• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of aemPermissionsJournals.ASM.orgJournalAEM ArticleJournal InfoAuthorsReviewers
Appl Environ Microbiol. Aug 2011; 77(15): 5352–5360.
PMCID: PMC3147445

Large-Scale Comparative Phenotypic and Genomic Analyses Reveal Ecological Preferences of Shewanella Species and Identify Metabolic Pathways Conserved at the Genus Level [down-pointing small open triangle]

Abstract

The use of comparative genomics for the study of different microbiological species has increased substantially as sequence technologies become more affordable. However, efforts to fully link a genotype to its phenotype remain limited to the development of one mutant at a time. In this study, we provided a high-throughput alternative to this limiting step by coupling comparative genomics to the use of phenotype arrays for five sequenced Shewanella strains. Positive phenotypes were obtained for 441 nutrients (C, N, P, and S sources), with N-based compounds being the most utilized for all strains. Many genes and pathways predicted by genome analyses were confirmed with the comparative phenotype assay, and three degradation pathways believed to be missing in Shewanella were confirmed as missing. A number of previously unknown gene products were predicted to be parts of pathways or to have a function, expanding the number of gene targets for future genetic analyses. Ecologically, the comparative high-throughput phenotype analysis provided insights into niche specialization among the five different strains. For example, Shewanella amazonensis strain SB2B, isolated from the Amazon River delta, was capable of utilizing 60 C compounds, whereas Shewanella sp. strain W3-18-1, isolated from deep marine sediment, utilized only 25 of them. In spite of the large number of nutrient sources yielding positive results, our study indicated that except for the N sources, they were not sufficiently informative to predict growth phenotypes from increasing evolutionary distances. Our results indicate the importance of phenotypic evaluation for confirming genome predictions. This strategy will accelerate the functional discovery of genes and provide an ecological framework for microbial genome sequencing projects.

INTRODUCTION

The Shewanella genus is composed of facultative anaerobic bacteria known for their distinctive capability of utilizing a variety of electron acceptors, such as NO3, U, Cr, Tc, Pu, and nitroaromatic compounds (12). Members of this genus have also been regarded for their role as drivers of global biogeochemical cycles of C, N, and S in redox interfaces of marine environments (3, 28).

Since its members are found in different environments, such as salt water, freshwater, sediments, and subsurface formations, it is not surprising that the Shewanella genus developed its hallmark respiratory capability of utilizing many different electron acceptors. This diversity in respiratory phenotypes is a reflection of the genetic makeup of the members of this genus. The sequenced genome of Shewanella oneidensis strain MR-1 shows a large percentage of genes dedicated to the cell's electron transport system, including genes for cytochromes, reductases, iron-sulfur proteins, and quinones (13). As revealed by the genome sequencing of 22 additional Shewanella species and strains of the same species (10), the genetic diversity in this genus is significant, with fewer than half of the genes being shared among 10 of the sequenced Shewanella genomes (21).

Recently, several studies have used comparative genomics to systematize the genomic content into two groups: the core genome, comprised of genes present in all strains, and the accessory genome, consisting of unique or strain-specific genes (21, 39). This approach has allowed for putative determination of the total number of genes and operons that might be involved in the ecological fitness of strains subjected to a specific environmental condition (18, 19, 24, 33). It is less clear, however, how this genomic diversity is translated into phenotypic traits and what their implications are for the ecological success of the species. Traditionally, a particular genotype has been linked to a phenotype through the development and characterization of mutants (23). Based on the 862 genes (19.2%) that still remain to be characterized in the genome of the model microorganism Escherichia coli strain K-12 (36), the above procedure is not only labor-intensive but also a time-consuming activity.

High-throughput phenotype arrays can be used as an alternative approach to expedite the functional characterization of genes. The Biolog assay uses tetrazolium violet to monitor cell respiration, assuming that oxidation of the nutrient source will lead to respiration and hence to purple dye formation (1). High-throughput phenotype arrays have been used extensively to characterize knockout mutants of single microorganisms (16, 43) but have yet to be tested for comparative analysis of phenotypes in light of genome sequence data (2).

In this study, we sought to gain access to the ecology of members of the Shewanella genus through a large-scale comparative analysis of phenotypes. We took advantage of five fully sequenced Shewanella genomes and compared them to high-throughput phenotype arrays containing 561 nutrient sources. We established genotype-phenotype relationships, expanded the number of genes associated with specific phenotypes, and showed that there is a limit in predicting phenotypes from increased phylogenetic distances.

MATERIALS AND METHODS

Strains used in this study.

Microorganisms (GenBank accession numbers) used in this study were Shewanella oneidensis strain MR-1 (AE014299 and AE014300), Shewanella sp. strain MR-4 (CP000446), Shewanella sp. strain MR-7 (CP000444 and CP000445), Shewanella sp. strain W3-18-1 (CP000503), and Shewanella amazonensis strain SB2B (CP000507). Strain selection was based on the following criteria: (i) representation of an evolutionary gradient with strains of the same species and different species and (ii) availability of genomes that were curated manually. A description of habitat conditions at the time of sampling is presented in Table 1.

Table 1.
Habitats of isolation and genome information for the Shewanella strains used in this study

High-throughput phenotypic comparisons.

