A typical sugar utilization pathway includes a transport system for sugar uptake and a set of intracellular enzymes that perform biochemical transformations. Although the latter processes vary from sugar to sugar (and, to a degree, from organism to organism), most of them are performed by a rather narrow spectrum of enzymatic activities such as kinases, isomerases, oxidoreductases, hydrolases, and aldolases. Sugar kinases or phosphoenolpyruvate:sugar phosphotransferase systems (PTS) catalyze an essential phosphorylation step (as all intermediates of central carbon pathways to the point of pyruvate are anchored by one or two phosphates). We also included in this analysis associated transcriptional regulators of sugar utilization pathways that mediate specific induction of a catabolic pathway. For utilization of various natural polysaccharides, many pathways are equipped by upstream hydrolytic enzymes producing oligo-, di- or mono-saccharides that are then transported into the cell and further metabolized. These and other upstream and auxiliary (e.g. involved in sensing and chemotaxis) components of the carbohydrate utilization machinery were typically excluded from our analysis, except for those that share genomic context with other sugar utilization components.

Despite their functional diversity, many characterized components of sugar utilization pathways occur in a limited number of protein families containing multiple paralogs with varying substrate specificities. Representatives of these families can be recognized by homology-based genomic searches, thus providing a source of gene candidates for pathway reconstruction. On the other hand, homology-based methods, taken alone, often fail to reliably identify the exact substrate specificity. This can be accomplished by additional reasoning based on the analysis of functional and genomic context (functional coupling) 10 .

Bacterial genes involved in sugar catabolism are often organized in compact operons and/or regulated by committed transcription factors. Therefore, genome context analysis techniques 6 based on the identification of conserved chromosomal clusters (operons) and shared regulatory sites (regulons) are particularly efficient for accurate functional assignment of previously uncharacterized genes of sugar utilization pathways.

Based on the above considerations we developed the bioinformatic workflow that was applied for the analysis of 19 complete Shewanella genomes (Fig. 1). The key starting point of this workflow was a collection of manually curated subsystems in the SEED genomic database 7 capturing a substantial fraction of known sugar utilization pathways projected across many bacterial species. A compilation of ~480 groups of isofunctional homologs from this collection (see additional data file 1) was used as a source of queries for homology-based scanning of 19 Shewanella genomes integrated in the SEED database. The underlying assumption was that any sugar utilization pathway, even extremely divergent or fully unknown, should contain at least one protein recognizably homologous to some of the query sequences from this collection. Therefore, all of the revealed homologs would be considered potential "kernels" of sugar utilization pathways, even though a specific function of some of them might be unclear (and distinct from the query). During further analysis many of the initially identified gene candidates whose functional roles were deemed unrelated to sugar utilization (e.g. involved in biosynthetic pathways) or remained dubious (lacking any evidence for functional assignment) were rejected.

By applying this workflow we tentatively defined the first draft of the Genomic Encyclopedia of Sugar Utilization (or, shortly, sugar catabolome) in 19 Shewanella spp as comprised of 170 isofunctional protein families (FIGfams) 11 . The complete list and some statistics of the analyzed genomes are provided in Table 1. The detailed results of this analysis are captured in the SEED subsystem 'Sugar catabolome in Shewanella species' available online http://theseed.uchicago.edu/FIG/subsys.cgi and summarized in additional files 2 and 3. Functional annotation of the identified protein families was conducted using simultaneous genome context analyses, and homology searches combined with the metabolic reconstruction and identification of missing steps in the putative catabolic pathway.

Specific functional assignments have been proposed for 157 FIGfams comprising 17 peripheral sugar utilization pathways and the key components of the central carbon metabolism (CCM) pathways (Table 2, Fig. 2). A total of 21 FIGfams involved in CCM upstream of pyruvate are conserved in all analyzed Shewanella spp. They comprise complete canonical pentose phosphate (PP) and Entner-Doudoroff (ED) pathways. However, an essential enzyme of glycolysis, 6-phosphofructokinase (Pfk), is missing in all of these species suggesting that sugar utilization in Shewanella can proceed only through the ED or PP pathways. This is in agreement with previous metabolic reconstruction reports 1 3 and biochemical studies of CCM enzymes in cell extracts of certain Shewanella strains 12 . All other canonical enzymes of glycolysis/gluconeogenesis are present in all analyzed Shewanella genomes pointing to the presence of intact gluconeogenetic route. The peripheral pathways are comprised of 136 FIGfams showing a mosaic distribution among 19 compared species. The total number of proteins that comprise the peripheral sugar utilization machinery of individual species varies broadly, from 18 proteins comprising 3 pathways in S. sediminis up to 74 proteins comprising 10 pathways in S. baltica OS223.

13 FIGfams have not been assigned a specific function in any sugar utilization pathway due to the lack of any suggestive genomic context (see additional data files 2 and 3). Among them there are three genes encoding putative sugar kinases of unknown specificity, and genes from a hypothetical sugar utilization gene cluster (for more details see additional data file 4).

