A type VI secretion-related pathway in Bacteroidetes mediates interbacterial antagonism

Summary

Bacteroidetes are a phylum of Gram-negative bacteria abundant in mammalian-associated polymicrobial communities, where they impact digestion, immunity and resistance to infection. Despite extensive competition at high cell density that occurs in these settings, cell contact-dependent mechanisms of interbacterial antagonism, such as the type VI secretion system (T6SS), have not been defined in this group of organisms. Herein we report the bioinformatic and functional characterization of a T6SS-like pathway in diverse Bacteroidetes. Using prominent human gut commensal and soil-associated species, we demonstrate that these systems localize dynamically within the cell, export antibacterial proteins, and target competitor bacteria. The Bacteroidetes system is a distinct pathway with marked differences in gene content and high evolutionary divergence from the canonical T6S pathway. Our findings offer a potential molecular explanation for the abundance of Bacteroidetes in polymicrobial environments, the observed stability of Bacteroidetes in healthy humans, and the barrier presented by the microbiota against pathogens.

Introduction

The Bacteroidetes are a phylum of Gram-negative bacteria that can be isolated from diverse natural habitats (Thomas et al., 2011). Though they include agriculturally and medically relevant pathogens, as well as representatives that play important roles in critical environmental processes such as bioremediation, the phylum is relatively poorly studied. Bacteroidetes are perhaps most appreciated for their intimate association with humans and other mammals, as abundant residents of the gastrointestinal (GI) tract. In this ecosystem, bacteria form dense microbial communities that can exceed 10¹³ cells per milliliter (Lozupone et al., 2012; Qin et al., 2010; Smith et al., 2006). Bacteroidetes constitute 20–80% of the fecal microbiota of most adult humans. and are largely represented by two genera, Bacteroides and Prevotella (Human Microbiome Project Consortium, 2012). Members of this phylum generally act as mutualists by aiding in the digestion of complex carbohydrates, promoting gut development, modulating the immune system, and protecting against colonization by pathogens (Round and Mazmanian, 2009; Smith et al., 2006; Thomas et al., 2011). As metabolically pliable organisms, Bacteroidetes also help to support a diverse gut community through syntrophic interactions with other microbes (Fischbach and Sonnenburg, 2011; Rey et al., 2010).

Evidence suggests that the capacity of a bacterium to survive in a polymicrobial environment is related to the elaboration of interbacterial antagonistic factors. Studies performed primarily on Proteobacteria have shown that Gram-negative organisms can utilize soluble products as well as contact-dependent mechanisms to compete with other bacteria (Hayes et al., 2014; Riley and Wertz, 2002). Although Bacteroidetes occupy numerous polymicrobial niches, including the human gut, to our knowledge contact-dependent antagonistic pathways have not yet been characterized in this phylum.

The type VI secretion system (T6SS) is a pathway that grants Gram-negative bacteria the capacity to translocate substrates to a wide range of recipient cells (Coulthurst, 2013). Initially speculated to participate strictly in host cell interactions, it is now clear that the more common function of the system is to deliver proteins from the cytoplasm of a donor cell to the periplasm of a Gram-negative recipient (Schwarz et al., 2010a). Substrates transported in a T6S-dependent manner include antibacterial effectors with diverse activities such as phospholipases, peptidoglycan hydrolases, nucleases, and membrane pore-forming proteins (Benz and Meinhart, 2014; Russell et al., 2014). The pathway appears to lack a mechanism for discriminating self from non-self; thus, bacteria with active T6SSs possess immunity proteins that interact with, and inactivate, the effector molecules (Hood et al., 2010; Russell et al., 2011). These interactions are allele specific, and cognate effector–immunity (E–I) pairs are generally encoded adjacently within predicted operons (English et al., 2012; Russell et al., 2012).

The T6S pathway requires the functions of 13 core proteins; unique subsets of these appear to have evolutionary relatedness to type IV secretion system (T4SS) components or to bacteriophage (Boyer et al., 2009; Cascales and Cambillau, 2012). Proteins within the subsets are generally encoded adjacent to each other and interact extensively, suggesting that although each of the 13 core genes is essential, the complete system may be composed of modular, functionally distinct sub-complexes. The T4S-related components, TssL and TssM, are integral membrane proteins that form a trans-envelope complex with an outer-membrane lipoprotein, TssJ (Aschtgen et al., 2010; Felisberto-Rodrigues et al., 2011; Ma et al., 2009). The bacteriophage-like protein TssC, in conjunction with TssB, forms a dynamic filamentous assembly with gross structural similarity to the bacteriophage sheath complex (Basler et al., 2012; Bonemann et al., 2009). Two other bacteriophage-related proteins, VgrG and Hcp, interact with non-overlapping sets of effectors, forming the basis for genetically distinct pathways for T6S-dependent substrate export (Shneider et al., 2013; Silverman et al., 2013; Whitney et al., 2014). VgrG is a phage tail spike-like protein that interacts with effectors via conserved adaptor domains, whereas Hcp is ring-shaped, bears structural homology to the major phage tail tube protein gpV, and interacts with effectors within its pore. Supporting the relationship of Hcp to gpV, Hcp rings have been observed to form higher order head-to-tail stacked structures in vivo, analogous to those observed in bacteriophage (Brunet et al., 2014).

