Main

The genome sequence of the coastal bacterioplankter S. pomeroyi DSS-3 was determined by the random whole-genome shotgun method and found to contain 4,283 predicted coding sequences (CDS, Table 1). The megaplasmid contains 10.4% of the CDS, with biases towards genes involved in energy metabolism, transport and regulation (see Supplementary Table S1). Compared with other α-Proteobacteria for which complete genome sequences are available, S. pomeroyi has the highest proportion of genes coding for signal transduction (1.6% versus 0.23 ± 0.39% (± s.d.) for 14 other α-Proteobacteria) and transport/binding proteins (12.1% versus 8.1 ± 3.0%) (see Supplementary Fig. S1), potentially reflecting an enhanced ability to sense and respond to conditions outside the cell.

Table 1 General features of the S. pomeroyi genome

Two operons encoding aerobic carbon monoxide dehydrogenases for the oxidation of CO to CO2 are present in the S. pomeroyi genome (coxSML). The absence of ribulose bisphosphate carboxylase or other complete pathways for autotrophy suggests that the bacterium gains energy, but not carbon, through carboxidotrophy. As all previously characterized carboxidotrophs are autotrophic8, CO oxidation was experimentally confirmed in S. pomeroyi and found to occur at CO concentrations typical of surface sea water (10 nM in coastal and 2 nM in oceanic regions) (see Supplementary Fig. S2). Silicibacter-like lithoheterotrophs might act as a microbial CO sink in the surface ocean—this poorly characterized process consumes 10–60 teragrams of C (as CO) annually and buffers the partial pressure of this greenhouse gas in the ocean9. A cluster encoding reduced inorganic sulphur oxidation (soxRSVWXYZABCDF) provides another mechanism for lithoheterotrophic growth in Silicibacter. The presence of a reduced inorganic sulphur compound was experimentally confirmed to enhance biomass production in acetate-grown S. pomeroyi cultures by 45% relative to cultures receiving no sulphur (see Supplementary Fig. S3). Analysis of codon usage in the S. pomeroyi genome indicates that both coxL and soxB have codon adaptive indices (CAIs) higher than >90% of the genes (0.36 and 0.35, respectively). CAIs have been found to correlate positively with levels of transcription in other genomes.

Because cultured bacteria might not be appropriate analogues of their uncultured relatives, the importance of lithoheterotrophy among marine bacterioplankton was assessed by an analysis of gene stoichiometry in the Sargasso Sea environmental shotgun library10. coxL was present in the library at an abundance equivalent to one per 14 bacterial cells and soxB at one per 10 cells (Table 2). As the Roseobacter clade accounted for only 3% of Sargasso 16S ribosomal RNA genes (see Supplementary Table S2), lithoheterotrophic strategies involving CO and inorganic sulphur compounds probably have distribution beyond this taxonomic group. Both coxL and soxB are also represented in a smaller coastal environmental genomics library11. CO is ubiquitous in marine surface waters owing to photo-oxidation of dissolved organic matter (DOM)9, and sulphide has been measured in reducing microzones of marine snow12 and might be generated from the degradation of organic sulphur compounds. Energy derived from oxidation of reduced inorganic compounds could allow greater heterotrophic efficiency for marine bacterioplankton by decreasing the fraction of assimilated organic carbon respired and hence increasing bacterial growth yield. Alternatively, lithotrophy could enhance anaplerotic CO2 fixation via pyruvate carboxylase, as has been found for Sulfitobacter pontiacus—a member of the Roseobacter clade—in the presence of thiosulphate13. Either alternative would influence the pathways and efficiencies of carbon flow through the marine microbial loop. Along with recent discoveries of bacterial photoheterotrophy in marine surface waters14,15, they imply that mixed metabolic schemes might be particularly successful in the ocean.