Using the genome context analysis combined with metabolic reconstruction we inferred specific functional assignments for 62 families of isofunctional homologs whose functions were previously unknown or defined only at the level of general class (Table 2, see also additional file 5). The novel functional assignments, including 34 components of transport systems, 11 transcriptional regulators (and corresponding DNA motifs), 12 metabolic enzymes, and 5 auxiliary proteins, are supported by the overall internal consistency of the entire reconstruction where all of the pathways are complete (no gaps or missing genes), and all of the annotated genes are associated with complete pathways (no genes out of context). Selected functional assignments mostly for novel sugar transporters from those pathways that are present in most of the analyzed Shewanella species were tested by targeted biochemical and genetic experiments (for details see next section and additional file 6). An ultimate validation of the entire reconstruction was attained by the experimental testing of growth phenotypes (an ability of a given strain to grow on a specific sugar substrate as a sole source of carbon and energy) predicted by the presence or absence of respective reconstructed pathways. The pathway presence was defined based on the presence of all components of the pathway in a particular genome. This phenotype profiling was performed using a "matrix" approach where 14 representative Shewanella species were profiled against a panel of 18 diagnostic sugar substrates (Table 3).

(Glc) is utilized by bacteria using either (i) hexose permease and glucokinase, or (ii) PTS^Glc system. The inability to ferment glucose as a carbon source under aerobic conditions was originally attributed to the Shewanella genus, whereas later studies have identified several species that were able to utilize glucose such as S. baltica and S. frigidimarina 13 14 15 . In this study we have tentatively identified respective genes, glucose transporters glcP ^Bgl and glcP ^Mal and glucokinase glk ^II, that are conserved in most Shewanella genomes, where they are clustered on the chromosome with the genes from β-glucoside (Bgl) and maltodextrin (Mal) utilization pathways, respectively (Fig. 3A). These catabolic pathways include several secreted glucosidases (e.g. BglA^II, CgA, SusB) generating extracytoplasmic glucose that can be utilized via associated glucose transporters and the glucokinase (see below). The predicted glucose transporters belong to the glucose-galactose permease (GGP) family, and are most similar to the glucose and galactose transporter GluP from Brucella abortus 16 and fucose permease FucP from E. coli. The predicted glucokinase Glk^II belongs to the ROK family of sugar kinases 17 , and is similar to fructokinase Mak, D-allose kinase AlsK, and N-acetylglucosamine kinase NagK from E. coli, and glucokinase GlcK from B. subtilis. Orthologs of both glcP ^Mal/glcP ^Bgl and glk ^II are absent from E. coli and other Enterobacteria. The GlcP^Bgl and GlcP^Mal transporters in Shewanella were found only in the context of the Bgl and Mal pathways, whereas Glk^II seems to play a general housekeeping role being conserved in all analyzed Shewanella genomes (see additional data files 2 and 3). In 9 of 19 genomes glk ^II was also found within the Bgl utilization gene cluster, and in 6 of these cases it is a second copy of the gene in the genome.

To confirm the functional assignment of GlcP, we constructed the ΔglcP ^Mal knockout mutant in Shewanella sp. ANA-3, and tested it for glucose-dependent growth. In contrast to the wild-type ANA-3, the ΔglcP ^Mal strain was unable to grow on D-glucose as a sole carbon and energy source, confirming the glucose transporter functional assignment (see additional data file 6 for original experimental data). A representative of the predicted glucokinase subfamily (Glk^II) from S. baltica OS155 was experimentally characterized as a part of our analysis of the Bgl pathway (see below). To assess the functionality of the predicted glucose catabolic pathway we performed phenotypic characterization of Shewanella for growth on D-glucose as a sole carbon and energy source (see additional data files 7 and 8 for the original physiological growth data). Among 14 Shewanella strains tested only S. oneidensis and two S. putrefaciens strains were unable to grow on glucose. These results are consistent with the distribution of glcP transporters in these Shewanella genomes (Table 3, see also additional file 2). The inability of S. oneidensis MR-1 to grow on glucose is most likely due to a frameshift in the glcP ^Mal gene 18 .

The second (PTS-driven) route of glucose utilization is not expected to support the growth of bacteria (such as Shewanella spp) with the incomplete glycolytic pathway due to the inability of the ED or PP pathways to generate enough phosphoenolpyruvate to compensate for its consumption in the phosphotransferase reaction. This assertion is supported by the observed impaired growth of the E. coli pfk mutant on glucose using PTS^Glc 19 . Therefore, the actual physiological role of Shewanella genes orthologous to the components of the E. coli PTS^Glc system (ptsHI-crr and ptsG) remains a mystery. It would be tempting to speculate that this PTS system is used by Shewanella for glucose consumption in the presence of other substrates contributing to nonglycolytic phosphoenolpyruvate generation. However, our attempts to grow MR-1 on the mixture of lactate and glucose did not confirm this conjecture (data not shown).

