Free-Living Enterobacterium Pragia fontium 24613: Complete Genome Sequence and Metabolic Profiling.

Pragia fontium is one of the few species that belongs to the group of atypical hydrogen sulfide-producing enterobacteria. Unlike other members of this closely related group, P. fontium is not associated with any known host and has been reported as a free-living bacterium. Whole genome sequencing and metabolic fingerprinting confirmed the phylogenetic position of P. fontium inside the group of atypical H2S producers. Genomic data have revealed that P. fontium 24613 has limited pathogenic potential, although there are signs of genome decay. Although the lack of specific virulence factors and no association with a host species suggest a free-living style, the signs of genome decay suggest a process of adaptation to an as-yet-unknown host.

Introduction

To date, the Enterobacteriaceae family contains 55 genera and 248 species (www.bacterio.net, September 1, 2016). Most of the enterobacteria live in the vertebrate intestine, whereas several other enterobacterial genera/species represent plant pathogens or invertebrate endosymbionts.[1] Other enterobacteria are believed to live only in the environment, eg, Pragia, Saccharobacter, Obesumbacterium, Shimwellia,[1] Mangrovibacter,[2] and Biostraticola.[3] However, it is possible that their pathogenic/symbiotic potential will be revealed in the future, as it was for Budvicia.[4,5]

Pragia fontium is a gram-negative, mesophilic, rod-shaped, motile bacterium. The genus Pragia contains only 1 species, P. fontium, which was described in 1988.[6] A total of 18 strains were isolated in Czechoslovakia between 1982 and 1986. All strains, except 1, were isolated from water wells and water pipes, whereas 1 strain was obtained from the stool of a healthy woman. Another set of Pragia strains was isolated in Ukraine between 1996 and 1997.[7] They were mostly isolated from water (9 strains) and other environmental material (5 strains), although 2 strains came from human clinical material; their relatedness to the Czechoslovakia strains varied from 84% to 95% (based on DNA-DNA hybridization). To date, only strains from these 2 locations have been characterized, and the exact ecological niche and pathogenic potential of Pragia remains unclear.

Pragia fontium, as well as Budvicia spp. and Leminorella spp., is a closely related atypical enterobacterial species. Their common feature is hydrogen sulfide production, with Budvicia diplopodorum being the only known exception.[5] These H2S-producing enterobacteria share several metabolic features including reduced metabolic activity that results in utilization of a limited set of substrates. The optimal growth temperature for Pragia and Budvicia is 25°C, whereas Leminorella is capable of growing at temperatures up to 42°C.[6] Pragia fontium can be differentiated from Budvicia spp. based on a positive (Simmons) citrate utilization test and from Leminorella spp. by its motility, tartrate utilization, tyrosine clearing, and inability to grow at 42°C.[6] In addition, a whole-cell protein pattern analysis of P. fontium, B. aquatica, and Leminorella spp. was determined and the data supported the delineation of these genera.[8] On the DNA level, Pragia strains were most closely related to Budvicia (based on DNA-DNA hybridization, relatedness 20%-37%) but barely related to other genera, eg, relatedness to Escherichia coli K12 was about 3%.[9]

To date, 485 completed enterobacterial genome sequences, covering 21 genera and 47 species, have been deposited in the Genomes OnLine Database (GOLD, https://gold.jgi.doe.gov/). Attention has been focused mainly on clinically and agriculturally important bacteria (eg, Escherichia, Salmonella, Klebsiella, and Yersinia), leaving the remaining genera relatively unexplored.

The whole genome sequence and the pilot assembly of P. fontium 24613 were published in 2015.[10] In this study, we characterized P. fontium based on genomic data, including the relationship of Pragia to other genera, and compared metabolic pathways with the results of phenotypic metabolic fingerprinting.

Materials and Methods

Bacterial strains and cultivation conditions

The strains used in this study came from the collection of the Department of Biology, Masaryk University, Brno, Czech Republic (P. fontium 24613, originally stored at the National Institute of Public Health, Prague, Czech Republic); from the Czech National Collection of Type Cultures, Prague, Czech Republic (Budvicia aquatica CNCTC 6285T); and from the Czech Collection of Microorganisms, Brno, Czech Republic (Leminorella grimontii CCM 4003T). Pragia fontium 24613 came from the same set of strains as P. fontium DSM 5563T6. Strains were cultivated in TY medium (8 g casein, 5 g yeast extract, 5 g sodium chloride, pH 7.5; HiMedia, Mumbai, India) at 30°C for 24 hours.