Here, we report the bioinformatic and functional characterization of a T6S-like pathway in the phylum Bacteroidetes. We demonstrate that this pathway has the capacity to mediate cell contact-dependent intra- and inter-phyla bacterial antagonism. Although the pathway lacks conserved elements essential to the well characterized Proteobacterial T6SS, and components that are shared with Proteobacteria display considerable sequence divergence, we provide evidence that they function in a mechanistically similar manner. Several genera that possess the Bacteroidetes T6S-like pathway, including Bacteroides, Prevotella, and Porphyromonas, are abundant members of human-associated polymicrobial communities, suggesting that the pathway may have an important role in defining the composition of the microbiome (Falagas and Siakavellas, 2000).

Results

Bioinformatic characterization of a T6S-like gene cluster in Bacteroidetes

A generally applicable diagnostic secretion signal for T6S effectors is not available; however, genes encoding these proteins can often be identified by sequence-based approaches. Our group and others have noted that in many cases T6S effector and vgrG genes occur proximally and co-directionally on bacterial chromosomes (Barret et al., 2011; Russell et al., 2013; Zhang et al., 2012). We recently exploited this observation to identify a large superfamily of T6S-exported phospholipases. Interestingly, a search for homologs of this class of effectors revealed their presence in Bacteroidetes – a bacterial phylum that does not possess a characterized T6S pathway (Russell et al., 2013). Moreover, like the Proteobacterial effectors, those found in the Bacteroidetes reside in close proximity to apparent vgrG genes and adjacent to open reading frames (ORFs) encoding predicted periplasmic immunity determinants (Figure S1). Our observations concerning phospholipase effector distribution are supported by an exhaustive study of polymorphic toxin domains conducted by Aravind and colleagues, which found genes encoding putative T6 effectors of various catalytic classes represented in Bacteroidetes (Zhang et al., 2012).

Given the abundance and widespread nature of antibacterial T6S effector genes in Bacteroidetes, we postulated that these proteins participate in interbacterial interactions via a yet uncharacterized T6S-like pathway. Proteobacterial T6S gene clusters often include effector loci; thus, to identify a T6S-like pathway in Bacteroidetes, we searched in the vicinity of putative effector genes for elements that could constitute a secretion system. This led to the identification of a cluster of twelve genes, including vgrG, with orthologs invariantly found in species with predicted effectors (Figure 1A–C and Table S1). Supporting the hypothesis that this gene cluster encodes a T6-like pathway, among the products of the twelve genes, we found a predicted ATPase with domain architecture similar to the Proteobacterial T6S core ATPase, ClpV (Figure 1D) (Schlieker et al., 2005).

Figure 1. T6S-like gene clusters are found within the Bacteroidetes.

(A–C) Gene content and conservation both within and between selected representative members of T6SSⁱ (A), T6SSⁱⁱ (B), and T6SSⁱⁱⁱ (C). Genes with commonly used tss nomenclature are abbreviated to a single letter. The Francisella tularensis FPI is depicted owing to its status as the only T6SSⁱⁱ to be characterized; however, this system lacks clear homologs of tssA and tssJ, which are present in representative T6SSⁱⁱ systems such the F. novicida gene cluster shown. Locus tags of the regions shown: B. thailandensis E264 BTH_I2705 and BTH_I2954-2968 (A); F. novicida U112 FTN_0037-0054, F. tularensis SCHU FTT_1344-1361c (B); F. johnsoniae UW101 Fjoh_3254-328, B. fragilis NCTC 9343 BF9343_1918-1943, P. veroralis F0319 HMPREF0973_02422-02423 and HMPREF0973_02441-02466. Genes encoding F. johnsoniae T6SSⁱⁱⁱ substrates identified in this study (dark grey) and a validated immunity locus (light grey) are labeled. For sequence alignments of T6SSⁱⁱⁱ TssB,C,E,F,G,K proteins, including those depicted, see File S1.

(D) Comparison of domain organization of ClpV homologs from T6SSⁱ (B. thailandensis, Bt) and T6SSⁱⁱⁱ (F. johnsoniae, Fj) pathways. Colors denote homologous domains: blue, Clp N; red, AAA+; purple, ClpB D2. Sequences highlight the conservation of motifs implicated in ATP binding and hydrolysis within the Walker A and B motifs.

(E) SDS-PAGE analysis of purified Fjoh_3262 bearing a C-terminal hexahistidine tag (Fj Hcp1–H₆). Proteins were visualized by Coomassie Blue staining.

(F) F. johnsoniae Hcp1 is a ring-shaped molecule with dimensions similar to Proteobacterial Hcp proteins. Transmission electron micrograph of purified Fj Hcp1–H₆ (E) negatively stained with uranyl formate. The arrowhead indicates a typical ring-like assembly. Scale bar, 40 nm.

Homology searches of the remaining conserved elements of the Bacteroidetes gene cluster failed to identify corresponding Proteobacterial T6S proteins. Given the evolutionarily distance between Bacteroidetes and Proteobacteria, we posited that conservation between the constituents of this putative secretion system and the Proteobacterial T6SS might be below the detection limit of non-iterative approaches. By applying iterative search algorithms such as jackHMMER and PSI-BLAST (Altschul et al., 1997; Johnson et al., 2010), we found that six additional genes in the Bacteroidetes cluster encode proteins bearing distant homology to core elements of the Proteobacterial T6SS, TssB, C, E, F, G, and K (Figure 1C and File S1).

