Comparative Genome Analysis of Enterobacter cloacae

The Enterobacter cloacae species includes an extremely diverse group of bacteria that are associated with plants, soil and humans. Publication of the complete genome sequence of the plant growth-promoting endophytic E. cloacae subsp. cloacae ENHKU01 provided an opportunity to perform the first comparative genome analysis between strains of this dynamic species. Examination of the pan-genome of E. cloacae showed that the conserved core genome retains the general physiological and survival genes of the species, while genomic factors in plasmids and variable regions determine the virulence of the human pathogenic E. cloacae strain; additionally, the diversity of fimbriae contributes to variation in colonization and host determination of different E. cloacae strains. Comparative genome analysis further illustrated that E. cloacae strains possess multiple mechanisms for antagonistic action against other microorganisms, which involve the production of siderophores and various antimicrobial compounds, such as bacteriocins, chitinases and antibiotic resistance proteins. The presence of Type VI secretion systems is expected to provide further fitness advantages for E. cloacae in microbial competition, thus allowing it to survive in different environments. Competition assays were performed to support our observations in genomic analysis, where E. cloacae subsp. cloacae ENHKU01 demonstrated antagonistic activities against a wide range of plant pathogenic fungal and bacterial species.


Introduction
Enterobacter cloacae is a gram-negative Proteobacterium belonging to the Enterobacteriaceae family. Within this family, Enterobacter is most closely related to, and is grouped in a sub-clade with, Klebsiella. The two groups of bacteria diverged from a sub-clade consisting of Escherichia, Citrobacter and Salmonella [1].
The E. cloacae species comprises an extremely diverse group of bacteria that has been found in diverse environments, ranging from plants to soil to humans. Plant pathogenic strains of E. cloacae have been reported to cause Enterobacter bulb decay in onion plants and bacterial wilt in mulberry [2,3]; endophytic E. cloacae strains have been shown to colonize and benefit plant growth in various crops, such as soybean, cucumber, corn, rice and ginger [4,5]. Previous biological studies of several plant-origin isolates have shown that E. cloacae has antagonistic effects against the oomycete pathogen Pythium ultimum [6], the fungal pathogens Fusarium moniliforme and Fusarium oxysporum [5,7] and the bacterial pathogen Ralstonia solanacearum [8]. Additionally, several strains of E. cloacae are considered to be plant growthpromoting rhizobacteria (PGPR). The genomic features that underlie the antagonistic characteristics have been demonstrated in previous genome analyses of model PGPR species, such as the gram-negative Pseudomonas and the gram-positive Bacillus [9,10,11], but these features have not been studied in Enterobacter spp.
E. cloacae is best known as a human opportunistic pathogen that is commonly found in hospitals and causes a wide range of infections, such as lower respiratory tract infections, urinary tract infections and meningitis [12]. Outbreaks usually occur in Intensive Care Units, primarily affecting patients in vulnerable age groups and patients who are hospitalized for a prolonged period. E. cloacae is clinically significant, particularly because its strains usually carry multiple antibiotic resistance genes [13,14]. The complete genome sequence was published for E. cloacae subsp. cloacae ATCC13047, a classic strain isolated in the last century from human brain fluids, but no detailed analysis or supporting data were reported [15].
Compared with other genera in the family of Enterobacteriaceae such as Escherichia, Salmonella and Yesinia, which are well-known for their association with pathogenicity in humans and animals, there is scarcity of genomic data for Enterobacter. Eight 638. These genome sequencing projects mostly emphasized the potential for the use of Enterobacter in industrial applications, such as lignin degradation [1,16] and 2,3-butanediol production [17,18,19]. A detailed genome study illustrated the synergistic interactions between the poplar tree host and the growth promoting endophyte Enterobacter sp. 638 [18]. Four of the available Enterobacter genomes belong to E. cloacae, which include human opportunistic pathogens E. cloacae subsp. cloacae NCTC9394 and E. cloacae subsp. cloacae ATCC13047 [15], a plant pathogen E. cloacae EcWSU1 [20] and a 2,3-butanediol producing E. cloacae subsp. dissolvens SDM [19].
Using the complete genome sequence of the plant growthpromoting endophytic E. cloacae subsp. cloacae ENHKU01 and the available complete genome sequence data from different strains of E. cloacae [15,19,20,21], we have performed a comparative genome analysis between E. cloacae strains that originated from diverse environments. In this study, we identified crucial conserved genomic factors that support the antagonistic functions of E. cloacae. Genomic factors differentiating the various strains of E. cloacae were also investigated. Also, this study is the first comprehensive comparative genome analysis for E. cloacae genomes.

