Ten GEIs were identified on the chromosome of R. litoralis (Figure 1, tagged with Arabic numerals) making up ~665 kb (14.8%). In R. denitrificans, in contrast, only ~300 kb (7.1%) were identified as genomic islands. The excess of 365 kb of alien genetic material in R. litoralis corresponds to the larger chromosome size (~372 kb, see also Table 1) of the organism. Thus, the additional genetic material of R. litoralis was most likely acquired via horizontal gene transfer.

Typically, GEIs contain a G+C content and a Codon Adaptation Index (CAI) different from the average 17 . Furthermore, transposases within the islands and tRNAs flanking the GEIs are indicators for translocation processes 17 18 . Many of the genes located in GEIs are of unknown function and several do not exhibit significant similarities to other genes in the databases (orphan genes). These orphan genes are thought to be phage-derived genetic material 19 . Although no complete prophages are present in the genome of R. litoralis, in some of the islands phage-like genes were identified, e.g. in island 8 three putative phage tail proteins are located (RLO149_c037250 - RLO149_c037270).

In other GEIs, however, genes were identified that were probably derived from other bacterial species. Frequently, amino acid and carbohydrate transport and metabolism genes are present in the GEIs, providing R. litoralis with additional abilities for substrate utilization. For example island 5 contains genes for rhamnose transport and degradation (RLO149_c023060 - RLO149_c023140) that have been described in Rhizobia 20 . Genes involved in nitrogen metabolism were identified in islands 1, 6 and 9 including different amidases (RLO149_c009550, RLO149_c040080, RLO149_c040170, RLO149_c028370), a second uncommon urease gene cluster (RLO149_c028310 - RLO149_c028360) and assimilatory nitrite and nitrate reductases (RLO149_c039830 - RLO149_c039850). Island 4 contains a carbon monoxide dehydrogenase encoding gene (CODH, RLO149_c017450 - RLO149_c017470), which is not present in R. denitrificans. The carbon monoxide dehydrogenase was believed to be common in the Roseobacter clade, as all members contain the corresponding genes 21 22 23 . However, recently it was shown that only a small proportion of Roseobacter clade members, among those R. litoralis, are able to oxidise carbon monoxide and that a wide variety lacks an essential subunit of the CODH-complex 24 . R. denitrificans is not able to oxidise carbon monoxide, but instead, has genes coding for a nitrate reductase that enables the organism to reduce nitrate under anaerobic growth conditions. Island 8 contains genes for coenzyme PQQ biosynthesis (RLO149_c036920 - RLO149_c036960) that are also located in a GEI of R. denitrificans. In island 10, genes for antigen biosynthesis were identified, e.g. UDP-N-acetylglucosamine 2-epimerase WecB (RLO149_c044390) and UDP-4-amino-4-deoxy-L-arabinose--oxoglutarate aminotransferase ArnB (RLO149_c044340). The latter is also similar to the perosamine synthetase from Brucella melitensis, with GDP-perosamine being part of the O-antigen of the organism 25 . Antigens are polysaccharides and lipopolysaccharides that define the structure of the bacterial cell surface. Genes important for cell envelope biosynthesis are often found in islands of environmental bacteria 26 . The cell surface structure is important for biofilm formation and host association of the organisms and structural alteration can provide niche adaption and phage defence 19 26 27 28 .

Several species-specific genes, of which the majority is associated with heavy metal resistance (Table 2), are located on plasmid pRLO149_83. Therefore, the two Roseobacter species were compared with respect to zinc and copper tolerance. R. litoralis showed a higher tolerance of zinc, whereas R. denitrificans showed a higher copper tolerance. R. litoralis could grow without impairment up to 0.08 mM of zinc, but was inhibited in its growth in the presence of low copper concentrations (0.04 mM). In contrast, R. denitrificans could not grow with 0.02 mM zinc added to the medium, but was able to grow with 0.1 mM of copper. The higher zinc tolerance of R. litoralis could be due to the Zn-Cpx-type ATPase and/or the putative cobalt-zinc-cadmium resistance protein CzcD (Table 2), a member of the cation diffusion efflux (CDF) family 29 . Substrates of CDF proteins can be various cations 29 , but mainly Zn²⁺-transporting CDFs such as ZitB from Escherichia coli 30 are also known.

