The Tip of the VgrG Spike Is Essential to Functional Type VI Secretion System Assembly in Acinetobacter baumannii.

The type VI secretion system (T6SS) is a critical weapon in bacterial warfare between Gram-negative bacteria. Although invaluable for niche establishment, this machine represents an energetic burden to its host bacterium. Acinetobacter baumannii is an opportunistic pathogen that poses a serious threat to public health due to its high rates of multidrug resistance. In some A. baumannii strains, the T6SS is transcriptionally downregulated by large multidrug resistance plasmids. Other strains, such as the clinical isolate AbCAN2, express T6SS-related genes but lack T6SS activity under laboratory conditions, despite not harboring these plasmids. This suggests that alternative mechanisms exist to repress the T6SS. Here, we used a transposon mutagenesis approach in AbCAN2 to identify novel T6SS repressors. Our screen revealed that the T6SS of this strain is inhibited by a homolog of VgrG, an essential structural component of all T6SSs reported to date. We named this protein inhibitory VgrG (VgrGi). Biochemical and in silico analyses demonstrated that the unprecedented inhibitory capability of VgrGi is due to a single amino acid mutation in a widely conserved C-terminal domain of unknown function, DUF2345. We also show that unlike in other bacteria, the C terminus of VgrG is essential for functional T6SS assembly in A. baumannii Our study provides insight into the architectural requirements underlying functional assembly of the T6SS of A. baumannii We propose that T6SS-inactivating point mutations are beneficial to the host bacterium, since they eliminate the energy cost associated with maintaining a functional T6SS, which appears to be unnecessary for A. baumannii virulence.IMPORTANCE Despite the clinical relevance of A. baumannii, little is known about its fundamental biology. Here, we show that a single amino acid mutation in VgrG, a critical T6SS structural protein, abrogates T6SS function. Given that this mutation was found in a clinical isolate, we propose that the T6SS of A. baumannii is probably not involved in virulence; this idea is supported by multiple genomic analyses showing that the majority of clinical A. baumannii strains lack proteins essential to the T6SS. We also show that, unlike in other species, the C terminus of VgrG is a unique architectural requirement for functional T6SS assembly in A. baumannii, suggesting that over evolutionary time, bacteria have developed changes to their T6SS architecture, leading to specialized systems.

analyses provide insight into the characteristics that differentiate an inhibitory VgrG from a canonical VgrG. Furthermore, we demonstrate a previously unappreciated role for the C terminus of VgrG proteins in T6SS assembly in A. baumannii.

RESULTS
Clinical strain AbCAN2 encodes a T6SS locus. AbCAN2 is a coccygeal isolate collected at the University of Alberta Hospital, Canada, with an inactive T6SS under standard laboratory conditions (39). To gain a better understanding of why the T6SS is inactive in this strain, we determined the whole-genome sequence of AbCAN2 (GenBank accession number CP045428). We found that this strain does not harbor an LCP. In addition, it encodes all genes required for a functional T6SS. In fact, the T6SS locus of AbCAN2 is syntenic with that of other A. baumannii strains, including the lab strain A. baumannii ATCC 17978 (Ab17978) (Fig. 1) (39). The T6SS genes of AbCAN2 and Ab17978 present remarkably high levels of sequence conservation, ranging from 98 to 100% nucleotide identity (see Table S1 in the supplemental material). Besides encoding 12 of the 13 structural T6SS proteins conserved among Proteobacteria (Acinetobacter species lack a TssJ homolog [39]), the T6SS locus of AbCAN2 also contains genes coding for accessory proteins TagF, TagN, PAAR, and TagX. In addition, the T6SS locus harbors genes that code for three hypothetical proteins that are conserved among Acinetobacter species, some of which have been shown to be essential for T6SS function (46,47). The specific roles of these proteins are yet to be elucidated.