(Nag) and chitin catabolic pathway (Fig. 2, see also additional file 9) conserved in most Shewanella spp contains several novel functional roles (Table 2). Previously we have experimentally confirmed the predicted enzymatic activities of novel Nag kinase (NagK) and glucosamine-6-phosphate deaminase (NagB^II) and reconstituted the entire three-step biochemical conversion of Nag to fructose-6-phosphate in vitro 9 . We have also confirmed that all of the tested Shewanella strains can grow on Nag as a single source of carbon and energy with the exception of S. frigidimarina, which does not contain these genes (Table 3). Here we present additional experimental results supporting functional assignments of the predicted transporter NagP in S. oneidensis MR-1. NagP belongs to the GGP family of transporters and has a limited similarity to the glucose permeases GlcP^Bgl and GlcP^Mal. We have constructed the S. oneidensis ΔnagP targeted deletion mutant and demonstrated the loss of its ability to grow on Nag (see additional data file 6).

The Nag catabolic genes shows their wide distribution in the Alteromonadales lineage, suggesting the presence of the Nag pathway in the common ancestor of this lineage. The absence of Nag pathway in S. frigidimarina is attributed to a loss of the entire chromosomal gene cluster. The observed wide distribution of Nag utilization pathway in Shewanella could be related to their ability to utilize chitin, a highly abundant constituent of aquatic invertebral exoskeleton.

(Grt) utilization pathway in Shewanella species involves D-glycerate kinase GarK, transcriptional regulator SdaR, and a novel D-glycerate permease termed here GrtP (Fig. 2, see also additional file 9). In E. coli, GarK is involved in the glucarate/galactarate catabolic pathway, generating 2-phosphoglycerate as product, and SdaR is a common transcriptional regulator of this pathway 20 . Most of the glucarate/galactarate utilization genes are absent in Shewanella species. The only exception is a garK gene ortholog, which was found in 17 Shewanella strains in a conserved operon with a hypothetical gene encoding the predicted D-glycerate transporter GrtP (see additional data files 2 and 3). The grtP-garK operon is clustered on the chromosome with an sdaR gene ortholog. A comparative genomic reconstruction of the SdaR regulon in γ-proteobacteria allowed us to predict a candidate SdaR-binding site located upstream of the grtP-garK operon in Shewanella genomes (Fig. 3A).

The predicted D-glycerate uptake transporter GrtP belongs to the gluconate permease family and has orthologs in Pseudomonadales and Vibrionales but not in E. coli or other Enterobacteria. The predicted function of GrtP was confirmed by the loss of ability of the S. oneidensis ΔgrtP strain to grow on D-glycerate as a sole carbon and energy source (see additional data file 6). Of the 14 Shewanella strains, all but 2 (S. denitrificans and S. amazonensis) could grow on D-glycerate (see additional data files 7 and 8), which is fully consistent with the distribution of the grtP-garK operon among the analyzed Shewanella genomes (Table 3). The natural source of D-glycerate in aquatic environments is yet to be elucidated.

'Bgl' utilization pathway revealed in 9 Shewanella species involves β-glucanase LamA, two β-glucosidases BglA^I and BglA^II, two novel β-glucoside transporters BglT and Omp^Bgl, a novel LacI-type transcriptional regulator BglR, glucose permease GlcP^Bgl and ROK-type glucokinase Glk^II (Fig. 2, see additional data files 2, 3 and 9). The previously described Bgl catabolic pathways use PTS-type transport systems and 6-phospho-β-glucosidases (as in B. subtilis and E. coli) or ABC-type transport systems and β-glucosidases (as in Streptomyces spp and Archaea) 21 . The predicted β-glucoside uptake transporter BglT is a member of the glycoside-pentoside-hexuronide (GPH) transporter family, and has orthologs in other Alteromonadales species. Homologs of BglT in Enterobacteria (e.g. YicJ and YagG from E. coli that show 40-44% sequence similarity to BglT) are predicted xyloside transporters regulated by the xylose activator XylR 22 . Omp^Bgl belongs to the TonB-dependent outer membrane transporter (TBDT) family, which includes proteins involved in high-affinity binding and energy-dependent uptake of various substrates (including oligosaccharides) into the periplasm 23 . The predicted Bgl-specific transporter Omp^Bgl is a nonorthologous replacement of the Bgl-specific outer membrane porin BglH from E. coli. Comparative genomic reconstruction of the BglR regulon allowed us to predict candidate BglR-binding sites located upstream of the divergently transcribed bglA^I-bglT-bglR and glcP ^Bgl -bglA ^II operons, and upstream of the omp ^Bgl gene (Fig. 3A). Thus the predicted novel transporters are positionally clustered and co-regulated with other Bgl catabolic genes.

The reconstructed Bgl pathway in Shewanella involves three glucosidases, two of which, LamA and BglA^II, are predicted to be secreted outside of the cell and to the periplasm, respectively, whereas BglA^I is likely a cytoplasmic enzyme. We propose that β-glucoside-containing glucans are first degraded by extracellular endo-β-1,3-glucanase LamA, the resulting oligo-β-glucosides are transported to the periplasm by Omp^Bgl, and subsequently utilized by BglA^II to produce D-glucose and shorter β-glucosides (e.g., cellobiose, gentibiose). The latter products are taken up by the predicted BglT transporter into the cytoplasm where they are finally hydrolyzed by the BglA^I enzyme. D-glucose is taken up by the predicted glucose transporter GlcP^Bgl and phosphorylated by the Glk^II glucokinase (Fig. 2).

