CdiA Effectors Use Modular Receptor-Binding Domains To Recognize Target Bacteria

ABSTRACT Contact-dependent growth inhibition (CDI) systems encode CdiA effectors, which bind to specific receptors on neighboring bacteria and deliver C-terminal toxin domains to suppress target cell growth. Two classes of CdiA effectors that bind distinct cell surface receptors have been identified, but the molecular basis of receptor specificity is not understood. Alignment of BamA-specific CdiAEC93 from Escherichia coli EC93 and OmpC-specific CdiAEC536 from E. coli 536 suggests that the receptor-binding domain resides within a central region that varies between the two effectors. In support of this hypothesis, we find that CdiAEC93 fragments containing residues Arg1358 to Phe1646 bind specifically to purified BamA. Moreover, chimeric CdiAEC93 that carries the corresponding sequence from CdiAEC536 is endowed with OmpC-binding activity, demonstrating that this region dictates receptor specificity. A survey of E. coli CdiA proteins reveals two additional effector classes, which presumably recognize distinct receptors. Using a genetic approach, we identify the outer membrane nucleoside transporter Tsx as the receptor for a third class of CdiA effectors. Thus, CDI systems exploit multiple outer membrane proteins to identify and engage target cells. These results underscore the modularity of CdiA proteins and suggest that novel effectors can be constructed through genetic recombination to interchange different receptor-binding domains and toxic payloads.

B acteria have long been known to compete with one another using diffusible antibiotics and bacteriocins (1,2). More recently, type IV, V, and VI secretion systems have been found to mediate interbacterial competition in a proximitydependent manner (3)(4)(5)(6). This phenomenon was first described for Escherichia coli isolate EC93, which uses a type Vb secretion system to bind and inhibit the growth of other E. coli cells (3). Because this inhibition requires direct cell-to-cell contact, the CdiA receptor-binding domains are modular. The putative receptor-binding region of CdiA EC93 (from Ser1379 to Tyr1636) shares only~24% sequence identity with the corresponding region of CdiA EC536 (Gln1377 to Trp1668) from uropathogenic E. coli 536 ( Fig. 2; see also Fig. S1 in the supplemental material). Therefore, we reasoned that sequence divergence over this region could determine receptor specificity. To test this hypothesis, we replaced residues Ser1347 to Tyr1636 of CdiA EC93 with Ala1345 to Trp1668 from CdiA EC536 (Fig. 2) and determined the receptor-binding specificity of the resulting chimera using a flow cytometry-based cell-cell adhesion assay (40). In this assay, CdiA-expressing inhibitor cells are labeled with green fluorescent protein (GFP) and mixed at a 5:1 ratio with DsRed-labeled target bacteria. After the populations are allowed to interact, the cell suspension is analyzed by flow cytometry for events that exhibit both GFP and DsRed fluorescence (Fig. 3A), which are quantified as inhibitortarget cell aggregates. Control experiments show that Ͻ5% of target bacteria adhere nonspecifically to mock (CDI Ϫ ) inhibitor cells ( Fig. 3A and B). In contrast, cells that express CdiA bind to target bacteria in a receptor-dependent manner. CdiA EC93expressing cells bound~80% of bamA Eco -target bacteria but failed to aggregate with bamA ECL targets ( Fig. 3A and C). Similarly, inhibitor cells that express CdiA EC536 bind a substantial fraction of ompC EC536 target bacteria but exhibited only background adhesion with ΔompC targets (Fig. 3A and D). As predicted, chimeric CdiA EC93 containing the putative receptor-binding region from CdiA EC536 recognized target cells in an ompCdependent manner ( Fig. 3A and E). Moreover, because this chimera was expressed at approximately the same level as CdiA EC536 (Fig. S2), the grafted domain appears to bind OmpF-OmpC with the same avidity as in its native context. These data, together with the in vitro binding results, indicate that the central regions of CdiA EC93 and CdiA EC536 are responsible for receptor recognition.