Table 2 Stoichiometry of selected genes in the Sargasso Sea shotgun library*

High numbers of peptide transporters (16 ATP-binding cassette (ABC)-type systems), branched chain amino acid transporters (ten ABC-type systems) and amino acid efflux proteins (seven RhtB family systems) suggest that proteins are an important carbon source for Silicibacter. Six ABC-type transporter systems for putrescine and spermidine are present (no other sequenced genome has more than three); these compounds are produced in marine phytoplankton and zooplankton to regulate cell proliferation and bloom formation16. Transport systems for algal osmolytes feature significantly in the S. pomeroyi genome, and include five systems for transporting glycine betaine and/or dimethylsulphoniopropionate (DMSP) (OpuA and OpuD). Only the marine bacterium Oceanobacillus iheyensis has more such systems (with 16, see Supplementary Table S3). There is also a transport system and degradation pathway for taurine (tauABCR, xsc, pta, tpa). Organic matter from living plankton and detrital particles in the surface mixed layer is composed of ≥50% protein17, and osmolytes such as DMSP can comprise 20% of the cytoplasm of some marine phytoplankton18. Several genome features suggest that organic matter derived from plankton and marine snow are important substrates for Silicibacter-like organisms. Members of the Roseobacter clade are commonly found in association with algal cells19 and have a depth distribution that matches that of phytoplankton5. S. pomeroyi is motile and is potentially able to position itself in favourable microniches in association with plankton and particles (31 genes encode elements for motility via a polar complex flagellum20). However, there are no open reading frames (ORFs) with strong homology to known chemotaxis proteins or methyl-accepting chemotaxis transducers, suggesting either that the organism is motile but not chemotactic or that chemotaxis occurs by an unknown mechanism, either of which is unusual.

Analysis of flanking genes of an uncommonly high number of TRAP transporter systems (26 systems; no other sequenced genome has more) indicates an ability of S. pomeroyi to transport carboxylic acids of the type produced in surface waters during photo-oxidation of DOM21, including glyoxylate and acetate. Other secondary transporters target malonate and formate. Carboxylic acids together with CO form a pool of biologically labile DOM photoproducts that can support 20% of bacterioplankton biomass production in near-surface coastal waters at temperate latitudes21. While it is anticipated that many marine bacteria assimilate selected photoproducts, Silicibacter-like organisms are genetically positioned to make thorough use of the mixture of compounds formed by DOM photo-oxidation.

Two putative quorum-sensing systems (luxI/luxR homologue pairs) are present on the S. pomeroyi chromosome. The presence of these luxI homologues is consistent with detection of at least three acylhomoserine lactone (AHL)-type signal molecules in culture extracts and distinct AHL-type activities for each homologue when expressed in Escherichia coli (see Supplementary Fig. S4). The presence of two quorum-sensing systems is in contrast to what is expected for an oligotrophic marine bacterium with conservative metabolic strategies. Quorum sensing would provide regulatory flexibility to manage a complex suite of metabolic pathways, possibly associated with switching between particle-associated (high population density, high substrate availability) and free-living (low population density, low substrate availability) stages22. The presence of three rRNA operons in the S. pomeroyi genome is consistent with the potential to respond rapidly to changes in resource availability23, and a pathway for polyhydroxyalkanoic acid synthesis (phaZCP, phbAB) provides a mechanism for carbon and energy storage when substrates are abundant.

Four transporters for ammonium (amt) and one for urea (urtABC) are present in the S. pomeroyi genome, as are genes for ammonium (glnA, gdhA) and urea (ureABCDEFG) assimilation. Genes for assimilating nitrate and nitrite were not identified. Experiments confirmed that ammonium and urea are used as nitrogen sources, whereas nitrate and nitrite are not (see Supplementary Fig. S5). Nitrite is toxic at high concentrations (> 500 µM), potentially because NO accumulates through the partial denitrification pathway (NO2 to N2) encoded on the S. pomeroyi megaplasmid (norFEDQBC, nnrS, nirSECF,D-L fusion,GHJN, nosRZDFYL). Like CO and other DOM photodegradation products, ammonium and urea are regenerated in marine surface waters. Thus, nutrient, energy and carbon acquisition strategies in S. pomeroyi coincide with Roseobacter depth distributions (highest abundance in surface waters)2,5. Although a single cultured organism cannot represent the breadth of features embodied in a diverse taxon that is broadly distributed over distinct hydrographical regions of the ocean24, the emerging picture of S. pomeroyi physiology fits distribution patterns previously observed for Roseobacter clade members.

This genome sequence from a major bacterioplankton clade reveals an organism equipped to take advantage of transient occurrences of high-nutrient niches within a bulk low-nutrient environment. Living and dead plankton and microscale ‘hot spots’ of the surface ocean25 might provide such niches. Lithoheterotrophic growth could allow Silicibacter-like bacterioplankton to use a greater proportion of organic carbon for biomass production as it becomes available. Although most ecologically relevant marine heterotrophs were previously assumed to be oligotrophs that subsist on dilute organic substrates dissolved in sea water26, an ‘opportunitroph’ strategy might be a successful alternative. The available metagenomic data from coastal11 and oceanic10 sites (Table 2) indicate that such a strategy is not atypical among marine bacterioplankton. The genome sequence of this cultured marine microbe provides a window into the ecological strategies that maintain a significant fraction of the prokaryotes in the ocean.