Pragia fontium 24613 genome sequencing and annotation

In our previous study, protocols for DNA extraction, whole genome sequencing, and annotation of P. fontium 24613 were described in detail.[10] For additional gene mining and genome comparisons, annotation was manually curated based on results of a RAST (Rapid Annotation using Subsystem Technology) pipeline[11] and DOE-JGI (US Department of Energy-Joint Genome Institute) Microbial Genome Annotation Pipeline.[12] Detected proteins were assigned to Clusters of Orthologous Group (COG) categories based on DOE-JGI results. Methylome was characterized using PacBio single-molecule real-time sequencing (1× SMRT cell) of kinetic data collected during the genome sequencing process.[13] SMRT analysis version 2.3, using the “RS_Modification_and_Motif_Analysis.1” protocol, was used for genome-wide base modification and detection of the affected motifs. Regarding sequencing coverage, a default quality score value of 30 (corresponding to a P value of .001) was used for motif determination. The detected motifs were uploaded and further analyzed using the REBASE database.[14] The complete genome was also scanned for homologues of restriction-modification system genes (using a Basic Local Alignment Search Tool [BLAST] search, with the BLASTX algorithm) against the REBASE and GenBank databases.

Phylogenetic position of P. fontium

The genome sequence of P. fontium 24613 was compared with other enterobacterial genera on a genome-wide level. Whole genome sequences were downloaded from the GOLD (https://gold.jgi.doe.gov/); their accession numbers are listed in Table S1. Each genus was represented by 1 sequence (except for Pragia where both the type strain DSM 5563T and strain 24613 were used). If available, the sequence of the type strain was used. For genera Biostraticola, Cosenzaea, Gibbsiella, Mangrovibacter, Obesumbacterium, Saccharobacter, and Samsonia, no sequences were available. A whole genome phylogenetic analysis was built using PhyloPhlAn 0.99,[15] which compared more than 400 selected protein sequences conserved across bacterial domains. The genes were identified using an internal PhyloPhlAn database by translated mapping with USEARCH 8.1.[16] The topology was computed using the neighbor-joining algorithm in conjunction with the Jukes-Cantor evolution model. Moreover, the CAT model, with gamma correction, was used to optimize and rescale the tree. The final tree was reconstructed, using FastTree 2.1,[17] from protein subsequences of the genes concate-nating their most informative amino acid positions, and each was aligned using MUSCLE 3.8.[18] The tree was visualized in MEGA 6.06.[19] Dot plot diagrams between genomes were constructed using the Integrated Microbial Genome platform.[12] The core genome of P. fontium, B. aquatica, and L. grimontii was determined based on orthologous clusters produced by OrthoVenn[20] using a modified OrthoMCL heuristic approach. Default parameters (E-value 1e−5 and inflation value 1.5) were used. Metabolic pathway analysis of P. fontium 24613, Wigglesworthia glossinidia (acc. no. CP003315), and Buchnera aphidicola G002 (acc. no. CP002701) was performed using the KEGG PATHWAY database,[21] which is part of KEGG Web services (http://www.genome.jp/kegg/).

Analyses of metagenomics data

Data from the Human Microbiome Project database (http://hmpdacc.org) and EBI Metagenomics database (https://www.ebi.ac.uk/metagenomics/) were searched with BLASTN 2.2.22[22] using a consensus sequence of 7 16S ribosomal RNA (rRNA) genes of P. fontium 24613. The first database contained a complete set of human microbiome data (associating data from several human sites), and the latter database covered data from different environmental sources.

Substrate diversity studies

The Biolog GN2 MicroPlate analysis platform (Biolog, Inc., Hayward, CA, USA) was used for determination of the biochemical profiles of P. fontium 24613, B. aquatica CNCTC 6285T, and L. grimontii CCM 4003T cultivated on Biolog Universal Growth (BUG) agar at 30°C for 24 hours. Utilization of 95 carbon sources was tested[23] (Table S2). Media and all reagents were supplied by Biolog and used according to the manufacturer’s protocol. Plates were incubated in parallel under aerobic and anaerobic conditions and tests were read after 24 hours of incubation.