The CVR is required for OmpC-dependent toxin delivery. Though the receptorbinding region chimera supported robust OmpC-dependent adhesion, this effector protein did not inhibit target cell growth in competition cocultures (Fig. 4D). In contrast, inhibitors expressing CdiA EC93 outcompeted target cells approximately 10 5 -fold in cocultures (Fig. 4B), and cells that deploy CdiA EC536 exhibited a greater than 100-fold advantage (Fig. 4C). In each instance, CDI ϩ inhibitors only outcompeted target cells that express the appropriate CdiA receptor ( Fig. 4B and C). Together, these results suggest that the chimera is defective for toxin delivery. We noted that the CdiA EC93 region spanning Ala1910 to Gly2205 also diverges significantly with CdiA EC536 and that CdiA EC536 contains a 69-residue insertion in this region ( Fig. 2 and Fig. S1). Because this region is conserved between CdiA EC536 homologues that bind OmpC-OmpF, we refer to this sequence as the "covarying region" (CVR). We explored the function of this  sequence by replacing CdiA EC93 residues Ser1347 to Gly2205 with the receptor-binding and covarying regions of CdiA EC536 (Ala1345 to Gly2310) (Fig. 2). The resulting chimera was expressed stably (Fig. S2) and supported ompC-dependent cell-cell adhesion comparably to wild-type CdiA EC536 (Fig. 3A and G). Moreover, CdiA EC93 carrying the analyzed by flow cytometry for dual green/red fluorescent events. CdiA protein identity is indicated schematically in the left margin, and target cell genetic backgrounds are indicated along the top. (B to G) CdiA-dependent cell-cell adhesion was quantified for each effector protein: mock CDI Ϫ (B), CdiA EC93 (C), CdiA EC536 (D), CdiA EC93 with the receptor-binding region (RBR) from CdiA EC536 (E), CdiA EC93 with the covarying region (CVR) from CdiA EC536 (F), and CdiA EC93 with both RBR and CVR from CdiA EC536 (G). The fraction of target bacteria bound to inhibitor cells was quantified for two independent experiments. Data are presented as averages Ϯ standard errors. UPEC, uropathogenic E. coli.

FIG 4 Competition cocultures.
Target bacteria with the indicated bamA and ompC alleles were cocultured with inhibitor strains that express the following CdiA effectors: mock CDI -(A), CdiA EC93 (B), CdiA EC536 (C), CdiA EC93 with the receptor-binding region (RBR) from CdiA EC536 (D), CdiA EC93 with the covarying region (CVR) from CdiA EC536 (E), and CdiA EC93 with both RBR and CVR from CdiA EC536 (F). Viable inhibitor and target cells were enumerated as CFU milliliter Ϫ1 , and the competitive index was calculated as final inhibitor-to-target cell ratio divided by the initial ratio for each coculture. Competitive indices are reported as averages Ϯ standard errors for two independent experiments. UPEC, uropathogenic E. coli. heterologous receptor-binding and covarying regions also inhibited target bacteria in an ompC-dependent manner (Fig. 4F). In contrast, another CdiA EC93 chimera containing only the covarying region from CdiA EC536 (Pro1669 to Gly2310) retained BamA-binding specificity ( Fig. 3A and F) and inhibited target cells in a bamA Eco -dependent manner (Fig. 4E). Collectively, these data suggest that CdiA EC536 residues Pro1669 to Gly2310 may be important for toxin delivery through the OmpC-OmpF receptor pathway.