In total, our sequence-based searches identified eight of the thirteen putative functional orthologs of the Proteobacterial T6SS encoded within the Bacteroidetes gene cluster (Table S1). Included in these eight components are each of the identified proteins of the bacteriophage-like module of the T6SS, with the exception of Hcp. In the Proteobacterial T6SS, Hcp proteins are required for effector recognition and export, thus the apparent absence of this conserved component was unexpected (Silverman et al., 2013). To ensure we had not overlooked a cryptic Hcp functional ortholog, we turned to structural prediction algorithms, which can identify highly divergent proteins, or convergent proteins, by their common folds. Indeed, using the Phyre remote homology modeling server, we found that one of the remaining four conserved genes in the Bacteroidetes cluster is predicted to encode a protein that adopts a structure with a high degree of similarity to Proteobacterial Hcp proteins (>90% confidence) (Kelley and Sternberg, 2009). To further probe this predicted relatedness, we heterologously expressed and purified a member of this putative Hcp-like protein family encoded within the T6S-like gene cluster of Flavobacterium johnsoniae, a soil-dwelling member of the Bacteroidetes phylum(Figure 1E) (McBride et al., 2009). Visualization of this protein by negative stain transmission electron microscopy demonstrated that it adopts the characteristic ring shape and approximate dimensions of Proteobacterial Hcp proteins (Figure 1F). Together, our data suggest that a conserved gene cluster within the Bacteroidetes phylum encodes a T6S-like pathway. Henceforth, we refer to this pathway as T6SS subtype 3 (T6SSⁱⁱⁱ), with the intent to distinguish it from the general Proteobacterial and Francisella pathogenicity island-like systems, herein constituting subtypes 1 and 2 (T6SS^i,ii), respectively (Figure 1A–C) (Boyer et al., 2009; Broms et al., 2010).

In order to better understand the relationship of the three T6SS groups, we performed phylogenetic analyses on the shared elements TssC and TssF. Phylogenetic trees generated from conserved regions of these proteins exhibited similar topologies, suggesting that the genes encoding them have been co-inherited (Figures 2A and S2A and Files S1–3). In each tree, all Bacteroidetes components comprise a unique clade, distinct from Proteobacterial T6S homologs. While some proteins encoded by species in the phylum Acidobacteria are also found in this clade, analysis of the genomic context of these homologs indicates they reside in gene clusters that lack conserved Bacteroidetes components (Figure S2B). We therefore restrict our definition of T6SSⁱⁱⁱ to those systems that reside in Bacteroidetes (Figure 2B and Files S1 and S4). Interestingly, T6SSⁱⁱⁱ gene clusters lack homologs of the T6SSⁱ proteins, TssA, L, M, and J. A gene encoding a putative TssM protein was recently suggested to reside in a B. cellulosilyiticus T6S gene cluster(Coyne et al., 2014). However, we note that homologs of this gene are neither generally found within or adjacent to T6SSⁱⁱⁱ clusters across the Bacteroidetes phylum, nor are they encoded in the genomes of all organisms that possess the T6SSⁱⁱⁱ pathway (data not shown). Our data suggest the entirety of the T6SSⁱ trans-envelope sub-complex – including tssM – is absent from T6SSⁱⁱⁱ. In summary, our data suggest that a phylogenetically distinct T6S-like pathway – composed of an assemblage of proteins distinct from those required for the function of Proteobacterial T6SSs – is found within members of the phyla Bacteroidetes.

Figure 2. T6SSⁱⁱⁱ is phylogenetically distinct from T6SSⁱ and T6SSⁱⁱ.

(A) Maximum likelihood (ML) phylogenetic tree generated from a partial alignment of 686 representative TssC sequences spanning the diversity present in T6SSⁱ, T6SSⁱⁱ, and T6SSⁱⁱⁱ gene clusters. Phyla represented by each system are indicated. Branch support values derived from aBayes analysis for the T6SSⁱⁱⁱ clade are shown. Scale bar represents amino acid changes per site. A similar tree for TssF is provided in Figure S2a.

(B) ML tree created from a partial alignment of TssC sequences found within T6SSⁱⁱⁱ gene clusters. The tree is rooted with Acidobacterial TssC sequences. Lengths do not reflect evolutionary distance. Colors trace the Class from which each sequence is derived. Nodes representing TssC sequences of organisms discussed in the text and those of particular significance are marked with an asterisk. Branch support values were generated by aBayes analysis. Partial sequence alignments and branch supports corresponding to phylogenetic trees in panels (A) and (B) and Figure S2 are provided in Files S1–4.