AbCAN2 also encodes two vgrG gene clusters located remotely from its T6SS locus ( Fig. 1). We arbitrarily named the VgrGs of each cluster VgrG1 and VgrG2. Each VgrG is encoded upstream of an open reading frame (ORF) containing a predicted rearrangement hotspot A (RhsA) domain. T6SS effectors of the RhsA family have been previously reported to possess DNase activity (48)(49)(50); thus, it is possible that the proteins coded by these ORFs possess nuclease activity. Given their genetic proximity to vgrG1 and vgrG2, we expect these proteins to be T6SS effectors. Thus, we named these ORFs type 6 effector 1 (tse1) and tse2, respectively. Downstream of each tse1 and tse2 is an ORF that we predict to be the cognate immunity proteins of these effectors. Immunity proteins are commonly encoded downstream of their cognate effector, and they inactive their cognate effector by specifically binding it and obstructing its active site (10). Thus, we named these ORFs type six immunity 1 (tsi1) and tsi2, respectively.
A VgrG homolog inhibits the T6SS of AbCAN2. Our finding that AbCAN2 lacks T6SS activity despite encoding all genes required for T6SS assembly and not harboring an LCP led us to hypothesize that AbCAN2 encodes a novel T6SS repressor. To identify the putative novel repressor of the T6SS, we generated a transposon mutant library and screened mutants for Hcp secretion, using our high-throughput Hcp colony blot assay (see Materials and Methods) (51). We hypothesized that genetic disruption of a T6SS inhibitor would result in appreciable T6SS activity in a mutant strain. Of the ϳ3,000 colonies screened, we detected Hcp in culture supernatants of four unique mutant strains, indicating that these transposon mutants possess an active T6SS (Fig. 2a). A VgrG Homolog Inhibits the A. baumannii T6SS ® Surprisingly, the gene disrupted in all four mutant strains was vgrG1 (Fig. 2b). VgrG is a conserved structural component of all T6SSs reported to date. It is a modular protein containing N-terminal domains structurally similar to spike complex proteins of bacteriophage contractile tails (gp27 and gp5), as well as C-terminal DUF2345 and transthyretin (TT)-like domains, whose roles are yet to be fully understood (20,21,(52)(53)(54). VgrGs are widely regarded as essential to T6SS assembly and function, being implicated in roles such as baseplate formation, proper Hcp tube assembly, and effector delivery (18,22,24,55). Thus, our result that a VgrG homolog inhibits the T6SS is against the current paradigm of VgrG function. To confirm our results, we generated a clean deletion mutant of vgrG1 (hereafter referred to as vgrGi for inhibitory VgrG) and compared its T6SS activity to that of wild-type (WT) AbCAN2, using Hcp secretion and bacterial killing assays. Consistent with our previous result, Hcp was detected in the supernatant fraction of ⌬vgrGi but not WT AbCAN2 or the complemented strain (vgrGi ϩ ), indicating that VgrGi inhibits the T6SS of AbCAN2 ( Fig. 2c and Fig. S1a). Similarly, AbCAN2⌬vgrGi, but not WT or vgrGi ϩ , demonstrated significant levels of interbacterial killing when coincubated with E. coli (Fig. 2d). Altogether, our data indicate that VgrGi inhibits T6SS activity in AbCAN2.
VgrGi acts independently to inhibit the T6SS. A previous report in Vibrio cholerae suggests that T6SS effectors may play a critical role in T6SS assembly (56). Thus, we tested whether AbCAN2 effectors Tse1 and Tse2 or VgrG2 were involved in T6SS inhibition. To this end, we generated mutant strains of AbCAN2 lacking tse/tsi1, tse/tsi2, vgrG2, or both vgrGi and vgrG2 (⌬vgrGi,2) and determined their T6SS activity by Hcp secretion assay. We did not detect Hcp in the supernatant fraction of any of the mutant strains, indicating that VgrGi-mediated T6SS inhibition is independent of any proteins encoded within the two vgrG clusters of AbCAN2 (Fig. 3). Importantly, we also found that unlike ⌬vgrGi, ⌬vgrG2, and ⌬vgrGi,2 lack the ability to secrete Hcp and kill E. coli ( Fig. 3 and Fig. S1b). Together, these results demonstrate that VgrG2 acts as a canonical VgrG; it is required for proper T6SS function and effector delivery into bacterial competitors.