Phenotype microarray assays were performed at Biolog Inc. (Hayward, CA) as previously described (1). Briefly, Shewanella strains were grown on R2A plates and incubated overnight at 22°C. Colony swabs were used to suspend cells in IF-0 GN medium, and the sample was diluted until 85% transmittance of the cell suspension was achieved. A volume of 100 μl was added to each plate well. Plates were incubated in an OmniLog reader at 22°C. Monitoring of color changes was recorded for 48 h for all plates. Only positive results for three replicate plates are reported. For the purpose of this work, a nutrient was defined as being used as the sole source for the element, e.g., when glycine was defined as the N source, another C source was provided in the well as a substrate to be oxidized, providing electrons for dye reduction.

Gene analysis and pathway reconstructions.

Protein sequences encoded by the five Shewanella genomes were analyzed and compared. Functional predictions for the gene products were obtained from the Shewanella Knowledgebase (17). Many of these predictions resulted from careful manual curation as previously described (17, 35, 37). Furthermore, the gene products were assigned to metabolic pathways according to the MetaCyc schema (5) and the primary literature. In addition, we made use of the orthologous relationships previously determined for 10 Shewanella strains, including the 5 strains used in our analysis (21). The sets of orthologs were identified based on their sequence similarity and genome neighborhoods and included genes present in one or more of the Shewanella genomes. The cutoff criteria for gene presence or absence were compiled by three different methods: (i) protein-protein pairwise reciprocal BLAST-P analysis, (ii) pairwise alignments with a percent accepted mutation (PAM) score of 100, and (iii) reciprocal tBLAST-N analysis (21). An ortholog table for the five Shewanella strains is available in Table ST2 in the supplemental material. Metabolic pathway predictions and the complete Shewanella ortholog data set are also available through the Shewanella Knowledgebase (17).

Phenotypic and genome clustering analyses.

Data analysis was performed using the kinetic and parametric modules of OmniLog v1.2 software. Tested compounds yielding negative or nonreproducible results for all Shewanella strains were removed from the analysis. Results for three replicates were averaged and subtracted from the control intensity value. Next, substrates with intensity values below the threshold of 10 were discarded as a conservative measure of respiration. Phenotypic profiles were converted into a matrix by using the following parameters: 1 for values below 50, 2 for values from 50 to 100, and 3 for values above 100. Clustering was performed with Cluster 3.0 software (6), using a single-linkage scheme and Euclidean distances.

Genomic clustering for all five strains was calculated based on gene orthologs identified in the genomes. When a gene ortholog was observed in the genome, it was scored as present (1 in the matrix), whereas a gene that was not observed was scored as absent (0 in the matrix). Clustering analysis was carried out using a single-linkage scheme and Euclidean distances. Pairwise average nucleotide identity (ANI) values between genomes were calculated according to the method of Konstantinidis and Tiedje (20).

RESULTS

Large-scale comparative phenotypic analysis.

Five Shewanella strains were tested for metabolic diversity, using 561 sources of carbon (C), nitrogen (N), sulfur (S), or phosphorus (P). A positive phenotype for at least one strain was observed for a total of 441 sources (see Fig. SF1 in the supplemental material). Among the different substrates tested, 55 (92%) of the P, 293 (77%) of the N, 21 (58%) of the S, and 67 (35%) of the C sources were utilized by one or more of the strains. The majority of the nutrient sources were used by two to four of the strains, with only 15% and 16% being utilized by all strains and by one of the strains, respectively.

Carbon utilization patterns.

Only 15 of the 190 carbon sources tested were utilized by all five strains (Fig. 1A). Growth on six of these substrates (l-lactate, adenosine, inosine, pyruvate, and N-acetyl-d-glucosamine) has been confirmed in other studies (7, 31, 42). The remaining compounds (2′-deoxyadenosine, methyl pyruvate, uridine, Tween 80, Tween 40, Tween 20, gelatin, and three dipeptides) have not been shown to support growth of the five Shewanella strains and remain targets for further experimental studies. Hydrolysates of gelatin and Tween detergents have previously been reported for S. loihica strain PV-4 (11) and for S. affinis strains KMM 3586 and KMM 3587 (14), respectively, and it is possible that these phenotypes are shared by the entire genus. The three dipeptides utilized by all strains had an N-terminal glycine (Gly-Glu, Gly-Pro, and Gly-Asp). A fourth dipeptide, with a C-terminal glycine (Ala-Gly), was degraded only by SB2B, suggesting the need for a separate transporter or peptidase to utilize this dipeptide. The majority of the substrates degraded by all strains enter the central metabolism at the 2- or 3-carbon compound level (Fig. 2). This agrees with earlier predictions made in an analysis of the S. oneidensis MR-1 genome (37) where many of these pathways were outlined. It was also noted that MR-1 contained a smaller number of iso-enzymes for the degradation of 5- and 6-carbon compounds than did Escherichia coli, an organism that is capable of using a plethora of 5- and 6-carbon compounds. Instead, the MR-1 genome was found to encode iso-enzymes for the utilization of 3-carbon compounds, i.e., 3-glyceraldehyde dehydrogenases. Escherichia coli also contains over 40 phosphotransferase systems (PTSs) for import of 4- to 6-carbon carbohydrates, while MR-1 has only one such system. These trends hold up for the five Shewanella strains included in this study. An additional mannose-specific PTS was found on a mobile island in Shewanella sp. strain W3-18-1.

