Introduction Bacteria have evolved many mechanisms of defense against competitors and predators in their environment. Some of these, such as type III secretion systems (T3SSs) and bacteriocins, provide specialized protection against eukaryotic or bacterial cells, respectively [1], [2]. Gene clusters encoding apparent type VI secretion systems (T6SSs) are widely dispersed in the proteobacteria; however, the general roles of these systems in eukaryotic versus bacterial cell interactions are not known [3], [4]. To date, most studies of T6S have focused on its role in pathogenesis and host interactions [5], [6], [7]. In certain instances, compelling evidence for the specialization of T6S in guiding eukaryotic cell interactions has been generated. Most notably, the systems of Vibrio cholerae and Aeromonas hydrophila were shown to translocate proteins with host effector domains into eukaryotic cells [8], [9]. Evidence is also emerging that T6SSs could contribute to interactions between bacteria. The Pseudomonas aeruginosa HSI-I-encoded T6SS (H1-T6SS) was shown to target a toxin to other P. aeruginosa cells, but not to eukaryotic cells [10]. Unfortunately, analyses of the ecological niche occupied by bacteria that possess T6S have not been widely informative for classifying their function [3], [4]. These efforts are complicated by the fact that pathogenic proteobacteria have environmental reservoirs, where they undoubtedly encounter other bacteria. The observation that many bacteria possess multiple evolutionarily distinct T6S gene clusters–up to six in one organism–raises the intriguing possibility that each system may function in an organismal or context-specific manner [3]. The T6SS is encoded by approximately 15 core genes and a variable number of non-conserved accessory elements [4]. Data from functional assays and protein localization studies suggest that these proteins assemble into a multi-component secretory apparatus [11], [12], [13]. The AAA+ family ATPase, ClpV, is one of only a few core proteins of the T6S apparatus that have been characterized. Its ATPase activity is essential for T6S function [14], and it associates with several other conserved T6S proteins [15], [16]. ClpV-interacting proteins A and B (VipA and VipB) form tubules that are remodeled by the ATPase, which could indicate a role for the protein in secretion system biogenesis. Two proteins exported by the T6SS are haemolysin co-regulated protein (Hcp) and valine-glycine repeat protein G (VgrG). Secretion of these proteins is co-dependent, and they may be extracellular components of the apparatus [10], [13], [17], [18], [19], [20]. Burkholderia pseudomallei is an environmental saprophyte and the causative agent of melioidosis [21]. Infection with B. pseudomallei typically occurs percutaneously via direct contact with contaminated water or soil, however it can also occur through inhalation. The ecological niche and geographical distribution of B. pseudomallei overlap with a relatively non-pathogenic, but closely related species, Burkholderia thailandensis (B. thai) [22]. The genomes of these bacteria are highly similar in both overall sequence and gene synteny [23], [24]. One study estimates that the two microorganisms separated from a common ancestor approximately 47 million years ago [24]. It is postulated that the B. pseudomallei branch then diverged from Burkholderia mallei, which underwent rapid gene loss and decay during its evolution into an obligate zoonotic pathogen [25]. As closely related organisms that represent three extremes of bacterial adaptation, this Burkholderia group offers unique insight into the outcomes of different selective pressures on the expression and maintenance of certain traits. B. pseudomallei possesses a large and complex repertoire of specialized protein secretion systems, including three T3SSs and six evolutionarily distinct T6SSs [3], [26], [27]. The genomes of B. thailandensis and B. mallei contain unique sets of five of the six B. pseudomallei T6S gene clusters; thus, of the six evolutionarily distinct “Burkholderia T6SSs,” four are conserved among the three species. Remarkably, T6SSs account for over 2% of the coding capacity of the large genomes of these organisms. For the current study, we have adopted the Burkholderia T6SS nomenclature proposed by Shalom and colleagues [28]. To date, only Burkholderia T6SS-5, one of the four conserved systems, has been investigated experimentally. The system was investigated in B. mallei based on its co-regulation with virulence determinants such as actin-based motility and capsule [27]. B. mallei strains lacking a functional T6SS-5 are strongly attenuated in a hamster model of glanders. Preliminary studies suggest that T6SS-5 is also required for B. pseudomallei pathogenesis [28], [29]. In one study, a strain bearing a transposon insertion within T6SS-5 was