Identification of the receptor for CdiA STECO31 . We surveyed all identifiable E. coli CdiA effectors in the NCBI protein database and found that they segregate into at least four classes based on receptor-binding region sequences. CdiA EC93 defines the BamAbinding class I effectors, and CdiA EC536 is the paragon for class II effectors that recognize OmpC-OmpF. Class III CdiA proteins are closely related to class I and II over the two-partner secretion (TPS) transport domain and filamentous hemagglutinin (FHA)-peptide repeat regions but diverge significantly at the receptor-binding region ( Fig. 5; see also Fig. S1). Class IV CdiA proteins appear to be more distantly related and share little sequence identity with the other three classes over the FHA-1 peptide repeat region (Fig. S3). To determine whether class III CdiA recognizes a unique receptor, we cloned the cdiBAI gene cluster from E. coli STEC_O31 and tested its activity in growth competitions. Class III CdiA STECO31 (NCBI reference sequence WP_001385946.1) carries a C-terminal EndoU RNase toxin domain, and cells deploying this effector outcompeted target bacteria after 4 hours of coculture (Fig. 6A). Having established growth inhibition activity, we used a genetic approach to identify the receptor for CdiA STECO31 . We subjected target cells to mariner transposon mutagenesis and selected for CDI STECO31resistant (CDI r ) mutants. We isolated 12 CDI r clones from three independently prepared mutant pools and identified transposon insertion sites. Four of the CDI r mutants carried independent insertions in the ptsG gene, which encodes the fused IIB and IIC components of the glucose phosphotransferase system (PTS). We previously reported that ptsG null alleles also confer resistance to CDI toxins from E. coli 3006 and NC101 (14). CdiA-CT STECO31 , CdiA-CT 3006 , and CdiA-CT NC101 all share the same "translocation" domain, which is thought to bind PtsG and mediate toxin transport from the periplasm into the target cell cytosol (14). The remaining eight CDI r mutants were disrupted in tsx (Fig. 6A), which encodes an outer membrane nucleoside transporter. To confirm the role of Tsx in CDI STECO31 , we tested target cells carrying a nonpolar Δtsx mutation and found that they were resistant to growth inhibition (Fig. 6A). Moreover, complementation of Δtsx target cells with plasmid-borne tsx restored sensitivity to growth inhibition (Fig. 6A). Finally, we used flow cytometry to demonstrate that Tsx is required for cell-cell adhesion between target bacteria and CdiA STECO31 -expressing inhibitor cells (Fig. 6B). Together, these results show that the distinct receptor-binding region of class III CdiA STECO31 recognizes Tsx.

DISCUSSION
The results presented here show that CdiA effectors from E. coli strains bind target bacteria using at least three cell surface receptors. Receptor specificity for the different CdiA classes is determined by an unannotated stretch of~300 residues located between the FHA-1 and FHA-2 peptide repeat regions (see Tables S1, S2, S3, and S4 in the supplemental material). CdiA EC93 is the paragon for class I effectors, which bind specifically to BamA from E. coli (28). Class II effectors recognize heterotrimeric OmpC-OmpF osmoporins (31), and here, we show that class III effectors bind the outer membrane nucleoside transporter Tsx. The receptor-binding regions of class I, II, and III effectors share 24 to 27% pairwise sequence identity, and each domain contains a conserved central FHA-1 element (Fig. S1). These observations suggest that the classes diverged from a common ancestral sequence and that the three receptor-binding domains adopt similar conformations. The domains also exhibit some mutational drift within each class, though it is not clear whether these minor sequence variations influence receptor specificity. Class I effectors carry nine different receptor-binding coli Δtsx mutants are resistant to CdiA STECO31 -mediated growth inhibition. Eight independent transposon insertions within tsx were identified in selections for CDI-resistant mutants. Inhibitor cells (CDI STECO31 or CDI Ϫ mock) were cocultured with tsx ϩ or ⌬tsx target bacteria in shaking broth for 4 hours as described in Materials and Methods. Viable inhibitor and target cells were enumerated as CFU, and the competitive index was calculated as the final inhibitor-to-target cell ratio divided by the initial ratio. Competitive indices are presented as averages Ϯ standard errors. (B) Tsx is required for CdiA STECO31 -dependent cell-cell adhesion. The fraction of red fluorescent target bacteria bound to green fluorescent inhibitor cells was quantified for two independent experiments. Data are presented as averages Ϯ standard errors. sequences that arise from combinations of 17 missense changes (Table S1). Class II receptor-binding regions are less polymorphic with only nine missense variations (Table S2), and there are 14 distinct class III receptor-binding sequences (Table S3). Class IV E. coli CdiA effectors have diverged from the other three classes but still share the same overall domain structure (Fig. S3). Given the similarities in effector architecture, we predict that the receptor-binding domain of class IV effectors also resides between the FHA-1 and FHA-2 regions. Notably, class IV cdi gene clusters also encode a predicted HlyC-type lysyl acyltransferase (14). The functional significance of this acyltransferase has not yet been explored, but it presumably modifies CdiA and/or CdiB to promote cell adhesion and growth inhibition activities. Additionally, the database contains three CdiA proteins from E. coli isolates SWW33 (WP_001764992.1), upec-172 (WP_052432358.1), and 696_ECOL (WP_049080366.1) that do not fit into the four major classes. These proteins are most closely related to class IV CdiA but have longer FHA-1 repeat regions and diverge over the central (putative) receptor-binding region (Fig. S4). Moreover, the three effectors are encoded by cdi gene clusters that lack the characteristic hlyC acyltransferase of class IV loci. These observations suggest that these three CdiA proteins constitute a fifth effector class that probably recognizes yet another cell surface receptor.
