Coping with cold: The genome of the versatile marine Antarctica bacterium Pseudoalteromonas haloplanktis TAC125

Genome organization

As in many marine γ-proteobacteria, the P. haloplanktis TAC125 genome is made of two chromosomes (Table 1; Supplemental Table 2). The replication origin of chromosome (chr) I maps near dnaA (McLean et al. 1998; Lobry and Louarn 2003) in a region that is highly conserved in γ-proteobacteria (Fig. 1). However, in remarkable contrast with the genomes of the vibrios (Okada et al. 2005), the second chromosome does not display a standard GC skew (Supplemental Fig. 1). The pattern observed is likely to be caused by unidirectional replication. To our knowledge, this is the first time that such a system would be uncovered in an authentic bacterial chromosome. This is supported by the signature of R1 plasmid replication (del Solar et al. 1998): the tus and repA-like genes, the repA and dnaA boxes, and the parA and parB genes (Fig. 1). In addition, kisB and kidB coding for a typical plasmid maintenance system have also been found in chrII. We chose the start at the centre of a TATATA palindrome near the genes coding for the partition system. The G+C content and gene density of chrII match that of chrI (Supplemental Table 2). It contains the essential genes hisS and gcpE, in addition to a series of genes ubiquitous in γ-proteobacteria (Supplemental Table 3). A third of chrII genes have orthologs in Escherichia coli. Remarkably, the whole metabolism of histidine is coded in chrII, in a highly conserved gene cluster (Fig. 1). Nineteen percent of the P. haloplanktis chrII genes show high similarities with plasmid-encoded genes, further suggesting that this replicon was a plasmid recruited to become a chromosome encoding essential genes (Fig. 1; Supplemental Table 2).

Genes around the origin of replication in chromosome display a high level of synteny with genes of other known proteobacteria. chrI codes for nine rDNA clusters (23S, 5S, and 16S RNAs, one operon has two copies of the 5S RNA gene), a large number compared with that found in most sequenced γ-proteobacteria (Ussery et al. 2004). The genome sequence of P. haloplanktis TAC125 shows some variability in the rRNA genes of ∼1% (interestingly, most variations correspond to compensating mutations in regions coding for double stranded RNA) (data not shown). However, this probably does not influence phylogenies (Cilia et al. 1996), and the rRNA sequences are in line with the established phylogeny, placing TAC125 near vibrios and Shewanella (Ivanova et al. 2004). In the same way, the number of tRNA genes is quite high (106 genes), a feature in common with that in vibrios and in Photobacterium profundum (Table 1). These genes are organized in long runs of repeated sequences. The longest contains 19 tRNA genes in a row, seven of which coding for an identical tRNA^glu (TTC anticodon), suggesting a slipped mispair mechanism of expansion in situations of rapid growth. Genomes with an origin of replication display a protein-coding gene distribution bias, with most essential genes located in the leading replication strand (Rocha and Danchin 2003). Interestingly, while the number of CDSs located in the leading strand of the chromosome is 61%, 72% of the tRNA genes are located in that strand. Because the speed of transcription/translation must be limited at low temperature, the large number of rRNA and tRNA genes may participate in the adaptation, allowing fast growth of the organism in the cold. TAC125 is similar to other bacteria in terms of number of tandem repeats. Several genes relevant to adaptation to cold conditions are clustered together in approximate repeats (Supplemental Fig. 2): genes coding for cold-shock proteins, nine paralogs of cspA, as in E. coli (four in chrII, three of them clustered together), as well as genes coding for a class of putative short secreted proteins that could bind calcium, next to a divalent metal exporter system, most likely used in calcium export. Calcium is known to be involved in cold adaptation and formation of exopolysaccharides (EPS) in bacteria (Kierek and Watnick 2003; Dominguez 2004). The duplication in chrII of the ast operon may be involved in adaptation to cold and high osmolarity.