A T6SSⁱⁱⁱ pathway exports antibacterial effectors

T6SSs are functionally versatile and can deliver effectors to other bacteria, to eukaryotic host cells, or to both of these cell types. We identified a number of predicted antibacterial effectors associated with T6SSⁱⁱⁱ gene clusters in Bacteroidetes, suggesting that this system might possess the capacity to mediate interbacterial antagonism (data not shown). To first establish whether the T6SSⁱⁱⁱ gene cluster encodes a secretory pathway responsible for the export of effector proteins, we conducted secretome measurements using F. johnsoniae. We chose this organism because it is genetically tractable, easily maintained under aerobic conditions in the laboratory (McBride et al., 2009; Rhodes et al., 2011), and it possesses a T6SSⁱⁱⁱ gene cluster that is highly representative of the system in other Bacteroidetes, including the human gut-associated commensal species B. fragilis, B. vulgatus, and B. eggerthii (Figure 2B).

To determine the contribution of the T6SSⁱⁱⁱ pathway to the secretome of F. johnsoniae, we compared the culture supernatant of a strain bearing an in-frame deletion of its predicted tssC homolog, Fjoh_3266, to the wild-type parental strain using mass spectrometry. This analysis identified six proteins that were undetected in F. johnsoniae ΔtssC, but met our criteria for inclusion in the wild-type secretome (Table 1 and Table S2). Strikingly, the two most abundant of these proteins were Fjoh_3262 and Fjoh_3260, the Hcp and VgrG homologs encoded within the T6SSⁱⁱⁱ gene cluster, respectively (Figure 1C). This finding parallels similar secretome studies of T6SSⁱ pathways, which generally find Hcp- and VgrG-family proteins as the major components of the T6SS-dependent substrate pool (Fritsch et al., 2013; Hood et al., 2010; Russell et al., 2012).

Table 1.

Secreted proteins not detected the F. johnsoniae ΔtssC secretome

Locus Taga	Name	Abundanceb	Unique peptides detected	Signal peptidec	T6SSiii clusterd	Molecular weight	Predicted/determined function
Fjoh_3262	Hcp	133 ± 3.5e	10	N	Y	14.5	Secreted T6S structural component
Fjoh_3260	VgrG	13.3 ± 1.5	5	N	Y	66.0	Secreted T6S structural component
Fjoh_3206	RemH	11.7 ± 2.3	4	Y	N	16.9	Gliding motility
Fjoh_0984	RemF	5 ± 1.0	4	Y	N	16.8	Gliding motility
Fjoh_3274	–	4 ± 1.0	3	N	Y	102	T6SSiii effector
Fjoh_3257	Fte1	3 ± 1.7	2	N	Y	62.7	T6SSiii effector

^aLocus tags derived from F. johnsoniae UW101 (NCBI Accession NC_009441.1).

^bAverage spectral counts of three technical replicates.

^cSignal peptide predicted using SignalP 4.1 (http://www.cbs.dtu.dk/services/SignalP).

^dThe T6SSⁱⁱⁱ gene cluster comprises two apparent divergently transcribed operons (Fjoh_3254-Fjoh_3281), with the terminal gene of each operon encoding a conserved T6SS^III element.

^eStandard deviation of three technical replicates.

Of the remaining four proteins, two are encoded by hypothetical ORFs present within the T6SSⁱⁱⁱ gene cluster (Fjoh_3257 and Fjoh_3274), whereas the remaining two are implicated in gliding motility and possess predicted signal sequences (Figure 1C, Table 1) (Rhodes et al., 2011). While the latter may have yet unrecognized roles relevant to T6SSⁱⁱⁱ, for the purposes of identifying secreted effectors we focused on Fjoh_3257 and Fjoh_3274, which do not contain predicted signal peptides. Notably, these proteins both possess domains found in known or predicted T6SSⁱ effectors. Fjoh_3257 contains an HEXGH motif found in zinc metalloproteases fused to T6SSⁱ-exported VgrG proteins and Fjoh_3274 contains both a central glycoside hydrolase domain and a C-terminal zinc-dependent peptidoglycan endopeptidase motif (Figure 3A) (Pukatzki et al., 2007; Russell et al., 2012). Based on these data, we hypothesized that Fjoh_3257 and Fjoh_3274 are T6SSⁱⁱⁱ-exported effectors that exert toxicity in the periplasm of target cells.

Figure 3. F. johnsoniae T6SSⁱⁱⁱ exports an antibacterial protein that is encoded adjacent to a cognate immunity determinant.

(A) Domain organization of the putative substrates of F. johnsoniae T6SSⁱⁱⁱ. PAAR-like (blue), zinc-dependent metalloprotease (red), glycoside hydrolase (green), and zinc-dependent peptidoglycan amidase (purple) domains are indicated. Expanded amino acid sequences in each domain correspond to conserved motifs and invariant or critical catalytic residues (red).

(B–C) Growth of E. coli strains harboring the indicated expression vectors. Empty vectors (control) and vectors that introduce an N-terminal Sec signal peptide (peri) are indicated. Cells were induced to express predicted immunity proteins (C) at time 0 and Fte1 at the indicated time point (arrow). Type VI secretion immunity protein 1 (Tsi1) is used as a non-cognate immunity control. Error bars represent ± standard deviation (SD) (n = 3). Expression data for (B) are provided in Figure S3.

(D) Intercellular self-intoxication of the indicated F. johnsoniae strains as measured by propidium iodide staining. Liquid cultures were grown with vigorous shaking, which inhibits the formation of the prolonged cell–cell contacts required for T6-mediated interactions (Hood et al., 2010; Leroux et al., 2012). Error bars represent ± standard deviation (SD) (n = 3). Samples differing significantly from parent as measured by a two-tailed T-test are indicated by asterisks (p < 0.01).