Methods

Isolation of S. pomeroyi (ATCC700808)

Sea water from southeastern US coastal waters (salinity = 31) was diluted with filter-sterilized sea water and enriched with 10 µM DMSP for two weeks. A small white colony was selected from a low-nutrient seawater plate spread with 0.1 ml of the enriched sea water20.

Sequencing and annotation

The complete genome sequence of S. pomeroyi was determined using the whole-genome shotgun method as described27, with libraries of 1–2 kb and 12–15 kb. Closure of physical and sequencing gaps was performed by primer walking, sequencing of transposon-tagged libraries of large-insert clones and multiplex polymerase chain reaction (PCR). Assembly was performed with the TIGR Assembler27, and repeats were identified using RepeatFinder. G+C skew and oligoskew analyses27 identified a putative origin of replication, and base pair 1 was assigned adjacent to the glucose-inhibited division protein A gene (gidA).

An initial set of ORFs predicted to encode proteins was identified using GLIMMER. ORFs consisting of <30 codons or containing overlaps were eliminated, and frameshifts and point mutations were corrected or designated authentic. Functional assignment of genes, identification of membrane-spanning domains, determination of paralogous gene families and identification of regions of unusual nucleotide composition were performed27. Phylogenomic analysis was used to assist with functional predictions27, and comparative genome analyses were performed using the Comprehensive Microbial Resource (http://www.tigr.org/tigr-scripts/CMR2/CMRHomePage.spl). CAI index was calculated using the CAIC package28.

Carbon monoxide oxidation

Three millilitres of culture (absorbance at 600 nm (A600) of 1) were incubated in triplicate 160-ml serum bottles at 30° C, shaking at 250 r.p.m. CO was added at 10 p.p.m. and uptake was monitored to lower limits of 0.13 ± 0.05 p.p.m. using headspace gas chromatographic (GC) analysis (see Supplementary Fig. S2).

Thiosulphate assay

One millilitre of overnight culture was added to 100 ml of basal medium plus acetate (2, 5 or 10 mM) with or without 10 mM thiosulphate. Cultures were incubated at 30 °C and shaken at 250 r.p.m. in the dark and A600 was measured (see Supplementary Fig. S3).

Nitrogen assays

Cultures were grown as described for thiosulphate assays but with 20 mM acetate and 10 mM nitrogen supplied as urea, nitrate, nitrite or ammonium (see Supplementary Fig. S5).

Acylhomoserine lactone analysis

Dichloromethane extracts of S. pomeroyi cultures fractionated by reverse phase thin-layer chromatography were analysed using an Agrobacterium tumefaciens bioassay29. Expression of the LuxI homologues designated SilI1 and SilI2 was engineered by fusing a PCR-amplified complete coding sequence in line with a PLac promoter in pCR2.1-TOPO (Invitrogen). The plasmids were introduced into E. coli XL1-Blue cells expressing the Lac repressor lacIQ. Dichloromethane extracts from an E. coli culture expressing SilI1 contained three AHL activities with the same migration as those identified from S. pomeroyi cultures, and synthesis was elevated several fold by IPTG. E. coli expressing SilI2 had multiple IPTG-inducible AHL activities that migrated to positions different from those activities observed in extracts of S. pomeroyi and from E. coli expressing SilI1, but still in the same general area of the TLC plate (see Supplementary Fig. S4).

Environmental library BLAST analysis

Homologues of S. pomeroyi genes were identified in the Sargasso Sea shotgun library10 by BLAST analysis of trace files downloaded from NCBI using a cutoff for E of 10-95 (coxL) and 10-30 (soxB, nirS, recA). coxL candidates were further screened to identify genes clustering with known CO dehydrogenases8 (see Supplementary Fig. S6). The number of recA homologues was divided by the number of lithoheterotrophy homologues to estimate frequency in cells, assuming one gene copy per cell. 16S rRNA genes from the Sargasso Sea library were assigned to taxonomic groups based on similarity in Smith–Waterman alignments with representative sequences. In the Monterey Bay BAC-end sequence library11, BLAST searches yielded one coxL, one soxB, eight recA and no nirS homologues.

Phylogenetic tree construction

A phylogenetic tree of rRNA gene sequences from cultured and uncultured members of major marine taxa was generated using the neighbour-joining method (positions 226–878, E. coli numbering) excluding positions with <50% conservation. Uncultured archaeon ‘KTK 31A’ (GenBank accession number AJ133625) served as the outgroup.