Materials and Methods
Discovery of E. cloacae subsp. cloacae ENHKU01 E. cloacae subsp. cloacae ENHKU01 was isolated from a pepper plant infected by R. solanacearum in Hong Kong in May 2010. Three single colony subcultures were performed to obtain a pure isolate on TTC (2,3,5-triphenyl-tetrazolium chloride) medium w/v 1.5% agar plates at 28°C. An inoculation and re-isolation experiment using 4-week old tomato and pepper seedlings was carried out to confirm its endophytic characteristic in plants. Unlike R. solanacearum, the bacterial isolate did not cause wilting symptoms in these plants following inoculation. Inoculation on potato tubers and onion bulbs also indicated that the isolate did not cause disease in these plants. Sequencing and BLAST analysis of partial 16S rRNA and housekeeping genes (fusA, gyrB, hsp60, rpoB) [22] was conducted for classification, and the bacterial strain was shown to belong to E. cloacae and was most closely related to E. cloacae subsp. cloacae ATCC13047.

Genome sequencing and comparative analysis
Whole genome sequencing and annotation of E. cloacae subsp. cloacae ENHKU01 was performed as described in Liu et al., 2012 [21]. In brief, de novo shotgun sequencing and paired-end sequencing strategies were applied to produce the whole genome sequence of ENHKU01 using the 454 GS  Junior platform (454 Life Sciences, Branford, CT, USA). Newbler Assembler software (454 Life Sciences, Branford, CT, USA) was used to construct a draft genome with one scaffold containing 36 contigs [23]. To complete the whole-genome sequence of E. cloacae subsp. cloacae ENHKU01, sequence gaps were filled by PCR, primer walking and Sanger sequencing.
Gene annotation and analysis was performed using NCBI Prokaryotic Genomes Automatic Annotation Pipeline (PGAAP) [24]. Annotation was performed using BLASTP, and the protein sequences of predicted genes were searched against all proteins from complete microbial genomes and aligned with the best BLAST-hit [25]. Genome sequence data have been deposited to NCBI, GenBank accession number is CP003737.1, and it is available for download at NCBI: ftp:// ftp.ncbi.nlm.nih.gov/genomes/Bacteria/ Enterobacter_cloacae_ENHKU01_uid172463/.
For comparative genome analysis, complete genome sequence data of Enterobacter species: E. aerogenes KCTC2190 (GenBank accession CP002824), E. cloacae subsp. cloacae ATCC13047 (GenBank accession CP001918, CP001919 and CP001920), E. cloacae subsp. dissolvens SDM (GenBank accession CP003678), E. lignolyticus SCF1 (named as E. cloacae SCF1 in NCBI database) (GenBank accession CP002272), E. asburiae LF7a (CP003026, CP003027 and CP003028), E. cloacae EcWSU1 (GenBank accession CP002886 and CP002887) and Enterobacter sp. 638 (GenBank accession CP000653 and CP000654) were obtained from NCBI (ftp://ftp.ncbi.nlm.nih.gov/genomes/Bacteria/). E. cloacae subsp. cloacae NCTC9394 sequence data were not available for download; thus, it was not included in this study. Scaffolds of the Enterobacter species involved in this study were also uploaded to the Rapid Annotations using Subsystems Technology (RAST) server for SEED-based automated annotation, whole-genome sequence-based comparative analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathway for comparative analysis [26]. Efficient Database framework for comparative Genome Analyses using BLAST score Ratios (EDGAR) was used for core genome, pan genome and singleton analysis, and Venn diagram construction using E. cloacae subsp. cloacae ENHKU01 as a reference genome [27]. Further comparative analysis was performed for specific regions and genes-ofinterest by BLASTN, BLASTX and BLASTP.