Next, we tested whether VgrGi-mediated T6SS inhibition is due to an intrinsic property of AbCAN2 or whether VgrGi can also inhibit the T6SS of a heterologous host. To this end, we expressed plasmid-borne 6ϫHis-tagged VgrGi in Ab17978 lacking LCP pAB3 (hereafter referred to as WT Ab17978), which possesses constitutive T6SS activity under standard laboratory conditions (40), and probed for Hcp secretion. We found that WT Ab17978 expressing VgrGi secreted similar levels of Hcp compared to the vector control ( Fig. 4a and Fig. S2). This result suggested that VgrGi is unable to inhibit the T6SS of a noncognate host. Nonetheless, it is noteworthy that unlike AbCAN2, which encodes one canonical VgrG, VgrG2, Ab17978 encodes four VgrG homologs, each sufficient for assembling a functional T6SS (46). We reasoned that it was possible that VgrGi was unable to inhibit the T6SS of Ab17978 due to functional redundancy between the four VgrGs. Therefore, we then expressed VgrGi in mutant strains of Ab17978 lacking one to three VgrGs. We found that VgrGi expression in Ab17978⌬vgrG1 resulted in a considerable reduction of Hcp detected in culture supernatants (Fig. 4b). Remarkably, VgrGi completely abolished Hcp secretion in Ab17978  Taken together, our results show that VgrGi can inhibit the T6SS of its native strain, as well as that of Ab17978 mutant strains with limited VgrGs. Given that AbCAN2 and Ab17978 possess a different arsenal of effectors, these findings provide further evidence that VgrGi-mediated T6SS inhibition is independent of the effectors expressed.
VgrGi-mediated T6SS inhibition is due to a leucine-to-arginine mutation. VgrGi is, to the best of our knowledge, the first VgrG homolog with an ability to inhibit the T6SS. Thus, we sought to identify unique characteristics that differentiate VgrGi from other VgrGs. However, protein sequence and structural prediction analyses revealed that VgrGi contains domains similar to those found in canonical VgrGs of diverse bacteria, such as VgrG1 of P. aeruginosa PAO1 (VgrG1 Pa ) and VgrG1 of EAEC 17-2 (VgrG1 Ec ) (52,54,57). (Despite their similar names, VgrG1 Pa and VgrG1 Ec are two distinct proteins.) VgrGi contains two domains at its N terminus that closely resemble bacteriophage T4 tail spike proteins gp27 (residues 1 to 426) and gp5 (residues 537 to 666) (Fig. 5a). These domains are linked by an oligosaccharide-binding (OB)-fold domain (residues 427 to 536). The C terminus of VgrGi is composed of a DUF2345 domain (residues 688 to 834), followed by a TT-like domain (residues 860 to 907) and a region of ϳ130 residues with no predicted functional domains. Finally, the N-and C-terminal segments of VgrGi are linked by a predicted coiled-coil region (residues 667 to 687) (Fig. 5a).
Given that we identified no differences in the predicted functional domains of canonical VgrGs and those of VgrGi, we reasoned that VgrGi may possess critical differences in its primary sequence that are responsible for its inhibitory ability. VgrGs encoded by A. baumannii strains were recently grouped into six distinct phylogenetic groups (58). Based on the amino acid identity (Ͼ95%), we determined that VgrGi belongs to class 1 VgrGs. Interestingly, despite such remarkable levels of sequence conservation, we found that VgrGi contains a leucine-to-arginine substitution in position 749 compared to all other class 1 VgrGs (Fig. 5b). We hypothesized that the leucine-to-arginine mutation was responsible for the unprecedented ability of VgrGi to inhibit the T6SS of A. baumannii. To test our hypothesis, we expressed plasmid-borne 6ϫHis-tagged VgrGi or a VgrGi revertant mutant containing an arginine-to-leucine mutation (hereafter referred to simply as R749L) in AbCAN2⌬vgrGi and probed for Hcp secretion. As expected, we detected Hcp in the supernatant of the vector control but not in the vgrGi ϩ strain (Fig. 5c). Remarkably, ⌬vgrGi expressing R749L (vgrGi R749L ϩ ) retained T6SS activity ( Fig. 5c and Fig. S3a), indicating that the mutation R749L abrogated the inhibitory capability of VgrGi. In fact, vgrGi R749L ϩ showed higher levels of Hcp secretion compared to the vector control (Fig. 5c), suggesting that R749L likely participates in T6SS assembly. To test this, we expressed plasmid-borne 6ϫHis-tagged VgrGi or R749L in AbCAN2⌬vgrGi,2. This strain lacks T6SS activity ( Fig. 3) but encodes all other genes necessary to assemble a functional T6SS. Thus, we hypothesized that if R749L participates in T6SS assembly, heterologous expression of this protein in ⌬vgrGi,2 would result in T6SS activity. Indeed, ⌬vgrGi,2 expressing R749L but not VgrGi or the empty vector secretes Hcp ( Fig. 5d and Fig. S3b), indicating that the mutation R749L converts VgrGi to a canonical VgrG essential to T6SS assembly.