Fig. 1.
Two-dimensional map representation of a large-scale comparative analysis of nutrient utilization for five Shewanella strains. Differences in color shade intensities represent differences in metabolic activity for transformed data, as follows: black, no ...
Fig. 2.
Carbon source utilization by five Shewanella strains. The substrates and their entry points into central metabolism via key metabolic intermediates are shown. Substrates utilized by all five strains are highlighted in bold. Abbreviations: PP, pentose ...

We observed a variation in the growth phenotype for sugar and polysaccharide utilization where each substrate was degraded by one to three strains. Several of the sugars tested included dimers or polymers of glucose subunits (i.e., α-, β-, and γ-cyclodextrin, dextrin, maltose, maltotriose, and sucrose). Since all of the glucose multimers were utilized by Shewanella amazonensis SB2B, Shewanella sp. strain MR-4, and Shewanella sp. strain MR-7 and not by W3-18-1 or MR-1, it is likely that some of the same enzymes (pathways) and transporters are shared by different strains and involved in the degradation of the above compounds.

SB2B showed the most phenotypic versatility for carbon sources compared to the other species. This strain was able to utilize 60 different substrates, while strain W3-18-1 used only 25. The limited substrate utilization by W3-18-1 included an inability to utilize the many glucose-based multimers as well as several amino acids. In fact, none of the single amino acids tested were degraded by this strain. When the two most evolutionarily related strains were compared, namely, strains MR-4 and MR-7, isolated from the Black Sea water column, their carbon utilization patterns showed larger-than-expected differences. Both strains shared the ability to metabolize 31 carbon substrates, with an additional 13 sources being utilized solely by strain MR-4, while strain MR-7 used another 6 substrates. MR-4 and MR-7 have an average nucleotide identity of 97%, suggesting that they belong to the same species.

Nitrogen utilization patterns.

A total of 50 nitrogen sources were utilized by all five strains (a subset of sources is presented in Fig. 1B and in Fig. SF1 in the supplemental material). These substrates were all dipeptides, with a pronounced preference for amino acids with polar uncharged side chains, i.e., serine, threonine, asparagine, and glutamine. Amino acids with electrically charged side chains (positive or negative) or hydrophobic side chains were of limited use for these strains, with the exception of alanine and leucine. In analyzing nitrogen substrates utilized by all Shewanella strains but one, the number of substrates increased to 126, with the inclusion of purine bases such as xanthine and adenine.

The Black Sea isolates MR-4 and MR-7 were able to utilize 285 and 242 substrates, respectively, followed by strain MR-1, with a pattern of substrate utilization corresponding to 203 nitrogen sources. Inversely proportional to the carbon utilization profile, strain SB2B had the smallest number of positive Biolog phenotypes (81) for nitrogen sources. The low N source diversity displayed by SB2B reflects its inability to utilize 194 (71%) of the dipeptides and tripeptides, and these made up 89% of the N sources tested. Furthermore, the only amino acids or amines utilized by SB2B were l-tyrosine and N-acetylglucosamine.

Phosphorus and sulfur utilization patterns.

AMP was the sole phosphorus source utilized by all five strains. Of the 55 phosphorous compounds degraded by Shewanella, 39 were utilized only by MR-7 and MR-1. In fact, these two strains yielded larger numbers of positive phenotypes for both the phosphate and sulfur sources tested. MR-7 was able to utilize 52 phosphate and 15 sulfur sources, and strain MR-1 used 48 and 19 substrates, respectively (Fig. 1C and D). The remaining strains degraded a very limited number of phosphate substrates besides AMP: W3-18-1 utilized TMP, AMP, GMP, CMP, UMP, and pyrophosphate, while MR-4 utilized phosphogluconic acid and carbamylphosphate. Strain SB2B was able to utilize only TMP in addition to AMP (Fig. 1C). While W3-18-1 tested positive for 3 of the sulfur compounds (l-cysteine sulfinic acid, glutathione, and Glu-Met), MR-4 and SB2B did not utilize any sulfur compounds under the conditions tested (Fig. 1D).

Small-scale comparisons of Biolog phenotypes.

We sought to explain some of the Biolog phenotypes by using known phenotypes, genome contents, and orthologous relationships of the encoded genes (see Table ST1 in the supplemental material). The utilization of N-acetyl-d-glucosamine as a carbon source agrees with the published literature (42). Lactate and pyruvate degradation via acetyl-coenzyme A (acetyl-CoA) to acetate or the tricarboxylic acid (TCA) cycle has been shown with growth experiments (31). Also, the use of nucleotides and nucleosides, including adenosine and inosine, has previously been shown for MR-1 (7) and can be inferred for the other strains based on the presence of genes encoding the degradation enzymes.