identified in a screen for B. pseudomallei mutants with impaired intercellular spreading in cultured epithelial cells [29]. The authors also showed that this insertion caused significant attenuation in a murine infection model. Herein, we set out to systematically define the function of the Burkholderia T6SSs. Our study began with the observation that well-characterized examples of eukaryotic and bacterial cell-targeting T6SSs segregate into distant subtrees of the T6S phylogeny. We found that Burkholderia T6SS-5 clustered closely with eukaryotic cell-targeting systems, and was the only system in B. thai that was required for virulence in a murine model of pneumonic melioidosis. The remaining systems clustered proximally to a bacterial cell-targeting T6SS in the phylogeny. One of these, T6SS-1, displayed a profound effect on the fitness of B. thai in competition with several bacterial species. The function of T6SS-1 required cell contact and its absence caused sensitivity of the strain to stasis induced by competing bacteria. In flow cell biofilm assays initiated with 1∶1 mixtures of B. thai and Pseudomonas putida, wild-type B. thai predominated, whereas the ΔT6SS-1 strain was rapidly displaced by P. putida. Our findings point toward an important role for T6S in interspecies bacterial interactions. Results Phylogenetic analysis of T6SSs We conducted phylogenetic analyses of all available T6SSs to examine the evolutionary relationship between eukaryotic and bacterial cell-targeting systems. The phylogenetic tree we constructed was based on VipA, as this protein is a highly conserved element of T6SSs that has been demonstrated to physically interact with two other core T6S proteins, including the ClpV ATPase [15]. In the resulting phylogeny, the systems of V. cholerae and A. hydrophila, two well-characterized eukaryotic cell-targeting systems, clustered closely within one of the subtrees, whereas the bacteria-specific P. aeruginosa H1-T6SS was a member of a distant subtree (Figure 1 and see Figure S1) [8], [9], [10]. In an independent analysis, Bingle and colleagues observed a similar T6S phylogeny, and termed these subtrees “D” and “A,” respectively [3]. 10.1371/journal.ppat.1001068.g001 Figure 1 The Burkholderia T6SSs cluster with eukaryotic and prokaryotic-targeting systems in a T6S phylogeny. (A) Overview of the B. thai T6SS gene clusters. Burkholderia T6SS-3 is absent from B.thai. Genes were identified according to the nomenclature proposed by Shalom and colleagues [28]: tss, type six secretion conserved genes; tag, type six secretion-associated genes variably present in T6SSs. Genes are colored according to function and conservation (dark grey, tss genes; light grey, tag genes; color, experimentally characterized tss or tag genes; white, genes so far not linked to T6S). Brackets demarcate genes that were deleted in order to generate B. thai strains ΔT6SS-1, -2, -4 -5 and -6 and their assorted combinations. Locus tag numbers are based on B. thai E264 genome annotations. (B) Neighbor-joining tree based on 334 T6S-associated VipA orthologs. The locations of VipA proteins from T6SSs discussed in the text are indicated. Each line represents one or more orthologous T6SSs from a single species. Lines are colored based on bacterial taxonomy of the corresponding organism. Indicated bootstrap values correspond to 100 replicates. This phylogeny is available in expanded format in Figure S1. A key for the coloring scheme is also present in Figure S1. Next we examined the locations of the six Burkholderia T6SSs. Interestingly, T6SS-5, the only Burkholderia system previously implicated in virulence, clustered within the substree containing the V. cholerae and A. hydrophila systems (Figure 1). Four of the remaining Burkholderia systems clustered within the subtree that included the H1-T6SS, and the final system was found in a neighboring subtree. These data led us to hypothesize that T6SSs of differing organismal specificities are evolutionarily distinct. Apparent contradictions between organismal specificity based on our phylogenetic distribution and studies demonstrating T6S-dependent phenotypes were identified, however these instances are difficult to interpret because specificity was not measured and cannot be ascertained from available data. T6SS-5 is required for virulence; systems 1, 2, 4 and 6 are dispensible We chose B. thai as a tractable model organism in which to experimentally investigate the role of the Burkholderia T6SSs. Due to our limited knowledge regarding the function and essentiality of each gene within a given T6SS cluster, we reasoned it prudent to inactivate multiple conserved genes for initial phenotypic studies. Strains lacking the function of each of the five B. thai T6SSs (Burkholderia T6SS-3 is absent in B. thai) were prepared by removing three to five genes, including at least two that are highly