Comparative genomics

Of the 3488 identified protein coding genes (CDSs) (Table 1), a biological function, based on a classification scheme adapted from Riley (Riley 1993; Fang et al. 2005), has been assigned for >65.6% (36.7% with a final assignment and 28.9% with a putative role assignment). More than 63% of the P. haloplanktis CDSs are similar to Shewanella oneidensis genes (Heidelberg et al. 2002), of which 47% are found in synteny groups, making the comparison with this aquatic organism particularly revealing (Fig. 2A). These mesophilic marine bacteria, together with Vibrio vulnificus, share with P. haloplanktis several sodium-type flagellar proteins (mot genes), sodium-dependent transporters, gene clusters of type IV pilus and flagellin proteins, and some iron ABC transporter and tonB system-dependent transport proteins (Fig. 2B; Supplemental Table 4C). A putative cold-shock RNA methyltransferase (PSHAb0516) and a cold-shock regulated carbon starvation protein A (PSHAb0210) have also been found in S. oneidensis and V. vulnificus. Other P. haloplanktis TonB-dependent receptors and calmodulin-like proteins have counterparts in S. oneidensis only, while V. vulnificus shares with P. haloplanktis several integrases/transposases, putative potassium channel proteins and metabolite exporters (Supplemental Table 4A,B). Gene content comparisons have also been performed with the two closest psychrophilic γ-proteobacteria, Idiomarina loihiensis and Photobacterium profundum (Fig. 2B): apart from the set of genes shared also with the marine bacteria, many TonB-dependent receptors are specific to P. haloplanktis and I. loihiensis, and two sodium-dependent carbohydrate transporters (a permease and a transporter) have been found in P. profundum only (Supplemental Table 4D,E). The set of genes common to P. haloplanktis and these two psychrophilic bacteria contain two copper resistance protein genes (copA and copB genes in chrII) (Supplemental Table 4F). These genes belong to a synteny group made of six genes with I. loihiensis and, interestingly, with the megaplasmid of Ralstonia solanacearum. They are also found in other distant cold marine bacteria such as Psychrobacter sp. and Silicibacter pomeroyi (in its megaplasmid; Supplemental Table 5).Two synteny groups (a potassium efflux system and an urease operon) are found with S. pomeroyi only. A total of 133 P. haloplanktis genes (73 hypothetical), including several insertion sequences and genes coding for TonB-dependent receptors, have not been found in the genome of the cold-adapted marine bacteria used for comparison (Supplemental Table 6). In addition, a prophage region (50 kb long, between genes PSHAa1505 and PSHAa1558) is specific to P. haloplanktis. We also identified one specific region in chrI coding for several calcium-dependent proteins, as well as a specific gene in chrII that may regulate cell volume and resistance to cold conditions (PSHAb0555).

General features of the proteome

Global properties of the proteins at the level of individual amino acids and motifs integrate all kinds of selection pressure associated to adaptation to cold. The pattern of amino acid distribution in γ-proteobacteria from different biotopes (Supplemental Fig. 3) displays an overall trend similar in the various genomes of interest, with leucine (L) being most abundant, while tryptophan (W), cysteine (C), histidine (H), and to a lesser extent methionine (M) are used infrequently. The proteome of the thermophilic genomes looks significantly different from that of the mesophilic and psychrophilic counterparts (strong avoidance of glutamine [Q] in thermophilic species, preference for alanine [A] in mesophilic and psychrophilic species, except in Oceanobacillus iheyensis). The amino acid distribution in mesophilic and psychrophilic species display a few noteworthy differences specifically relevant to growth in the cold (in particular in the relative abundance of N and Q) (Supplemental Fig. 3). Using correspondence analysis (CA) (Benzécri 1984) coupled to dynamic clustering (Delorme and Henaut 1988), to identify subtle differences in this cluster of related objects, five classes with close amino acid composition were found (Fig. 3): (1) integral inner membrane proteins (∼12%); (2) proteins involved in the metabolism of small molecules (25%); (3) associated to information transfer pathways (21.5%); (4) associated to the outer membrane or secreted (21.5%); and (5) with unknown functions, or likely to be of phage origin (20%). This biological consistency demonstrates that a relationship exists between amino acid composition and the role of the protein inside the cell. The two first biases have been previously identified: They are driven by the membrane compartmentalization of some proteins and by the G+C-content of codons (Pascal et al. 2005). In contrast, the bias scattering proteins along the third factorial axis was unexpected. It discriminated proteins in P. haloplanktis TAC125 according to their asparagine (N) content. This small hydrophilic uncharged amino acid carries an amide group that is often thermolabile (Zhou et al. 2000; Stratton et al. 2001; Weintraub and Manson 2004). Analysis of this remarkable N-driven bias was further submitted to CA using the pool of two psychrophilic (Desulfotalea psychrophila and P. haloplanktis), two mesophilic (E. coli K-12 and Bacillus subtilis) and two thermophilic (Aquifex aeolicus and Thermotoga maritima) bacteria. An N-driven bias was observed in proteins from the psychrophiles, differentiating them from their mesophilic and thermophilic counterparts (Supplemental Fig. 4). The proteins responsible for the bias belong to the following structures or processes: motility of the organism, cell wall, outer membrane, transport, sensor activity, adaptation to atypical conditions, and secretion. Some proteins of DNA metabolism (replication, packaging, segregation, restriction/modification, proof-reading, and repair) and RNA metabolism are also present. Asparagine residues often undergo deamidating cyclisation, a process extremely sensitive to temperature (Daniel et al. 1996). The P. haloplanktis TAC125 proteome is enriched in N residues compared with counterparts that prefer growth at higher temperature, making it an organism of choice for foreign protein production when deamidation ought to be put to a minimum (Weintraub and Manson 2004).