The newly predicted β-glucoside BglT transporter was experimentally confirmed by genetic complementation in E. coli ΔbglF mutant (a knockout of cellobiose PTS system). This strain, when transformed by an expression vector containing an operon bglA^I - bglT cloned from S. baltica OS155 gained the ability to grow on minimal medium with cellobiose as the sole carbon and energy source. In the same conditions, no growth was observed for this strain transformed by vector only or by the plasmids containing single genes, bglA^I or bglT (see additional data file 6).

The predicted glucokinase activity of the S. baltica OS155 glk ^II gene product was confirmed by an in vitro enzyme assay. Recombinant purified Glk^II exhibited a broad substrate specificity at 37°C with the highest activity with D-glucose (22 μmol/mg/min), and an appreciable activity with some other tested hexoses including D-mannose, D-fructose, D-glucosamine, and D-mannosamine (see additional data file 6).

The results of the phenotype profiling of 14 Shewanella strains showed that only half of them (four S. baltica species, S. amazonensis, S. frigidimarina, and S. denitrificans) were able to grow on cellobiose as a sole carbon and energy source (see additional data files 7 and 3). This result is consistent with the distribution of the respective genes in the analyzed Shewanella genomes (Table 3).

The Bgl catabolic genes are sparsely distributed among Shewanella species (Fig. 4) and many other species from the Alteromonadales lineage. Therefore, it might be an ancestral pathway predating the speciation of Alteromonadales that could have been independently lost several times within the Shewanella genus. The species-specific loss of Bgl pathway by many Shewanella could be linked to a particular ecophysiology of their habitat. For example, S. pealeana which colonizes the accessory nidamental gland of the squid may not have access to Bgl substrate and thus lost the respective gene cluster from its chromosome 15 .

(Scr) utilization pathway in 8 Shewanella species involves sucrose phosphorylase ScrP, fructokinase ScrK, two novel sucrose transporters (ScrT^II and Omp^Scr), and a novel LacI-type transcriptional regulator ScrR^II (Fig. 2; see also see additional data files 2, 3 and 9). This pathway variant is quite different from the canonical pathway known in Enterobacteria, which includes the sucrose-specific PTS-driven uptake and phosphorylation followed by sucrose-6-phosphate hydrolase 24 . An alternative variant of the Scr pathway previously described in Bifidobacterium spp includes sucrose permease ScrT from the MelB melibiose transporter family and sucrose phosphorylase 25 . The major disctinction in Shewanella is a nonorthologous replacement of ScrT by a predicted transporter ScrT^II of the GGP family of sugar transporters. The novel sucrose regulator ScrR^II in Shewanella is another nonorthologous replacement of the previously characterized ScrR repressor from other bacteria. A genomic reconstruction of the ScrR^II regulon allowed us to predict its candidate binding sites upstream of the divergently transcribed scrP and scrT^II genes, as well as upstream of the omp^Scr-scrK operon (Fig. 3A). The predicted Scr-specific TBDT Omp^Scr is functionally equivalent to the Scr-specific outer membrane porin ScrY from Enterobacteria, and is presumably involved in the uptake of sucrose into the periplasm 26 .

The ScrT^II transporter was validated by complementation in E. coli K-12, which lacks the Scr pathway and is unable to utilize sucrose (see additional data file 6). Two divergently transcribed genes from S. frigidimarina, scrT^II and scrP, were expressed in E. coli under the control of an endogenous promoter in their common intergenic region. As a negative control, we used an empty vector as well as single scrT^II or scrP genes expressed in the same strain. The cell growth was monitored in a minimal medium with sucrose as the only carbon and energy source. We have found that only the cells carrying both scrT^II and scrP were able to grow on sucrose providing an experimental verification of the reconstructed Scr utilization pathway.

The growth phenotype profiling of 14 Shewanella species demonstrated that eight of them (Shewanella sp. ANA-3, MR-4, MR-7, four S. baltica strains, S. frigidimarina) are able to grow on sucrose (see additional data files 7 and 8). These results are in agreement with the presence of the respective genes in these 8 Shewanella genomes (Table 3).

The Scr catabolic genes are present in a large group of Shewanella species (Fig. 4) and in some other species from the Alteromonadales lineage. The most parsimonious scenario suggests that the Scr pathway was present in a common ancestor of the Shewanella genus and that it has been independently lost several times in the evolutionary history of Shewanella. Distribution of the Scr pathway among the Shewanella species provides another possible link to their ecophysiology, the availability of plant-derived sucrose in their respective ecologic niches.

(Mal) utilization gene locus identified in 14 Shewanella species contains the amy, cga, susA, susB, and malZ genes encoding orthologs of previously characterized starch and maltodextrin hydrolytic enzymes, as well as genes encoding novel transporters termed MalT, GlcP^Mal and Omp^Mal, and a novel LacI-type transcriptional regulator MalR (see additional data file 2). The reconstructed Mal catabolic pathway in Shewanella (Fig. 2) uses a predicted novel sodium:solute symporter MalT (a homolog of the melibiose permease MelB of E. coli) instead of the maltose ABC transporter MalEFGK previously described in E. coli 27 . The outer membrane maltose/maltodextrin uptake in Shewanella is presumably mediated by a novel TBDT Omp^Mal, the functional equivalent of E. coli maltoporin LamB 26 . The novel glucose transporter GlcP^Mal is a close paralog (67% identity) of the predicted glucose transporter GlcP^Bgl from the Bgl utilization pathway in Shewanella. We propose that D-glucose is produced in the periplasm by the action of maltodextrin α-glucosidases and transported into the cytoplasm by GlcP^Mal. A comparative genomic reconstruction of the MalR regulon allowed us to predict candidate MalR-binding sites located upstream of the divergently transcribed omp ^Mal and malZ-glcP ^Mal operons, as well as upstream of the susB and malT-cga operons (Fig. 3A).

