The genome sequence of Geobacter metallireducens: features of metabolism, physiology and regulation common and dissimilar to Geobacter sulfurreducens

Background The genome sequence of Geobacter metallireducens is the second to be completed from the metal-respiring genus Geobacter, and is compared in this report to that of Geobacter sulfurreducens in order to understand their metabolic, physiological and regulatory similarities and differences. Results The experimentally observed greater metabolic versatility of G. metallireducens versus G. sulfurreducens is borne out by the presence of more numerous genes for metabolism of organic acids including acetate, propionate, and pyruvate. Although G. metallireducens lacks a dicarboxylic acid transporter, it has acquired a second putative succinate dehydrogenase/fumarate reductase complex, suggesting that respiration of fumarate was important until recently in its evolutionary history. Vestiges of the molybdate (ModE) regulon of G. sulfurreducens can be detected in G. metallireducens, which has lost the global regulatory protein ModE but retained some putative ModE-binding sites and multiplied certain genes of molybdenum cofactor biosynthesis. Several enzymes of amino acid metabolism are of different origin in the two species, but significant patterns of gene organization are conserved. Whereas most Geobacteraceae are predicted to obtain biosynthetic reducing equivalents from electron transfer pathways via a ferredoxin oxidoreductase, G. metallireducens can derive them from the oxidative pentose phosphate pathway. In addition to the evidence of greater metabolic versatility, the G. metallireducens genome is also remarkable for the abundance of multicopy nucleotide sequences found in intergenic regions and even within genes. Conclusion The genomic evidence suggests that metabolism, physiology and regulation of gene expression in G. metallireducens may be dramatically different from other Geobacteraceae.


Background
Geobacter metallireducens is a member of the Geobacteraceae, a family of Fe(III)-respiring Delta-proteobacteria that are of interest for their role in cycling of carbon and metals in aquatic sediments and subsurface environments as well as the bioremediation of organic-and metal-contaminated groundwater and the harvesting of electricity from complex organic matter [1,2]. G. metallireducens is of particular interest because it was the first microorganism found to be capable of a number of novel anaerobic processes including: (1) conservation of energy to support growth from the oxidation of organic compounds coupled to the reduction of Fe(III) or Mn(IV) [3,4]; (2) conversion of Fe(III) oxide to ultrafine-grained magnetite [3]; (3) anaerobic oxidation of an aromatic hydrocarbon [5,6]; (4) reduction of U(VI) [7]; (5) use of humic substances as an electron acceptor [8]; (6) chemotaxis toward metals [9]; (7) complete oxidation of organic compounds to carbon dioxide with an electrode serving as the sole electron acceptor ( [10]; and (8) use of a poised electrode as a direct electron donor [11]. Although the complete genome sequence of the closely related Geobacter sulfurreducens is available [12] and can provide insights into some of the common metabolic features of Geobacter species, G. metallireducens and G. sulfurreducens are significantly different in many aspects of their physiology. G. sulfurreducens is known to use only four carbon sources: acetate, formate, lactate (poorly) and pyruvate (only with hydrogen as electron donor), whereas G. metallireducens uses acetate, benzaldehyde, benzoate, benzylalcohol, butanol, butyrate, p-cresol, ethanol, p-hydroxybenzaldehyde, phydroxybenzoate, p-hydroxybenzylalcohol, isobutyrate, isovalerate, phenol, propionate, propanol, pyruvate, toluene and valerate [2]. Therefore, in order to gain broader insight into the physiological diversity of Geobacter species, the genome of G. metallireducens was sequenced and compared to that of Geobacter sulfurreducens [12]. Both genome annotations were manually curated with the addition, removal and adjustment of hundreds of protein-coding genes and other features. Phylogenetic analyses were conducted to validate the findings, including homologs from the finished and unfinished genome sequences of more distantly related Geobacteraceae. This paper presents insights into the conserved and unique features of two Geobacter species, particularly the metabolic versatility of G. metallireducens and the numerous families of multicopy nucleotide sequences in its genome, which suggest that regulation of gene expression is very different in these two species.

Contents of the two genomes
The automated annotation of the G. metallireducens genome identified 3518 protein-coding genes on the chromosome of 3997420 bp and 13 genes on the plasmid (designated pMET1) of 13762 bp. Manual curation added 59 protein-coding genes plus 56 pseudogenes to the chromosome and 4 genes to the plasmid. Ten of the chromosomal genes were reannotated as pseudogenes and another 22 were removed from the annotation. In addition to the 58 RNA-coding genes in the automated annotation, manual curation identified 479 conserved nucleotide sequence features. Likewise, to the 3446 protein-coding genes in the automated annotation of the G. sulfurreducens genome [12], manual curation added 142 protein-coding genes and 19 pseudogenes. Five genes were reannotated as pseudogenes and 103 genes were removed from the annotation. In addition to the 55 RNAcoding genes in the automated annotation, manual curation identified 462 conserved nucleotide sequence features. Of the 3629 protein-coding genes and pseudogenes in G. metallireducens, 2403 (66.2%) had one or more fulllength homologs in G. sulfurreducens.
The nucleotide composition of the 3563 intact proteincoding genes of G. metallireducens was determined in order to identify some of those that were very recently acquired. The average G+C content of the protein-coding genes was 59.5%, with a standard deviation of 5.9%. Only three genes had a G+C content more than two standard deviations above the mean (> 71.2%), but 146 genes had a G+C content more than two standard deviations below the mean (< 47.7%), most of which lack homologs in G. sulfurreducens and may be recent acquisitions (Additional file 1: Table S1). Clusters of such genes (shaded in Additional file 1: Table S1) were often interrupted or flanked by transposons with higher G+C content. The functions of most of these genes cannot be assigned at present, but 23 of them are predicted to act in cell wall biogenesis.