Altogether, our results indicate that VgrGi lacks any functional domains that differentiate it from canonical VgrGs. Instead, its inhibitory capability is due to a single amino acid mutation.
Polar or charged residues within the SLFAAQ motif disrupt VgrG function. Our previous result suggests that a leucine residue in position 749 plays an important role in canonical VgrG function. A protein alignment analysis revealed that despite considerable levels of divergence among VgrGs of diverse A. baumannii strains, most VgrGs possess a conserved SLFAAQ motif (Fig. S4). The conservation of this motif further points to the leucine residue as important for VgrG function. Thus, we hypothesized that substitution of the conserved leucine for arginine would convert a canonical VgrG to an inhibitory VgrG. VgrG1 of Ab17978 (ACX60_17665) possesses a leucine residue within its SLFAAQ motif, L758 (Fig. S4), and is important for T6SS function (46). To test our hypothesis, we generated plasmid-borne 6ϫHis-tagged VgrG1 or VgrG1 bearing a L758R mutation, expressed them in Ab17978⌬vgrG1 (vgrG1 ϩ and vgrG1 L758R ϩ , respectively) (Fig. S5), and determined the resulting levels of Hcp secretion. In contrast with the vgrG1 ϩ strain, which had increased levels of Hcp secretion compared to the vector control, vgrG1 L758R ϩ , demonstrated a marginal level of Hcp secretion (Fig. 6). This result indicates that substitution of a conserved leucine residue for arginine confers on VgrG1 the ability to repress the T6SS of Ab17978.
Next, we tested whether VgrGi-mediated T6SS inhibition depends on a specific property of arginine or whether other amino acids yield similar phenotypes. To this end, A VgrG Homolog Inhibits the A. baumannii T6SS ® we generated pWH1266 vector constructs containing the following point mutants of VgrGi: R749K, R749D, R749N, and R749F. Then, we expressed these VgrGi variants in AbCAN2⌬vgrGi (Fig. S6) and determined whether expression of these constructs results in T6SS inhibition. We found that VgrGi variants containing polar or charged residues (i.e., R749K, R749D, and R749N) inhibit Hcp secretion (Fig. 7). In contrast, VgrGi variants containing nonpolar residues (i.e., R749L and R749F) do not inhibit Hcp secretion (Fig. 7). Similar results were obtained when these constructs were expressed in Ab17978⌬vgrG1,2 (Fig. S7). Collectively, our results indicate that polar or charged residues in the second position of the SLFAAQ motif disrupt the canonical function of VgrGs in T6SS assembly.
The C terminus of VgrG is essential to T6SS function in A. baumannii. The SLFAAQ motif lies within the C-terminal DUF2345 domain of VgrGi. Although present in VgrGs of diverse Gram-negative bacteria, the role of the DUF2345 domain in T6SS assembly remains poorly understood. Recent work in E. coli showed that the gp27 domain of VgrG1 Ec is sufficient for T6SS activity, indicating that the DUF2345 and TT domains are dispensable for T6SS assembly (52). Similarly, the C terminus of P. aeruginosa VgrG5 is not required to support VgrG4b secretion (53). Nonetheless, our results indicate that a point mutation within the C-terminal DUF2345 domain of A. baumannii VgrGs is sufficient to disrupt T6SS activity. Thus, we hypothesized that in A. baumannii, the DUF2345 domain plays an essential role in the assembly of a functional T6SS. To test our hypothesis, we expressed different truncations of R749L in AbCAN2⌬vgrGi,2 and determined the resulting levels of Hcp secretion. As shown in Figure 5d, ⌬vgrGi,2 lacks T6SS activity unless it expresses a functional VgrG (e.g., full-length R749L). The R749L constructs tested include the gp27 domain (gp27), a C-terminal truncation ending with the gp5 domain (C-gp5), a C-terminal truncation ending with the DUF2345 domain (C-DUF R749L ), and mutants lacking either the DUF2345 domain (ΔDUF) or the TT domain (ΔTT R749L ) but possessing all other domains ( Fig. 8a and Fig. S8). We found that in contrast to full-length R749L, the expression of none of the R749L truncations tested led to Hcp secretion (Fig. 8b). This result indicates that the C terminus of R749L is essential to the role of VgrG in assembling a functional T6SS.