According to the high-throughput phenotype assay, MR-1 was the only strain unable to utilize the C4 dicarboxylates l-malate, succinate, and fumarate. When these results were observed in light of the genome sequences, we found that the gene encoding the dicarboxylate carrier AbgT is absent from MR-1 but present in the other strains (MR4_3833, MR7_3926, W3181_3971, and Sama_3559). Furthermore, an oxaloacetate decarboxylase gene is absent in MR-1 and present in the other strains (MR4_2984-7, MR7_3066-9, W3181_3133-6, and Sama_1054-1). This enzyme decarboxylates oxaloacetate to pyruvate and may be involved in the conversion of C4 dicarboxylic acids via pyruvate and the gluconeogenesis pathway to C5 and C6 essential metabolic intermediates. MR-1 also tested negative for l-arabinose utilization. An arabinose (and arabinoside) metabolism locus, including genes for a TonB arabinose receptor, an ABC arabinose transporter, and enzymes degrading l-arabinose to xylose-5-phosphate, was found in the genomes of strains MR-4 (MR4_1977-2001), MR-7 (MR7_1997-1974), and W3-18-1 (W3181_1944-1966) but not in MR-1, as also observed previously (34). Strain SB2B, which also had an l-arabinose-utilizing phenotype, does not contain the above locus and may use another, hitherto unknown pathway to degrade arabinose. Two additional sugars, d-mannose and d-fructose, were utilized by MR-7 and SB2B. A locus containing genes with similarity to the mannose utilization pathway is present in these organisms (MR7_3383-8 and Sama_0565-60), with a second mannose transporter and utilization locus detected in SB2B (Sama_0303-4). While fructose can be degraded via the mannose degradation pathway, a transporter specific for fructose was not identified, agreeing with previous results (34).

In another example, N-acetyl-d-galactosamine was degraded by MR-4, MR-7, and SB2B. This phenotype agrees with the presence of the aga operon for uptake and degradation of N-acetyl-d-galactosamine in these strains (MR4_2530-6, MR7_2597-2603, and Sama_1199-3). The Biolog assay also showed degradation of citric acid by the same three strains. Based on a recent paper describing citric acid utilization by Corynebacterium glutamicum (4), we identified genes for a citrate-sensing two-component regulator and a three-component citrate transporter of the tricarboxylate transporter (TTT) family in MR-4 and MR-7 (MR4_3099-5 and MR7_0873-7). Genes encoding these products were not found in the SB2B genome, suggesting that another citrate utilization path was taken in this organism.

Expanding testable predictions.

The high-throughput genotype-phenotype comparison among different strains allowed us to make targeted predictions for further experimental analysis. For example, degradation of Tween 80 has been shown to involve an outer membrane esterase in Pseudomonas aeruginosa (29). While no close homolog to the P. aeruginosa enzyme was found in the 22 Shewanella genome sequences, we did identify candidate genes present in the Tween-degrading strains. Specifically, three candidate genes, encoding a surface-expressed lipase (SO_2934, MR4_1469, MR7_1535, W3181_1613, and Sama_2120), an outer membrane phospholipase (SO_0428, MR4_0432, MR7_3595, W3181_0530, and Sama_0379), and a cold-adapted lipase (SO_1994, MR4_2269, MR7_2341, W3181_1692, and Sama_2029), were conserved among the strains. The last gene encodes a protein with sequence similarity to a cold-active lipase isolated from a deep-sea sediment metagenome (15). We also identified two proton-dependent (oligo)peptide transporter genes (SO_0002, MR4_3938, MR7_4030, W3181_4066, and Sama_2411 for one and SO_3195, MR4_1313, MR7_1380, W3181_1459, and Sama_2266 for the other) that might be associated with the use of Gly-Glu, Gly-Pro, or Gly-Asp as a carbon source by Shewanella.

Phenotype assays yielded positive results for strains MR-1, MR-4, and SB2B tested with the heteropolysaccharide pectin. We searched the genomes for genes present in these three strains and absent in the MR-7 and W3-18-1 genomes and noted an outer membrane TonB-dependent receptor gene (SO_1822, MR4_2467, and Sama_1252). The involvement of TonB receptors in the uptake of sugars and their derivatives has been shown for MR-1 and others (8, 22, 42). We also identified a gene coding for a sugar-binding periplasmic protein that may be linked to the degradation of dextrins, maltose, and sucrose in MR-4 (MR4_0355), MR-7 (MR7_3671), and SB2B (Sama_3287).

Linking genotypic and phenotypic changes.

To gain insight into the phenotypic variation and its relationship to the genome, we plotted the evolutionary distances (defined as % ANI) and the percentages of identical phenotypes among the five strains (Fig. 3; see Tables ST2 and ST3 in the supplemental material). The regression line for % rRNA similarity versus % ANI was selected to represent a conserved trait (correlation coefficient [R2] = 0.89). The % conservation for orthologs was also included, and it decreased more sharply as the evolutionary distance increased (R2 = 0.76). We also observed a correlation between N source utilization and evolutionary distance (R2 = 0.86) (Fig. 3). However, no such correlation was detected for C source utilization (R2 = 0.05). Based on the clustering of the P and S utilization phenotypes (Fig. 1C and D), it was evident that neither of these phenotypes would be conserved according to evolutionary distance. Because the number of phenotypes (294) for N sources was significantly higher than those for the other compounds (67 C, 55 P, and 21 S phenotypes), we randomly selected sets of 60 N sources to calculate the % similarity versus % ANI in order to test for sample size bias. These data sets also showed a decrease in % similarity with increasing evolutionary distance, suggesting that the N source phenotypic results are not dependent on the number of chemicals compared (data not shown).