Plasmid pMET1 of G. metallireducens consists of a series of six predicted transcriptional units on one strand, tentatively attributed to the mobilization (Gmet_A3575-Gmet_A3574-Gmet_A3573-Gmet_A3572-Gmet_A3643), entry exclusion (Gmet_A3571), addiction (Gmet_A3570-Gmet_A3579-Gmet_A3642), partition (Gmet_A3568-Gmet_A3641), transposition (Gmet_A3567), and replication (Gmet_A3566-Gmet_A3565) functions of the plasmid, and one operon on the opposite strand, comprised of three genes of unknown function (Gmet_A3576-Gmet_A3577-Gmet_A3644). The predicted origin of replication, located 3' of the repA gene (Gmet_A3565), includes four pairs of iterons and a set of six hairpins, suggesting that pMET1 replicates by a rolling-circle mechanism, although it is significantly larger than most such plasmids [13]. Among the fifteen other nucleotide sequence features identified on the plasmid during manual curation was a palindromic putative autoregulatory site (TTTGTTATACACGTATAACAAA) located 5' of the addiction module. Other than the potential toxicity of the addiction module, the impact of pMET1 on the physiology of G. metallireducens is unknown.

Metabolism of acetate and other carbon sources
Acetate is expected to be the key electron donor supporting Fe(III) reduction in aquatic sediments and subsurface environments [14], and Geobacter species quickly become the predominant bacterial species when acetate is injected into subsurface environments to promote in situ bioremedation of uranium-contaminated groundwater [15,16]. Surprisingly, the initial activation of acetate by ligation with coenzyme A (CoA) in G. sulfurreducens occurs by two reversible pathways [17] (Figure 1), indicating that acetate may be inefficiently utilized at low concentrations. These two pathways are also present in G. metallireducens, along with a third, irreversible reaction that may permit efficient activation of acetate at low concentrations. The first pathway of acetate activation ( Figure 1a) occurs through either of two succinyl:acetate CoA-transferases that can convert succinyl-CoA to succinate during oxidation of acetate by the tricarboxylic acid (TCA) cycle pathway, in the same capacity as succinyl-CoA synthetase but conserving energy in the form of acetyl-CoA rather than GTP or ATP [17]. Microarray data from both species suggest that expression of one succinyl:acetate CoA-transferase isoenzyme (Gmet_1730 = GSU0174) is constant and expression of the other (Gmet_3044 = GSU0490) is induced during acetate-fueled growth with electron acceptors other than sol-uble Fe(III), such as Fe(III) oxides, nitrate, or fumarate (D. Holmes, B. Postier, and R. Glaven, personal communications). The second pathway (Figure 1b) consists of two steps: acetate kinase (Gmet_1034 = GSU2707) converts acetate to acetyl-phosphate, which may be a global intracellular signal affecting various phosphorylation-dependent signalling systems, as in Escherichia coli [18]; and phosphotransacetylase (Gmet_1035 = GSU2706) converts acetyl-phosphate to acetyl-CoA [17]. G. metallireducens possesses orthologs of the enzymes of both pathways characterized in G. sulfurreducens [17], and also has an acetyl-CoA synthetase (Gmet_2340, 42% identical to the Bacillus subtilis enzyme [19]) for irreversible activation of acetate to acetyl-CoA at the expense of two ATP ( Figure  1c). Thus, Geobacteraceae such as G. metallireducens may be better suited to metabolize acetate at the low concentrations naturally found in most soils and sediments.
Three enzymes distantly related to the succinyl:acetate CoA-transferases are encoded by Gmet_2054, Gmet_3294, and Gmet_3304, for which there are no counterparts in G. sulfurreducens. All three of these proteins closely match the characterized butyryl:4-hydroxybutyrate/vinylacetate CoA-transferases of Clostridium species [20]. However, their substrate specificities may be different because the G. metallireducens proteins and the Clostridium proteins cluster phylogenetically with different CoA-transferases of Geobacter strain FRC-32 and Geobacter bemidjiensis (data not shown). The presence of these Pathways of acetate activation in G. metallireducens   CoA-transferases indicates that G. metallireducens has evolved energy-efficient activation steps for some unidentified organic acid substrates that G. sulfurreducens cannot utilize.