RNA folding and metabolism

The genome of P. haloplanktis TAC125 contains 19 genes presumably coding for known RNA binding proteins or RNA chaperones. The most unexpected feature of the genome is the prominent absence of a RNA/nucleoid-associated cold-shock gene ubiquitous in γ-proteobacteria, hns. Analysis in silico failed to uncover a H-NS–like protein gene even when using the highly relaxed comparison criteria that uncovered counterparts in bacteria phylogenetically distant from E. coli (Tendeng and Bertin 2003). In contrast, the genes coding for all other nucleoid-associated proteins such as HU, IHF, FIS, and Hfq are present in the genome. The existence of a complete set of genes (including hns) was also observed in other marine and cold-adapted bacteria, including S. oneidensis (Heidelberg et al. 2002) and D. psychrophila (Rabus et al. 2004). In vivo complementation of a hns defect of E. coli at room temperature led us to repeatedly isolate clones coding for the counterpart of csrA as an efficient complementation of the hns defect of E. coli. The lack of the hns-encoding gene in the psychrophilic P. haloplanktis genome shows that H-NS is not sufficient to promote growth at low temperature and that its role is connected to that of the regulatory protein/RNA complex CsrA/CsrB (Fig. 4). The importance of control of RNA folding and degradation at low temperature is visible in the presence of many RNA helicases (three copies of rhlE, two in chrI and one in chrII, and possibly a fourth one, PSHAa0641, and two copies of srmB, instead of one in E. coli). Interestingly, in contrast with the situation reported for Oleispira antarctica (Ferrer et al. 2004) the groES groEL genes from TAC125 did not permit growth of E. coli at low temperature. Other factors are therefore important for cold adaptation in the present organism.

Several RNA motifs indicating the presence of RNA-coding genes have been found in chrI using the Rfam databank (see Methods): a tmRNA at position 2,231,297 bp, the RNA component of RNase P at position 2,674,641 bp, the t44 RNA (of unknown function), and three riboswitch structures (Nudler and Mironov 2004), a lysine riboswitch just in front of dapA gene (at position 189,092 bp); the RFN element, located at 1283,710 near ribB gene; and the “ubiquitous” TPP riboswitch (THI element) at 496,922 bp (upstream of the thiC gene).

Response to oxygen and reactive oxygen species

The solubility of gasses increases rapidly at low temperature. This is the case of dioxygen, which is a very reactive molecule. We expected that the proteome would comprise a vast arsenal of enzymes active against H₂O₂ and superoxide. This was the more so because the organism is indeed remarkably resistant to H₂O₂ (Supplemental Fig. 5). Surprisingly, we found only the gene counterpart coding for the iron superoxide dismutase (sodB) and only one clear catalase (katB) located in chrII, with a possible paralog in chrI (PSHAa1737). Furthermore, while the oxygen responding OxyR control is present, the SoxR regulation is absent.