To test the hypothesis that the T6SSⁱⁱⁱ pathway exports antibacterial effectors, we further investigated Fjoh_3257. When directed to the periplasm of Escherichia coli, Fjoh_3257 induced significant cell lysis, whereas the native protein – predicted to localize to the cytoplasm – did not exhibit apparent toxicity (Figures 3B and S3). T6SSⁱ delivers effectors directly to the periplasm of recipient cells. If T6SSⁱⁱⁱ functions similarly, our data suggest Fjoh_3257, henceforth referred to as Flavobacterium type VI secretion effector 1 (Fte1), could promote intercellular toxicity and thus necessarily associate with a cognate immunity determinant. Moreover, we would expect this immunity protein to reside in the periplasmic compartment, as T6SSⁱ effector inactivation invariably occurs via direct interaction with immunity (Benz et al., 2012; Dong et al., 2013a; Russell et al., 2011; Zhang et al., 2013). We identified a gene encoding a protein matching the predicted immunity criteria directly downstream of fte1, Fjoh_3256 (hereafter referred to as Flavobacterium type VI secretion immunity 1, or fti1). Co-expression of Fti1 specifically abrogated the lytic effects of Fte1, indicating that Fte1-Fti1 comprise an antibacterial effector–immunity pair of T6SSⁱⁱⁱ in F. johnsoniae (Figure 3C).

To test whether Fte1 exerts antibacterial activity in a T6SSⁱⁱⁱ-dependent manner, we measured the cellular integrity of F. johnsoniae strains lacking fti1. When propagated on a solid substratum, a condition conducive to prolonged cell contact, we observed increased membrane permeability in the Δfti1 strain (Figure 3D). This phenotype was abrogated by concomitant deletion of tssC, fte1, or by growth in liquid media. Together with our secretome studies, these data strongly suggest the capacity of T6SSⁱⁱⁱ to participate in interbacterial interactions through the export of antibacterial effectors.

T6SSⁱⁱⁱ mediates interspecies bacterial antagonism

Interbacterial T6SSⁱ has been observed to be a crucial determinant of fitness during polymicrobial growth. Under contact-promoting conditions, its inactivation generally leads to significant defects in the capacity to outcompete other organisms in co-culture. To determine whether T6SSⁱⁱⁱ also functions in interspecies antagonism we grew wild-type F. johnsoniae and derivative strains with either Burkholderia thailandensis or Pseudomonas putida under T6-conducive conditions. The inactivation of T6SSⁱⁱⁱ by a deletion of tssC in F. johnsoniae greatly impacted the outcome of these growth competitions, allowing for significant expansion of the competitor population (Figure 4A and 4B). This phenotype could be complemented by the introduction of an extra-chromosomal copy of tssC, demonstrating that the observed change in fitness was not due to mutant polarity. Moreover, wild-type and ΔtssC displayed equal fitness in liquid growth medium, consistent with the known requirement for intimate cell–cell contact in T6S-dependent interactions. We further examined interspecies co-cultures containing F. johnsoniae lacking the T6SSⁱⁱⁱ-restricted component, tssN (Fjoh_3277, Table S1). F. johnsoniae ΔtssN antagonized B. thailandensis and P. putida to an equivalent degree as ΔtssC or a strain bearing deletions in both tssC and tssN, consistent with our hypothesis that these genes encode essential elements of the same pathway (Figure 4A and 4B). Overall, our data strongly suggest that the T6SSⁱⁱⁱ pathway mediates interbacterial antagonism in a manner analogous to T6SSⁱ, yet using a distinct complement of proteins.

Figure 4. F. johnsoniae utilizes a dynamic T6SSⁱⁱⁱ apparatus to target competitor organisms.

(A–B) Growth experiments measuring fitness of the indicated F. johnsoniae (Fj) strains in co-culture for 20 hours with fluorescently-labeled competitors B. thailandensis (A, Bt) or P. putida (B, Pp). Competitive index is defined as the change in F. johnsoniae/competitor colony forming units between initiation and harvest of the co-culture. Experiments were performed under contact-promoting (solid) and contact-inhibiting (liquid) conditions. Qualitative analysis of competition outcome by photographic (P) and fluorescence (F) imaging is shown for corresponding samples grown under contact-promoting conditions after 48 hours of co-culture. Error bars represent ± SD (n = 3).

(C) Micrograph series depicting a 10s time course of wild-type F. johnsoniae clpV–gfp. Phase and GFP channels are presented separately. Three dynamic foci were observed over the duration of the experiment (arrowheads), and the presence or absence of each focus in each frame is schematized in the phase micrographs with white or unfilled circles, respectively. Full-length movies that include the region represented are available as Supplemental Files (Movies S1 and S2). See also Figure S4.