Most of the other putative heavy metal resistance genes on plasmid pRLO149_83 have weak similarities to known copper and silver efflux proteins (Table 2). But since no higher copper tolerance of R. litoralis compared to R. denitrificans was observed, the efflux systems might be involved in transport of other cations. Two other members of the Roseobacter clade, Dinoroseobacter shibae DFL-12 and Roseovarius nubinhibens ISM, have orthologous heavy metal resistance genes on their plasmids (Figure 2B).

Plasmids are important mobile genetic elements and therefore often contain recently acquired genetic material. Thus, the occurrence of species-specific and also alien genes on two of the plasmids (Figure 2B and 2C) was not surprising. Plasmid pRLO149_94, however, is an exception as no alien genes or genomic islands were identified on the plasmid. Nearly the entire genetic information of pRLO149_94 was found on the chromosome of R. denitrificans, with approximately 78 kb being syntenic. Also in R. denitrificans the area was not identified as GEI. Only nine ORFs on the plasmid are not present in the genome of R. denitrificans. The genes for the photosynthetic apparatus comprise ~45 kb and are part of the syntenic area with the plasmid replication genes located amidst the photosynthesis genes in R. litoralis (Figure 2A). The remaining 33 kb that are syntenic in both organisms are located upstream of the photosynthesis genes in R. denitrificans and encode other functions. The genes downstream of the syntenic region to plasmid pRLO149_94 in R. denitrificans are also present in R. litoralis and are located approximately 100 kb upstream of dnaA. At the position on the chromosome of R. litoralis that corresponds to the position of the first gene associated with the photosynthesis apparatus of R. denitrificans (idi, RD1_0147) several transposases are encoded (Figure 1, upstream of island 10), suggesting this region to be a genomic hot spot. These findings suggest a translocation of the genetic material from the chromosome to the plasmid. This is further supported by the fact that photosynthesis genes are usually located on the chromosomes of AAnPs 5 and are thought to be rather vertically than horizontally acquired genetic material in Roseobacter clade bacteria 23 .

To identify the genes specific for the genus Roseobacter, the genomes of both species were compared to 38 genomes of other Roseobacter clade bacteria. The results of the comparison are shown for two representatives of each phylogenetic subgroup of the Roseobacter clade outlined by Newton et al. 22 in Figure 1. Six hypervariable regions (HVR I-VI), areas of low conservation in the Roseobacter-clade, were found adjacent to the genomic islands predicted on the chromosome of R. litoralis (Figure 1). The HVRs of R. litoralis are characterized by a mosaic-like structure, with regions conserved in all Roseobacter clade bacteria alternating with genus-unique but also species-unique genes. Frequently, tRNAs were found flanking the HVRs, but not many transposases were present inside the areas indicating that the regions are permanently anchored in the chromosome 31 . The G+C-content and also the codon-adaptation index vary inside the HVRs.