This unexpected finding was explained when we discovered that P. haloplanktis TAC125 lacks a series of activities that result in reactive oxygen species (ROS) production. Despite the availability of molybdate in sea water (Hille 2002), P. haloplanktis TAC125 entirely lacks molybdopterin metabolism: not only are the biosynthetic and transport gene absent but genes coding for enzymes using the cofactor are also missing (e.g., TMAO reductase, xanthine oxidase, biotin sulfoxide reductase, or the novel oxido-reductase YedY) (Loschi et al. 2004). All related organisms such as the vibrios or Shewanella have molydopterin metabolism, as do almost all Bacteria and Eukarya. The cells, however, must cope with increased oxygen solubility and inevitable interaction with reduced iron, leading to the deleterious Fenton reaction and ROS. A way for the cells to protect their metabolism against those is to use dioxygen directly. This is seen in the large number of putative dioxygenases present in both chromosomes (Table 2). To this list one should add the fatty acids desaturases (see below) that also play a role to increase membrane fluidity at low temperature.

Furthermore, the protective role of methionine against ROS is enhanced by the existence of two MsrA proteins, instead of one as is usual, together with a fused MsrA-MsrB protein, a rare situation, found in Nitrosomonas europaea. Interestingly, one MsrA copy, PSHAa2274, is present in all related genomes, while the second copy, PSHAa1583, not present in P. profundum, exists in Shewanella and vibrios, and two copies in D. psychrophila that develop preferentially at low temperature. A further sign that the bacteria cope with the specific problems posed by ROS is the number of proteins involved in scavenging chemical groups affected: peroxiredoxins such as alkyl hydroperoxide reductase AhpC, AhpCB, thiol-specific cytoplasmic peroxidase BcpA; thiol-specific periplasmic peroxidase Tpx, and their coupled flavoproteins: AhpF, TrxB, PSHAa0892, Gor. All these activities would protect against the inevitable consequences of the Fenton reaction, and the needed sulfur metabolism genes, involving glutathione and S-adenosylmethionine, are all present. Furthermore, a control system that would limit the concentration of copper in the cell (Cu²⁺ is particularly reactive toward oxygen) involves at least two transport systems to expel copper out of the cell (CusA, CopAB, in chrI and PSHAb0008, PSHAb0009, CopB, CopA, and PSHAb0325 in chrII), a periplasmic disulfide reductase (DipZ), and a periplasmic putative laccase (PcoA) that could chelate copper ions while also acting as a dioxygenase.

Finally, an interesting gene cluster involved in fatty acid metabolism PSHAa0894–0910 is absent from other γ-proteobacteria, but is largely present in Mycobacterium tuberculosis (where it may have a protective role against ROS, contributing to virulence). In short, P. haloplanktis TAC125 is remarkably well adapted to protection against ROS under cold conditions, a feature that could be very useful for expression of foreign proteins in the cold.

Metabolic features

Marine bacteria are facing a medium generally unbalanced in terms of carbon, nitrogen, and phosphorous supply (Moran et al. 2004) but are not depleted in sulfur sources. Strain TAC125 is adapted to fast growth, suggesting that it regularly encounters a fairly rich medium (this is probably due to its propensity to make a water/air biofilm [see below], allowing it to live in region full of plankton debris). Excess of several easily metabolized carbon sources present simultaneously is unlikely, making catabolite repression the exception rather than the rule. Indeed, P. haloplanktis TAC125 is lacking the cAMP-CAP complex that regulates carbon availability in related organisms such as vibrios and Shewanella. Furthermore, in contrast to many γ-proteobacteria (including vibrios), it does not possess a phosphoenolpyruvate-dependent phosphotransferase system for the transport and first metabolic step of carbohydrate degradation. This accounts for its lack of growth on glucose, likely to be phosphorylated by glucokinase (PSHAa1364). When oxygen is present at a high level, the presence of glucose-phosphate isomerase might drive the Embden Meyerhof pathway and subsequently activate NADPH-dependent aldose reductase (PSHAa2392). This would affect the polyol pathway, lowering the amount of NADPH available for the reduction of oxidized glutathione by glutathione reductase. Indeed, growth is inexistent on glucose unless tyrosine is supplemented to the medium (data not shown). The counterpart of PSHAa2392 has been implicated in a process controlling tyrosine bradytrophy in E. coli (Timms and Bridges 1998).