Numerous other enzymes of acyl-CoA metabolism are predicted from the genome of G. metalllireducens but not that of G. sulfurreducens (Additional file 2: Table S2), including six gene clusters, three of which have been linked to degradation of aromatic compounds that G. metallireducens can utilize [6,[21][22][23] but G. sulfurreducens cannot [24]. All seven acyl-CoA synthetases of G. sulfurreducens have orthologs in G. metallireducens, but the latter also possesses acetyl-CoA synthetase, benzoate CoAligase (experimentally validated [23]), and seven other acyl-CoA synthetases of unknown substrate specificity. The G. metallireducens genome also includes eleven acyl-CoA dehydrogenases, three of which are specific for benzylsuccinyl-CoA (69% identical to the Thauera aromatica enzyme [25]), glutaryl-CoA (experimentally validated [26]) and isovaleryl-CoA (69% identical to the Solanum tuberosum mitochondrial enzyme [27]), whereas none can be identified in G. sulfurreducens. G. metallireducens also has nine pairs of electron transfer flavoprotein genes (seven of which are adjacent to genes encoding iron-sulfur cluster-binding proteins) that are hypothesized to connect acyl-CoA dehydrogenases to the respiratory chain, whereas G. sulfurreducens has only one. None of the seventeen enoyl-CoA hydratases of G. metallireducens is an ortholog of GSU1377, the sole enoyl-CoA hydratase of G. sulfurreducens. G. metallireducens also possesses eleven acyl-CoA thioesterases, of which G. sulfurreducens has orthologs of five plus the unique thioesterase GSU0196. Of the ten acyl-CoA thiolases of G. metallireducens, only Gmet_0144 has an ortholog (GSU3313) in G. sulfurreducens. BLAST searches and phylogenetic analyses demonstrated that several of these enzymes of acyl-CoA metabolism have close relatives in G. bemidjiensis, Geobacter FRC-32, Geobacter lovleyi and Geobacter uraniireducens, indicating that their absence from G. sulfurreducens is due to gene loss, and that this apparent metabolic versatility is largely the result of expansion of enzyme families within the genus Geobacter (data not shown). The ability of G. metallireducens and other Geobacteraceae to utilize carbon sources that G. sulfurreducens cannot utilize may be due to stepwise breakdown of multicarbon organic acids to simpler compounds by these enzymes.
Growth of G. metallireducens on butyrate may be attributed to reversible phosphorylation by either of two butyrate kinases (Gmet_2106 and Gmet_2128), followed by reversible CoA-ligation by phosphotransbutyrylase (Gmet_2098), a pathway not present in G. sulfurreducens, which cannot grow on butyrate [24]. These gene products are 42-50% identical to the enzymes characterized in Clostridium beijerinckii and Clostridium acetobutylicum [28,29].
An enzyme very similar to succinyl:acetate CoA-transferase is encoded by Gmet_1125 within the same operon as methylisocitrate lyase (Gmet_1122), 2-methylcitrate dehydratase (Gmet_1123), and a citrate synthase-related protein hypothesized to be 2-methylcitrate synthase (Gmet_1124) [30] (Figure 2a), all of which are absent in G. sulfurreducens. This arrangement of genes, along with the ability of G. metallireducens to utilize propionate as an electron donor [31] whereas G. sulfurreducens cannot [24], suggests that the Gmet_1125 protein could be a succinyl:propionate CoA-transferase that, together with the other three products of the operon, would convert propionate (via propionyl-CoA) and oxaloacetate to pyruvate and succinate ( Figure 2b). Upon oxidation of succinate to oxaloacetate through the TCA cycle and oxidative decarboxylation of pyruvate to acetyl-CoA, the pathway would be equivalent to the breakdown of propionate into six electrons, one molecule of carbon dioxide, and acetate, followed by the succinyl:acetate CoA-transferase reaction (Figure 2b). In a phylogenetic tree, the hypothetical succinyl:propionate CoA-transferase Gmet_1125 and gene Geob_0513 of Geobacter FRC-32, which is also capable of growth with propionate as the sole electron donor and carbon source (M. Aklujkar, unpublished), form a branch adjacent to succinyl:acetate CoA-transferases of the genus Geobacter (data not shown). In a similar manner, the hypothetical 2-methylcitrate synthase Gmet_1124 and gene Geob_0514 of Geobacter FRC-32 form a branch adjacent to citrate synthases of Geobacter species (data not shown), consistent with the notion that these two enzyme families could have recently evolved new members capable of converting propionate via propionyl-CoA to 2methylcitrate.
Gmet_0149 (GSU3448) is a homolog of acetate kinase that does not contribute sufficient acetate kinase activity to sustain growth of G. sulfurreducens [17] and has a closer BLAST hit to propionate kinase of E. coli (40% identical sequence) than to acetate kinase of E. coli. Although it does not cluster phylogenetically with either of the E. coli enzymes, its divergence from acetate kinase (Gmet_1034 = GSU2707) is older than the last common ancestor of the Geobacteraceae (data not shown). This conserved gene product remains to be characterized as a propionate kinase or something else.
The proposed pathway for growth of G. metallireducens on propionate ( Figure 2) is contingent upon its experimentally established ability to grow on pyruvate [31]. G. sulfurreducens cannot utilize pyruvate as the carbon source unless hydrogen is provided as an electron donor [17]. Oxidation of acetyl-CoA derived from pyruvate in G. sul-furreducens may be prevented by a strict requirement for the succinyl:acetate CoA-transferase reaction (thermodynamically inhibited when acetyl-CoA exceeds acetate) to complete the TCA cycle in the absence of detectable activity of succinyl-CoA synthetase (GSU1058-GSU1059) [17]. With three sets of succinyl-CoA synthetase genes (Gmet_0729-Gmet_0730, Gmet_2068-Gmet_2069, and Gmet_2260-Gmet_2261), G. metallireducens may produce enough activity to complete the TCA cycle.
G. sulfurreducens and G. metallireducens may interconvert malate and pyruvate through a malate oxidoreductase fused to a phosphotransacetylase-like putative regulatory domain (maeB; Gmet_1637 = GSU1700), which is 51% identical to the NADP + -dependent malic enzyme of E. coli [32]. G. sulfurreducens has an additional malate oxidoreductase without this fusion (mleA; GSU2308) that is 53% identical to an NAD + -dependent malic enzyme of B. subtilis [33], but G. metallireducens does not.