Of special interest is HVR I as many genes identified in this area seem to be connected to the relation of R. litoralis to the algal host. For example, several genes for the degradation and transport of algal osmolytes like taurine (RLO149_c007790 - RLO149_c007880) and sarcosine (RLO149_c006600 - RLO149_c006630) were identified in HVR I. Furthermore, the genes for vitamin B12 biosynthesis (RLO149_c006160 - RLO149_c006260) are present. Vitamin B12 was shown to be important for the symbiosis of D. shibae with its dinoflagellate host 32 . A degradation pathway of erythritol is also located in HVR I (RLO149_c008260 - RLO149_c008410). The pathway is only present in R. litoralis and R. denitrificans and is known from Rhizobia 33 34 . In Rhizobium leguminosarum erythritol catabolism is important for competitiveness of the organism in the nodulation of pea plants 34 . Thus, erythritol catabolism might also be associated with the algal host relation of the Roseobacter species. The mosaic-like structure of HVR I resembles the symbiosis islands of Rhizobia 18 35 . Two different areas can be identified on the chromosome of R. denitrificans that correspond to HVR I of R. litoralis. In the alignment of the chromosomes of the two species, the HVRs of the two organisms form an X-like structure (Figure 3) which is probably due to the tendency of genes in closely related organisms to be located in the same distance from the origin 36 .

Uncharacterized sugar and amino acid transporters are frequently found in the HVRs, e.g. in HVRs II and V. Genes for glucoside (RLO149_c021710 - RLO149_c021770) and galactose/arabinose (RLO149_c021930 - RLO149_c021980) transport and degradation as well as an arsenite resistance system (RLO149_c022030 - RLO149_c022040) are located in HVR III. The genes for sulfur oxidation (RLO149_c031760 - RLO149_c031920) and the denitrification genes (RLO149_c031300 - RLO149_c031570) were identified in HVR IV. Due to the lack of the nitrate reductase in R. litoralis the organism is not able to reduce nitrate; however, genes encoding all other enzymes required for denitrification are present. Thus, it is possible that the organism is able to reduce nitrite to molecular nitrogen under anoxic conditions. HVR V contains genes for formaldehyde degradation that are more common in the mixotrophic than in the heterotrophic group of Roseobacter clade members 22 .

A rfb-gene cluster essential for the development of O-antigens, i.e. lipopolysaccharides of the outer membranes of Gram-negative bacteria 37 , was identified on plasmid pRLO149_63 of R. litoralis. Many of the Roseobacter clade bacteria have rfb-genes in their genomes and in about half of those the genes are located on plasmids. In R. denitrificans the rfb-gene cluster is encoded on a plasmid of 69 kb (pTB2), a size similar to pRLO149_63. Approximately 50% of the genes on pTB2 and pRLO149_63 are orthologs (Figure 2C). The remaining parts contain unique genes for each organism, and many of these are associated with cell envelope biosynthesis. These findings suggest that the plasmids are important for the cell surface structure of the two Roseobacter species and may have originated from a common ancestor.

In E. coli the O-antigens are known to interact with the host defences and are therefore important virulence factors of pathogenic bacteria 37 . Many other Roseobacter clade bacteria with plasmid-located rfb-genes were also isolated from surfaces of algal or animal hosts. Therefore, we investigated whether a correlation between the replicon location of the rfb-genes and host-association exists. The genome sequences of marine Rhodobacterales species available in the IMG database 38 39 were searched for rfb-genes and the replicon location was determined if possible (see Additional File 2). For the unfinished genomes, co-occurrence of plasmid replication genes with the rfb-genes on the same DNA-contig was regarded as an indicator for plasmid localisation. Chromosome location was confirmed by rRNA clusters or chromosome partitioning genes on the DNA-contigs that harbour the rfb-genes. In 24 of the 39 genomes rfb-genes were present and their location could be identified. Chromosomal rfb-genes were found in 12 organisms and 11 of those were isolated from the water column. Nine of the 12 plasmid-located rfb-gene clusters were found in organisms that were isolated either from host surfaces, aquacultures, algae blooms or the like. Thus, the replicon analysis of the rfb-genes of 20 of the 24 organisms supports the hypothesis that plasmid-location of the rfb-genes is coherent with host-association of Roseobacter clade bacteria.