An essential step for biomass construction is formation of pyruvate, which, because of the absence of the PTS, must go through an alternative pathway starting with pyruvate kinase. In contrast to E. coli, with two such enzymes, and vibrios, with three, TAC125 possesses only one pyruvate kinase. Interestingly, it is homologous to the cold-adapted PykA enzyme of E. coli. This is further in line with gluconate as a preferred carbon source, providing pyruvate directly through the Entner-Doudoroff pathway (Edd, Eda), which also provides the level of NADPH needed for protection against oxygen toxicity. As in vibrios, phosphoenolpyruvate synthase (ppsA) and phosphoenolpyruvate carboxylase (ppc) genes are located in chrII. However, ppsA is located in chrI in P. profundum.

TAC125 grows in minimal medium under anaerobic conditions (data not shown), in line with the presence of the fnr gene, while the putative aspartate ammonia lyase PSHAa0048 would provide the needed fumarate electron acceptor.

In contrast, the metabolism of nitrogen appears to be highly similar to that in phylogenetically related organisms and be controlled by a phosphorylation cascade involving PtsP (a homolog of the PTS enzyme PtsI) that phosphorylates PtsO and the regulator PtsN, controlling all sigma54-dependent operons. Arginine catabolism could provide a direct source of ammonia under nitrogen-limiting conditions, through the AST pathway, while providing metabolites for adapation to cold (Schneider et al. 1998). The organism can metabolize N-acetyl-glucosamine, a carbon and nitrogen source ubiquitously present in marine environment (Riemann and Azam 2002). In the same way, phosphate input in metabolism is controlled by the counterparts of PhoB, PhoR, and PhoU, with several putative transport systems, including one of high affinity.

Most coenzymes can be synthesized in the organism except coenzyme B12 and, as discussed, molybdopterin. There does not appear to exist a selenium metabolism, in line with the high reactivity of that atom toward oxygen. Other marine bacteria (including the closest one, S. oneidensis) do have a selenium metabolism.

Growth, yield, and adaptation to salt

A remarkable feature of TAC125 is that, when provided with sufficient nutrients and aeration, it grows to very high density under laboratory settings, even at 0°C. The very high growth yield indicates that respiration must be particularly efficient in this bacterium. The cells are well adapted to salt, and although they can grow in low osmolarity media, optimal growth is between 1.5% and 3.5% NaCl. We looked for systems that would account for controlling osmolarity in the cell. The common trehalose system does not seem to be present. In contrast, chrII harbors two copies of a choline dehydrogenase (PSHAb0261 and PSHAb0418) for synthesis of glycine betaine, an extremely efficient osmoprotectant (Felitsky et al. 2004). Along the same line, GOGAT glutamate synthase is coded in that same chromosome, allowing synthesis of glutamate that, as potassium glutamate, is the most common response of bacteria to increased osmolarity (Table 2; Lee and Gralla 2004).

All autonomous organisms have at least one pathway to degrade S-adenosylhomocysteine (AdoHcy). The MtnN(Pfs)/LuxS pathway leads to synthesis of the quorum sensing (QS) effector autoinducer-2 (AI-2). Xanthomonadales and Pseudomonadales aside, all γ-proteobacteria sequenced use that pathway. LuxS, responsible for the last enzymatic step of AI-2 synthesis is present in bacteria closest to TAC125, and AI-2 produced by heterologous organisms triggers luminescence in reporter strains. The mtnN gene is present in TAC125. We failed, however, to identify the luxS gene. We looked further for other genes involved in AdoHcy degradation and recycling to homoserine lactones or autoinducer CAI-1. Among other systems, involving quinolones, cyclic dipeptides, or indole, none appears to be present, and in assays, using Photorhabdus luminescens as reporter TAC125 supernatant did not trigger luminescence. This does not, however, exclude the presence of less well known QS systems: gene PSHAa0159 codes for a multidomain putative aconitate hydratase that may use aconitate as an iron-sulfur cluster-dependent signal in stationary phase (Kiley and Beinert 2003). Furthermore, TAC125 possesses several enoyl-CoA hydratase-like genes that may be involved in synthesis of a diffusible signaling factor as in plant pathogenic γ-proteobacteria (Barber et al. 1997).