Results

Genome analyses of P. fontium 24613

Complete genome sequence of P. fontium 24613

A complete genome sequence for P. fontium 24613 represents a single circular chromosome with a length of 4 094 629 bp.[10] The P. fontium 24613 genome was compared with 3 draft genomes of related bacteria, including the draft genome of P. fontium DSM 5563T (Table 1). Both the B. aquatica DSM 5075T genome and the L. grimontii DSM 5078Tgenome were larger in size and gene count compared with the complete genome sequence of P. fontium 24613. Moreover, the proportion of pseudogenes was larger in P. fontium (4.1%) than in the draft genomes of other H2S-producing enterobacteria (ie, 2.8% and 1.6% for B. aquatica and L. grimontii, respectively), suggesting genome decay in P. fontium. In addition, a clearly higher GC content was found in the L. grimontii DSM 5078T (L. grimontii) genome. The draft status was likely responsible for the lower number of predicted rRNA and transfer RNA genes in the P. fontium DSM 5563T, B. aquatica, and L. grimontii genomes.

Table 1.

Genome features of Pragia fontium 24613 in comparison with the draft genome of the type strain and the draft genomes of closely related hydrogen sulfide producers.

Feature	P. fontium 24613	P. fontium DSM 5563T	B. aquatica DSM 5075T	L. grimontii DSM 5078T
Genome status	Complete	Draft	Draft	Draft
Genome size	4 094 629 bp	3 950 845 bp	5 670 930 bp	4 222 128 bp
GC content	45.38%	45.23%	45.68%	53.86%
No. of CDS	3579	3464	5130	3878
No. of rRNA genes	22 (8–7–7)	10 (2–5–3)	7 (5–2–0)	16 (8–6–2)
No. of tRNA genes	72	58	57	57
No. of pseudogenes	146 (4.1%)	NA	144 (2.8%)	62 (1.6%)
No. of genes with predicted function	2809 (78.49%)	2862 (82.62%)	3896 (75.95%)	3083 (79.50%)
No. of genes assigned to COG	2601 (72.67%)	2613 (75.43%)	2601 (72.67%)	2804 (72.31%)
No. of genes assigned to KEGG pathways	1160 (32.41%)	1172 (33.83%)	1419 (27.66%)	1217 (31.38%)

Abbreviations: CDS, coding sequences; COG, Clusters of Orthologous Group; KEGG, Kyoto Encyclopedia of Genes and Genomes; rRNA, ribosomal RNA; tRNA, transfer RNA.

Accession numbers of the whole genome sequences of the type strains are listed in Table S1. Order of the rRNA genes in parentheses: 5S-16S-23S. NA—data not available in the GenBank and Genomes OnLine databases.

Phylogenetic position of P. fontium

A whole genome phylogenetic approach was used to compare the genome sequence of P. fontium 24613 with genome sequences of other enterobacterial genera. The relevant part of the Enterobacteriaceae tree is shown in Figure 1. Strong support was found for a close relationship among Pragia and other atypical H2S producers, including Budvicia and Leminorella. The high similarity among genomes was also supported by a dot plot analysis of H2S producer genomes (Figure S1). Another related genus was Plesiomonas, an oxidase-positive genus recently reclassified into the Enterobacteriaceae family.[24] A sister clade contains a cluster of genera occurring frequently in the (1) environment (Providencia, Moellerella, Proteus, and Morganella), (2) genera associated with nematodes (Xenorhabdus, Photorhabdus), and (3) endosymbionts (Arsenophonus, Buchnera, and Wigglesworthia). Except for the delineation of Proteus vs Morganella and endosymbionts Buchnera vs Wigglesworthia, all other branches were supported by bootstrap values higher than 99%.

Figure 1.

Phylogenetic position of the genus Pragia based on a whole genome sequence tree. Only the relevant part of the Enterobacteriaceae tree is shown (the whole tree is depicted in Figure S2). All branches are supported with high bootstrap values. The tree was drawn to scale; the scale bar represents the estimated number of amino acid changes per site per unit of branch length.