Phylogenomic analysis of Enterobacter spp
An in-house pipeline has been developed for the phylogenomic analysis of Enterobacter spp. Scaffolds and genome data of the chromosome of eight Enterobacter genomes and three Pantoea genomes: Pantoea sp. At-9b (NC_014837), P. vagans C9-1 (NC_014562) and P. ananatis LMG20103 (NC_013956), were downloaded from NCBI (ftp:// ftp.ncbi.nlm.nih.gov/genomes/). Gene clustering was performed using OrthoMCL 2.0 under default parameters [28]. A minimum length coverage filter of 85% was applied to further confirm orthology and the families that passed were aligned by MUSCLE v3.8.31 [29]. The corresponding amino acid alignments of 1732 core genes were created and then concatenated to construct a "supergene" for the reconstruction of the phylogenomic tree of Enterobacter spp. by MrBayes v3.2 using the WAG model. Ten million generations were performed with four chains; burn-in was set to 10,000 generations [30].
Phylogenetic analysis. To analyze the phylogeny of specific genes-of-interest, partial-or full-length CDS of E. cloacae subsp. cloacae ENHKU01 was used as a bait to search for orthologs in the NCBI database by BLASTP/ BLASTX using an E-value cut-off of 1e-10. Protein sequences of the top 100 hits were obtained. MEGA5 was used to perform phylogenetic analysis, multiple alignments of protein sequences were built by CLUSTALW and a 100 bootstrap replicate Neighbor-joining tree was constructed for phylogenetic analysis [31].
Phylogenetic analysis of the Type VI Secretion System (T6SS) was performed using a similar strategy as described in previous studies with modifications [32]. T6SS component genes were searched in the Integrated Microbial Genome database version IMG 3.4 in Joint Genome Institute (http:// img.jgi.doe.gov) with 1350 finished bacterial genomes. T6SS clusters containing core component genes COG0542 (clpV) with at least four of the following loci: COG3516 (ImpB), COG3517 (ImpC), COG3519 (VasA), COG3520 (VasB) and COG3522 (ImpJ/VasE) were included for further analysis. T6SS clusters with fewer than five core component genes were discarded. As a result, a total of 346 T6SS clusters were identified in 230 bacterial genomes, mostly belonging to the five subdivisions of Proteobacteria. Two T6SSs of ENHKU01 were manually identified and reconfirmed by BLAST and included in the analysis. MEGA5 was used to perform phylogenetic analysis for the ClpV orthologs in a total of 348 T6SS [31]. Multiple alignments of ClpV protein sequences were built by CLUSTALW, and a Neighbor-joining tree was constructed for evolutionary analysis.