Fig. 3.
Phenotypic and genotypic changes over evolutionary time. Five Shewanella strains were compared in a pairwise manner for their similarity in 16S rRNA level (green), number of gene orthologs (purple), nitrogen utilization phenotype (blue), and carbon utilization ...

In order to test whether the observed phenotypic diversity paralleled the gene content diversity, we performed clustering analyses with each of the data sets (Fig. 4 A and B). Similar clustering branches were observed for closely related strains MR-4 and MR-7, but the same correspondence was not seen with greater evolutionary distances. The phenotypic diversity for closely related strains was much greater (longer branch lengths in Fig. 4B) than their gene content diversity (shorter branch lengths in Fig. 4A).

Fig. 4.
Cladograms showing genomic and phenotypic comparisons among five Shewanella strains. (A) Gene content clustering based on orthologs identified among the five genomes. (B) Phenotypic transformed data clustering based on the presence/absence of respiratory ...

DISCUSSION

Historically, the establishment of a direct link between a genotype and its phenotype has been performed through the study of mutants. However, the generation of thousands of mutants can be costly and laborious. In this study, we explored a different avenue for developing such a link by making use of hundreds of Biolog phenotypes and taking advantage of predicted genes and pathways through comparative genomics. Owing to the high level of phenotypic diversity within the genus Shewanella and the availability of many sequenced genomes, we selected five isolates representing a gradient of evolutionary distances within the genus for this study.

The phenotype assay is not a measure of growth on the different substrates but rather an indication of respiration when a nutrient source is provided. The presence of a transport system and catabolic pathway for a specific chemical compound leads to the production of NADH that then reduces a tetrazolium dye (1). Assays measuring respiration have an advantage over growth assays because the microbial cell metabolic response to a chemical compound is detected even when growth support is not observed. Hence, in our study, respiration is a reflection of the ability to utilize different substrates. Likewise, we used five genome sequences and their gene contents to reflect the genetic potential of the Shewanella genus, without taking into account the many regulatory levels that determine whether a gene is expressed or whether the protein is synthesized and active. By analyzing the predicted functions of the encoded proteins as well as their presence or absence relative to a given phenotype, we were able to successfully confirm a series of genome predictions of a specific metabolism and to identify a series of target genes for further genetic analysis.

Because the comparative phenotype analysis dealt with different numbers of C, N, P, and S sources, we normalized the utilization profiles based on the percentage of positive results among the total number of compounds tested. The emerging pattern of phenotypes suggested that Shewanella is capable of utilizing a variety of N compounds (77% of the tested N sources), including several amino acids and dipeptides (Fig. 1). The particular ability to grow on several amino acids was observed previously for strain MR-1 (32), but the use of amino acids as a sole N source was not tested extensively in the Shewanella genus. This preference for N sources might explain the isolation of different species of this genus from a variety of environments, such as dead fish (9), chicken breasts (9), clinical samples (41), and marine sediments (26, 27, 40), where DNA and polypeptides are known to accumulate. Despite the large number of N sources tested, only 50, mainly dipeptides (90%), were utilized by all five strains. N-Acetyl-d-glucosamine was used by these five strains as both an N and a C source (Fig. 2). While N-acetyl-d-glucosamine has previously been characterized as a carbon and energy source for Shewanella (42), its use as an N source has not been reported. We identified 15 compounds that could be used as both N and C sources by at least one of the strains. Interestingly, while SB2B was the only strain able to utilize l-threonine as a C source, the other four strains utilized it as an N source. Overall, there was a decrease in the number of N compounds shared among the strains as the ANI value increased. Fewer shared compounds might indicate a decrease in ecological niche overlap once speciation has taken place. To address this possibility, we first compared the N utilization profiles for the most closely related isolates, namely, MR-4 and MR-7. These Shewanella strains, with 97.05% ANI, are strains of the same species (21) and shared the largest number of N compounds utilized (227). A comparison of these two strains to S. oneidensis showed the number of shared N sources decreasing to 187 for MR-4 versus MR-1 (87.74% ANI) and to 175 for MR-7 versus MR-1 (87.70% ANI). Strains MR-4 and MR-7 came from the same environment, where niches have more similar conditions (Table 1). Accordingly, strain SB2B, which came from an environment differing significantly from the others, had the least number of overlaps in N source utilization, with 81 positive phenotypes, compared to the other strains. None of these N sources were utilized uniquely by SB2B. However, upon inspection of the SB2B orthologs (see Table ST1 in the supplemental material), we identified a stretch of six genes (Sama_0018 to Sama_0023) absent from the other four strains that could be linked to nitrogen metabolism. Two of these genes, Sama_0022 and Sama_0023, encode a predicted urea carboxylase cleaving urea-1-carboxylate to CO2 and ammonia, with the possibility of the latter being assimilated as a nitrogen source. The other colocalized genes included genes encoding two membrane proteins and a LamB family protein, but their specific role in nitrogen metabolism remains unknown.