The T6SSⁱⁱⁱ apparatus exhibits dynamic behavior

Antibacterial effectors released by the T6SSⁱ pathway are operative on Gram-negative recipients only if they are delivered across the outer membrane by the translocation machinery. Owing to this feature of the system, the apparatus must behave dynamically in order to sample localizations that orient the system toward competitor cells. Green fluorescent protein (GFP) fusions to the C-terminus of ClpV proteins have served as a convenient means to visualize this behavior of T6SSⁱ systems. Since an apparent ClpV ortholog is identifiable in T6SSⁱⁱⁱ gene clusters, we sought to monitor the subcellular localization and dynamic behavior of this protein as a way to further interrogate the mechanistic similarity between T6SSⁱ and T6SSⁱⁱⁱ pathways. We began by generating a strain of F. johnsoniae bearing a functional clpV–gfp fusion at the native clpV locus (Figure S4). Visualization of this strain using time-lapse fluorescence microscopy revealed punctate foci appearing, disappearing, and frequently reappearing at different subcellular locations, on a rapid time scale (Figure 4C and Movies S1 and S2). Inactivation of the system in F. johnsoniae through the deletion of tssC abrogated these foci, similar to what has been observed in the T6SSⁱ pathway. It is worth noting that T6SSⁱⁱⁱ ClpV exhibits punctate localization and dynamic behavior in the absence of an apparent TssM homolog, whereas in T6SSⁱ systems, TssM proteins are required for ClpV dynamics. Taken together with our bioinformatic, secretomic, and phenotypic data, these findings strongly suggest that the T6SSⁱⁱⁱ pathway functions in a manner mechanistically similar to T6SSⁱ despite a highly divergent and unique assemblage of core components.

Bacteroides fragilis targets B. thetaiotaomicron via T6SSⁱⁱⁱ

Motivated by our characterization of the T6SSⁱⁱⁱ pathway in F. johnsoniae, we sought to explore the relevance of our findings to human-associated Bacteroidetes. Our analyses indicate T6SSⁱⁱⁱ gene clusters are present in many members of the genus Bacteroides, including numerous prominent human gut residents (Figure 2B and File S1). To probe the potential for the T6SSⁱⁱⁱ pathway to influence the behavior of these organisms in a physiological setting, we colonized germfree mice with a community containing B. fragilis, B. eggerthii, and the Proteobacterium E. coli, and measured T6SSⁱⁱⁱ expression. Quantitative reverse transcriptase (RT)-PCR of the cecal contents from these mice revealed the tssC gene in both organisms is expressed, at levels approaching (B. eggerthii) or exceeding (B. fragilis) the housekeeping transcript rpoD (Figure 5A).

Figure 5. The T6SSⁱⁱⁱ pathway is expressed and active in the genus Bacteroides.

(A) Quantitative RT-PCR analysis of the in vivo expression of tssC for the indicated bacteria. Expression was measured in cecal samples from germfree mice colonized for one week with B. fragilis, B. eggerthii, and E. coli and normalized using species-specific primers for the housekeeping sigma factor, rpoD.

(B) Growth competitions measuring viability of B. thetaiotaomicron after 24 h in the presence of the indicated B. fragilis strains on solid media. B. thetaiotaomicron growth was normalized to values obtained in the absence of B. fragilis. Viability was determined by colony forming units. Error bars represent ± SD (n = 3). Significance as indicated by asterisks was measured by a two-tailed T-test (p<0.002), n.s.; no statistical difference.

Expression of tssC in the mammalian gut environment led us to hypothesize that the T6SSⁱⁱⁱ pathway could be employed by Bacteroides to target other Gram-negative human gut microbes, including other species of Bacteroides. To test this hypothesis, assessed the ability of wild-type B. fragilis to inhibit growth of the prominent human gut commensal B. thetaiotaomicron, which lacks a T6SS. Growth competition experiments revealed that B. fragilis reduces B. thetaiotaomicron growth by approximately two orders of magnitude (Figure 5B). Strikingly, this activity is almost entirely T6SSⁱⁱⁱ-dependent, as an in-frame, unmarked deletion of tssC (BF9343_1941) in B. fragilis renders this species largely unable to reduce B. thetaiotaomicron growth. These data demonstrate the capacity of T6SSⁱⁱⁱ to act between prominent human gut-associated members of the genus Bacteroides.

Discussion

With the bioinformatic and functional characterization of T6SSⁱⁱⁱ, it is now evident that the Bacteroidetes possess a means for contact-dependent interbacterial antagonism. This is in-line with the observation that Bacteroidetes frequently occupy contact-promoting, polymicrobial niches (Thomas et al., 2011). Indeed, many of the organisms we identify the T6SSⁱⁱⁱ pathway within, including Porphyromonas, Prevotella and Bacteroides spp, are highly adapted host-associated obligate anaerobes that predominate – as pathogens or commensals – within the most densely populated polymicrobial sites in the human body (Falagas and Siakavellas, 2000; Smith et al., 2006). Thus, within sites such as the GI tract, oral cavity, and the vagina, where bacteria with T6SSⁱⁱⁱ are abundant, the pathway may play a broad role in defining community composition.