Isolation of other bacterial, fungal isolates and antagonistic bioassays
Bioassays were designed to investigate the antagonistic activities of E. cloacae subsp. cloacae ENHKU01 against various plant bacterial and fungal pathogens. All biological samples involved in this study were collected within a 200 km zone in Hong Kong SAR and Guangdong, China. Plant pathogenic fungi Alternaria sp., Choanephora infundibulifera, Colletotrichum capsici, Didymella bryoniae, Fusarium oxysporum, Sclerotinia sclerotiorum and Sclerotinia rolfsii were isolated from infected host plants and purified by subsequent subculture on potato dextrose agar (PDA) plates at 28°C; next, pure isolates were obtained and strain identification was confirmed by cloning and sequencing (Clover Seed Co., Ltd., Hong Kong). For fungal antagonistic assays, each challenging fungus was grown for 4 days at 28°C. A sample composed of 0.5 cm of agar with hyphae was cut and placed at 2 cm from the edge of a new potato dextrose agar (PDA) plate. An overnight culture of ENHKU01 was streaked across the middle of the plate [33]. Controls were set up for challenging fungus without E. cloacae subsp. cloacae ENHKU01. Fungal growth was monitored for two weeks. Three independent experiments were performed for each E. cloacae subsp. cloacae ENHKU01fungal antagonistic assay.
The bacterial isolate, Ralstonia solanacearum, was isolated from the same pepper plant as the E. cloacae subsp. cloacae ENHKU01 isolate. Three single-colony subcultures were performed to obtain a pure isolate on TTC medium w/v 1.5% agar plates at 28°C. Identification of R. solanacearum was confirmed by cloning and sequencing of partial 16S rRNA and fulfilling Koch's Postulates (Clover Seed Co., Ltd., Hong Kong) [34]. E. cloacae subsp. cloacae ENHKU01-R. solanacearum competition assays were performed in both planktonic and biofilm cultures. To prepare for the competition assay, bacterial cultures were grown overnight in TTC medium and adjusted to OD 600 =0.1(~1 x 10 8 c.f.u/ml). For the planktonic culture, equal volumes (25 ml) of ENHKU01 and R. solanacearum were mixed in a 1:1 ratio in a 250 ml flask and incubated in a shaking incubator (250rpm at 28°C). One milliliter of culture was collected at 2, 4, 6 and 24 hours after incubation for quantification. For the biofilm culture, ENHKU01 and R. solanacearum were mixed in a 1:10 ratio. 20 µl of ENHKU01 culture was placed on a TTC w/v 1.5% agar plate and incubated at 28°C. Biofilm samples were collected for quantification using 1 ml pipette tips and re-suspended in water after 24 hours of incubation. Controls consisted of samples containing only ENHKU01 or R. solanacearum for both planktonic and biofilm competition assays. For quantification (including initial concentration of ENHKU01 and R. solanacearum), serial dilutions were prepared for each sample to a desirable concentration and plated on TTC w/v 1.5% agar. C.F.U.s were counted after incubation at 28°C for 48 hours. ENHKU01 and R. solanacearum can be distinguished by morphological differences in size, color and texture. Four replications were performed for each assay.

Phylogenomic analysis of Enterobacter
The classification of Enterobacter spp. using 16S rRNA and house-keeping genes has often been inconsistent [22]. It has also been a great challenge for researchers to distinguish the environmental and the pathogenic strains of E. cloacae in previous studies [22]. Phylogenomic analysis was performed using 1732 core genes of eight Enterobacter spp. and three Pantoea spp. The four functionally distinctive strains of E. cloacae (the plant growth promoting E. cloacae subsp. cloacae ENHKU01 (ENHKU01) ,the plant pathogen E. cloacae EcWSU1 (EcWSU1), the opportunistic human pathogen E. cloacae subsp. cloacae ATCC13047 (ATCC13047) and the 2,3-butanediol producing E. cloacae subsp. dissolvens SDM (SDM)) were clustered together, and this observation is consistent with previous studies of the phylogenetic relationships of E. cloacae [22] (Figure 1).

General properties of the E. cloacae genomes
The general genome properties of ENHKU01, EcWSU1, ATCC13047 and SDM are shown in (Table 1). The genome of E. cloacae subsp. cloacae ENHKU01 is composed solely of a single 4.72-Mbp chromosome, and a total of 4338 protein coding regions (CDS) were predicted, with 87% being connected to Clusters of Orthologous Groups (COGs). The genome size, total number of genes and predicted CDS of ENHKU01 is slightly smaller but similar to that of EcWSU1 and SDM, while ATCC13047 appears to have a larger genome of 5.31 Mb. The genome expansion of ATCC13047 is due to the addition of two plasmids and more than 20 genomic variable regions in its chromosome. EcWSU01 contains one plasmid, and SDM lacks plasmids. There is an absence of sequence similarity between the plasmids in different strains of E. cloacae. The genomic GC content of the strains ranges between 54.5-55.1%. MAUVE analysis showed an overall collinear relationship across E. cloacae strains, despite a large scale chromosomal reorganization that occurred in ATCC13047 that was conferred by a single recombination event and resulted in the inversion of a genomic region ( Figure  2) [35].
Major subsystems and metabolic pathways are conserved between E. cloacae strains; however, the number of genes is increased in ATCC13047 in several functional categories compared to other E. cloacae strains (Figure 3). E. cloacae strains, similar to other Enterobacter, are characterized by their ability to use a wide range of carbon sources through their diverse carbohydrate metabolic pathways and transport systems [18]. Over 640 of annotated genes, accounting for 13-15% of the E. cloacae genomes, had a designated role for carbohydrate utilization (Figure 3; Data S1), this number is independently. Functional roles of RAST annotated genes were assigned and grouped in subsystem feature categories as shown in the figure [26], and colored bars indicate the number of genes assigned to each category. Details for subsystem and functional role assignment for the genes of each strain are listed in Data S1. doi: 10.1371/journal.pone.0074487.g003 comparable to that of the related genomes in the family of Enterobacteriacae [36,37,38].
Pan-genome analysis was performed for eight Enterobacter species, and a further comparison was performed for E. cloacae strains (Table S1). As shown in the Venn diagram ( Figure 4; Data S2), the four E. cloacae strains shared 3540 CDS in their core genome, corresponding to approximately 64%-82% of all CDS in these genomes. A relatively small portion of CDS were shared between two or three E. cloacae. Both ENHKU01 and SDM had approximately 6% of unique CDS that are absent in the other E. cloacae genomes evaluated, and the unique CDS percentage was 12% and 20% for EcWSU1 and ATCC13047, respectively. The majority of singletons were mostly associated with hypothetical proteins (Data S3), and this observation is consistent with singleton analysis of other species [39,40]. However, a larger number of functional singletons, most likely contributing to virulence, were identified in ATCC13047 (File 4).