To confirm the functionality of the plasmid-encoded photosynthesis apparatus in R. litoralis, the photosynthetic activity of the strain was measured via oxygen consumption (Figure 4). Whereas almost no reaction to light was observed during growth, cells in the stationary growth phase were highly responsive to light and showed a reduced respiration rate when exposed to light (Figure 4). Even though pigmentation occurred already during the exponential growth phase, the organism did not use the photosynthetic apparatus until the culture reached the stationary growth phase. The use, but not the expression, of the photosynthesis apparatus might therefore be influenced by nutrient availability in R. litoralis, as stationary phase cultures are nutrient depleted. We obtained similar results for R. denitrificans, with cells in the late stationary phase showing a stronger response to light than cells from the exponential growth phase (data not shown). For the alpha-Proteobacteria Labrenzia alexandrii DFl-11 and Hoeflea phototrophica DFL-43 periodic nutrient starvation has been reported to trigger bacteriochlorophyll-a production whereas only slight effects were observed for D. shibae 40 . Obviously, the regulation mechanisms differ between the aerobic phototrophic bacteria and so does the architecture of their photosynthesis genes. In Additional File 3, the organization of the photosynthesis gene clusters of the organisms mentioned above is compared, showing that organisms with similar physiological traits also have similar gene organizations. The suggestion that the organization of genes within purple bacterial photosynthesis gene clusters reflects regulatory mechanisms, evolutionary history, and relationships between species was also made by other authors 7 41 42 . In the oligotrophic environment of the ocean, the use of the photosynthesis apparatus during nutrient depletion may be an important advantage for Roseobacter species in the competition with non-photosynthetic organisms.

Growth on different substrates was tested for R. litoralis to substantiate the reconstruction of metabolic pathways based on the genomic analyses. The results of the growth experiments are shown in Additional File 4. For each substrate the growth characteristics of R. litoralis and the genomic data were combined and the putative degradation pathways for the substrates allowing growth are shown in Figures 5 and 6. For the use of amino acids and amino acid derivatives, the predictions from the genome are mainly consistent with the experimentally achieved data (Additional File 4). However, for the degradation of sugars, sugar derivatives and algal osmolytes, genomic and experimental data differ in several cases (Additional File 4). Selected examples are discussed below.

For D-mannose and D-glucosamine the process of phosphorylation could not be revealed from the genomic data (Figure 5). In other bacteria, a phosphotransferase system (PTS) mediates phosphorylation of monosaccharides already during transport (for reviews see 43 44 ). As for the majority of the Roseobacter clade members, no complete PTS is encoded in the genome of R. litoralis. Present are genes for an EIIA component, Hpr and HprK, but no permease was identified. The function of the incomplete PTS is to date unknown. It rather exhibits a regulatory function as proposed for the spirochaetes Treponema pallidum and Treponema denticola 45 and as shown for Sinorhizobium meliloti 46 , which also lack the permease component of the PTS. Therefore, a PTS-independent system for transport and phosphorylation of D-mannose and D-glucosamine must be present in R. litoralis.

R. litoralis grew with D-galactose, L-arabinose and D-fucose, but no complete pathways could be assigned to the degradation of these monosaccharides. A known mechanism for D-galactose degradation in bacteria is the DeLey-Doudoroff pathway (DD-pathway, 47 ). Parts of the DD-pathway are encoded in the genome of R. litoralis, namely the genes coding for 2-dehydro-3-deoxygalactokinase (DgoK, RLO149_c015740) and 2-dehydro-3-deoxy-6-phosphogalactonate aldolase (DgoA, RLO149_c015730). The other genes required for the degradation of D-galactose via the DD-pathway are not present in this gene cluster (cluster 1, see Table 3). However, a gene with high similarity to galactose 1-dehydrogenase (Gal, RLO149_c021980) that catalyzes the first step of the pathway 47 was found elsewhere in the genome (cluster 2, see Table 3). This protein has also a high similarity to L-arabinose 1-dehydrogenase of Azospirillum brasilense. In this organism a pathway for L-arabinose degradation was described that is analogue to the DD-pathway for galactose degradation 48 49 50 . Beside L-arabinose 1-dehydrogenase, also arabonate dehydratase of A. brasilense has an ortholog in R. litoralis (RLO149_c021970) which is also located in cluster 2 (Table 3). The other proteins of the pathway known in A. brasilense cannot be assigned to proteins of R. litoralis. Thus, both pathways are incomplete in R. litoralis; however, if the two clusters are combined, a degradation mechanism based on the DD-pathway may be functional for both sugars (Figure 5). Correspondingly, the regulators as well as the sugar-binding periplasmic protein of the transport system in cluster 2 are known to interact with D-galactose, L-arabinose and D-fucose 51 52 53 54 , and also other enzymes are known to act on all three sugars 55 56 . As no mechanism for D-fucose degradation was identified in the genome of R. litoralis and only weak growth was observed with this sugar, it is possible that the enzymes also act on D-fucose but with a lower affinity.