DISCUSSION
The T6SS is a critical weapon for interbacterial warfare (12). Seminal studies have provided general principles that are largely applicable to the T6SS of various bacterial species. However, it is becoming increasingly clear that the T6SS is a diversified machine, with important structural and functional differences depending on the bacterial host (31,59,60). Here, we identified a VgrG homolog with an unprecedented ability to inhibit the T6SS of the clinical isolate AbCAN2. Our efforts to characterize the mechanism underlying VgrGi-mediated T6SS inhibition led us to determine that the C terminus of VgrG is essential for functional T6SS assembly in A. baumannii.  (21,22,54). Despite its relevance, VgrG remains largely understudied compared to other T6SS proteins, likely due to its inherent insolubility (61). Only one full-length VgrG protein, VgrG1 Pa , has been structurally characterized (57 One of the functional domains present in longer VgrGs is the DUF2345 domain. In Escherichia coli, this domain has been shown to stabilize the interaction between the TT domain of VgrG1 and effector Tle1, but it is dispensable for T6SS assembly and dynamics (52,54). Despite its seeming dispensability for T6SS assembly, this domain is present in the VgrGs of a wide range of Gram-negative bacteria, including V. cholerae (VCV52_2925), Klebsiella pneumoniae (BN49_3373), P. aeruginosa (PA0262), and Salmonella enterica (FJR63_23580). In fact, VgrGs were originally described as possessing a COG4253 functional domain, which belongs to the DUF2345 superfamily (60). Our results suggest that the DUF2345 domain may play a role in T6SS assembly in A. baumannii. However, since C-DUF R749L is not sufficient for T6SS assembly (Fig. 8), the DUF2345 domain may require additional elements present at the C terminus of VgrG for proper function. Moreover, a recent bioinformatic analysis of 73 VgrGs across 22 A. baumannii strains, as well as the soil bacterium A. baylyi ADP1, showed that all VgrGs contain a DUF2345 domain (58). The high levels of conservation suggest that the It is noteworthy that our results are not in conflict with previous results indicating that the C terminus of VgrG is dispensable for T6SS function (52,53). Instead, we propose that over evolutionary time, bacteria developed changes to their T6SS architecture, leading to specialized systems. Indeed, it was recently reported that TssA homologs carry out three diverse functions depending on their C-terminal domain (62). Moreover, homologs of membrane complex protein TssM possess a Walker A motif; however, no general role for this motif has been described. In Agrobacterium tumefaciens, for instance, TssM exhibits ATP binding and hydrolysis, as well as ATP-bindingdependent conformational changes, all of which are implicated in T6SS function (63,64). In contrast, the Walker A motif of the Edwardsiella tarda TssM, is dispensable for T6SS activity (65). Thus, it seems likely that bacteria have co-opted conserved structural proteins to better suit their needs.