Virulence associated genes
The virulence genes of pathogenic bacteria are often associated with pathogenicity islands that encode a Type III secretion system (T3SS) or a Type IV secretion system (T4SS) acquired by horizontal gene transfer [41,42]. These genomic factors related to pathogenesis and virulence were identified in  Figure 2; Table S1). T4SS is associated with pathogenesis in plants and mammalian bacterial pathogens and contributes to genome plasticity in bacteria [43,44]. DNA, effector proteins and virulence factors are delivered to the target host cells by the conjugative transfer machine, although several nonpathogenic strains also carry the membrane transfer system [43,45]. The plasmid pECL_A also contains multiple heavy metal resistance operons for copper, tellurium and mercury that are not conserved with other Enterobacter species but share notably high homology to Cronobacter sakazakii, Klebseilla pneumoniae and Escherichia coli (data not shown), which are all species commonly known as human pathogens or opportunistic pathogens. ATCC13047 is believed to acquire multiple heavy metal resistance genes horizontally from other human microflora associated bacteria, thereby contributing to the adaptation and fitness of the bacteria in heavy metal-rich environments, such as sewage [15]. The pathogenesis of many plant pathogens often involves T3SS and its associated effector proteins [46]. Although E. cloacae EcWSU1 was reported to be a plant pathogen that causes Enterobacter bulb decay in the onion, no T3SS or T4SS was observed in either the plasmid or the chromosome. It remains largely unclear how EcWSU1 became a plant pathogen.

Diversity of fimbriae contributes to variation in colonization and host determination of different E. cloacae strains
Adhesion to and colonization of the host is one of the key factors for host determination for symbiotic and endophytic bacteria and determining the success of pathogenesis for pathogenic bacteria [47,48]. Fimbriae, also known as pili, are widely distributed among the Proteobacteria and are known to be critical in adhesion and the specific binding to tissues of preferred hosts [49]. Located in the outer membrane, fimbriae subunits are usually assembled into filament structures using the chaperone/usher pathway [47]. We compared the fimbriae in different strains of E. cloacae and observed significant diversity. Nine to thirteen fimbrial protein-encoding loci were identified in each strain, but only four of them were conserved across all four E. cloacae strains ( Table 2; Table S1). ATCC13047 and EcWSU1 each had two unique fimbrial clusters, and four of the ENHKU01 fimbrial clusters are absent in other E. cloacae (Table 2; File 2). The presence of multiple fimbriae genes indicates that E. cloacae are likely able to colonize a rather wide range of hosts or environments. Variation in fimbriae between different strains of E. cloacae is expected to alter the choice of hosts or environmental niches that E. cloacae strain can colonize and thus to contribute to the diversity of the species.
One of the fimbrial loci in ATCC13047 and SDM, encoding the Colonization Factor Antigen I (CFA/I) fimbrial proteins (ECL_00070-00074), has drawn our attention. The locus is located in a cluster of genes encoding the virulence related resistance-nodulation-division (RND) efflux system within a variable region that is absent in the plant-associated ENHKU01 and EcWSU1strains [50]. CFA/I fimbriae, as a member of alpha-fimbriae belonging to Class 5 fimbriae, have been shown to play a critical role in colonizing the epithelia of the human intestine in the enterotoxigenic strains of Escherichia coli (ETEC) [51]. BLAST and alignment analyses further demonstrated that the fimbria are highly associated with human-related Enterobacter species [52] (Figure S1), but they are rarely found in plant-associated strains. The Enterobacter CFA/I fimbriae is most closely related to Escherichia coli ( Figure S1), which illustrates that the CFA/I fimbrial gene cluster is a viable candidate for facilitating Enterobacter adhesion and colonization in humans. Such a specific correlation was not observed in the other fimbriae found in E. cloacae.