R. litoralis was not able to grow with glycogen as a carbon source, probably due to the fact that no genes exist in the predicted secretome coding for extracellular glycogen cleavage enzymes, and also no putative transporters for glycogen were identified. Intracellular glycogen, however, was detected in cells of both Roseobacter species when grown in mineral medium. In contrast, no glycogen formation was observed when the cells were grown in complex medium. The glycogen biosynthesis and degradation genes in the two organisms are therefore probably involved in intracellular glycogen production under limiting conditions. The proposed pathway for glycogen metabolism in R. litoralis is shown in Figure 5.

Glycogen formation is often induced by nitrogen starvation 57 58 59 ; however, both Roseobacter species were not nitrogen-starved. Thus, another limiting factor must be the inducer of glycogen production in the mineral medium.

In contrast to the other Roseobacter clade members, in both Roseobacter species the genes for glycogen biosynthesis and degradation are co-located with essential genes of the ED-pathway. It is known from other studies that Roseobacter clade members use the ED-pathway for sugar breakdown 60 61 . The co-location of the genes coding for the ED-pathway with those of the glycogen biosynthesis and degradation indicates a close relation of the central carbon metabolism and glycogen storage in the Roseobacter species.

The algal osmolytes tested in this study all served as carbon and nitrogen sources for both Roseobacter species, except for dimethylglycine. Additionally, R. litoralis used taurine as sulfur source. It was possible to reconstruct the pathways for algal osmolyte degradation from the genome of R. litoralis with the three exceptions creatinine, glycine betaine and putrescine (Figure 6, Additional File 4).

Whereas the degradation mechanism for creatinine is clear in R. denitrificans, in R. litoralis the enzyme encoding the initial step of the pathway is not encoded in the genome (Figure 6). Nevertheless, no differences were observed between the two organisms when grown on creatinine. Furthermore, the enzymes mediating the second step of both possible pathways for creatinine degradation are present in the genome of R. litoralis (Figure 6). Thus, the organism is able to degrade creatinine, but the mechanism cannot be completely reconstructed from the genome.

The gene coding for the enzyme converting glycine betaine to dimethylglycine (betaine-homocysteine S-methyltransferase, BHMT) was not identified in the genomes of the two Roseobacter species. However, a potential candidate gene is RLO149_c039100 which is co-located with the genes for dimethylglycine dehydrogenase (DMGDH4, RLO149_c039110) and a sarcosine dehydrogenase (SARDH, RLO149_c039090) that are both directly involved in osmolyte degradation (Figure 6) and are annotated according to eukaryotic enzymes. The domain homocysteine S-methyltransferase (InterPro database 62 , entry IPR003726) that was identified in the protein sequence of RLO149_c039100 is known to be part of mammalian BHMTs 63 , but the predicted protein of R. litoralis shares only 26% identical amino acids with the mammalian enzymes (E-value 6e-09). A similar ORF (RD1_0018, 96% identity to RLO149_c039100) is encoded in the genome of R. denitrificans that is located in the same genomic neighbourhood as RLO149_c039100.