Future work will characterize the mechanism by which VgrGi inhibits the T6SS, since it may provide insight into the role of the C terminus of A. baumannii VgrGs. Our finding that VgrGi inhibits the T6SS of Ab17978 only when endogenous VgrGs are limited suggests that VgrGi acts at the posttranslational level. We propose that VgrGi likely inhibits the T6SS by preventing its assembly. VgrGs can form heterotrimeric assemblies (21,53,66,67). Thus, it is possible that VgrGi binds canonical VgrGs and prevents essential protein-protein interactions with other structural components of the T6SS. PAAR proteins, for instance, bind VgrG proteins at their C termini (68). In A. baylyi, PAAR proteins are essential for effective T6SS firing (68). Thus, VgrG trimers containing VgrGi may be unable to bind PAAR proteins, which would be expected to disrupt T6SS assembly. Moreover, it is possible that VgrGi possesses conformational changes that abrogate critical baseplate-membrane complex interactions. Alternatively, it is possible that VgrGi acts post-T6SS assembly. The T6SS is expected to undergo two large conformational changes to enable deployment of the spiked tube. First, conformational changes in the baseplate are proposed to trigger sheath contraction (22,69,70). Second, the membrane complex is expected to undergo large conformational changes to enable the passage of the spiked tube (19,71). It is possible that a leucine-to-arginine mutation causes aberrant interactions between VgrG and components of the membrane complex or baseplate, which prevent the necessary conformational changes for T6SS firing. In this instance, VgrGi may prove to be an invaluable tool to lock the dynamic T6SS in a static conformation and facilitate the structural characterization of the Acinetobacter T6SS.
Finally, it is noteworthy that we found an inactivating mutation of the T6SS in a clinical isolate. Although we cannot say with certainty whether this mutation was acquired during culturing or in the human host, it is tempting to speculate that there may be a selective pressure against the T6SS of medically relevant strains. As previously mentioned, the T6SS is an energetically expensive machine. Thus, there may be benefits to inactivating this machine, for example, to conserve energy. Moreover, T6SS structural components have been shown to be immunogenic (13,72). It is possible that T6SSinactive strains are selected for in the human host, since they are more likely to bypass detection by the immune system. In fact, most of the characterized virulence factors of A. baumannii involve evading immune detection or overcoming nutritional immunity (38). Thus, eliminating the T6SS may represent yet another strategy to bypass immune detection. Nonetheless, the T6SS confers a competitive advantage to nonpathogenic bacteria living in polymicrobial communities. Thus, it is likely that the environment exerts selective pressure in favor of a silenced T6SS in clinical A. baumannii strains and a constitutively active T6SS in nonpathogenic strains, as was previously proposed for V. cholerae (29). Indeed, clinical isolates of A. baumannii often have a nonfunctional T6SS due to genetic disruptions or absence of T6SS genes (44,(73)(74)(75). Our work indicates that point mutations at critical residues of T6SS structural proteins could constitute an underappreciated mechanism by which the T6SS is genetically disrupted.
Collectively, we have demonstrated that unlike in other bacteria, the C terminus of VgrG is essential to functional T6SS assembly in A. baumannii. The specific roles of the DUF2345 and other C-terminal domains of VgrGs warrant further investigation.

MATERIALS AND METHODS
Bacterial strains and growth conditions. All strains and plasmids used in this study are listed in Table S2. Strains were grown in Luria-Bertani (LB) broth at 37°C with shaking. Antibiotics were added to the media when appropriate (see below).
Transposon mutagenesis and screen T6SS-active mutants. Our method for transposon mutagenesis has been described previously (41). Plasmid pSAM::OmpApϩTn903 was introduced to AbCAN2 from E. coli BW19851 by biparental mating. Briefly, overnight cultures were pelleted, washed three times with fresh LB medium, and resuspended at an optical density at 600 nm (OD 600 ) of 1.0. The cultures were then mixed at a 1:1 ratio and spotted onto a dry LB agar plate. After a 5-h incubation at 37°C, the spot was resuspended in 1 ml of LB broth, and 10-fold dilutions were plated on LB agar containing kanamycin (20 g/ml) plus chloramphenicol (12.5 g/ml), followed by incubation overnight at 37°C. The transposon mutants obtained were subsequently subjected to a colony blot (51) to identify mutants with an active T6SS. Briefly, colonies were transferred to a nitrocellulose membrane and allowed to dry at room temperature for 30 min. The dried membrane is washed twice with Tris-buffered saline-Tween and treated as a normal Western blot, probing for Hcp and RNAP (see below). Candidates were then isolated and verified using a standard Hcp secretion assay (see below).
Identification of transposon insertion sites was carried out using a method adapted from Goodman et al. (76). Transposon mutant gDNA was digested with MmeI and separated on a 0.7% agarose gel by electrophoresis. Fragments ϳ500 and ϳ1,600 bp in length were extracted and appended with doublestranded oligonucleotide sequencing adapters by ligation. Finally, the transposon plus two 16-bp genomic sequences flanking the transposon was amplified by PCR and sequenced. The insertion site was verified by amplifying a genomic fragment of the putative insertion site by PCR and sequencing that fragment as well.