In attempting to make in-depth connections between the ecology and genomic information of microorganisms, environmental metadata are generally missing or not fully described. Although this was not different in our study, the limited habitat description for the Shewanella isolates in congruence with the large-scale comparative phenotype approach provided insights into the functional strategies devised by the five strains in their environment. Shewanella amazonensis strain SB2B was capable of utilizing a full range of C compounds (60 of the 67 tested), varying from well-known growth substrates for Shewanella, such as pyruvate and lactate, to previously unknown compounds, such as the Tween series and laminarin. Since SB2B was isolated from the coastal muddy waters of the Amazon River, a low-salinity environment (Table 1), it is likely exposed to a full plethora of C compounds washed out from the Amazon forest (25). In contrast to the SB2B C utilization profile, strain W3-18-1 yielded positive results for only 25 C compounds. This might be a reflection of the genetic adaptation of W3-18-1 to nutrient limitation imposed in the marine sediments of the Pacific Ocean (38).

Together, the five Shewanella strains selected for this study carry a total of 6,790 genes. Approximately 39% of the genes are present in all strains, while the unique genes comprise 21 to 40% of the genomes. On average, 550 genes per strain were identified as unique, a number in accordance with the previously reported 468 unique genes when 10 Shewanella genomes were compared (21). In our high-throughput genotype-phenotype analyses, we were able to link a few of these unique genes to specific phenotypes. In addition, pathways believed to be missing in these Shewanella strains (e.g., degradation of galactose, ribose, and trehalose) were confirmed as missing using the phenotype assay.

It is noteworthy that not all observed phenotypes were explained through the use of comparative genomics. For example, MR-7 was the only strain unable to utilize acetate and thymidine as C sources. From the ortholog table, we identified only four proteins that were missing in MR-7 and present in the other four strains: two lipopolysaccharide biosynthesis proteins and a transcriptional regulator with an adjacent multidrug transporter. It is not evident how any of these functions could explain the acetic acid and thymidine growth phenotypes. Likewise, W3-18-1 was the only strain unable to degrade l-glutamate and l-glutamine, yet no genes encoding proteins for transport or degradation of either compound could be found among the 137 genes absent from W3-18-1 and present in the other four strains. Sequences encoding glutamate transporters and degrading enzymes (i.e., glutamate decarboxylase, glutamate racemase, and mutase) were detected in all five genome sequences. We observed a glutamate racemase gene (W3181_0949) and a glutamine amidotransferase gene (W3181_0925) among those encoding proteins unique to W3-18-1, suggesting that this organism may have a different strategy for metabolizing glutamine and glutamate. There are two possible explanations for the above differences. First, there is a large fraction of genes without function predictions (37). Presently, 21% of the protein coding genes in the five strains remain of unknown function. Novel enzyme variants may also carry some of the “missing” activities, as uncovered in studies focused on revaluating the genomic content of S. oneidensis strain MR-1 (31, 42). Genome reconstruction allowed identification of novel enzymes for degradation of both l-lactate (lactate dehydrogenase [lldEGF]) and N-acetylglucosamine (glucosamine-6-phosphate deaminase [nagB] and N-acetylglucosamine kinase [nagK]). Second, regulation of expression or activity at the gene or protein level affects whether a phenotype is exhibited, leading to weak phenotypic expression or a false-negative result in our analysis.

In order to validate our high-throughput phenotype approach, we compared our results to recent work describing the reconstruction of Shewanella carbohydrate utilization pathways (34). Rodionov et al. determined the C degradation pathways for 19 Shewanella genomes by using orthologous relationships, regulon predictions, and growth phenotypes. The increased number of genomes as well as the inclusion of growth phenotypes and regulatory networks allowed for a thorough analysis of a subset of C sources compared to our analyses. Growth phenotype data were obtained for eight of the sugars (arabinose, cellobiose, fructose, glucose, mannose, maltose, N-acetylglucosamine, and sucrose) tested in our high-throughput phenotype assay. A comparison between both studies indicated an agreement of 75% of the phenotypes (W3-18-1 was not included in their growth studies). Although the data set used for comparison contains only a small number of sugars, we did not detect bias with regard to a particular substrate or strain, attesting the validity of our approach. We also found agreement in the utilization of amino acids by MR-1 between the Biolog phenotype and growth (30) for six of the seven amino acids tested in both assays.

The five strains selected for this study had various degrees of relatedness as measured by the ANI. The two closest strains in the data set, namely, MR-4 and MR-7, shared the largest fraction of genes (93%) relative to the other genome pairs (e.g., 80% between MR-4 and MR-1 and 72% between MR-4 and SB2B). However, this high sequence similarity between MR-4 and MR-7 was not reflected throughout the phenotypic results (Fig. 4). These strains were among the most similar for the carbon sources (88% similarity) and nitrogen sources (77% similarity), but they were unable to both utilize any of the 21 S sources tested and could both use only 1 of the 55 P sources tested. Rather, MR-1 and MR-7 were able to utilize 87% and 95% of the P sources and 90% and 76% of the S sources, respectively. The remaining strains utilized only 14% or less of the P and S compounds. Considering that strains MR-1 and MR-7 have a more distant evolutionary history (87.70% ANI), their similar growth phenotypes for P and S sources might reflect the importance of whole-cell networks and expression in regulating ecological fitness under certain environmental conditions. High-throughput phenotypic arrays were not designed to capture the interplay between whole-cell gene expression and environmental conditions. While it is clear that the present study succeeds in deciphering novel phenotypic traits from a combination of high-throughput phenotypes and comparative genomics, future work on linking the evolution of the phenotypes to environmental fitness will need to include genome-wide analyses.