G. metallireducens possesses orthologous genes for all three pathways that activate pyruvate or oxaloacetate to phosphoenolpyruvate in G. sulfurreducens (Figure 3a): phosphoenolpyruvate synthase (Gmet_0770 = GSU0803), pyruvate phosphate dikinase (Gmet_2940 = GSU0580) and GTP-dependent phosphoenolpyruvate carboxykinase Gmet_2638 = GSU3385) [17]. It also encodes a homolog of the ATP-dependent phosphoe-  Gmet_1125  Figure 3b) and contribute to the observed futile cycling of pyruvate/oxaloacetate/phosphoenolpyruvate [34] if not tightly regulated. Thus, control of the fate of pyruvate appears to be more complex in G. metallireducens than in G. sulfurreducens.

Evidence of recent fumarate respiration in G. metallireducens
The succinate dehydrogenase complex of G. sulfurreducens also functions as a respiratory fumarate reductase, possi-Potential futile cycling of pyruvate/oxaloacetate and phosphoenolpyruvate in G. metallireducens bly in association with a co-transcribed b-type cytochrome [35]. G. metallireducens has homologous genes (Gmet_2397-Gmet_2395 = GSU1176-GSU1178), but is unable to grow with fumarate as the terminal electron acceptor unless transformed with a plasmid that expresses the dicarboxylic acid exchange transporter gene dcuB of G. sulfurreducens [35], which has homologues in Geobacter FRC-32, G. bemidjiensis, G. lovleyi, and G. uraniireducens. Surprisingly, G. metallireducens has acquired another putative succinate dehydrogenase or fumarate reductase complex (Gmet_0308-Gmet_0310), not found in other Geobacteraceae, by lateral gene transfer from a relative of the Chlorobiaceae (phylogenetic trees not shown), and evolved it into a gene cluster that includes enzymes of central metabolism acquired from other sources ( Figure 4). Thus, G. metallireducens may have actually enhanced its ability to respire fumarate before recently losing the requisite transporter.

Nitrate respiration and loss of the modE regulon from G. metallireducens
G. metallireducens is able to respire nitrate [4], whereas G. sulfurreducens cannot [24]. The nitrate reductase activity of G. metallireducens is attributed to the narGYJI genes ( Figure  5a; Gmet_0329-Gmet_0332), which are adjacent to the narK-1 and narK-2 genes encoding a proton/nitrate symporter and a nitrate/nitrite antiporter (Gmet_0333 and Gmet_0334, respectively) predicted according to homology with the two halves of narK in Paracoccus pantotrophus [36]. A second narGYI cluster (Figure 5b; Gmet_1020 to Gmet_1022) is missing a noncatalytic subunit (narJ), and its expression has not been detected (B. Postier, personal communication). The first gene of both operons encodes a unique diheme c-type cytochrome (Gmet_0328 and Gmet_1019), suggesting that the nitrate reductase may be connected to other electron transfer components besides the menaquinol pool, perhaps operating in reverse as a nitrite oxidase. The product of the ppcF gene (Gmet_0335) in the intact nar operon, which is related to a periplasmic triheme c-type cytochrome involved in Fe(III) reduction in G. sulfurreducens [37], may permit electron transfer to the nitrate reductase from extracellular electron donors such as humic substances [38] or graphite electrodes [11]. The final two genes of the intact nar operon (Gmet_0336-Gmet_0337), encode the MoeA and MoaA enzymes implicated in biosynthesis of bis-(molybdopterin guanine dinucleotide)-molybdenum, an essential cofactor of the nitrate reductase.
Phylogenetic analysis indicates that the moeA and moaA gene families have repeatedly expanded in various Geobacteraceae (data not shown). G. sulfurreducens has a single copy of each, but G. metallireducens has three closely related isoenzymes, of which moeA-1 (Gmet_1038 = GSU2703, 40% identical to the E. coli protein [39]) and moaA-1 (Gmet_0301 = GSU3146, 36% identical to the E. coli protein [40]) occupy a conserved location among other genes of molybdopterin biosynthesis (Table 1, Figure 6). A possible reason for the expansion in G. metallireducens and other Geobacteraceae is a need to upregulate molybdopterin biosynthesis for specific processes: moeA-2 and moaA-2 (Gmet_0336-Gmet_0337, 38% and 33% identity to the E. coli proteins) may support nitrate reduction; moaA-3 (Gmet_2095, 35% identity to E. coli) may function with nearby gene clusters for catabolism of benzoate [23] and p-cresol [22]; and moeA-3 (Gmet_1804, Acquisition of a second fumarate reductase/succinate dehydrogenase by G. metallireducens Figure 4 Acquisition of a second fumarate reductase/succinate dehydrogenase by G. metallireducens. (a) The ancestral gene cluster. (b) The gene cluster acquired from a relative of the Chlorobiaceae, located near other acquired genes relevant to central metabolism: an uncharacterized enzyme related to succinyl-CoA synthetase and citrate synthase (Gmet_0305-Gmet_0306) and phosphoenolpyruvate carboxylase (Gmet_0304). Conserved nucleotide sequences (black stripes) were also identified in the two regions. 37% identity to E. coli) may aid growth on benzoate, during which it is upregulated [21]. G. metallireducens differs from G. sulfurreducens in other aspects of molybdenum assimilation as well (Table 1): notably, G. sulfurreducens possesses a homolog of the moaE gene (GSU2699) encoding the large subunit of molybdopterin synthase, but lacks homologs of the small subunit gene moaD and the molybdopterin synthase sulfurylase gene moeB, whereas G. metallireducens lacks a moaE homolog but possesses homologs of moaD (Gmet_1043) and moeB (Gmet_1042). Comparison with the genomes of other Geobacteraceae suggests that these differences are due to loss of ancestral genes. How the nitrate reductase of G. metallireducens can function with the molybdopterin synthase complex being apparently incomplete is unknown.