The growth phenotype characterization of 14 Shewanella species demonstrated that 11 of them (all except MR-1 and two S. putrefaciens strains) are able to grow on maltodextrin as a sole carbon and energy source (see additional data files 7 and 8). These phenotypic results are in agreement with the genomic reconstruction of the Mal pathway (Table 3). The inability of S. oneidensis MR-1 to grow on maltodextrin is attributed to significant genetic perturbations within the Mal utilization gene loci (e.g. glcP ^Mal and susA are pseudogenes with frameshifts, whereas omp ^Mal is interrupted by an insertion element 18 ).

The Mal catabolic genes are widely distributed among Shewanella species (Fig. 4) and several other species from the Alteromonadales lineage. Such phyletic pattern suggests the presence of an ancestral pathway in the common ancestor of the Shewanella genus and independent species-specific pathway losses. Interestingly, while the Mal pathway in S. oneidensis MR-1 demonstrates genetic signs of decay (see above), two closely related S. putrefaciens strains have completely lost the Mal pathway genes. The loss of the Mal catabolic genes in some Shewanella spp. is most likely caused by a significant shift in available nutrient composition in their habitat, e.g. the absence of plant materials as a source of maltodexrins.

(Ara) and arabinoside utilization gene locus identified in six Shewanella species contains more than 20 genes. Nearly half of these genes are similar to previously characterized Ara catabolic genes whereas others are novel (see additional data files 2 and 9). The reconstructed Ara catabolic pathway in Shewanella involves the following predicted functional roles: arabinose ABC transporter AraUVWZ, two arabinoside transporters AraT and Omp^Ara, a novel GntR-type transcriptional regulator AraR^II, L-arabinose mutarotase AraM, and Ara 1-dehydrogenase AraY (Table 2, Fig. 2).

Two alternative routes of Ara utilization are known in bacteria. The major one present in E. coli and B. subtilis depends on L-arabinose isomerase AraA, L-ribulokinase AraB, and L-ribulose-phosphate epimerase AraD, whereas the alternative Ara pathway characterized in Azospirillum brasiliense proceeds through L-arabinose dehydrogenase, arabinolactonase, and L-arabonate dehydratase 28 . In addition to the araBAD genes encoding the conventional Ara utilization pathway through L-ribulose, the Ara metabolic locus in Shewanella contains two genes from the alternative Ara pathway, namely L-arabonate dehydratase araC and arabinolactonase araL, although other genes from the alternative pathway are missing (see additional data file 9). Based on genome context and distant homology analysis we have predicted the gene araY, which is similar to D-xylose 1-dehydrogenase from Caulobacter crescentus (45% similarity), to be the missing L-arabinose 1-dehydrogenase. A member of aldose 1-epimerase family encoded in the Ara gene cluster was tentatively assigned the functional role Ara mutarotase (AraM), which interconverts alpha and beta anomers of L-arabinose.

The predicted ABC-type arabinose uptake transporter system AraUVWZ is similar to the hypothetical sugar transporter YtfQRST from E. coli (58% similarity) and to the ribose transport systems in E. coli and B. subtilis. The predicted TBDT Omp^Ara and the GPH-family transporter AraT are presumably involved in the uptake of arabinosides through the outer and inner membrane, respectively. Comparative genomic reconstruction of the AraR^II regulon allowed us to predict its candidate binding sites located in the likely regulatory regions of most operons in the ara locus (Fig. 3A). Thus, the predicted novel transporters are both positionally clustered and co-regulated with the other Ara catabolic genes (see additional data files 2 and 3).

The functionality of the predicted arabinose utilization pathway is supported by the growth phenotype profile of Shewanella spp on L-arabinose (see additional data files 7 and 8), which correlates perfectly with the presence/absence of the Ara catabolic genes (Table 3).

The Ara catabolic genes are present in a large group of Shewanella species (Fig. 4), among which MR-1 and three S. baltica strains have lost the complete ara gene cluster. Outside of the Shewanella genus the most similar ara catabolic gene clusters were found in two other γ-proteobacteria, the polysaccharide-degrading marine bacterium Saccharophagus degradans and the plant cell wall-degrading soil bacterium Cellvibrio japonicus. Arabinose is an important component of plant cell wall. Therefore, the distribution of the Ara catabolic pathway among Shewanella species may also reflect differences in the availability of plant-derived arabinose in their environments.