By analogy with T6SSⁱ and T6SSⁱⁱ, it is reasonable to speculate that T6SSⁱⁱⁱ has the capacity to mediate host cell interactions in addition to its now established role in interbacterial antagonism. Certain T6SSⁱ and T6SSⁱⁱ pathways appear to specialize in either bacterial or host cell targeting, whereas others can act on both cell types (Hood et al., 2010; MacIntyre et al., 2010; Pukatzki et al., 2007; Schwarz et al., 2010b). Target range appears to be dictated, at least in part, by the specific complement and corresponding activities of the effectors transported by a system. For example, recent reports suggest that by virtue of structural conservation among the phospholipid constituents of cellular membranes, T6S effectors belonging to the Tle phospholipase superfamily can confer both anti-bacterial and anti-eukaryotic activity (Dong et al., 2013b; Jiang et al., 2014; Russell et al., 2013). While members of the Tle superfamily are among the many apparent effectors of T6SSⁱⁱⁱ, the preponderance of predicted effectors that target peptidoglycan, a molecule found exclusively in bacteria, indicates that interbacterial antagonism is likely the basal function of the T6SSⁱⁱⁱ pathway.

Despite lacking several T6SSⁱ core components, including TssJ, L, and M, our observations suggest that T6SSⁱⁱⁱ functions in a fundamentally analogous manner. Both systems exhibit dynamic behavior, target effectors to the periplasm of recipients, and abundantly export VgrG and Hcp-family proteins. There are several conceivable explanations for these observations. One possibility is that the unique T6SSⁱⁱⁱ components, TssN, TssO, and TssP functionally substitute for the missing components. This model is supported by the prediction that these components, like TssL and M, are integral membrane proteins (Ma et al., 2009) (Aschtgen et al., 2012). However, TssJ is a predicted lipoprotein that requires localization to the outer membrane for function, and so far a T6SSⁱⁱⁱ-conserved predicted outer membrane-localized protein has not been identified. It is worth noting that TssJ, L, and M interact stably to form a trans-envelope complex (Cascales and Cambillau, 2012). While it has been postulated that this complex facilitates the passage of bacteriophage-like proteins and effectors out of the recipient cell, there are little experimental data to support this notion. It is therefore not yet possible to rule out a model whereby the components shared between T6SS^i-iii – namely those belonging to the bacteriophage-like sub-complex – represent the minimal structural assemblage of the T6SS. Distinguishing essential structural components from proteins with critical regulatory roles, for example, can be challenging (Hsu et al., 2009; Silverman et al., 2011). Understanding the functional significance of the varied complement of core elements associated with T6SS^i-iii will ultimately require both detailed biochemical approaches aimed at defining more precisely the roles of the individual proteins and ultra-structural characterization of the system.

While T6SSⁱ and T6SSⁱⁱⁱ are divergent, predicted effector proteins in T6SSⁱⁱⁱ-encoding organisms are often closely related to homologs in organisms possessing T6SSⁱ. For example, homologs of Fte1 are readily identified in Acinetobacter spp as well as strains of E. coli. These homologs, like Fte1, are encoded adjacent to predicted periplasmic immunity proteins as well as VgrG, suggesting that they likely possess common modes of toxicity and export. The relative ease with which homologs of effectors that transit the T6SSⁱ and T6SSⁱⁱⁱ pathways can be identified is further indicative of the similarity between these two systems and suggests they might share a common pool of potential effectors exchanged through horizontal gene transfer.

The gene encoding one of the putative substrates identified in our study, Fjoh_3274, is found adjacent to a locus that encodes a small protein possessing a DUF4280 domain (Fjoh_3275). In Fjoh_3274 homologs found in other species, these open reading frames are often fused, suggesting their function is linked. Structure prediction algorithms indicate a strong likelihood that DUF4280 adopts a fold closely related to the PAAR domain, which forms a pyramidal structure that is thought to recruit effector proteins to the apparatus via interaction with VgrG (Shneider et al., 2013). Interestingly, proteins bearing DUF4280 are found in Gram-positive bacteria, a division of bacteria not known to possess a T6S-like pathway. Moreover, the genes encoding these proteins are often found adjacent to VgrG-like proteins. Our finding herein that the T6S pathway extends to the Gram-negative phylum Bacteroidetes raises the possibility that other organisms, even Gram-positive bacteria, may also possess related systems that have yet to be identified. By analogy, the antibacterial nature of the C-terminal polymorphic toxin domains of YD-repeat proteins was initially discovered in Gram-negative bacteria; however, more recent studies have found homologs of these toxin domains participate in interbacterial antagonism in Gram-positive organisms (Koskiniemi et al., 2013).

Colonization resistance is a property of the gut microbiota whereby it acts as a coherent, resilient entity that exhibits resistance to invading microbes (Stecher and Hardt, 2011). The importance of this property is exemplified by the enhanced susceptibility to pathogens observed following either depletion or dysbiosis of the gut microbial community during antibiotic treatment (Lawley and Walker, 2013). Notably, recent studies have also shown that individuals carry the same commensal strains in their gut microbiomes for years or decades, and that members of the Bacteroidetes exhibit the greatest stability (Faith et al., 2013). A complete molecular explanation for these observations will likely include metabolic exclusion (Turnbaugh et al., 2009), colonization of critical niches (Lee et al., 2013), and the production of diffusible antimicrobials such as bacteroicins (Pujol et al., 2011). We posit that antagonistic contact-dependent interactions mediated by the T6SSⁱⁱⁱ pathway are another important contributor to colonization resistance and commensal stability. Interestingly, Turnbaugh and colleagues recently identified a transcript corresponding to a core T6SSⁱⁱⁱ element (vgrG) derived from Bacteroidales in a metatranscriptome of fresh human fecal samples (Maurice et al., 2013). Moreover, another study showed that the physical environment of the gut is conducive to T6S-mediated interbacterial interactions (Fu et al., 2013). While experiments involving T6SSⁱⁱⁱ mutants within gut colonization models will be needed in order to directly establish its role in this environment, taken together with our demonstration of the antibacterial nature of the system, these findings are consistent with the hypothesis that contact-dependent interbacterial interactions occur among commensals in the human gut.