Membranes, motility, biofilms, and secretion

P. haloplanktis TAC125 deals with the membrane fluidity challenge at low temperature by lipid desaturases, which simultaneously protect against dioxygen (Table 2). Two clusters of genes, absent in the closest genomes, may be involved in the degradation of steroids or hopanoids, membrane rigidifying molecules present in the environment of heterotrophic bacteria. Protein export from the cytoplasm is similar to that of proteobacteria, with the long form of SecE. Type II secretion is functional (GSP proteins are present) as is the TAT export system. In contrast, type III secretion is absent. TAC125 possesses two gene clusters in chrI and one in chrII for the biosynthesis of type IV pili and of curli, respectively. Furthermore, chrII encodes elements of the new pathway involved in secretion of a specific amylase composed of a signal peptide, the mature enzyme, and a long C-terminal propeptide without foldase function or action on amylase activity (Claverie et al. 2003). Amylase secretion required two accessory proteins, PSHAb0130 (possibly an outer membrane associated protein) and PSHAb0132, coding for a conserved secreted protein present in several phytopathogenic γ-proteobacteria. Interestingly, the neighboring glutamate synthesis genes were essential to allow amylase secretion when reconstructed in E. coli.

P. haloplanktis has several genes and operons that may play an important role in colonization of both biotic and abiotic surfaces. In particular, up to 16 genes involved in the synthesis of mannose-sensitive agglutinin are located on chrI. These genes have been recently demonstrated to facilitate adhesion to the chitin surface in V. cholerae (Meibom et al. 2004). When investigating the formation of biofilms on solid surfaces (many genes are compatible with synthesis of a biofilm such as production of EPS), we did not observe synthesis of a strong biofilm. In contrast, the air–water interface was rapidly occupied by a dense layer of compact cells (Supplemental Fig. 6), suggesting that, for this organism, this is the normal way to concentrate cells and occupy a biotope. The formation of such a biofilm compatible with life in water and scavenging of organic particles that concentrate in the foam of waves has been recently demonstrated in V. parahaemolyticus (Enos-Berlage et al. 2005). Genes for the synthesis of polar flagellum are present, and these appendages are visible under the microscope. In salty water the cells are highly motile. However, in contrast to the situation with several vibrios, pseudomonads, and related bacteria, the cells have a reduced motility in low salt media, while they still grow well under such conditions (Supplemental Fig. 7). In minimal medium, the strain grew in a large range of NaCl concentration (0% to up 11% NaCl). In rich media, however, the growth of the strain is remarkably sensitive to the presence of salt. At 20°C, no growth occurs in the absence of NaCl. In remarkable contrast, slow but significant growth is observed at 4°C in the absence of salt, suggesting some adaptation to ice or melting ice water.

Conclusions

P. haloplanktis TAC125 has found many unexpected solutions to cope with cold. Not only does it grow fast under such conditions, but it displays remarkable resistance to ROS. Moreover, as seen in silico with its proteome composition, it provides a way to resist to the aging features involving asparagine cyclisation and deamidation. This makes this bacterium not only a model for the study of adaptation to cold marine conditions but also a promising tool for biotechnology production of proteins.

Bacterial strains, growth media, and assays

The P. haloplanktis strain TAC125 is deposited and available at the Institut Pasteur Collection (CIP). Escherichia coli strains used in this work are strains MG1655, FB8, and its hns defective counterpart BE1410 (Laurent-Winter et al. 1997). TAC125 bacteria were grown in TYP rich media: 16 g/L yeast extract (DIFCO) and 16 g/L bacto-tryptone (DIFCO) supplemented with NaCl as required. For growth on specific carbon sources (0.4%), the following minimum medium was used: 10 g/L Na₂HPO₄, 3 g/L KH₂PO₄, 1 g/L K₂SO₄, 20 g/L NaCl, 0.4 g/L MgSO₄ · 7H₂O, 0.1 g/L CaCl₂, 0.018 g/L FeSO₄ · 7H₂O, and 3 g/L NH₄Cl (pH 7). Motility assay of E. coli cells with plasmids carrying P. haloplanktis genes was performed on semisolid plates with 0.5% NaCl 16 h. For CsrA-dependent storage of glycogen analysis, wild-type MG1655 E. coli colonies growing on Kornberg medium with 1% glucose and 100 μg/mL ampicillin with (pDIA616) and without (pcDNA2.1) csrA were assayed twice for glycogen accumulation. The plates were stained with iodine vapor after 48 h growth at 25°C as in Liu and Romeo (1997). Oxidative stress adaptation was assayed as follows: after overnight growth in TYP medium with 1% NaCl, bacteria are washed in fresh TYP medium with 1% NaCl and diluted to OD₆₀₀ = 0.15. When the OD₆₀₀ reached 0.5, the culture was separated into four equal parts and then exposed to 0, 15, 20, and 25 mM of H₂O₂. The OD₆₀₀ was measured at various times during growth. Experiments were performed twice at 37°C for MG1655 E. coli strain and at 15°C for P. haloplanktis TAC125. For electron microscopy cells were stained with 0.1% (v/v) osmium tetroxide prepared in water: 20 μL cells were deposited onto 300-mesh copper grids coated with Formvar (Electron Microscopy Sciences). Excess sample was removed by using Whatman 3MM paper. Bacteria were examined at 75 keV under a Hitachi H600 transmission electron microscope. Images' acquisition was performed with a CCD Advantage HR Hamamatsu camera and the AMT 542 software (Advanced Microscopy Techniques).