conserved (Figure 1A). When possible, polar effects were minimized by deleting from a central location in each cluster. To probe the role of the Burkholderia T6SSs in virulence, we utilized a recently developed acute pneumonia model of melioidosis [30]. The survival of mice infected with approximately 105 aerosolized wild-type or mutant bacteria was monitored over the course of ten days. Consistent with previous studies implicating T6SS-5 in B. mallei and B. pseudomallei pathogenesis, mice infected with ΔT6SS-5 survived the course and displayed no outward symptoms of the infection (Figure 2A) [27], [29]. On the other hand, those infected with the wild-type strain or strains bearing deletions in the other T6SSs succumbed by three days post infection (p.i.). 10.1371/journal.ppat.1001068.g002 Figure 2 Of the five B. thai T6SSs, only T6SS-5 is required for virulence in a murine acute melioidosis model. C57BL/6 wild-type mice were infected by the aerosol-route with 105 c.f.u./lung of B. thai strains and monitored for survival for 10–14 days post infection (p.i.). Survival of mice after exposure to B. thai (A) wild-type and strains harboring gene deletions in individual T6SS gene clusters (n = 5 per group), (B) wild-type and a strain bearing an in-frame tssK-5 deletion (ΔtssK-5) or its complemented derivative (ΔtssK-5-comp; n = 7, 7 and 8, respectively), (C) or a strain with inactivating mutations in T6SS-5 or in four T6SSs (ΔT6SS-1,2,4,6; n = 6 and 8, respectively). The B. thai T6SS-5 locus is adjacent to bsa genes, which encode an animal pathogen-like T3SS. Inactivation of the bsa T3SS secretion system also leads to dramatic attenuation of B. thai in the model we utilized [26]. The regulation of these secretion systems appears to be intertwined; a recent study in B. pseudomallei showed that a protein encoded within the bsa cluster strongly activates T6SS-5 of that organism [31]. To rule out the possibility that attenuation of ΔT6SS-5 was attributable to polar effects or changes in regulation of the bsa T3SS, we generated a strain bearing an in-frame deletion of a single gene in the cluster, tssK-5 (Figure 1A). A tssK-5 ortholog is readily identified in nearly all T6S gene clusters and it shares no homology with known regulators. Like the T6SS-5 deletion, ΔtssK-5 completely attenuated the organism (Figure 2B). Genetic complementation of this phenotype further confirmed that T6SS-5 is an essential virulence factor of the organism. To investigate whether the retention of virulence in the ΔT6SS-1,2,4 and 6 strains could be attributed to either compensatory activity or redundancy, we next constructed a strain bearing inactivating mutations in all four clusters and measured its virulence in mice. Mice infected with this strain succumbed to the infection with similar kinetics to those infected with the wild-type, indicating that T6SS-5 is the only system of B. thai that is required for virulence in this model (Figure 2C). In summary, these data indicate that T6SS-5 is a major virulence factor for B. thai in a murine acute melioidosis model, whereas the remaining putative T6SSs of the organism are dispensible for virulence. Burkholderia T6SS-5 plays a specific role in host interactions To more closely examine the requirement for T6SS-5 during infection, we monitored B. thai wild-type and ΔtssK-5 c.f.u. in the lung, liver, and spleen at 4, 24, and 48 hours following inoculation with approximately 105 bacteria by aerosol. At 4 hours p.i., no differences were observed in c.f.u. recovered from the lung (Figure 3A). After this initial phase, lung c.f.u. of ΔtssK-5 gradually declined, whereas wild-type populations expanded approximately 100-fold. Both organisms spread systemically, however significantly fewer ΔtssK-5 cells were recovered from the liver and spleen at 24 and 48 hours p.i. (Figure 3B). 10.1371/journal.ppat.1001068.g003 Figure 3 B. thai ΔtssK-5 shows a replication defect in the lung of wild-type mice but is highly virulent in MyD88−/− mice. Mice were exposed to 105 c.f.u./lung aerosolized B. thai wild-type or ΔtssK-5 bacteria and c.f.u. were monitored in the (A) lung after 4, 24, and 48 h (n = 6 per time point), and in the (B) liver and spleen after 24 and 48 h (n = 6 per time point). (C) C57BL/6 wild-type (n = 6) and MyD88−/− mice (n = 7) were infected with the ΔtssK-5 strain and survival was monitored for 14 days. Error bars in (A) and (B) are ± SD. Thus far, our findings did not distinguish between a specific role for T6SS-5 in host interactions, such as escaping or manipulating the innate immune system, versus the alternative explanation that T6SS-5 is generally required for growth in host tissue. To discriminate between these possibilities, we compared the virulence of ΔtssK-5 in wild-type mice to a strain with compromised innate immunity, MyD88−/− [32], [33]. Mice lacking MyD88 were unable to control