Sequence alignments of class I, II, and III CdiA effectors predict that their receptorbinding domains are interchangeable. This hypothesis is supported by data showing that chimeric CdiA EC93 carrying a class II receptor-binding region is endowed with OmpC-binding activity. Though this chimera appears to bind OmpC-OmpF with the same avidity as wild-type CdiA EC536 , it does not inhibit target bacteria, indicating that toxin delivery function is compromised. Growth inhibition activity is restored when the covarying region is grafted together with the receptor-binding region. The latter result suggests that the covarying region of CdiA EC536 is critical for toxin delivery through the OmpC pathway. However, chimeric CdiA EC93 carrying the covarying region from Cdi-A EC536 inhibits target bacteria in a BamA Eco -dependent manner, showing that this sequence is not solely dedicated to the OmpC-OmpF pathway. In the context of domain modularity, these observations raise the possibility that specific toxin families may be excluded from certain receptor-mediated delivery pathways. Surveys of CdiA proteins show that some toxin families tend to be paired with specific receptor-binding domains. For example, Ntox28 RNase domains are found only on class II effectors in E. coli (Table S2). However, Ntox28 toxins are very effective at killing target bacteria when experimentally grafted onto class I CdiA EC93 (13,14,41). Thus, the apparent bias in naturally occurring effectors does not necessarily reflect domain incompatibility. In fact, we have shown that CdiA EC93 can deploy over a dozen heterologous toxin domains (9, 13, 14, 19-21, 41, 42), even though many of the grafted CdiA-CT sequences are not found on class I effectors. Moreover, the database shows significant combinatorial flexibility for some toxin families. For example, endonuclease NS_2 toxin domains are associated with each of the four major effector classes in E. coli (Tables S1 to S4). Comprehensive analysis of E. coli CdiA proteins reveals that each effector class is associated with multiple different toxin families (Tables S1 to S4). Taken together, these observations suggest that new CdiA effectors are assembled through genetic recombination to exchange receptor-binding domains and toxic payloads.
The location of the receptor-binding region has implications for the structure and presentation of CdiA on the inhibitor cell surface. CdiA effectors are thought to be structurally similar to the FhaB adhesins of Bordetella species because the protein families share related domain architectures (3,12). FhaB is synthesized initially as a 370-kDa precursor, from which the C-terminal "prodomain" is removed to yield a mature filamentous hemagglutinin (FHA) fragment of~220 kDa (43,44). FHA is monomeric and extends about 50 nm in length (45). The central shaft of FHA corresponds to the FHA-1 peptide repeat region, which is predicted to form a right-handed ␤-helix with a 4.8-Å pitch per 20-residue repeat (45,46). According to this model, the FHA-1 repeat region of CdiA EC93 should form a filament 30 to 35 nm in length. It is less clear that the CdiA FHA-2 repeat region is homologous to the FhaB prodomain. Though the prodomain contains a short FHA-2 region (Phe2927 to Gly3086), CdiA and FhaB share little sequence identity over this region. Moreover, FhaB lacks a C-terminal toxin domain and does not mediate interbacterial competition (47). Nevertheless, FhaB and CdiA are processed in a similar manner, with immunoblot analysis revealing a stable N-terminal CdiA fragment of~230 kDa (see Fig. S2). This processed fragment almost certainly retains adhesin activity, because truncated CdiA EC93 lacking the FHA-2 and CdiA-CT regions is still exported to the cell surface and mediates BamA Eco -dependent adhesion (38). These observations raise questions about the location of the FHA-2/ CdiA-CT regions relative to the cell