Antagonistic potential of E. cloacae: Microbial competition for resources
Microbes compete with each other for limited resources within their communities [53,54]. Certain microbes, primarily demonstrated for Pseudomonas spp., compete with other microbes for acquiring ferric ion (iron), an essential growth element from soil. These microbes produce higher affinity siderophores and indirectly suppress the growth of competing fungi, which produce lower affinity fungal siderophores, and non-siderophore producing bacteria [55,56]. Resource competition mechanisms thus lead to antagonistic effects in the rhizosphere. A conserved pathway for enterobacin synthesis was found across Enterobacter spp. In addition, E. cloacae strains possess an extra siderophore assembly kit for aerobactin siderophore biosynthesis, which is absent in certain other Enterobacter spp. (Table 3; Table S1).
Other resources subjected to microbial competition include fatty acids. Earlier functional studies demonstrated that E. cloacae competes for the seed exudate linoleic acid against P. ultimum, which suppresses sporangium germination of the oomycete pathogen and causes the damping-off of seedlings in many crop plants [57]. In this study, we have identified two candidate genes in the linoleic acid metabolic pathway that could play an important role in fatty acid competition. Phospholipase A1 Precursor (E.C. 3.1.1.4) and Lysophospholipase L2 (EC 3.1.1.5) are expected to hydrolyze lecithin (Table 3; File 2), which is an exudate lipid molecule composed of choline, linoleic acid, phosphorus and inositol, and initiate the fungal/ oomycete responses of seeds. Both genes are located adjunct to a multiple sugar transporter and are found across different Enterobacter genomes [57].

Chitinases
The ability of microbes to produce a wide range of antimicrobial compounds, including lytic agents, antibiotics, bacteriocins, protein exotoxins and other secondary metabolites, is critical to their success in antagonistic activities. In bacterial-fungal interactions, cell wall hydrolases, such as certain proteases and chitinases, are produced and secreted extracellularly by bacteria to target fungal cell walls [58,59,60]. Chitinases are lytic enzymes that break down glycosidic bonds in chitin, a major component of fungal cell walls, and have been demonstrated to effectively inhibit fungal growth. Three conserved putative chitinase genes have been identified across E. cloacae genomes. Two of these genes are located adjacent to a type II secretion system [61], while the third one is located elsewhere ( Table 3; Table S1). The Type II secretion system and the associated chitinases are absent in Enterobacter sp. 638 and E. lignolyticus SCF1. Phylogenetic analysis has shown that ECENHK_07430 and the synthetic Enterobacter chitinases are associated with the well-characterized Chitinase A in Serratia marcescens [62,63] (Figure 5), which has been demonstrated to control several important plant pathogenic fungi including Botrytis spp., Rhizoctonia solani, Fusarium oxysporum f. sp. cyclaminis and Sclerotinia spp. In fact, a functional Chitinase A that effectively controls Rhizoctonia solani has been identified in E. agglomerans [64]. The other two chitinases have not been reported in any E. cloacae isolates. Orthologs of ECENHK_08915, can also be found in several related species of Klebsiella, Citrobacter and Serretia, but it is absent in other family members of Enterobacteriaceae.
Interestingly, phylogenetic analysis further demonstrated the association between the Enterobacter's chitinase and Chitinase A1 of the gram-positive Bacillus spp. (Figure 5). It is notable that Bacillus spp. are well-known as plant growth-promoting rhizobacteria (PGPR), as they have antagonistic effects against other microorganisms and are used commercially as biological control agents for various fungal pathogens [65,66].