Only eight bacterial enzymes annotated as BHMT are present in the UniProt database 64 . An enzyme of S. meliloti has been described to mediate this step in glycine betaine degradation 65 , but the protein sequence of RLO149_c039100 shows only weak similarity to this protein. The genomic location, the functional domain and the fact that both Roseobacter strains are able to grow with glycine betaine as carbon and nitrogen sources point to RLO149_c039100 and RD1_0018 being involved in the degradation of glycine betaine in Roseobacters, representing a new class of bacterial BHMTs.

Even though several genes coding for dimethylglycine dehydrogenases (DMGDH) were identified in the genomes of both Roseobacter species, the strains were neither able to utilize dimethylglycine (DMG) as carbon nor as nitrogen source. It is possible that the DMGDHs are only mediating the degradation of intracellular DMG derived from the cleavage of glycine betaine.

The degradation of the osmolyte putrescine is likely to occur via 4-aminobutyrate, as some of the genes of the pathway are present (Figure 6). However, the enzymes mediating step three and four of the pathway were not predicted in the genome of R. litoralis and also the aminotransferase responsible for the last step of the pathway, i.e. the degradation of 4-aminobutyrate to succinate-semialdehyde, was also not identified. Steps three and four might be mediated by RLO149_c025400 (putative gamma-glutamyl-gamma-aminobutyrate hydrolase) and RLO149_c025390 (putative D-beta-hydroxybutyrate dehydrogenase). The last step of the pathway might be carried out by one of the uncharacterized aldehyde aminotransferases encoded in the genome of R. litoralis.

Two different pathways are postulated for the transport and degradation of taurine by microorganisms 66 . Apparently both pathways with different transport systems are present in R. litoralis. A taurine TRAP transport system in combination with a taurine dehydrogenase (TDH) also occurs in R. denitrificans, D. shibae, Rhodobacter sphaeroides and Paracoccus denitrificans, whereas a taurine ABC transporter and a taurine:pyruvate aminotransferase (Tpa) are present in most of the other genome sequenced Roseobacter clade members. Only D. shibae, R. denitrificans and R. litoralis have both uptake and degradation systems. Common features of these three organisms are their photosynthetic activity and the association with algal hosts 1 4 32 67 68 . The common feature of all taurine TRAP transporter-containing organisms is their ability to grow anaerobically 2 3 69 70 71 . The exception is R. litoralis, for which no anaerobic growth was reported yet. Nevertheless, the ability to grow anaerobically is indicated by the presence of genes for nitrite reduction (see above) and for dimethyl sulfoxide (DMSO)/TMAO reductases (RLO149_c007970, RLO149_c001820-RLO149_c001840). Anoxic conditions can occur during the collapse of algal blooms 72 which might also be the situation when high amounts of taurine and other algal osmolytes become available.

The strains R. litoralis OCh149 and R. denitrificans OCh114 were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig Germany). The genome sequencing of R. litoralis was carried out at the J. Craig Venter Institute (Rockville, MD, USA) within the Microbial Genome Sequencing Project by a Sanger sequencing-based approach. For details see the Microbial Genome Sequencing Project 73 . The Sanger-based sequencing resulted in 7.97 coverage of the genome sequence. Gap closure and all manual editing steps were carried out at the Göttingen Genomics Laboratory (University of Göttingen, Germany) using the Gap4 (v 4.11) software of the Staden package 74 . Remaining gaps in the sequences were closed by primer walking on PCR products. Sequences were obtained using the Big Dye 3.0 chemistry (Applied Biosystems), customized sequencing primers and ABI3730XL capillary sequencers (Applied Biosystems).