surface. Current models assume that the CdiA C terminus projects away from the inhibitor cell to facilitate toxin transfer to target bacteria. However, we propose that the receptor-binding domain should be positioned at the distal tip of the filament (Fig. 7), where it can easily interact with target bacteria. Studies on FhaB biogenesis from the Cotter lab provide further support for the topological model presented in Fig. 7 (48). Their work indicates that the C-terminal prodomain is retained in the periplasm during FhaB export and processing. Thus, both N and C termini of FhaB remain within an intracellular compartment while the FHA-1 region assembles into a ␤-helix on the cell surface (48). It is not clear how the FhaB chain is exported as a tethered loop; however, their work predicts that the CdiA receptor-binding domain should be located at the distal tip of the FHA-1 repeat filament. Further, the model suggests that the FHA-2/CdiA-CT region is sequestered within the inhibitor cell periplasm prior to receptor recognition (Fig. 7). The latter prediction is consistent with unpublished data from our laboratories showing that the C-terminal region is protected from extracellular protease, whereas the N-terminal TPS domain and FHA-1 region are rapidly degraded by this treatment. If this model is correct, then receptor recognition must induce an extraordinary change in CdiA conformation to transfer the CdiA-CT toxin domain from the inhibitor cell periplasm into the target bacterium.

MATERIALS AND METHODS
Bacteria, growth conditions, and competition cocultures. Bacteria used in this study are listed in Table 1. All strains were grown at 37°C in lysogeny broth (LB) or on LB agar unless otherwise indicated. Media were supplemented with antibiotics at the indicated concentrations: ampicillin (Amp), 150 g ml Ϫ1 ; chloramphenicol (Cm), 33 g ml Ϫ1 ; kanamycin (Kan), 50 g ml Ϫ1 ; rifampin (Rif), 200 g ml Ϫ1 ; and tetracycline (Tet), 15 g ml Ϫ1 . The ΔompC::kan allele from the Keio collection (49) was amplified using  (50). The ompC EC536 -kan allele (31) was introduced into E. coli CH9597 in the same manner. Kan resistance cassettes were subsequently removed with FLP recombinase. Plasmid exchange was used to displace pZS21amp-bamA ϩ with Kan-resistant pZS21-bamA or pZS21-bamA ECL (28). The Δtsx::kan gene disruption was transduced into E. coli EPI100 that expresses recA from pCH9674, which was subsequently cured by growing the resulting strain at 42°C. For competition cocultures, inhibitor and target bacteria were adjusted to an optical density at 600 nm (OD 600 ) of 0.05 and then mixed at a 10:1 ratio and incubated in shaking LB medium for 4 h at 37°C. Cell suspensions were diluted into 1ϫ M9 salts and plated onto antibiotic selective LB-agar to enumerate inhibitor and target cells as CFU per milliliter. Competitive indices were calculated as the ratio of target to inhibitor cells at 4 h divided by the ratio at the beginning of coculture.
Plasmid constructions. Plasmids used in this study are listed in Table 2. The bamA Eco gene from E. coli MG1655 was amplified with oligonucleotides bamA(Δss)-Spe (5=-TTT ACT AGT GAA GGG TTC GTA GTG AAA G) and bamA-rev-Sal (5=-TCC TTT GTC GAC AAC ACT TAC CAG GTT TTA CC), digested with SpeI/SalI, and ligated to SpeI/XhoI-digested plasmid pCH7277. The resulting pCH9216 construct overproduces BamA Eco under the control of the bacteriophage T7 promoter. The bamA ECL gene from Enterobacter cloacae ATCC 13047 was amplified with primer pair bamA(Δss)-Spe and Ecloa-bamA-Sal-rev (5=-TTT GTC GAC GAG AAT TAC CAG GTT TTA CC) and cloned in the same manner. Fragments of cdiA EC93 were amplified from plasmid pDAL660Δ1-39 using the following primer pairs: EC93-cdiA(R1358)-Spe . The resulting PCR products were digested with SpeI/XhoI and ligated to plasmid pCH7277 to generate constructs that overproduce His 6 -tagged CdiA EC93 fragments.