(Gal) metabolism genes encoding galactokinase GalK, mutarotase GalM, UDP-glucose epimerase GalE, and UTP-glucose-1-phosphate uridylyltransferase GalU are conserved in all Shewanella genomes (see additional data file 2). Utilization of extracellular galactose or galactosides requires specific transport systems and sugar hydrolases. The galKM locus in S. halifaxensis, S. loihica, S. pealeana, S. piezotolerans, and S. sediminis involves two additional genes, galP^II and lacZ, that are presumably involved in galactose and lactose catabolism. The predicted galactose permease GalP^II belongs to the SSF superfamily and is similar to the myo-inositol and glucose transporters from mammals (34% identity). The β-galactosidase LacZ has a candidate signal peptide cleavage site and is presumably a secreted enzyme. We propose that β-galactosides are degraded in the periplasm by LacZ, and the resulting D-galactose residues are taken up by GalP^II transporter (Fig. 2). The omp ^Gal -lacZ-galTK-galP ^II -galE operon in two S. baltica strains (OS185 and OS223) encodes an additional novel transporter from TBDT family, named Omp^Gal, which is likely involved in lactose uptake through the outer membrane.

The galTKP^II operon identified in S. woodyi is presumably involved in D-galactose monosaccharide utilization, because of the absence of candidate galactoside hydrolase genes in this genome. An ortholog of the transcriptional regulator gene galR from E. coli is clustered with the galTKP^II operon in S. woodyi and with the omp ^Gal -lacZ-galTK-galP ^II -galE operon in two S. baltica strains, and candidate GalR-binding sites were identified in their corresponding upstream regions (Fig. 3).

The growth phenotype characterization of 14 Shewanella species demonstrated that only three strains (S. loihica PV-4, and two S.batica strains) are able to grow on D-galactose (see additional data files 7 and 8). This pattern is consistent with the distribution of the predicted galactose permease gene galP ^II in the analyzed Shewanella genomes (Table 3).

A mosaic distribution of the galP ^II genes in two groups of the Shewanella spp., and the presence of their orthologs in several other marine bacteria from the Alteromonadales lineage suggest multiple gene loss events in the evolutionary history of the Gal catabolic pathway in the Shewanella genus. The Gal utilization genes could have been lost in those Shewanella species that do not share the same ecological niche with some of the marine animals that are thought to provide a natural source of galactose and β-galactosides.

strategies appear to be strikingly different between Shewanella and Enterobacteria. Overall, 34 gene families encoding novel versions of sugar uptake systems were identified in Shewanella spp. Most notably, in contrast to an extensive repertoire of 21 PTS systems actively used by E. coli for the uptake of various sugars 30 , most Shewanella species contain only one PTS system of unclear physiological role. As already discussed, the lack of Pfk genes blocks the conventional glycolytic route in Shewanella making the ED pathway the only feasible way of hexose utilization. The inability of the latter pathway to sustain energy requirements of PTS systems (due to phosphoenolpyruvate regeneration balance) provides a rationale for their absence in Shewanella spp. Thus, the uptake of Glc, Man, Scr, Tre, Nag, and Aga in E. coli is mediated by dedicated PTS systems, whereas in Shewanella species these sugars are transported by predicted novel permeases of the GGP family, GlcP, ManP, ScrT^II, TreT, NagP, AgaP^II, respectively. A combination of inner membrane transporters from the GGP and several other sugar permease families with committed outer membrane transporters of the TBDT family appears to be the predominant strategy of sugar uptake in Shewanella (Fig. 2). Committed TBDT outer membrane transporters were mapped for 10 of the 17 reconstructed sugar utilization pathways in Shewanella species based on their occurrence within respective operons and regulons (Table 2). The predicted Shewanella TBDT transporters are functionally equivalent to sugar-specific outer membrane porins of Enterobacteria (e.g. BglH, ScrY, LamB, NanC). However, in contrast to porins mediating transport across the outer membrane by passive diffusion, TBDT transporters employ a unique energizing mechanism utilizing the TonB complex and the proton-motive force of the cytoplasmic membrane. It is tempting to speculate that a relative abundance of TBDT transporters in Shewanella and some other environmental Proteobacteria reflects their rather limited access to carbohydrates as compared to Enterobacteria.

is another highly variable aspect of the sugar utilization machinery. Indeed, 11 of the 17 transcriptional regulators tentatively associated with Shewanella sugar utilization pathways are nonorthologous to their counterparts previously characterized in E. coli and other species (Table 2). Moreover, 9 of these transcription factors were recruited from structurally unrelated protein families, and they are predicted to bind to completely distinct DNA motifs. For example, the transcriptional repressor NagR of the Nag utilization pathway in Shewanella belongs to the LacI family, whereas a functionally equivalent NagC regulator of E. coli belongs to the ROK family. Therefore, a functional repertoire of co-regulated genes appears to be generally better conserved between species than the respective transcription factors and their DNA binding motifs 31 . Members of the LacI family are most abundant among the regulators of sugar catabolism in Shewanella. They control 10 of the 17 reconstructed pathways, whereas the remaining pathways are regulated by the members of GntR and DeoR families (Fig. 3). The majority of genes from sugar catabolic pathways were identified as candidate members of respective sugar catabolic regulons in Shewanella genomes (see additional data file 3).