The prediction of coding sequences (CDS) or open reading frames (ORFs) was done with YACOP 75 . The results were verified and improved manually by using criteria such as the presence of a ribosome-binding site, GC frame plot analysis, and similarity to known protein-encoding sequences using the Sanger Artemis tool 76 . Functional annotation of all ORFs was carried out with the ERGO software package 77 (Integrated Genomics, Chicago, IL, USA). The protein sequences of the predicted ORFs were compared to the Swiss-Prot and TrEMBL databases 78 . Functional domains, repeats and important sites were analysed with the integrated database InterPro using the Web-based tool InterProScan 79 .

The genomes of 38 Roseobacter clade members used for the comparison are the same as those used for the analysis of the rfb-genes and are listed in Additional File 2. Additionally, Roseobacter R2A57 and Loktanella sp. SE62 were included in the comparison. For comparative analysis, the BiBag software tool for reciprocal BLAST analyses as well as a global sequence alignment using the Needleman-Wunsch algorithm (pers. comm. Antje Wollherr and Heiko Liesegang, G2L Göttingen) was used. ORFs were considered as orthologs at a Needleman-Wunsch similarity-score of at least thirty percent and an E-value < 10e-21. Circular plots of DNA sequences were generated with the program DNAPlotter 80 . The comparative visualisation of the genetic organization of the photosynthesis genes of different organisms was realized with the GenVision software (DNASTAR, Inc., Madison, WI, USA). Whole genome alignments were performed and visualized with the Mauve Software Tool 81 .

The programs IslandViewer 82 and COLOMBO 83 were used for the detection of alien genes and genomic islands in the genomes of R. litoralis and R. denitrificans. To complete the computational prediction, the predicted regions were manually checked for elements commonly associated with GEIs like transposases, integrases, insertion sequence (IS)-elements, tRNAs and GC-content deviations 84 . For the prediction of the secretome of R. litoralis, PrediSi 85 was used.

The Codon Adaptation Index (CAI) measures the synonymous codon usage bias for a given DNA sequence by comparing the similarity between the synonymous codon usage of a gene and the synonymous codon frequency of a reference set. For the calculation of the CAI values the CAIcal server was used 86 . For the functional categorization of gene products, a BLAST search with all coding sequences was performed against the COG database 87 .

The metabolic pathways were reconstructed with the Pathway Tools Software 88 89 from the BioCyc Database collection 90 . The pathways were manually curated if necessary.

R. litoralis was cultured in 500 mL Erlenmeyer flasks containing 200 mL of modified PPES-II medium 1 (see Additional File 5). Cells were grown at 25°C on a rotary shaker at 120 rpm with a natural day-night-rhythm. After 40 hours of incubation (within the exponential growth phase) and 95 hours (stationary growth phase), respectively, the cultures were harvested by centrifugation (7000 rpm, 10°C, 15 minutes) and washed once with 100 mL of a salt solution containing 20 g/L NaCl, 13 g/L MgCl₂ × 6 H₂O and 11.18 g/L KCl. After an additional centrifugation step, the cell pellets were resuspended in 5 mL of the salt solution and the OD₆₀₀ was adjusted to 10. Respiration of the cells was measured via oxygen consumption 4 .

The heavy metal resistance tests were carried out on agar plates based on a modified medium described by Shioi 91 (see Additional File 5). Heavy metal stock solutions were prepared as aqueous solutions of 50 mM CuCl₂ × 2 H₂O and O₄SZn × 7 H₂O, respectively, and sterile filtrated. The stock solutions were added to the autoclaved medium directly before pouring the plates. Each agar plate contained 20 mL of medium and 0.02, 0.04, 0.06, 0.08 or 0.1 mM of one of the heavy metals. Tests were prepared in duplicates. Agar plates containing no heavy metals served as controls.

The pre-cultures were grown as follows: single colonies of the strains were transferred from agar plates to 20 mL of modified 70% Marine Broth medium (MB, Difco 2216, see Additional File 5) in 100 mL Erlenmeyer flasks. Cells were grown at 22°C on a rotary shaker (80 rpm) with a natural day-night-rhythm to an optical density (600 nm) between 0.3 and 0.4. The cell suspensions were diluted 10^-4 fold and 800 μL of the dilution were plated on each agar plate containing zinc or copper.