In G. sulfurreducens, putative binding sites for the molybdate-sensing ModE protein (GSU2964) have been identified by the ScanACE software [41,42] in several locations, and the existence of a ModE regulon has been predicted [43]. The genes in the predicted ModE regulon (Additional file 3: Table S3) include one of the two succinyl:acetate CoA-transferases, a glycine-specific tRNA (anticodon CCC, corresponding to 26% of glycine codons), several transport systems, and some nucleases. In G. metallireducens, there is no full-length modE gene, but a gene encoding the C-terminal molybdopterin-binding (MopI) domain of ModE (Gmet_0511) is present in the same location ( Figure 6). Phylogenetic analysis shows that the Gmet_0511 gene product is the closest known relative of G. sulfurreducens ModE, and that it has evolved out of the Geobacteraceae/Chlorobiaceae cluster of full-length ModE proteins by loss of the N-terminal ModE-specific domain (data not shown). The ScanACE software detected only one of the ModE-binding sites of G. sulfurreducens at the corresponding location in the G. metallireducens genome, but some vestigial sites were apparent when other syntenous locations were visually inspected (Additional file 3: Table S3), indicating that the ModE regulon once existed in G. metallireducens, but recent loss of the ModE N-terminal domain is allowing the regulatory sites to disappear gradually over the course of genome sequence evolution due to the absence of selective pressure for these sites to remain conserved. Thus, genes that may be controlled globally by ModE in G. sulfurreducens and other Geobacteraceae to optimize molybdenum cofactor-dependent processes have recently acquired independence in G. metallireducens.
For biosynthesis of lysine, threonine and methionine, G. metallireducens and other Geobacteraceae possess a linked The respiratory nitrate reductase operons Figure 5 The respiratory nitrate reductase operons. (a) The major (expressed) operon also encodes the nitrate and nitrite transporters (narK-1, narK-2), two c-type cytochromes including ppcF, and two genes of molybdenum cofactor biosynthesis (moeA-2, moaA-2). (b) The minor operon (expression not detected) also encodes the Rieske iron-sulfur component of nitrite reductase (nirD) and a c-type cytochrome, but lacks a narJ gene.
pair of aspartate-4-semialdehyde dehydrogenase genes: Pseudomonas aeruginosa-type Gmet_0603 (69% identity) [49] and Mycobacterium bovis-type Gmet_0604 (47% identity) [50], but G. sulfurreducens has only the former (GSU2878). A haloacid dehalogenase family protein (Gmet_1630 = GSU1694) encoded between two genes of the threonine biosynthesis pathway could be the enzyme required to complete the pathway, a phosphoserine:homoserine phosphotransferase analogous to that of P. aeruginosa [51], and may overlap functionally with the unidentified phosphoserine phosphatase required to complete the biosynthetic pathway of serine.
Conserved nucleotide sequences (possible promoters and riboswitches) were identified on the 5' sides of several biosynthetic operons ( Table 2). The lysine biosynthesis operon in G. sulfurreducens and other Geobacteraceae begins with a P. aeruginosa-type meso-diaminopimelate decarboxylase (GSU0158; 51% identity) [52], whereas G. metallireducens has two isoenzymes in other locations (Gmet_0219, 30% identical to the E. coli enzyme [53], with homologs in a few Geobacteraceae; Gmet_2019, 31% identical to the P. aeruginosa enzyme [52], unique to G. metallireducens). The recently identified L,L-diaminopimelate aminotransferase (dapL; Gmet_0213 = GSU0162) [54] is co-transcribed with the dapAB genes encoding the two preceding enzymes of lysine biosynthesis, but separated from them by a predicted short RNA element (Gmet_R1005 = GSU0160.1), also found in 23 G. sulfurreducens and G. metallireducens possess different genes for molybdenum cofactor biosynthesis Figure 6 G. sulfurreducens and G. metallireducens possess different genes for molybdenum cofactor biosynthesis. (a) G. sulfurreducens has the global regulator modE. (b) G. metallireducens has multiple copies of moeA, moaA, and mosC, and putative integration host factor binding sites (black stripes). Both genomes have conserved genes (dark grey) for molybdate transport (modABC) and molybdopterin biosynthesis (moeA, moaCB, mobA-mobB, mosC) alongside tup genes for tungstate transport (white), but neither genome has all the genes thought to be essential for bis-(molybdopterin guanine dinucleotide)-molybdenum biosynthesis (light grey). See also Table 1.  Figure S1, Additional file 5: Table S4).