the ΔtssK-5 infection and succumbed within 3 days (Figure 3C). The differences in virulence of the Δtssk-5 strain in wild-type and MyD88−/− infections suggest that T6SS-5 is required for effective defense of the bacterium against one or more innate immune responses of the host. Altogether, these data strongly support the conclusion that T6SS-5 has evolved to play a specific role in the fitness of B. thai in a eukaryotic host environment. T6S impacts the fitness of B. thai in co-culture with diverse bacterial species Earlier work by our laboratory has shown that T6S can influence intraspecies bacterial interactions. We showed that the H1-T6SS of P. aeruginosa targets a toxin to other P. aeruginosa cells [10], and that in growth competition assays, toxin-secreting strains are provided a fitness advantage relative to strains lacking a specific toxin immunity protein. Based on this information and the locations of the B. thai T6SSs within our phylogeny, we postulated that one or more of these systems could also play a role in interbacterial interactions. Preliminary studies indicated that T6S did not influence interactions between B. thai strains, thus we decided to test the hypothesis that the B. thai T6SSs play a role in interspecies bacterial interactions. Without information to guide predictions of specificity, we developed a simple and relatively high-throughput semi-quantitative assay to allow screening of a wide range of organisms for sensitivity to the B. thai T6SSs. The design of the assay was based on two key assumptions for T6S-dependent effects – that they are cell contact-dependent and that they impact fitness (as measured by proliferation). To facilitate measurement of T6S-dependent changes in B. thai proliferation in the presence of competing organisms, we engineered constitutive green fluorescent protein expression cassettes into wild-type B. thai and a strain bearing mutations in all five T6SSs (ΔT6S) [34]. Control experiments showed that the lack of T6S function did not impact growth or swimming motility (Figure 4A and 4B). To test the assay, we conducted competition experiments between the GFP-labeled wild-type and ΔT6S strains against the unlabeled wild-type strain. The GFP-expressing cells were clearly visualized in the mixtures, and, importantly, wild-type and ΔT6S competed equally with the parental strain (Figure 4C; BT). 10.1371/journal.ppat.1001068.g004 Figure 4 T6S plays a role in the fitness of B. thai in growth competition assays with other bacteria. (A) In vitro growth of B. thai wild-type and a strain bearing gene deletions in all five T6SSs (ΔT6S). The data presented are an average of three replicates. (error bars smaller than symbols). (B) B. thai wild-type and ΔT6S swimming motility in semi-solid LB agar (scale bar = 1.0 cm). (C) Fluorescence images of growth competition assays between GFP-labeled B. thai wild-type and ΔT6S strains against the indicated unlabeled competitor species. Competition assay outcomes could be divided into T6S-independent (AR, Agrobacterium rhizogenes; ATu, A. tumefaciens; AV, A. vitis; PD, Paracoccus denitrificans; RS, Rhodobacter sphaeroides; ATe, Acidovorax temperans; BT, B. thailandensis; BV, B. vietnamiensis; AC, Acinetobacter calcoaceticus; AH, Aeromonas hydrophila; PAt, Pectobacterium atrosepticum; FN, Francisella novicida; PAe, Pseudomonas aeruginosa; SM, Serratia marcescens; VC, Vibrio cholerae; VP, Vibrio parahaemolyticus; VV, V. vulnificus; XC, Xanthomonas campestris; XN, Xenorhabdus nematophilus; YP, Yersinia pestis LCR–; BC, Bacillus cereus; BS, B. subtilis; ML, Micrococcus luteus; SA, Staphylococcus aureus; SP, Streptococcus pyogenes), those with modest T6S-effects (BA, B. ambifaria; EC, E. coli; KP, Klebsiella pneumoniae; ST, Salmonella typhimurium) and those in which B. thai proliferation was strongly T6S-dependent (dashed boxes – PP, P. putida E0044; PF, P. fluorescens ATCC27663; SP, S . proteamaculans 568). This latter group of organisms is referred to as the T6S-dependent competitors (TDCs). We next screened the B. thai strains against 31 species of bacteria. Most of these were Gram-negative proteobacteria (5α; 3β; 18γ), however two Gram-positive phyla were also represented (4 Firmicutes; 1 Actinobacteria). Although we endeavored to screen a large diversity of bacteria, many taxa could not be included due to specific nutrient requirements or an unacceptably slow growth rate under the conditions of the assay (30°C, Luria-Bertani (LB) medium). The outcomes of most competition experiments were independent of the T6SSs of B. thai. T6S-independent outcomes varied; in most instances, B. thai flourished in the presence of the competing organism (Figure 4C). However, a small subset of species markedly inhibited B. thai growth (Figure 4C; PAt, PAe, SM, VP). Interestingly, B. thai proliferation was reproducibly affected in a T6S-dependent manner