Substrate tests were performed with R. litoralis to compare the experimental data with the genomic information. As genome comparisons of the two Roseobacter species revealed some putative differences, growth on algal osmolytes was tested for both organisms. All substrates tested as carbon, nitrogen or sulfur sources are listed in Additional File 4. Cells of R. litoralis and R. denitrificans were grown in sterile 22.5 mL metal-capped test tubes containing 5 mL of modified Marine Basal Mineral (MBM) medium 92 (see Additional File 5). Tests for taurine as sulfur source were carried out in modified MBM medium prepared without sulfate but with 100 μM taurine. For the substrates that were tested as nitrogen sources, Tris-HCl in the modified MBM medium was replaced by 0.19 g/L NaHCO₃ and no NH₄Cl was added. The pH of the medium was adjusted to 7.5 after autoclaving with sterile 100 mM HCl. Aqueous stock solutions of the substrates were prepared, sterile filtrated and stored at -20°C or at 4°C. Final concentrations for sugars, sugar derivatives, ethanol, glycogen and urea were 1 mg/mL, 2 mM for amino acids and amino acid derivatives, 10 mM for the algal osmolytes and 1 mL/L for glycerol. When the substrates served as nitrogen sources, the final nitrogen concentration was adjusted to 2.5 mM. All tests were carried out in parallels, one additional parallel was not inoculated and served as a control. For all substrates that were tested as carbon sources for R. litoralis, two additional parallels were supplemented with 1% of complex medium (modified 70% MB, see Additional File 5) to investigate whether an additional cofactor was needed by the strain. The requirement for supplements to the mineral medium was also described for other Roseobacter clade bacteria 93 94 . The cultures were inoculated (1% v/v) with cells grown in liquid complex medium.

For the tests of nitrogen and sulfur sources, mannitol was used as carbon source for R. litoralis and glycerol for R. denitrificans, as growth of the organisms with these substrates yielded similar optical densities and no addition of complex medium was necessary to support growth. As inoculum for the nitrogen tests 200 μL (4% v/v) of N-starved stationary phase cultures were used. All algal osmolytes that R. litoralis could use as carbon and nitrogen sources in separated experiments were additionally tested combined as carbon and nitrogen source within one experiment with a final concentration of 10 mM. In all tests, inoculated parallels that contained no carbon, nitrogen or sulfur source, respectively, served as negative controls. Growth was considered as positive if the optical density was higher than in the respective negative controls. All cultures were incubated at room temperature with a natural day-night-rhythm. At suitable time intervals the optical density at 600 nm (OD₆₀₀) was measured with a spectrophotometer (Bausch & Lomb). After reaching the stationary phase, cells were transferred into fresh medium containing the respective substrate to confirm growth. When taurine served as sulfur source, the cells were transferred twice to fresh medium because the sulfur-free cultures were not growth limited compared to the parallels with taurine in the first two passages. After reaching the stationary phase in the second growth passage, purity of each culture was tested on an agar plate.

R. litoralis and R. denitrificans were tested for intracellular production of glycogen when grown in complex medium (modified 70% MB, see Additional File 5) and in mineral medium using the method described by Bourassa and Camilli 57 . As mineral medium, modified MBM with NaHCO₃ as buffer, 5 mM ammonium and 1 mg/mL mannitol as carbon source (1 mL/L glycerol for R. denitrificans) was used. After growth the cells were pelleted, washed once with 1 × PBS (phosphate buffered saline) buffer and stored frozen until further processing.

The complete genome sequence of Roseobacter litoralis OCh149 was deposited in GenBank under the accession numbers [GenBank:CP002623]-[GenBank:CP002626].