sulfurreducens, however, the β2 gene trpB2 (Gmet_2493 = GSU2379, 60% identical to the T. maritima protein [67]) is the penultimate gene of the predicted trp operon and the trpB1 (Gmet_2482 = GSU2375, 66% identical to the Acinetobacter calcoaceticus protein [68]) and trpA (Gmet_2477 = GSU2371, 47% identical to the Azospirillum brasilense protein [69]) genes are separated from the 3' end of the operon and from each other by three or more intervening genes, most of which are not conserved between the two genomes (not shown). Next to the trpB2 gene of G. metallireducens is one of 24 pairs of a conserved nucleotide motif (Additional file 7: Figure S3, Additional file 5: Table S4) hypothesized to bind an unidentified global regulator protein. Other, evolutionarily related paired sites where another unidentified global regulator may bind (Additional file 8: Figure S4, Additional file 5: Table  S4) are found in 21 locations. Between the proBA genes of G. metallireducens, encoding the first two enzymes of proline biosynthesis (Gmet_3198-Gmet_3199 = GSU3212-GSU3211, 41% and 45% identical to the E. coli enzymes [70]), is one of eight pairs of predicted binding sites for yet another unidentified global regulator (Additional file 9: Figure S5, Additional file 5: Table S4). In G. sulfurreducens, the space between proBA is occupied by a different conserved nucleotide sequence (not shown), found only in four other places in the same genome. Overall, a comparison of the two genomes offers insight into unique features of amino acid biosynthesis and its regulation that deserve further study.

Nucleotide metabolism
Differences in nucleotide metabolism were identified in the two genomes. G. metallireducens has acquired a possibly redundant large subunit of carbamoyl-phosphate synthetase (Gmet_0661, 50% identical to the P. aeruginosa protein [71]) in addition to the ancestral gene (Gmet_1774 = GSU1276, 65% identity to P. aeruginosa), Both genomes encode a second putative thymidylate kinase (Gmet_3250 = GSU3301) distantly related to all others, in addition to the one found in other Geobacteraceae (Gmet_2318 = GSU2229, 41% identical to the E. coli enzyme [72]). G. sulfurreducens has evidently lost the purT gene product of G. metallireducens and several other Geobacteraceae (Gmet_3193, 58% identical to the E. coli enzyme [73]), which incorporates formate directly into purine nucleotides instead of using the folate-dependent purN gene product (Gmet_1845 = GSU1759, 46% identical to the E. coli enzyme [74]).
Notable differences between G. metallireducens and G. sulfurreducens are apparent in the biogenesis of c-type cytochromes, in biosynthesis of the heme group, and in reduction of disulfide bonds to allow covalent linkage to heme. In addition to the membrane-peripheral protoporphyrinogen IX oxidase of G. sulfurreducens and other Geobacteraceae, encoded by the hemY gene (Gmet_3551 = GSU0012, 38% identical to the Myxococcus xanthus enzyme [97]), G. metallireducens has a membrane-integral isoenzyme encoded by hemG (Gmet_2953, 43% identical to the E. coli enzyme [98]), with a homolog in Geobacter FRC-32. These two species also possess a putative disulfide bond reduction system not found in G. sulfurreducens and other Geobacteraceae, comprised of DsbA, DsbB, DsbE and DsbD homologs (Gmet_1380, Gmet_1381, Gmet_1383, Gmet_1384), encoded in a cluster alongside a two-component signalling system (Gmet_1378-Gmet_1379), an arylsulfotransferase (Gmet_1382), and a conserved protein of unknown function (Gmet_1385). Transcription of dsbA and dsbB is diminished during growth on benzoate [21], and phylogenetic analysis indicates that these DsbA and DsbB proteins belong to subfamilies distinct from those that have been characterized (R. Dutton, personal communication). Located apart from this cluster, DsbC/DsbG (Gmet_2250) of G. metallireducens has homologs in several Geobacteraceae, but not in G. sulfurreducens. However, CcdA/DsbD (Gmet_2451 = GSU1322) is present in both. Thus, the pathways of c-type cytochrome biogenesis may be significantly different in the two species and somehow linked to the degradation of aromatic compounds by G. metallireducens.
In both G. sulfurreducens and G. metallireducens, there are four c-type cytochrome biogenesis genes related to ResB of B. subtilis [99], each predicted to be co-transcribed with a gene encoding a ResC/HemX-like protein (hypothesized to be a heme transporter with eight predicted transmembrane segments) [100] and several multiheme c-type cytochrome genes (Additional file 10: Table S5). One more protein of the ResC/HemX-like family (Gmet_3232 = GSU3283) is encoded among enzymes of heme biosynthesis in both genomes. These gene arrangements suggest that each pair of c-type cytochrome biogenesis proteins may be dedicated to the efficient expression of the cytochromes encoded nearby. Two of the pairs are orthologously conserved (Gmet_2901-Gmet_2900 = GSU0613-GSU0614; Gmet_0592..Gmet_0594 = GSU2891-GSU2890); the other two pairs (Gmet_0572-Gmet_0573; Gmet_0578-Gmet_0579; GSU0704-GSU0705; GSU2881.1-GSU2880), which appear to derive from expansion of ancestral genes, may be relevant to the diversified c-type cytochrome repertoire of the two species. Interestingly, three of these gene pairs in G. metallireducens are arranged in proximity to each other in a cluster of ten operons with the same coding DNA strand (Gmet_0571 to Gmet_0601), suggesting that their expression may be co-ordinated by transcriptional readthrough (Additional file 10: Table S5). The purposes of various pairs of c-type cytochrome biogenesis proteins in Geobacteraceae remain to be determined.
The pili of G. sulfurreducens have been implicated in electron transfer [101,102] and biofilm formation [103]. Most genes attributed to pilus biogenesis in G. sulfurreducens have orthologs in G. metallireducens, suggesting that these roles of pili may be conserved. However, instead of the ancestral pilY1 gene found in G. sulfurreducens (GSU2038) and other Geobacteraceae, which may encode a pilus tip-associated adhesive protein [104], G. metallireducens possesses a phylogenetically distinct pilY1 gene in the same location (Gmet_0967; data not shown), surrounded by different genes of unknown function within a cluster of pilus biogenesis genes. Therefore, it remains possible that structural and functional differences between the pili of the two species will be identified in future.