The core genome of enterobacterial hydrogen sulfide producers

The core genomes of P. fontium, B. aquatica, and L. grimontii contain 2327 gene clusters (ie, at least 1 gene from each cluster was found in each genome; Figure 2). The number of gene clusters exclusively shared by 2 genomes was higher for the P. fontium and B. aquatica genomes (325) compared with the P. fontium and L. grimontii genomes (109), whereas there were 494 clusters shared by the L. grimontii and B. aquatica genomes. These data indicate a higher degree of relatedness between P. fontium and B. aquatica compared with P. fontium and L. grimontii. A set of 30 gene clusters was unique for the P. fontium genome; these clusters encoded homologues to fimbrial genes found in Serratia spp. and Proteus spp. and also homologues to pyocin S3 and its immunity protein–encoding genes. In total, 10 clusters encoded genes for hypothetical proteins.

The Venn diagram represents the core genome and pangenome of Pragia and the closely related atypical H2S producers. The numbers represent the gene clusters shared by corresponding group of genera. The diagram shows the close relationships among those inside the group of atypical H2S producers.

Genome-based metabolic and virulence analyses

Analysis of metabolic pathways in the P. fontium genome

Based on the genomic data analysis from KEGG PATHWAY and DOE-JGI, aerobic and facultative anaerobic metabolism of P. fontium 24613 was predicted. Oxidized nitrogen and sulfur compounds were capable of serving as alternative terminal electron acceptors under anaerobic conditions. Identification of thiosulfate reductase, responsible for H2S production, corresponded to previously detected enzyme activity.[6] The genes involved in glycolysis, citrate cycle, and pentose phosphate pathway could also be found in the P. fontium genome in addition to genes responsible for amino acid, fatty acid synthesis, lipid, and nucleotide metabolism. Pragia was found to be auxotrophic for l-tryptophan, l-histidine, and l-leucine and deficient in biotin synthesis. Compared with Budvicia and Leminorella, Pragia was able to synthetize l-arginine but lacked the genes for fatty acid degradation. In addition, the P. fontium genome contained fewer genes involved in carbohydrate metabolism compared with the L. grimontii and B. aquatica genomes (Table S3).

Genome methylation pattern

Analysis of PacBio sequencing data revealed 24 814 methylated positions of the m6A type, but only a single sequence motif (GATC) was found in all these modifications. More than 80% (21 735 of 26 606) of the GATC positions in the genome were methylated. Methylation type m4C was not found. Kinetic signatures of m5C were subtler than signatures of m6A and m4C and harder to detect using PacBio SMRT sequencing[25]; therefore, they were not assessed. The results of P. fontium genome methylation were deposited in the REBASE PacBio database (http://rebase.neb.com/cgi-bin/pacbiolist).[14] In total, 8 different putative restriction-modification systems, all of them type II, were predicted in the genome (Table S4). Seven of them consisted of only methyltransferases, whereas the last one modifying m5C consisted of methyltransferase, mismatch repair endonuclease, and restriction endonuclease.

Virulence and antimicrobial genes in the P. fontium genome

In silico analysis of virulence determinants of the P. fontium genome revealed genes involved in iron acquisition (encoding Fe2+ and Fe3+ transport systems), adhesion (encoding P pili and type I pili), secretion systems (T1SS and T6SS), and antibiotic resistance (encoding AmpC β-lactamase and several efflux pump) (see Table S5).

Production of tailocins, ie, R-type and F-type bacteriocins resembling phage tails, was previously detected in several Pragia strains.[26] Gene clusters similar to the phage genes were described as being responsible for production of these antimicrobial compounds.[27] A total of 6 clusters homologous to phage genes were predicted in the P. fontium genome, and one of them was likely responsible for tailocin production (see Table S5). The genome search also detected a gene encoding a colicin-like bacteriocin, a homologue of pyocin S3.