in competition experiments against 7 of the 31 species tested. All of these were Gram-negative organisms, and in each case, B. thai ΔT6S was less fit than the wild-type. T6S-dependent competition outcomes fell into two readily discernable groups; the first included three γ- and one β-proteobacteria (Figure 4C; BA, EC, KP, ST). In competition with these organisms, B. thai ΔT6S displayed only a modest decrease in proliferation relative to the wild-type. Differences in the size and morphology of assay “spots” containing wild-type or ΔT6S were noted in several instances for this group of organisms. Quantification of c.f.u. verified that these differences were reflective of a minor, but highly reproducible fitness defect of ΔT6S (data not shown). The second group consisted of three γ-proteobacteria: P. putida, P. fluorescens, and S. proteamaculans. The proliferation of B. thai grown in competition with these organisms appeared to be highly dependent on T6S (Figure 4C; PP, PF, SP). For further analyses, we focused on this latter group; henceforth referred to as the “T6S-dependent competitors” (TDCs). T6SS-1 is involved in cell contact-dependent interbacterial interactions The next question we addressed was whether one or more of the individual T6SSs were responsible for the TDC-specific proliferation phenotype of B. thai ΔT6S. To determine this, we inserted a GFP over-expression cassette into our panel of individual B. thai T6SS deletion strains, and performed plate competition assays against the TDCs. In competition with each TDC, ΔT6SS-1 appeared as deficient in proliferation as ΔT6S, whereas the other strains grew similarly to the wild-type (Figure 5A). The dramatic differences in the competition outcomes between the strains were also discernable by the naked eye. Competition experiments that included B. thai lacking T6SS-1 had a morphology similar to a mono-culture of the TDC, whereas co-cultures possessing an intact T6SS-1 were more similar in appearance to B. thai mono-culture. 10.1371/journal.ppat.1001068.g005 Figure 5 T6SS-1 is involved in cell contact-dependent interbacterial interactions. (A) Growth competition assays between the indicated GFP-labeled B. thai strains and the TDCs. Standard light photographs and fluorescent images of the competition assays are shown. (B) Fluorescence images of GFP-labeled B. thai wild-type and ΔT6SS-1 grown in the presence of the TDCs with (no contact, NC) or without (contact, C) an intervening filter. (C) Fluorescence images of growth competition assays between GFP-labeled B. thai ΔclpV-1 or complemented ΔclpV-1 with the TDCs. (D) Quantification of c.f.u before (initial) and after (final) growth competition assays between the indicated organisms. The c.f.u. ratio of the B. thai strain versus competitor bacteria is plotted. Error bars represent ± SD. It remained possible that the effects of T6SS-1 on the fitness of B. thai in competition with other bacteria were either non-specific or unrelated to its putative role as a T6SS. As mentioned earlier, one common observation from detailed studies of T6SSs conducted to date is that its effects require cell contact [8], [9], [10]. This has been postulated to reflect a conserved mechanism of the apparatus akin to bacteriophage cell puncturing [18]. To address whether the apparent fitness defect of ΔT6SS-1 involves a mechanism consistent with T6S, we probed whether its effects were dependent upon cell contact. A filter (0.2 µm pore diameter) placed between B. thai and TDC cells abrogated the T6SS-1-dependent growth defect (Figure 5B). In control experiments, the three TDCs were directly applied to an underlying layer of the B. thai strains. In each case, a zone of clearing was observed in the ΔT6SS-1 layer, while no effect on wild-type proliferation was noted. From these data we conclude that cell contact is essential for the activity of T6SS-1. We next sought to quantify the magnitude of T6SS-1 effects on B. thai fitness in competition with TDCs. To ensure the specificity of T6SS-1 inactivation in the strains used in these assays, we generated a B. thai strain bearing an in-frame clpV-1 deletion, and a strain in which this deletion was complemented by clpV-1 expression from a neutral site on the chromosome. In plate competition assays, the ΔclpV-1 strain displayed a fitness defect similar to ΔT6SS-1, and clpV-1 expression complemented the phenotype (Figure 5C). Measurements comparing B. thai and TDC c.f.u. in the competition assay inoculum to material recovered from the assays following several days of incubation confirmed that inactivation of T6SS-1 leads to a dramatic fitness defect of B. thai (Figure 5D). Depending on the TDC, the competitive index (c.i.; final c.f.u. ratio/initial c.f.u ratio) of wild-type B. thai was approximately 120-5,000-fold greater than that of the ΔclpV-1 strain. All TDCs out-competed ΔclpV-1 (0.0021