Solute transport systems
Although the substrates of most solute transport systems of G. metallireducens and G. sulfurreducens are unknown, several features distinguish the two species (Additional file 11: Table S6). One of two predicted GTP-dependent Fe(II) transporters of the Geobacteraceae (feoB-1 Gmet_2444 = GSU1380), located next to the ferric uptake regulator gene (fur Gmet_2445 = GSU1379), is present in G. metallireducens; the other (feoB-2 GSU3268), with two feoA genes on its 5' side (GSU3268.1, GSU3270) potentially encoding an essential cytosolic component of the transport system [105], is not. Phylogenetic analysis showed that the FeoB-2 proteins of Geobacteraceae are closely related to the characterized Fe(II)-specific FeoB proteins of Porphyromonas gingivalis [106] and Campylobacter jejuni [107], whereas the FeoB-1 proteins of Geobacteraceae cluster apart from them (data not shown). FeoB-1 proteins are not closely related to the manganesespecific FeoB of P. gingivalis [106] either, and so their substrate specificity cannot be assigned at present.
Several heavy metal efflux pumps are conserved between the two species, but their substrate specificity is uncertain. Transporters present in G. sulfurreducens but not G. metallireducens include that for uracil (GSU0932, 48% identical to the Bacillus caldolyticus protein [114]). Transporters present in G. metallireducens but not G. sulfurreducens include those for nitrate/nitrite (Gmet_0333-Gmet_0334) and chromate (Gmet_2732-Gmet_2731), which are each present as two paralogous genes rather than gene fusions such as their homologs that have been characterized in other bacteria [36,115].

Signalling, chemotaxis and global regulation
G. metallireducens possesses orthologs of the six sigma factors of RNA polymerase identified in G. sulfurreducens (Table 3), as well as a seventh factor (Gmet_2792) not found in other Geobacteraceae, related to the extracytoplasmic sigma-Z factor of B. subtilis [116]. Intriguingly, a particular anti-anti-sigma factor gene is frameshifted in both genomes: GSU1427 has frameshifts in the phosphatase domain, resulting in an in-frame protein, whereas the homologous Gmet_1229 is shifted out of frame in the kinase domain. These differences imply that global regulatory networks may be different in the two species.
The G. metallireducens genome encodes 83 putative sensor histidine kinases containing HATPase_c domains (Additional file 12: Table S7), of which 45 (54%) have orthologs among the 95 such proteins of G. sulfurreducens. There are 94 proteins with response receiver (REC) domains in G. metallireducens (Additional file 12: Table  S7), out of which 66 (70%) have orthologs among the 110 such proteins of G. sulfurreducens. Twenty-seven of the REC domain-containing proteins and another 101 genes and four pseudogenes (Additional file 12: Table S7) were predicted to be transcriptional regulators in G. metallireducens. There are 20 putative diguanylate cyclases containing GGDEF domains, of which 16 (80%) have orthologs among the 29 putative diguanylate cyclases of G. sulfurreducens (Additional file 13: Table S8). Overall, the portion of the genome dedicated to signalling and transcriptional regulation in G. metallireducens is slightly less than in G. sulfurreducens, but still considerable and significantly different in content.
Several protein factors involved in chemotaxis-type signalling pathways are conserved between the two genomes: G. sulfurreducens and G. metallireducens each possess four or five CheA sensor kinases and ten CheY response receivers, almost all of which are orthologous pairs (Additional file 14: Table S9). In contrast, 17 of the 34 methyl-accepting chemotaxis proteins (MCPs) of G. sulfurreducens have no full-length matches in G. metallireducens (Additional file 14: Table S9). Due to apparent gene family expansion in G. sulfurreducens, its remaining 17 MCPs correspond to only 13 MCPs of G. metallireducens (Additional file 14: Table S9). The other five MCPs of G. metallireducens lack full-length matches in other Geobacteraceae (Additional file 14: Table S9). Whereas G. sulfurreducens may use its closely related MCPs to fine-tune its chemotactic responses, G. metallireducens may accomplish response modulation by having twice as many MCP  Table S9).
Integration host factors (IHF) and histone-like (HU) DNA-binding proteins are global regulators of gene expression composed of two homologous proteins that bend DNA in specific locations [117]. IHF/HU binding sites are favoured by some mobile genetic elements for insertion. The genome of G. metallireducens encodes orthologs of the single HU protein, both IHF beta proteins, and one of two IHF alpha proteins of G. sulfurreducens (Table 4). Another HU gene and two additional IHF alpha genes are present in G. metallireducens but not G. sulfurreducens (  Figure S6, Additional file 5: Table S4) is similar to multicopy sequences in many other genomes. Two transposons (ISGme8 and ISGme9) were found inserted near putative IHF/HU-binding sites of Class 1 (Additional file 5: Table S4). No such putative global regulatory sequence elements were identified in G. sulfurreducens. However, pirin, a Fe(II)-binding protein that associates with DNA in eukaryotic nuclei [118,119], is present in G. sulfurreducens as GSU0825, but in G. metallireducens only as a frameshifted fragment, Gmet_3471. These genetic differences indicate that the proteins that decorate and bend the chromosome are very different in the two species.