Production of other antimicrobial compounds
Several bacteria and fungi exhibit antagonistic effects in microbial competition through the production of antibiotics [67]. In microbial-microbial competition, other bacteria also develop detoxification mechanisms and/or antibiotic resistance to  [37,68], these multiple antimicrobial mechanisms were also observed in E. cloacae (Table S2). One can expect, due to the wide range of antimicrobial resistance, multidrug resistance proteins and multidrug efflux systems benefit the survival of E. cloacae against both bacterial and fungal competitors in habitats with high microbial competition, and furthermore, this phenomenon explains how clinical strains of E. cloacae raise medical concerns [13,14]. The production of bacteriocins also plays an important role, particularly in bacteria-bacteria interactions [69]. Wellcharacterized mechanisms in gram-negative bacteria include the enteric bacteriocin, colicin [70] and the lipoprotein Entericidin [71]. Antimicrobial compounds identified in E. cloacae that are related to bacteriocin production involve the Colicin V and bacteriocin production cluster [69], tolerance to colicinE2, Entericidin A and B [71] and S-type Pyocin (Table 3;  Table S1). Additionally, other bacteriocin production-related genes were found to be associated with the Type VI secretion system (T6SS) in E. cloacae.

Type VI secretion systems and their roles in microbial competition
The T6SS is widely distributed in gram-negative bacteria. A T6SS usually involves approximately 15 conserved genes in a cluster with genes encoding a functional secretory apparatus that penetrates cell membranes and translocates effector proteins into their eukaryotic hosts or recipient cells [72,73]. Earlier studies have related the T6SS to virulence in humans and animals [74]; however, recent studies have demonstrated the directed function of the T6SS toward microbial interaction and fitness in microflora [75,76,77,78]. It is common to find more than one cluster of T6SS genes within a genome [32,79,80]; both ATCC13047 and ENHKU01 contain two clusters of T6SS, while SDM and EcWSU1 have only one. Each T6SS gene cluster is composed of sets of conserved core component genes and variable regions distinguished by hypothetical proteins (Figure 6). A recent discovery allowed us to understand more about the roles of these hypothetical proteins in microbial competition as anti-bacterial effectors. Genetics and characterization of the bacteriolytic effectors Tse1, Tse2 and Tse3 in Psuedomonas aeruginosa revealed that T6SS effector proteins target the peptidoglycan of prokaryotic cells [76]. Within the variable region of E. cloacae T6SS, we found RhsB (a rhs-like genetic element involved in bacteriocin production) [81,82,83] and LysM (involved in bacterial cell wall degradation) [84,85] (Figure 6). It is believed that bacteria deliver different effector proteins to their specific target counterparts among the microflora using a conserved T6SS apparatus [76,83].  [86]. Each clade is contributed by different bacterial families of Proteobacteria. Our result is consistent with a previous phylogenetic analysis of T6SS using 13 T6SS conserved component genes [86]. ClpV of E. cloacae are distributed in clades II and III of the phylogenetic tree and are clustered together with other strains and species in the family of Enterobacteriaceae possessing T6SS, thus forming sub-trees in each clade (indicated by gray circles). A simplified and enlarged version of the neighbor-joining tree with bootstrap values showing (B) the sub-tree of clade III formed by ClpV orthologs associated with ECENHK_13140 and (C) the subtree of clade II associated with ECENHK_15865. Color squares indicate the habitats of the corresponding bacterial species: plant/ soil-associated (light green), insect-associated (green) and Human/ animal associated (orange). doi: 10.1371/journal.pone.0074487.g007 The sequences and genetic organization of core component genes in each T6SS cluster are aligned across the genomes of Enterobacter, but the T6SS clusters within a genome are not associated with one another (Figure 6; Data S4). There is also  Figure  6). Such differentiation indicates that these T6SSs are not only evolutionarily distinct but also functionally and structurally different from each other [32]. Phylogenetic analysis of the representative T6SS component gene ortholog, ClpV, indicated that two evolutionary distinct T6SSs were introduced into E. cloacae independently before divergence, as ClpV orthologs of E. cloacae are distributed in two clades (Clade II and Clade III) ( Figure 7A). In clade III, orthologs are clustered with those found in other plant and soil-associated bacteria of Enterobacteriaceae, such as the plant pathogens Erwinia amylovora and Pectobacteria wasabiae, and soil-borne bacteria, such as Pantoea vagans C9-1 and Erwinia pyrifoliae Ep1/96 ( Figure 7B); also, there are many human and animal pathogenic/associated bacteria in this family, such as Escherichia coli, Klebsiella pneumoniae, Yersinia pestis are located in Clade II ( Figure 7C). Our observation of the correlation between the phylogeny of T6SSs and the ecological niches of bacteria is in line with a previous T6SS phylogenetic analysis study [86].
The presence of T6SS in E. cloacae strains is expected to allow the bacterial strain to enhance the survival of nonpathogenic strains in their habitats, thus competing with other microbes for the colonization of hosts [75]. E. cloacae strains possessing more than one distinct T6SS cluster, such as ATCC13047 and ENHKU01, are likely to have fitness advantages in a borader range of habitats and be able to survive in different environments.