Genome sequencing, assembly, and annotation

DNA was isolated from P. haloplanktis TAC125 grown in rich medium supplemented with 20 g/L NaCl. The complete genome sequence was determined by using the whole shotgun method (10× coverage, using two plasmid libraries and one BAC library to order contigs). Finishing was performed by PCR amplification from contigs' extremities. All rDNA clusters were sequenced individually. After a first round of annotation, regions of lower quality as well as regions with putative frameshifts were resequenced from PCR amplification of the dubious regions.

A first set of potential coding sequences (CDSs) was identified by using the AMIGene software (Annotation of MIcrobial Genes) (Bocs et al. 2003) trained with a set of CDSs >500 bp from the genomic sequence. Three gene models, computed from the three gene classes identified by codon usage analysis (see below), were then subsequently used together in the core of AMIGene with minimum CDSs length set to 60 bp. This second set of putative genes (made of 3488 CDSs) was submitted to functional annotation: exhaustive BLAST searches against the UniProt databank were performed to determine significant homology. Protein motifs and domains were documented by using the InterPro databank. In parallel, genes coding for enzymes were classified by using the PRIAM software (Claudel-Renard et al. 2003). PRO-TMHMM and PRODIV-TMHMM were used to identify transmembrane domains (Viklund and Elofsson 2004), and SignalP 3.0 was used to predict signal peptide regions (Bendtsen et al. 2004). Finally, tRNAs were identified by using tRNAscan-SE (Lowe and Eddy 1997).

Sequence data for comparative analyses were obtained from the National Center for Biotechnology Information (NCBI) databank. Putative orthologs between P. haloplanktis and the 228 other genomes were defined as genes showing a minimum of 30% identity and a ratio of 0.8 of the length of the smallest protein. Orthology relations were strengthened by synteny detection (i.e., conservation of the chromosomal colocalization between pairs of orthologous genes from different genomes) using the Syntonizer software, in which all possible kinds of chromosomal rearrangements are allowed (inversion, insertion/deletion). A “gap” parameter, representing the maximum number of consecutive genes not involved in a synteny group was set to five genes. Species-specific genes were identified as having no ortholog in the compared species. Specific regions are defined by at least two consecutive specific genes. Insertion of genes with similarities in the compared species was allowed. A gap parameter, representing the maximum number of consecutive genes with similarities, was set to two genes.

All the data (i.e., syntactic and functional annotations, and results of comparative analysis) were stored in a relational database (using the MySQL DBMS software). Each predicted gene was assigned a unique identifier prefixed with “PSHAa” for chrI and “PSHAb” for chrII. Manual validation of the automatic annotation was performed by using the Web interface MaGe (Magnifying Genomes), which allows graphic visualization of the P. haloplanktis annotations enhanced by a synchronized representation of synteny groups in other genomes chosen for comparisons. Translation start codons were corrected based on protein homology, proximity of ribosome-binding site, and relative position to predicted signal peptides when present. To this purpose, we used the Artemis sequence Viewer connected to the MaGe system. The P. haloplanktis nucleotide sequence and annotation data have been deposited at EMBL databank under accession number CR954246 and CR954247. The PseudoList database is constructed by using the MySQL DBMS, as previously described (Fang et al. 2005).

Correspondence analysis (Benzécri 1984) was used to analyze the data table with the relative synonymous codon usage values of each annotated gene as well as the table of distribution of amino acids in the proteome of P. haloplanktis. Clustering into consistent classes used a second method (dynamic clouds) (Delorme and Henaut 1988) that automatically clusters the objects located close to one another.