Hcp secretion assay and Western blot analysis. Overnight cultures were back-diluted in fresh LB medium to an OD 600 of 0.025 and grown at 37°C with shaking until they reached an OD 600 of 0.35 to 0.5 (AbCAN2) or 0.4 to 0.7 (Ab17978). The cells were then pelleted by centrifugation. The cells were resuspended in Laemmli buffer to a final OD 600 of 10, while the supernatant fraction was centrifuged once again (as above) to pellet residual cells. Supernatant proteins were subsequently precipitated with trichloroacetic acid, as previously described (39). Optical desity-normalized volumes of whole cells or supernatants were loaded onto 15% (for Hcp) or 8% (for VgrGs) SDS-PAGE gels for separation, transferred to a nitrocellulose membrane, and probed with polyclonal rabbit anti-Hcp (1:1,000) (39), polyclonal rabbit anti-6ϫHis (1:2,000; Invitrogen, Waltham, MA) or monoclonal mouse anti-RNA polymerase (1:2,600; BioLegend, San Diego, CA). Western blots were then probed with IRDye-conjugated anti-mouse and anti-rabbit secondary antibodies (both at 1:15,000; LI-COR Biosciences, Lincoln, NE) and visualized with an Odyssey CLx imaging system (LI-COR Biosciences).
Generation of mutants and pWH constructs. The primers used in this study are listed in Table S2. Marked mutant strains were generated by substitution of the gene of interest by a kanamycin resistance marker, as described previously (77). Selection was carried out using kanamycin (30 g/ml). To generate clean mutants, electrocompetent marked mutants were transformed with pAT03 to remove the FRTflanked kanamycin resistance cassette. Transformants were plated on LB agar containing 2 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) plus hygromycin (600 g/ml). Mutant strains were verified by PCR and sequencing.
Constructs pWH-vgrGi-6ϫHis and pWH-vgrG1-6ϫHis were generated by restriction cloning using the EcoRI and PstI sites. These plasmids were then used as templates to make point mutants of VgrGi and VgrG1, respectively. Specifically, point mutants were generated using the QuikChange II site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions. Constructs of VgrGi truncations ΔDUF and ΔTT R749L were generated by inverse PCR of pWH-vgrGi-6ϫHis, followed by blunt ligation. All other VgrGi truncation constructs were generated by restriction cloning, as described above. Selection was carried out using tetracycline (15 g/ml for AbCAN2 and 10 g/ml for Ab17978). All constructs were verified by PCR and sequencing.
Bacterial killing assay. Overnight cultures were washed three times with fresh LB and normalized to an OD 600 of 1. Predator and prey strains were mixed at a 10:1 ratio, respectively, and spotted on dry LB agar. After a 3.5-h incubation at 37°C, spots were resuspended in LB medium, and serial dilutions were spotted on LB agar with the appropriate antibiotic (kanamycin 50 g/ml for E. coli MG1655R/pBAV-gfp and tetracycline 15 g/ml for E. coli HB101/pWH1266). The CFU of surviving prey cells were enumerated after overnight incubation at 37°C.
VgrGi structural model and alignments. VgrGi C-gp5 (residues 1 to 666) was submitted to the iterative threading assembly refinement (I-TASSER) server (78), using the crystal structure of P. aeruginosa PAO1 VgrG1 (PDB 4MTK) as a template without alignment. The trimer assembly was constructed by aligning the VgrGi model to each VgrG1 Pa monomer in Biological Assembly 1 (PDB 4MTK), using the PyMOL Molecular Graphics System (v1.2r3pre; Schrödinger, LLC).
Domain designations were made according to Interpro (79) and HHPred (80) servers. The predictions for the coiled-coil domain and the DUF2345 domain overlap in residues 668 to 697; however, for the construction of the VgrGi truncations, preference was given to the DUF2345 domain. Sequence alignments were done using Clustal Omega (81).
Data availability. The genome of AbCAN2 (formerly named Ab1225) was deposited to GenBank under accession number CP045428.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.