 has two overlapping ORFs, insA and insB, and translational frameshift results in the production of InsAB, a functional IS1 transposase (27). There are four insAB-like genes in Salinibacter and a similar expression mechanism may exist in this bacterium. Genes with significant similarity to the Salinibacter insAB have been found only in cyanobacteria, Parachlamydia, and members of two archaeal orders, Methanosarcinales and Sulfolobales. This patchy phylogenomic distribution of the IS1 transposase is mirrored in the distribution of the KefB K⁺ efflux system proteins (28) and the Na-K-Cl cotransport proteins (29) present in the island, supporting the notion that these genes have passed as a single unit through an archaeal or cyanobacterial “host” before settling into the Salinibacter genome. The most striking examples of this observation are the Na-K-Cl cotransporters, which are found throughout the eukaryotes and function as integral membrane transport proteins (29). Outside the eukaryotes, we could identify orthologs in the genomes of only five prokaryotes, all of which also possessed a Salinibacter-like IS1 transposase. The origin of the prokaryotic Na-K-Cl cotransporters remains enigmatic, but it is possible that they have been exchanged among very distantly related organisms, including Salinibacter. In addition, these findings suggest the convergence on an aerobic hyperhalophilic lifestyle between haloarchaea and Salinibacter was mediated, in part, by interdomain lateral gene transfer.

Rhodopsins and Retinal. The discovery of the proton pump bacteriorhodopsin in the haloarchaea Halobacterium salinarum in the early 1970s launched a new fruitful field of bioenergetics and biophysics, and this protein and its relatives were seen as a defining invention of the archaea (30). More recently, rhodopsin-based photobiology has been found in other groups of prokaryotes (31, 32) and in unicellular eukaryotes (33, 34). The possibility that Salinibacter might also have rhodopsin genes, derived from the haloarchaea with which it lives, was one of the initial motivations behind the present work. Indeed, like Halobacterium sp. NRC-1 and Haloarcula marismortui, and unlike any other characterized bacterium, Salinibacter contains four rhodopsin genes (Fig. 4A ).

Three of the Salinibacter rhodopsins group with haloarchaeal rhodopsins in phylogenetic reconstructions and might be inferred, from the nature of neighboring genes in the Salinibacter genome, to have similar functions. Salinibacter SRU_2780 forms a basal member of the haloarchaeal halorhodopsin clade and, like them, is immediately upstream of an oxidoreductase gene (possibly a regulator of pump activity) (Fig. 4E ). A Na⁺/H⁺ antiporter is located only three genes away in Salinibacter and a Na⁺/substrate cotransporter is adjacent. In both archaeal genomes, trkA (a potassium symporter gene) is found nearby. This observation, together with relatively high sequence similarity, supports the same function, as an inward-directed chloride pump, for the Salinibacter protein.

Salinibacter has two presumptive sensory rhodopsin (SR) genes, 83 kb apart on the genome and distantly related to each other and to the haloarchaeal SRs, specifically SRI. In Halobacterium, two SRs (SRI and SRII) interact to produce a color-sensitive phototactic behavior. SRI is only synthesized under low oxygen tension (as are the proton pump BR and the chloride pump HR), mediating the attraction to orange light that drives the ion pumps (35). SRII is produced at high oxygen tensions when the bioenergetics of the cell are driven by the respiratory chain and mediates a photophobic response to blue light (35). In haloarchaea, the signaling function of both SRs depends on their tight association with transducers that block their potential ion transport activity and transform them into photosensors (36). Until very recently (see below), there has been no information about Salinibacter photobiology and the function of the two genes homologous to SRs still can be inferred only from the genomic context. Nevertheless, there seems little doubt that SRU_2579 is a photosensor, because the next ORF, 86 nucleotides downstream, is an Htr1 transducer, as in Halobacterium sp. NRC-1 (Fig. 4D ). The second putative SR in Salinibacter (SRU_2511) is also tightly linked to nearby signal transduction genes (Fig. 4C ), pointing to a photosensory function and potentially complex photobehavior in Salinibacter. Notably, although there is no doubt of the relatively close relationships among the sensory rhodopsins of Salinibacter and the haloarchaea, the transducers that are tightly linked at the genetic (and presumably functional) level are not closely related to those of haloarchaea (Fig. 8, which is published as supporting information on the PNAS web site). Their closest relatives outside the Salinibacter genome are transducers of other bacteria. Thus, physiological convergence between Salinibacter and haloarchaea likely has been effected through independent gene recruitment processes.

Downstream and tightly linked to the first Salinibacter SR gene discussed above (Fig. 4D ) (SRU_2579), there is a typically bacterial flagellar cluster that might be coregulated with it: SR function in phototaxis would be useful only when cells are actively motile; members of the Bacteroides/Chlorobi group typically lack flagella. However, a polar flagellum has been observed in Salinibacter's closest relative R. marinus (37). Phylogenetic analyses of the genes within this flagellar gene cluster indicate a mosaic structure, with many genes showing a close affinity to the δ/ε-proteobacteria. Perhaps the flagellum was a recent acquisition by Salinibacter/Rhodothermus. With respect to motility, a general, although not universal, character of the Bacteroides group is gliding motility. Studies in Flavobacterium johnsoniae have identified a suite of genes required for the gliding motility phenotype (gldABDFGH) (38). Salinibacter possesses homologs of the gldA, gldF, and gldG genes (SRU_1249-SRU_1251), which are thought to form an ATP-binding cassette transporter in F. johnsoniae, but it lacks the critical lipoprotein components (gldB, gldD, and glDH) that are required for gliding.