Chimeric cdiA EC93 genes were constructed by Red-mediated recombination and sucrose counterselection. The receptor-binding region of cdiA EC536 (encoding residues Ala1345 to Trp1668) was amplified with UPEC1345A (5=-GTC TGT GGG TAC AGA AGG ACG CTT CCG GCG GTG CAA ACA) and UPEC1668W (5=-TGT TAC CGG ATG GCA GGG GCC AGT CAT CAC TGA TAC CGG). The covarying region of cdiA EC536 (encoding residues Pro1669 to Gly2310) was amplified with UPEC1669P (5=-GGG CAA GTG TAA GCA GCT ATC CAC TGC CTT CCG GCA ACA A) and UPEC2310G (5=-TCG TGC GTT GTC TTA CTG CTG CCA ATG GTG AAA CCA ATA CC). The combined receptor-binding and covarying regions were amplified with primer pair UPEC1345A/UPEC2310G. Homology regions from cdiA EC93 were amplified with EC93-Chimera-for/ UPEC1345-rev (5=-TGT TTG CAC CGC CGG AAG CGT CCT TCT GTA CCC ACA GAC) encoding Asp1124 to Asp1346, EC93-Chimera-for/UPEC1669-rev (5=-TTG TTG CCG GAA GGC AGT GGA TAG CTG CTT ACA CTT GCC C) encoding Asp1124 to Tyr1636, UPEC1668-for (5=-CCG GTA TCA GTG ATG ACT GGC CCC TGC CAT CCG GTA ACA)/EC93-Chimera-rev encoding Pro1637 to Gly2508, and UPEC2311-for (5=-GGT ATT GGT TTC ACC ATT GGC AGC AGT AAG ACA ACG CAC GA)/EC93-Chimera-rev encoding Ser2207 to Gly2508. The resulting fragments were joined to cdiA EC536 amplicons using OE-PCR with primer pair EC93-Chimerafor/EC93-Chimera-rev. The final products were introduced into plasmid pDAL7912 by Red-mediated recombination, and chimeric recombinants were selected on no-salt LB agar supplemented with Amp and 5% sucrose. The identities of all plasmid constructs were confirmed by DNA sequence analysis. Protein purification. BamA Eco and BamA ECL were overproduced in E. coli CH2016 grown at 37°C in LB medium supplemented with Amp and 1.5 mM IPTG (isopropyl-␤-D-thiogalactopyranoside). Though the plasmid constructs encode N-terminal His 6 epitopes, we found that these tags were removed, precluding the use of Ni 2ϩ affinity chromatography. Therefore, we isolated BamA proteins from insoluble inclusion bodies. Bacteria were harvested by centrifugation, and the cell pellets were suspended in 5 ml of BugBuster reagent and incubated on a rotisserie at room temperature for 20 min. Next, cells were harvested by centrifugation at 6,000 ϫ g for 10 min, and the supernatant was decanted to remove soluble proteins. Cells were then resuspended in 5 ml of BugBuster reagent and broken by three passages through a French press. Lysates were diluted with 25 ml deionized water and vortexed vigorously. Inclusion bodies were collected by centrifugation at 15,000 ϫ g for 20 min and washed three times with 5 ml of 0.1ϫ BugBuster solution. Isolated inclusion bodies were dissolved in 0.5 ml of urea lysis buffer (8 M urea, 10 mM Tris-HCl [pH 8.0], 150 mM NaCl, 20 mM imidazole, 0.05% Triton X-100). Residual His 6 -tagged BamA was removed by overnight incubation with urea lysis buffer-equilibrated Ni 2ϩ -nitrilotriacetic acid (NTA) agarose resin. Unbound protein was collected by centrifugation through an 0.45-m nylon Costar Spin-X column and diluted into 10 mM Tris-HCl (pH 8.0)-0.5% Triton X-100. Refolding reaction mixtures were incubated on a rotisserie for 3 days at ambient temperature, followed by 3 weeks at 4°C. More than 95% of BamA was refolded as determined by heat-modifiable gel mobility as previously described (53).