Although no quorum sensing through N-acylhomoserine lactones (autoinducers) has ever been demonstrated for any Geobacteraceae, this kind of signalling may be possible for G. metallireducens because it possesses a LuxR family transcriptional regulator with an autoinducer-binding domain (Gmet_1513), and two divergently transcribed genes with weak sequence similarity to autoinducer synthetases (Gmet_2037 and Gmet_2038). Both Gmet_2037 and Gmet_2038 have atypically low G+C content (Additional file 1: Table S1) and may have been recently acquired by G. metallireducens. The presence of a conserved nucleotide sequence on the 5' side of Gmet_2037 and in 15 other locations on the chromosome (Additional file 16: Figure S7, Additional file 5: Table S4) suggests that Gmet_2037 may be an unusual autoinducer synthetase that is regulated by a riboswitch rather than an autoinducer-binding protein. This conserved sequence is also found on the 5' side of many genes (frequently c-type cytochromes) in the genomes of G. sulfurreducens, G. uraniireducens, and P. propionicus, and overlaps with predicted cyclic diguanylate-responsive riboswitches [120].
The genomes of G. metallireducens and G. sulfurreducens differ in several other aspects of regulation. Nine pairs of potential toxins and antitoxins were identified in the G. metallireducens genome (Additional file 17: Table S10), which may poison vital cellular processes in response to stimuli that interfere with their autoregulation. Only one of these was similar to one of the five potential toxin/antitoxin pairs of G. sulfurreducens. Both the CRISPR1 and CRISPR2 (clustered regularly interspaced short palindromic repeat) loci of G. sulfurreducens, thought to encode 181 short RNAs that may provide immunity against infection by unidentified phage and plasmids [121,122], have no parallel in G. metallireducens, which has CRISPR3 (also found in G. uraniireducens) instead, encoding only twelve putative short RNAs of more variable length and unknown target specificity (Additional file 18: Table S11). Another difference in RNA-level regulation is that a singlestranded RNA-specific nuclease of the barnase family (Gmet_2616) and its putative cognate inhibitor of the barstar family (Gmet_2617) are present in G. metallireducens but not G. sulfurreducens.
Several conserved nucleotide sequences were identified by comparison of intergenic regions between the G. sulfurreducens and G. metallireducens genomes, and those that are found in multiple copies (Additional file 19: Figure S8, Additional file 5: Table S4) may give rise to short RNAs with various regulatory or catalytic activities.

Conclusion
Inspection of the G. metallireducens genome indicates that this species has many metabolic capabilities not present in G. sulfurreducens, particularly with respect to the metabolism of organic acids. Many biosynthetic pathways and regulatory features are conserved, but several putative global regulator-binding sites are unique to G. metallireducens. The complement of signalling proteins is significantly different between the two genomes. Thus, the genome of G. metallireducens provides valuable infor-

Locus Tag G. metallireducens gene G. sulfurreducens gene
Gmet_1608 none *Gmet_3056 is frameshifted near the N-terminus, but may be expressed from an internal start codon. The functions and associations of the various IHF alpha (ihfA), IHF beta (ihfB), and HU (hup) genes are yet unknown, as is their correspondence to any of the predicted regulatory sites illustrated in Figures S3, S4, S5, and S6.
mation about conserved and variable aspects of metabolism, physiology and genetics of the Geobacteraceae.

Sequence analysis and annotation
The genome of G. metallireducens GS-15 [31]

Manual curation
The automated genome annotation of G. metallireducens was queried with the protein BLAST algorithm [126] using all predicted proteins in the automated annotation of the G. sulfurreducens genome [12] to identify conserved genes that aligned over their full lengths. The coordinates of numerous genes in both genomes were adjusted according to the criteria of full-length alignment, plausible ribosome-binding sites, and minimal overlap between genes on opposite DNA strands. The annotations of all other genes in G. metallireducens were checked by BLAST searches of NR. Discrepancies in functional annotation of conserved genes between the two genomes were also resolved by BLAST of NR and of the Swiss-Prot database. All hypothetical proteins were checked for similarity to previously identified domains, conservation among other Geobacteraceae, and absence from species other than Geobacteraceae. Genes that had no protein-level homologs in NR were checked (together with flanking intergenic sequences) by translated nucleotide BLAST in all six reading frames, and by nucleotide BLAST to ensure that conserved protein-coding or nucleotide features had not been missed. All intergenic regions of 120 bp or larger were also checked, which led to the annotation of numerous conserved nucleotide sequences numbered as follows: Gmet_R#### (for predicted RNAs and miscellaneous conserved sequences, a nonzero first digit indicating membership in a group of four or more sequences); Gmet_P#### (for conserved, putative regulatory sequences 5' of predicted operons, numbers corresponding to the first gene of the operon); Gmet_I [1][2][3][4]## [A, B] (for the four classes of putative global regulator binding sites, mostly found in pairs); Gmet_H4## (for putative global regulatory elements consisting of four tandem heptanucleotide repeats); and Gmet_C### (for the spacers of clustered regularly interspaced short palindromic repeats -CRISPR). Newly added features in the G. sulfurreducens genome were assigned unique numbers with decimal points (GSU####.#) in accordance with earlier corrections.

Phylogenetic analysis
Phylogenetic analysis of selected proteins was performed on alignments generated using T-COFFEE [127], manually corrected in Mesquite [128]. Phylogenetic trees were constructed by the neighbour-joining method using Phylip software [129], with 500 bootstrap replications.