The fourth Salinibacter rhodopsin gene (SRU_1500) appears to be most closely related to those found in cyanobacteria and forms a clade with organisms that encode diverse rhodopsin genes, which include the proteorhodopsins found among uncultured marine bacteria and known to act (at least when expressed heterologously in E. coli) as a proton pump (31). A recent study by Balashov et al. (39) confirmed the proton pumping activity of the Salinibacter protein (presumably encoded by SRU_1500) and identified a novel light-harvesting complex (xanthorhodopsin) that not only includes the chromophore retinal but utilizes a carotenoid antenna (salinixanthin) and enables the rhodopsin to absorb light across a greater absorbance spectrum than is possible with retinal alone. Salinixanthin, as noted in the Introduction, is responsible for the red color of Salinibacter, comprises nearly 100% of its carotenoid, and is structurally related to the major carotenoid pigment found in R. marinus (6). The gene encoding xanthorhodopsin is linked to two ORFs (crtY and crtO) (SRU_1501 and SRU_1502) that code for proteins responsible for important steps in both carotenoid and retinal synthesis (Fig. 4B ). Lycopene β-cyclase (CrtY) is required for the formation of β-carotene, a precursor to both retinal and subsequent carotenoid production, whereas β-carotene ketolase (CrtO) is likely responsible for an important step in salinixanthin biosynthesis (40, 41).

Until recently, the manner by which retinal was produced in bacteria was unknown. A metagenomic study of proteorhodopsin-containing BAC clones from marine waters (42) indicate that these organisms may be synthesizing retinal by a mechanism first identified in haloarchaea, which have been shown to use bacteriorhodopsin-related protein (Brp) and its paralog, bacteriorhodopsin-related protein like homolog (Blh) to produce retinal from symmetrical cleavage of β-carotene (43). A divergent homolog of Blh (20% amino acid identity with the archaeal version) was found linked to proteorhodopsin on BAC clones (42), whereas in Salinibacter, homologs of both brp and blh are present in the genome (unlinked to rhodopsin genes) and exhibit close relation to those found in haloarchaea rather than marine bacteria.

Discussion

Salinibacter has adapted to hypersalinity in three general ways. First, Salinibacter must have modified the sequences of many of its proteins. Whole and partial proteome studies of haloarchaea show that these hyperhalophilic organisms have accommodated high internal ionic strength by replacing neutral amino acids with acidic ones (44). This situation must also be the case for Salinibacter because an acidic proteome is not a property of its relatives among the Bacteroides/Chlorobi group (Fig. 2). Although some genes imported from haloarchaea may have produced proteins already adapted in this way, these comprise only a fraction of Salinibacter's genes. A comparative study of orthologous proteins in Salinibacter and haloarchaea and their closest nonhalophilic relatives should tell us much about convergent evolution at the level of protein structure.

Second, Salinibacter exhibits many cases of convergence at the level of physiology, in which proteins from different sources or with different original functions have been recruited to create a complex adaptation to hypersaline life that parallels a structure, pathway or behavior already known from haloarchaea. The mix-and-match of genes of haloarchaeal and bacterial origin in the postulated phototaxis system will likely provide a good example of analogous complex collective functions served by nonhomologous component parts. The recent discovery that xanthorhodopsin (Salinibacter's proteorhodopsin-like SRU_1500) functions in a light-harvesting complex with salinixanthin similarly provides a striking analogy with chlorophyll-based light-harvesting systems (39).

Third, some adaptations common among halophilic organisms have been passed between them by LGT. The rhodopsins shared by Salinibacter and the haloarchaea may be the best case in point. Overall, the specific functionally important residues in the Salinibacter rhodopsins match those of the haloarchaeal proteins to which the rhodopsins are homologous, and the two SR-transducer homologs both appear to be sensory rhodopsin photoreceptor transducers, like their haloarchaeal Htrl homologs, rather than chemoreceptor taxis proteins (John Spudich, personal communication). Unless we imagine that the last universal common ancestor had a full complement of rhodopsins, the presence of several identifiable classes of this protein in Salinibacter and haloarchaea is best explained by multiple events of LGT. Although proteorhodopsin-like genes have been described in several bacterial phyla (45), we note that Salinibacter is the only bacterium in which genes that cluster specifically with haloarchaeal rhodopsin genes are known. It might seem natural to assume that Salinibacter has derived these genes from archaea, by LGT, but we cannot be certain. Salinibacter's two sensory rhodopsins appear to have derived from one of the two haloarchaeal sensory rhodopsin paralogs (as if after the duplication in that lineage), but its halorhodopsin diverged before the diversification of haloarchaeal forms. Many of the other “haloarchaeal genes” found in Salinibacter could just as easily be “Salinibacterial genes” introduced by LGT into haloarchaeal genomes. Indeed, the respiratory chain and associated functions in haloarchaea are widely believed to have been of bacterial origin (46, 47), and Salinibacter's ancestors could have been one source. Bacterial and archaeal cells have likely shared the saltern habitat for millennia, with ample chance to exchange genes. The notion of a “habitat genome” (or a pool of genes useful for adaptation under a specific set of environmental constraints) is appealing. Such a gene pool could be analogous (in an evolutionary time scale) to the gene pool in a metazoan (or plant) zygote from which different tissues extract the required components (48).