Experimental Procedures

Bacterial strains, plasmids, and growth conditions

F. johnsoniae, B. thailandensis, P. putida, B. fragilis, B. eggerthii, and B. thetaiotaomicron used in this study were derived from the sequenced strains UW101, E264, KT2440, NCTC 9343, ATCC 27754, and VPI-5482 respectively. E. coli strains used in this study included DH5α for plasmid maintenance and tri-parental conjugation of plasmids into F. johnsoniae and B. fragilis, Rosetta 2(DE3) (EMD Millipore) for toxicity experiments, BL21(DE3) pLysS for the expression and purification of Fjoh_3262, and Nissle 1917 for mouse colonization experiments. Growth conditions for all strains and plasmid and strain construction details are described in Supplemental Experimental Procedures.

Informatic analyses

ClpV- and VgrG-like proteins were identified by automated annotation from NCBI blast servers. Hcp-like proteins were initially found within F. johnsoniae by PHYRE 2.0 (Kelley and Sternberg, 2009), and were thereafter identified by homology using blastp analysis. Other T6SS homologs were identified in T6SSⁱⁱⁱ gene clusters by the application of the iterative search algorithm jackHmmer on the RefSeq protein database using seed proteins obtained from F. johnsoniae (Finn et al., 2011). Alignments, domain prediction, phylogenetic trees, and subcellular localization were determined as described in the Supplemental Experimental Procedures.

Secretome preparation and MS analysis

The F. johnsoniae secretome was obtained using previously described methods with modifications indicated in the Supplemental Experimental Procedures (Hood et al., 2010). The UniProt F. johnsoniae UW101 database was used as a reference for peptide identification using MaxQuant v1.4.1.2 (Cox and Mann, 2008). Relative abundance of proteins was assessed using spectral counting (Liu et al., 2004). Proteins were filtered such that all had at least two unique peptides detected and possessed an average of three spectral counts in wild-type replicates.

Cellular toxicity assays

E. coli toxicity assays were performed as described previously with minor modifications (Russell et al., 2011). For the analysis of T6SSⁱⁱⁱ-dependent Fte1 toxicity in F. johnsoniae, strains were grown on a nitrocellulose surface as monocultures for 20 h before analysis by propidium iodide staining. Full details are provided in Supplemental Experimental Procedures.

Bacterial competitions

Bacterial co-cultures were prepared as described in Supplemental Experimental Procedures and either spotted on nitrocellulose placed on solid media or sub-inoculated into liquid media. After 20 h (F. johnsoniae) or 24 h (Bacteroides) of competition, cells were harvested and plated on selective media for quantification of each organism. Fluorescence images and photographs were acquired for the F. johnsoniae experiments after 48 h. Full details are provided in Supplemental Experimental Procedures.

Fluorescence microscopy

Microscopy was performed as described previously (Leroux et al., 2012). F. johnsoniae cells were prepared for microscopy after growth in conditions similar to bacterial competition experiments and were visualized on 1.5 % w/v agarose phosphate-buffered saline pads. Automated image acquisition was performed at 5s intervals for 6 minutes. Full details are provided in Supplemental Experimental Procedures.

In vivo expression of T6SSⁱⁱⁱ

All animal experiments were performed using protocols approved by the Yale University Institutional Animal Care and Use Committee. Germ-free Swiss Webster mice were maintained in flexible plastic gnotobiotic isolators with a 12-hour light/dark cycle. Mice (n = 5/group) were individually caged and were provided with standard autoclaved mouse chow (5K67 LabDiet, Purina) ad libitum. On day 0, mice were gavaged orally with 2x10⁸ CFU of each strain (B. fragilis, B. eggerthii, and E. coli). Animals were sacrificed on day 7 and samples were collected along the length of the gut. All samples were snap-frozen in liquid nitrogen and stored at −80°C.

RNA extraction from mouse cecal samples, cDNA synthesis, and quantitative RT-PCR were performed using standard methods. Expression for each T6SS gene was normalized to rpoD expression levels in the same organisms using species-specific primers. Genomic DNA samples used to generate standards for quantitative RT-PCR were obtained by published methods (Degnan et al., 2014). Full details are provided in Supplemental Experimental Procedures.

• Bacterial T6SS divides into three phylogenetically distinct subtypes (T6SS^i-iii)

• T6SSⁱⁱⁱ is restricted to Bacteroidetes and is composed of unique components

• T6SSⁱⁱⁱ targets toxic effectors to competing Proteobacteria and other Bacteroidetes

• Bacteroides fragilis T6SSⁱⁱⁱ targets B. thetaiotaomicron and is expressed in vivo