are far less variable between species than associated transporters or regulators. In case of Shewanella, the most notable variations are associated with sugar phosphorylation, which is at least partially due to a functional replacement of the uptake-coupled phosphorylation (characterstic of PTS) by a combination of permease and kinase or, less frequently, phopshorylase (for disaccharides). Thus, in the aforementioned Nag and Aga pathways, novel sugar kinases NagK and AgaK combined with respective permeases (NagP and AgaP) are employed in Shewanella spp rather than canonical PTS systems. In a novel variant of Scr utilization pathway predicted and experimentally validated in S. frigidimarina (see additional data file 6), the hydrolysis of sucrose upon ScrT-mediated uptake and phosphorylation of its glucose moiety is performed by the sucrose phosphorylase (ScrP). This single-enzyme transformation is a functional replacement of two consecutive enzymatic reactions, phosphotransferase by PTS (ScrA) followed by sucrose-6-phosphate hydrolase (ScrB), in a canonical version of the pathway described in Enterobacteria 24 . Nonorthologous replacements appear to be quite common for sugar kinases that are known to occur in a number of distinct protein families. For example, the glucokinase Glk^II of Shewanella belongs to the ROK family, whereas its functional counterpart in E. coli and other Enterobacteria belongs to the bacterial glucokinase family.

E. coli strains DH5α (Invitrogen, Carlsbad, CA) and the knockout mutant ΔbglF from the KEIO collection (a kind gift from Dr. Mori) 41 , were used for complementation analyses. E. coli BL21/DE3 (Gibco-BRL, Rockville, MD) was used for protein overexpression and purification. For expression of genes in E. coli, a pBAD-TOPO vector containing the arabinose promoter (Invitrogen, Carlsbad, CA) or a pET-derived vector containing the T7 promoter, His₆ tag, and tobacco etch virus-protease cleavage site were used. Enzymes for PCR and DNA manipulations were from New England Biolabs Inc. (Beverly, MA). Plasmid purification kits were from Promega (Madison, WI). PCR purification kits and nickel-nitrilotriacetic acid (Ni-NTA) resin were from QIAGEN Inc. (Valencia, CA). Oligonucleotides for PCR and sequencing were synthesized by Sigma-Genosys (Woodlands, TX). All other chemicals, including sucrose, cellobiose, NADH, ATP, phosphenolpyruvate, lactate dehydrogenase, pyruvate kinase, D-glucose, 2-deoxy-D-glucose, D-mannose, D-galactose, D-allose, D-fructose, L-sorbose, D-tagatose, D-glucosamine, D-mannosamine, D-galactosamine, N-acetylglucosamine, N-acetylmannosamine, N-acetylgalactosamine were purchased from Sigma-Aldrich (St. Louis, MO).

The 2.9-kb fragment from Shewanella frigidimarina NCIMB400, which contains two divergently transcribed genes, scrT^II and scrP (Sfri_3989 and Sfri_3990, respectively) and the intergenic region, was amplified using the primers: 5'-gcgTTATTTAGCGTCTGCGGTCAACAAATG (forward-1) and 5'-gagTTATCCGCTGTGTTTAGCCAGTAAATC (reverse-1). For cloning the fragment containing scrT^II only and the intergenic region, the primers of forward-1 and 5'-CGCTATACGATCTACGTAAGTGATCAGCTG were used. For cloning the fragment containing scrP only and the intergenic region, the primers of 5'-CTCATTATCCTTCATATGGTTAACCATAC and reverse-1 were used.

The 2.7-kb fragment from Shewanella baltica OS155, which contains the bglA ^I and bglT genes (Sbal_0544 and Sbal_0545, repectively), was amplified using the primers: 5'-gaggaataataaATGAAAATATCTTTACCAAAG (forward-2) and 5'-cacTTAGCTGGCTCTATTTAATTCCAG (reverse-2). For cloning Sbal_0544 only, the primers of forward-2 and 5'-gccTTATTTATTATTGTTATTAGTAAAGTGAGC were used. For cloning Sbal_0545 only, the primers of 5'-gaggaataataaATGATTAGCATAAAAGAAAAAATAG and reverse-2 were used. Nucleotides not present in the original sequence are shown in lowercase. PCR amplification was performed using the genomic DNAs of S. frigidimarina NCIMB400 and S. baltica OS155. The PCR fragments as describe above were cloned into the pBAD-TOPO expression vector.

The glk ^II gene from S. baltica OS155 (Sbal_1134) was amplified using the primers: 5'-ggcgcacATGTTACGAATTGGTATCGATCTTG (forward-3) and 5'-gcaacgtcgacTTAGCGTCCCCACAACCAAGC (reverse-3). Introduced restriction sites (PciI for forward-3 primer and SalI for reverse-3 primer) are shown in boldface. PCR fragment was cloned into the pET-derived expression vector cleaved by NcoI and SalI. Selected clones were confirmed by DNA sequence analysis.