The present study addressed the importance of bridging genomes and their counterpart phenotypes. The strategy devised here could accelerate the functional discovery of genes and provide an ecological framework for genome sequencing projects.

Supplementary Material

[Supplemental material]

ACKNOWLEDGMENTS

We express our gratitude to the following members of the Shewanella Federation: Jim Fredrickson, Kostas Konstantinidis, Ken Nealson, and Margie Romine.

The work conducted by the U.S. Department of Energy Joint Genome Institute is supported by the Office of Science of the U.S. Department of Energy under contract numbers DE-AC02-05CH11231 and DE-FG02-08ER64511.

Footnotes

Supplemental material for this article may be found at http://aem.asm.org/.

[down-pointing small open triangle]Published ahead of print on 3 June 2011.

REFERENCES

1. Bochner B. R. 2008. Global phenotypic characterization of bacteria. FEMS Microbiol. Rev. 33:191–205 [PMC free article] [PubMed]
2. Bochner B. R. 2009. New technologies to assess genotype-phenotype relationships. Nat. Rev. Genetics 4:309–314 [PubMed]
3. Brettar I., Moore E. R., Hofle M. G. 2001. Phylogeny and abundance of novel denitrifying bacteria isolated from the water column of the Central Baltic sea. Microb. Ecol. 42:295–305 [PubMed]
4. Brocker M., Schaffer S., Mack C., Bott M. 2009. Citrate utilization by Corynebacterium glutamicum is controlled by the CitAB two-component system through positive regulation of the citrate transport genes citH and tctCBA. J. Bacteriol. 191:3869–3880 [PMC free article] [PubMed]
5. Caspi R., et al. 2008. The MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Res. 36:D623–D631 [PMC free article] [PubMed]
6. de Hoon I. J., Imoto S., Nolan J., Miyano S. 2004. Open source clustering software. Bioinformatics 12:1453–1454 [PubMed]
7. Driscoll M. E., et al. 2007. Identification of diverse carbon utilization pathways in Shewanella oneidensis MR-1 via expression profiling. Genome Inform. 18:287–298 [PubMed]
8. Eisenbeis S., Lohmiller S., Valdebenito M., Leicht S., Braun V. 2008. NagA-dependent uptake of N-acetyl-glucosamine and N-acetyl-chitin oligosaccharides across the outer membrane of Caulobacter crescentus. J. Bacteriol. 190:5230–5238 [PMC free article] [PubMed]
9. Fonnesbech-Vogel B., Jorgernsen K., Christensen H., Olsen J. E., Gram L. 1997. Differentiation of Shewanella putrefaciens and Shewanella alga on the basis of whole-cell protein profiles, ribotyping, phenotypic characterization and 16S rRNA gene sequence analysis. Appl. Environ. Microbiol. 73:2189–2199 [PMC free article] [PubMed]
10. Fredrickson J. K., et al. 2008. Towards environmental systems biology of Shewanella. Nat. Rev. Microbiol. 6:592–603 [PubMed]
11. Gao H., et al. 2006. Shewanella loihica sp. nov., isolated from iron-rich microbial mats in the Pacific Ocean. Int. J. Syst. Evol. Microbiol. 56:1911–1916 [PubMed]
12. Hau H. H., Gralnick J. A. 2007. Ecology and biotechnology of the genus Shewanella. Annu. Rev. Microbiol. 61:237–258 [PubMed]
13. Heidelberg J. F., et al. 2002. Genome sequence of the dissimilatory metal ion-reducing bacterium Shewanella oneidensis. Nat. Biotechnol. 20:1118–1123 [PubMed]
14. Ivanova E. P., et al. 2004. Shewanella affinis sp. nov., isolated from marine invertebrates. Int. J. Syst. Evol. Microbiol. 54:1089–1093 [PubMed]
15. Jeon J. H., et al. 2009. Cloning and characterization of a new cold-active lipase from a deep-sea sediment metagenome. Appl. Microbiol. Biotechnol. 81:865–874 [PubMed]
16. Johnson D. A., et al. 2008. High-throughput phenotypic characterization of Pseudomonas aeruginosa membrane transport genes. PLoS Genet. 4:e1000211. [PMC free article] [PubMed]
17. Karpinets T. V., et al. 27 June 2010. Shewanella Knowledgebase: integration of the experimental data and computational predictions suggests a biological role for transcription of intergenic regions. Database doi: 10.1093/database/baq012 [PMC free article] [PubMed]
18. Kettler G. C., et al. 2007. Patterns and implications of gene gain and loss in the evolution of Prochlorococcus. PLoS Genet. 3:2515–2528 [PMC free article] [PubMed]
19. Keymer D. P., Miller M. C., Schoolnik G. K., Boehm A. B. 2007. Genomic and phenotypic diversity of coastal Vibrio cholerae strains is linked to environmental factors. Appl. Environ. Microbiol. 73:3705–3714 [PMC free article] [PubMed]
20. Konstantinidis K. T., Tiedje J. M. 2005. Genomic insights that advance the species definition for prokaryotes. Proc. Natl. Acad. Sci. U. S. A. 102:2567–2572 [PMC free article] [PubMed]