Metabolic profiling of P. fontium 24613

The carbohydrate utilization pattern resulting from the testing of various saccharides, carboxylic acids, alcohols, amino acids, aromatic compounds, and their derivatives was determined for P. fontium 24613, B. aquatica CNCTC 6285T, and L. grimontii CCM 4003T. In general, the data obtained from the Biolog assay revealed low levels of metabolic activity in all tested strains. Substrate utilization profiles differed for the 3 tested H2S producers in 17 substrates (Table S2). Pragia fontium 24613 was able to utilize 15 substrates (out of 95; 16%) under aerobic conditions and 22 (out of 95; 23%) under anaerobic conditions. Pragia utilized monosaccharides and their derivatives (α-d-glucose, α-d-glucose-1-phosphate, d-glucose-6-phosphate, N-acetyl-d-glucosamine, and β-methyl-d-glucoside), monocarboxylic acids (d,l-lactic acid, and d-gluconic acid), dicarboxylic acids (α-keto-glutaric acid, and l-glutamic acid), alcohols and their derivatives (glycerol, d,l-α-glycerol phosphate, myo-inositol, and xylitol), amino acids (d-serine), and aromatic compounds (uridine and thymidine). In addition to substrates utilized under aerobic conditions, anaerobically cultivated Pragia utilized l-arabinose, pyruvic acid methyl ester, d-glucuronic acid, bromosuccinic acid, l-aspartic acid, glycyl-l-aspartic acid, and l-serine. Although Budvicia utilized 16 substrates (17%) aerobically and 24 (25%) anaerobically, Leminorella utilized only 13 substrates (14%) aerobically and 18 (19%) anaerobically. Budvicia and Leminorella were able to metabolize several amino acids and their derivatives (l-asparagine, l-aspartic acid, and glycyl-l-aspartic acid) as well as derivatives of organic acids from Krebs cycle (pyruvic acid methyl ester, bromosuccinic acid), which were not utilized by Pragia. The complete results of this assay are shown in Table S2. In most of the substrate tests, which differed among H2S producers, the genes encoding corresponding catabolic enzymes or enzymes possibly involved in metabolism of these compounds were found (Table S6). The only exception was the B. aquatica genome, where some of the genes responsible for catabolism of uridine were not found.

Discussion

Pragia belongs to a relatively small group of H2S-producing enterobacteria containing P. fontium, Budvicia spp., and Leminorella spp. Although all members of this small group are closely related and have a relatively similar biochemical profile, they occupy quite different ecological niches. Although Budvicia was originally isolated from freshwater,[28] several other isolates have been described from the intestinal microflora of insects,[26,29] Diplopoda,[5] and salmonids.[30] A possible clinical relevance for B. aquatica was reported by Corbin et al[4] when this bacterium was isolated from a human clinical sample. Leminorella spp. have been exclusively isolated from human clinical specimens and no environmental sources have been reported. Although its clinical significance is unclear,[1] Leminorella spp. appear to be associated with urinary tract infections and other human nosocomial infections.[31] In contrast to Budvicia and Leminorella, Pragia has been isolated almost exclusively from environmental sources. Only 3 isolates originated from human clinical samples; there is no information on the role of these strains in infection or disease.[6,7] Because the prevalent habitat of other Pragia strains is drinking water, these cases likely reflect accidental isolations. Inspection of metagenomics data revealed the absence of Pragia 16S ribosomal DNA (rDNA) in both environmental and host-associated data sets (data not shown). From all the available data, Pragia appears to be the only H2S producer occupying environmental niches with no association with humans or other hosts.

A possible interaction between Pragia and a host species was examined by identification and analysis of genes encoding virulence factors. Several common virulence factors shared by most enterobacterial species (even saprophytic ones) were detected. Genes for adhesion, antibiotic resistance, iron uptake, and 2 secretion systems were found. Adhesion and the ability to acquire iron are key factors required for colonization and survival in a host (animal or plant).[32-34] These findings indirectly support an association between Pragia and an as-yet-unknown host. We can speculate that if a host organism exists, it will likely be similar to those of the closely related genus Budvicia, ie, nonvertebrate hosts such as insects or nematodes. Although the presence of Pragia has been detected in the intestines of freshwater salmon,[35] the much more frequent isolation from deepwater wells[6] tends to support a free-living lifestyle of Pragia. Both detected secretion systems, T1SS and T6SS, are widely distributed in gram-negative bacteria[36,37] and could mediate interaction with a host or with another bacterium.[38] Although the contribution of T6SS to pathogenesis has been described for several bacteria, eg, Pseudomonas[39] and E coli,[40] T6SS has also been found in saprophytic bacteria, where it was involved in interactions across the microbial community.[38] Several bacteriocin types have been suggested as putative virulence factors, whereas the importance of others was demonstrated in interactions across microbial community.[41,42] Although the function of P. fontium bacteriocins remains unknown, both tailocins and colicin-like homologues were found in the Pragia genome. The GATC methylation motif was found in the P. fontium genome, and because the corresponding gene for the restriction enzyme recognizing this motif was not found, methylation appears to be more connected to gene expression regulation[43] and not to degradation of foreign nucleic acid molecules.