Recombinant protein of glk ^II (Sbal_1134) from S. baltica OS155 was overexpressed as N-terminal fusion with a His₆ tag in E. coli strain BL21/DE3. Cells were grown on LB media to OD₆₀₀ = 0.8 at 37°C, induced by 0.2 mM IPTG, and harvested after 12 h shaking at 20°C. Protein purification was performed using rapid Ni-NTA agarose minicolumn. Briefly, harvested cells were resuspended in 20 mM HEPES buffer pH 7 containing 100 mM NaCl, 0.03% Brij 35, and 2 mM β-mercaptoethanol supplemented with 2 mM phenylmethylsulfonyl fluoride and a protease inhibitor cocktail (Sigma-Aldrich). Lysozyme was added to 1 mg/mL, and the cells were lyzed by freezing-thawing followed by sonication. After centrifugation at 18,000 rpm, the Tris-HCl buffer (pH 8) was added to the supernatant (50 mM, final concentration), and it was loaded onto a Ni-NTA agarose column (0.2 ml). After washing with the starting buffer containing 1 M NaCl and 0.3% Brij-35, bound proteins were eluted with 0.3 ml of the starting buffer containing 250 mM imidazole. Protein size, expression level, distribution between soluble and insoluble forms, and extent of purification were monitored by SDS-PAGE.

In all complementation analyses, cells were pre-cultured on LB media to exponential growth phase, harvested by centrifugation, and washed for three times with M9 minimal media without any carbon sources. All cultures were started with the same optical density at 600 nm (OD_{600 nm} = 0.03), and performed at 37°C in triplicates in 200 μl of the respective media. The cell growth was monitored spectrophotometrically at 600 nm using a microplate reader (ELx808, BioTek Inc., Winnoski, Vermont).

Recombinant proteins of S. frigidimarina NCIMB400 ScrT^II (Sfri_3989) only, ScrP (Sfri_3990) only, and ScrT^II-ScrP were expressed in E. coli K-12 strain DH5α under the control of endogenous promoter in the intergenic region. The empty pBAD-TOPO vector was expressed in the same strain and used as a negative control. The complementation analysis was performed on M9 minimal media supplemented with 50 μg/ml of thiamine and 20 mM of sucrose.

Recombinant proteins of S. baltica OS155 BglA^I (Sbal_0544) only, BglT (Sbal_0545) only, and BglA^I-BglT were expressed under the control of arabinose promoter in E. coli ΔbglF mutant. The empty pBAD-TOPO vector was expressed in the same strain and used as a negative control. The complementation analysis was performed on M9 minimal media supplemented with 0.15% of L-arabinose and 20 mM of cellobiose.

Glucokinase activity was assayed by coupling the formation of ADP to the oxidation of NADH to NAD⁺ via pyruvate kinase and lactate dehydrogenase and monitored at 340 nm. Briefly, 0.1-0.2 μg of purified glucokinase was added to 200 μL of reaction mixture containing 50 mM Tris buffer (pH 7.5), 10 mM MgSO₄, 1.2 mM ATP, 1.2 mM phosphoenolpyruvate, 0.3 mM NADH, 1.2 U of pyruvate kinase, 1.2 U of lactate dehydrogenase, and 5 mM D-glucose at 37°C. No activity was detected in a control experiment, in which an unrelated gene (SO3505) was expressed in the same vector and purified in parallel. The substrate specificities were examined by using the same assay method and exchanging glucose for 5 mM of other potential substrates: 2-deoxyglucose, D-mannose, D-galactose, D-allose, D-fructose, L-sorbose, N-acetyl-D-mannosamine, N-acetyl-D-glucosamine, N-acetyl-D-galactosamine, D-galactosamine, D-glucosamine, D-mannosamine. In the coupled assays, the change in NADH absorbance was monitored at 340 nm using a Beckman DTX-880 multimode microplate reader. An NADH extinction coefficient of 6.22 mM^-1cm^-1 was used for rate calculation.

Total 14 Shewanella strains were tested for their ability to grow on 20 different carbon sources as a sole carbon and energy source. D-/L-lactate mixture was used as a control. Growth conditions and other details of two different experimental techniques used for growth phenotype analysis are provided in additional files 7 and 8.

In-frame deletion mutagenesis of glyT (SO1771) or nagP (SO3503) was performed using previously published method 42 with minor modifications. Upstream and downstream fragments flanking the target locus were PCR amplified using S. oneidensis MR-1 genomic DNA and fused via overlap extension PCR. The fusion PCR amplicon was ligated into XcmI-digested pDS3.0. The resulting recombinant plasmids were used to transform E. coli ß-2155 or WM3063 and subsequently transferred to S. oneidensis strain MR-1 by conjugation. The primary integrants were selected by plating on LB medium containing 7.5 μg/ml gentamycin. The screening for a second round of homologous recombination to remove the plasmid sequence was accomplished by counter-selection on LB medium with 5% sucrose. Colonies growing in the presence of sucrose were screened for sensitivity to gentamycin (7.5 μg/ml) and then screened for deletion of the gene of interest using PCR. The resulting PCR amplicon was then used as the template for DNA sequencing of the deleted and flanking regions involved in the recombination events (ACGT, Inc. Wheeling, IL). In order to validate that the observed phenotype could be attributed to the targeted deletion, complementing plasmids were constructed by inserting appropriate genes downstream to the lac promoter encoded in pBBR1MCS-5.