Antagonistic activities of E. cloacae subsp. cloacae ENHKU01
Through genome analysis, the antagonistic potential of E. cloacae has been revealed in this study, and we further hypothesized that E. cloacae has advantages in microbial competition against other microbes within its environmental niche. To investigate the effect of antagonistic activities in E. cloacae strains further, we performed bacterial-fungal and bacterial-bacterial antagonistic assays using the plant growthpromoting endophytic ENHKU01 as a model. As expected, ENHKU01 showed significant antagonistic effects against a wide range of fungal species: Collectotricum capsici. Sclerotinia sclerotiorum, Alternaria sp., Didymella bryoniae, and Fusarium oxysporum by suppressing their growth at different levels ( Figure 8; Table S3). These fungal species are common plant pathogens, and C. capsici and S. sclerotiorum are direct competitors with ENHKU01, sharing the pepper plant as a host. Additionally, in our bacteria-bacteria competition assay, it was demonstrated that ENHKU01 effectively suppressed the growth of Ralstonia solanacearum, a devastating bacterial pathogen with a wide host range that is most well-known for causing bacterial wilt in Solanaceous crops, such as tomato, pepper and eggplant ( Figure 9). Taken together, these bioassays provide further evidence, drawn from comparative genome analysis, demonstrating the antagonistic potential of E. cloacae.

Conclusions
Comparative genome analysis reveals the antagonistic potential of E. cloacae. The multiple antagonistic mechanisms of E. cloacae are expected to contribute its success in competition against other microbes in various environmental niches, thus allowing different strains of E. cloacae to survive in diverse environments.

Supporting Information
Data S1. The functional roles of RAST annotated genes were curated into subsystems by SEED Viewer, and the details are listed in this file. The numbers of genes grouped in each subsystem feature category for each E. cloacae strain were counted as shown in the summary page. (XLSX) Data S2. A Venn diagram of the data generated by EDGAR [27]. (XLSX)

Data S3.
List of singletons of ENHKU01, EcWSU1, ATCC13047 and SDM generated by comparing the genomes of the four of E. cloacae strains using E. cloacae subsp. cloacae ENHKU01 as the reference genome [27]. (XLSX) Data S4. List of genes associated with Type VI Secretion Systems in Enterobacter, corresponding to Figure 6A and Figure 6B.  Table S1. Pan-genome analysis of Enterobacter by EDGAR.In brief, E. cloacae subsp. cloacae ENHKU01 was selected as reference genome and its gene contents were taken as the base set for stepwise comparison of the seven Enterobacter genomes involved in this study. For each round of all-againstall comparison based on a bidirectional best BLAST hit, every gene in a Enterobacter genome that has no ortholog in the base set was added to the reference set and compared with the next genome. The process was repeated step by step to obtain the pan-genome of Enterobacter [27]. (XLSX )   Table S2. List of genes involved in potential antibiotic resistance, multidrug resistance and multidrug efflux systems in Enterobacter. (XLSX)