Metabolic profiling revealed a metabolic pattern for Pragia, Budvicia, and Leminorella, which was quite distinct from other enterobacteria,[44] supporting the distinctness of enterobacterial H2S producers and also the close relationship of these bacteria within this group. Despite their overall similarity, H2S-producing enterobacteria revealed several differences in their ability to utilize substrates. Analyses of genomic data supported the metabolic findings, with only 1 case in which some of the genes encoding expected enzymatic activity were not found. This is likely a result of an incomplete genomic sequence in Budvicia. Surprisingly, all species were able to degrade multiple substrates under anaerobic conditions suggesting that alternative electron acceptors (nitrate, reduced sulfur compounds) could be used under anaerobic conditions. Nitrogen oxidation could be carried out using the “nitrite reduction to ammonium pathway” for which the corresponding genes were found in the P. fontium genome. This pathway is preferred for respiration under anaerobic conditions, and it is common across Enterobacteriaceae and in other facultatively anaerobic bacteria.[45]

A comparative genomics approach revealed that almost 80% of the gene clusters were shared by H2S-producing enterobacteria, whereas only 49% were shared when E coli K12 was added to the analysis. Analysis of the complete genome sequence of Pragia revealed that the genome contains genes involved in essential metabolic pathways, in nutrient metabolism, and also in the synthesis of most of the amino acids. However, the “fatty acid degradation pathway” is missing from the P. fontium 24613 genome. This pathway is present in most enterobacterial genomes but not in invertebrate endosymbionts with a reduced genome, such as Wigglesworthia and Buchnera. Nevertheless, when compared with these endosymbionts, the P. fontium genome is relatively large and also contains an additional set of genes, eg, those responsible for degradation of more complex polysaccharides. However, P. fontium 24613 has a relatively small genome in comparison with other enterobacteria, even in comparison with the genus Budvicia. In addition, the proportion of pseudogenes was larger in Pragia compared with other closely related bacteria (despite their draft status, which is prone to assembly errors). Larger proportions of pseudogenes have also been observed in bacteria that were associated with or dependent on eukaryotic hosts.[46] Nevertheless, this analysis comes from a limited number of genome sequences per species and it is known that the prevalence of pseudogenes is quite variable even among closely related strains.[47] A reduction in genome size and an increased number of pseudogenes are common signs of bacterial adaptation to a eukaryotic host. In addition, the P. fontium genome contains fewer genes involved in carbohydrate utilization compared with other H2S producers; a large battery of degradation enzymes is important mainly for free-living bacteria. The traces of genome decay (ie, small genome, absence of fatty acid degradation pathways, the small number of genes associated with carbohydrate utilization, and a larger proportion of pseudogenes) suggest an ongoing process of adaptation to a particular host organism. Although no such host has been identified for P. fontium, the recent progress in metagenome studies could help to answer this question in the near future.

Conclusions

Analysis of the complete genome sequence of P. fontium 24613 and metabolic profiling confirmed the close relatedness of this bacterium to other H2S-producing enterobacteria, Budvicia spp. and Leminorella spp., although for each genus a different environmental niche has been described. Virulence gene mining and the absence of Pragia 16S rDNA sequences in the human metagenomics data suggest limited pathogenic potential for Pragia, consistent with the previously described free-living lifestyle of this bacterium. On the contrary, reduced genome size, limited number of encoded enzymes for carbohydrate and fatty acid degradation, and frequent presence of pseudogenes suggest a process of adaptation to an as-yet-unknown host.