21. Konstantinidis K., et al. 2009. Comparative systems biology across an evolutionary gradient within the Shewanella genus. Proc. Natl. Acad. Sci. U. S. A. 106:15909–15914 [PMC free article] [PubMed]
22. Lohmiller S., Hantke K., Patzer S. I., Braun V. 2008. TonB-dependent maltose transport by Caulobacter crescentus. Microbiology 154:1748–1754 [PubMed]
23. Merlin C., McAteer S., Masters M. 2002. Tools for characterization of Escherichia coli genes of unknown function. J. Bacteriol. 184:4573–4581 [PMC free article] [PubMed]
24. Moran M. A., et al. 2007. Ecological genomics of marine roseobacters. Appl. Environ. Microbiol. 73:4559–4569 [PMC free article] [PubMed]
25. Moreira-Turcq P., Seyler P., Guyot J. L., Etcheber H. 2003. Exportation of organic carbon from the Amazon river and its main tributaries. Hydrol. Processes 17:1329–1344
26. Murray A. E., et al. 2001. DNA/DNA hybridization to microarrays reveals gene-specific differences between closely related microbial genomes. Proc. Natl. Acad. Sci. U. S. A. 98:9853–9859 [PMC free article] [PubMed]
27. Myers C. R., Nealson K. H. 1988. Bacterial manganese reduction and growth with manganese oxide as sole electron acceptor. Science 240:1319–1321 [PubMed]
28. Nealson K. H., Myers C. R., Wimpee B. B. 1991. Isolation and identification of manganese-reducing bacteria and estimates of microbial Mn(IV)-reducing potential in the Black Sea. Deep Sea Res. A 38:S907–S920
29. Ohkawa I., Shiga S., Kageyama M. 1979. An esterase on the outer membrane of Pseudomonas aeruginosa for the hydrolysis of long chain acyl esters. J. Biochem. 86:643–656 [PubMed]
30. Pinchuk G. E., et al. 2010. Constraint-based model of Shewanella oneidensis MR-1 metabolism: a tool for data analysis and hypothesis generation. PLoS Comput. Biol. 6:e1000822. [PMC free article] [PubMed]
31. Pinchuk G. E., et al. 2009. Genomic reconstruction of Shewanella oneidensis MR-1 metabolism reveals a previously uncharacterized machinery for lactate utilization. Proc. Natl. Acad. Sci. U. S. A. 106:2874–2879 [PMC free article] [PubMed]
32. Ringo E., Stenberg E., Strom A. R. 1984. Amino acid and lactate catabolism in trimethylamine oxide respiration of Alteromonas putrefaciens NCMB 1735. Appl. Environ. Microbiol. 47:1084–1089 [PMC free article] [PubMed]
33. Rocap G., et al. 2003. Genome divergence in two Prochlorococcus ecotypes reflects oceanic niche differentiation. Nature 424:1042–1047 [PubMed]
34. Rodionov D. A., et al. 2010. Genomic encyclopedia of sugar utilization pathways in the Shewanella genus. BMC Genomics 11:494–512 [PMC free article] [PubMed]
35. Romine M. F., Carlson T. S., Norbeck A. D., McCue L. A., Lipton M. S. 2008. Identification of mobile elements and pseudogenes in the Shewanella oneidensis MR-1 genome. Appl. Environ. Microbiol. 74:3257–3265 [PMC free article] [PubMed]
36. Serres M. H., et al. 2001. A functional update of the Escherichia coli K-12 genome. Genome Biol. 2:RESEARCH0035. http://genomebiology.com/2001/2/9/research/0035./ [PMC free article] [PubMed]
37. Serres M. H., Riley M. 2006. Genomic analysis of carbon source metabolism of Shewanella oneidensis MR-1: predictions versus experiments. J. Bacteriol. 188:4601–4609 [PMC free article] [PubMed]
38. Stapleton R., et al. 2005. Metal reduction at cold temperatures by Shewanella isolates from various marine environments. Aquat. Microbiol. 38:81–91
39. Tettelin, et al. 2005. Genome analysis of multiple pathogenic isolates of Streptococcus agalactiae: implications for the microbial “pan-genome.” Proc. Natl. Acad. Sci. U. S. A. 105:13950–13955 [PMC free article] [PubMed]
40. Wenkateswaran K., Dollholpf M. E., Aller R., Stackebrandt E., Nealson K. H. 1998. Shewanella amazonensis sp. nov., a novel metal-reducing facultative anaerobe from Amazonian shelf muds. Int. J. Syst. Bacteriol. 48:965–972 [PubMed]
41. Wenkateswaran K., et al. 1999. Polyphasic taxonomy of the genus Shewanella and description of Shewanella oneidensis sp. nov. Int. J. Syst. Bacteriol. 49:705–724 [PubMed]
42. Yang C., et al. 2006. Comparative genomics and experimental characterization of N-acetylglucosamine utilization pathway of Shewanella oneidensis. J. Biol. Chem. 281:29872–29885 [PubMed]
43. Zhou L., Lei X.-H., Bochner B. R., Wanner B. L. 2003. Phenotype micro-array analysis of Escherichia coli K-12 mutants with deletions of all two-component systems. J. Bacteriol. 185:4956–4972 [PMC free article] [PubMed]

Articles from Applied and Environmental Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links