His 6 -tagged CdiA EC93 fragments were overproduced in E. coli CH2016 grown at 37°C in LB medium supplemented with Amp and 1.5 mM IPTG. Bacteria were harvested by centrifugation, and the cell pellets were frozen at Ϫ80°C and then resuspended directly in 15 ml urea lysis buffer. Cells were disrupted by two freeze-thaw cycles. Insoluble debris was removed through two rounds of centrifugation at 15,000 ϫ g for 20 min. Ni 2ϩ -NTA agarose resin was equilibrated in urea lysis buffer and added to each clarified lysate, and the mixtures were incubated on a rotisserie for 90 min at ambient temperature. The resin was collected by centrifugation at 3,000 ϫ g for 30 s, resuspended in 5 ml urea lysis buffer, and transferred to a fresh tube. After three washes with 5 ml urea lysis buffer, resin-bound His 6 -CdiA EC93 fragments were refolded with three washes of 5 ml native lysis buffer (10 mM Tris-HCl [pH 8.0], 150 mM NaCl, 30 mM imidazole, 0.05% Triton X-100). After the final wash, resins were resuspended in 250 l of binding buffer (10 mM Tris-HCl [pH 8.0], 150 mM NaCl, 30 mM imidazole, 0.5% Triton X-100) for use in BamA-binding experiments.
BamA-binding assays. Refolded BamA Eco and BamA ECL were diluted to 10 M in 1 ml of binding buffer. Ni 2ϩ -NTA agarose resin (20 l) containing bound His 6 -CdiA EC93 fragments was added to the BamA solutions, and the mixtures were incubated on a rotisserie for 90 min at 4°C. Resins were collected by centrifugation at 3,000 ϫ g for 30 s, resuspended in 1 ml of binding buffer, and transferred to a fresh tube. The resins were washed three times with binding buffer and then eluted with binding buffer supplemented with 25 mM EDTA. Eluted proteins were run on SDS-10% polyacrylamide gels and blotted onto nitrocellulose for immunoblot analysis using polyclonal antibodies raised against BamA Eco . Blots were visualized using IRDye 680 (Li-Cor)-labeled anti-rabbit secondary antibodies and an Odyssey infrared imager as described previously (54).
Cell-cell adhesion assays. BamA Eco -binding studies were conducted using E. coli strain DL4259, which expresses gfp-mut3 from the papBA promoter (19). E. coli DL4259 cells were transformed with pDAL660Δ1-39 (CdiA EC93 ), pCH12352 (CdiA EC536 ), pDAL7718 (CdiA EC93 -RBR EC536 ), pDAL7936 (CdiA EC93 -CVR EC536 ), pDAL7720 (CdiA EC93 -RBR/CVR EC536 ), or pCH13604 (CDI STECO31 ), and the resulting strains were grown at 37°C in LB medium (supplemented with Amp or Tet) until the cells developed fluorescence. The expression of each CdiA effector was assessed by immunoblot analysis of total urea-soluble protein using polyclonal antibodies raised against the N-terminal TPS transport domain of CdiA EC93 (38). Target cells were fluorescently labeled with DsRed using plasmid pDsRedExpress2. For Tsx-dependent adhesion experiments, the target cells were grown overnight in LB supplemented with Tet, Amp, 0.4% arabinose, and 1 mM IPTG to induce expression of Tsx and DsRed prior to mixing with inhibitors. GFP-labeled inhibitor cells were mixed at a 5:1 ratio with DsRed-labeled target bacteria at a final OD 600 of 0.2. Cell suspensions were shaken at 30°C for 20 min, diluted into 1ϫ PBS, vortexed briefly, and then analyzed on an Accuri C6 flow cytometer using FL1 (533/30 nm, GFP) and FL2 (585/40 nm, DsRed) fluorophore filters. The fraction of target bacteria bound to inhibitor cells was calculated as the number of dual green/red fluorescent events divided by the total number of red fluorescent events.
Transposon library construction and selection for CDI r mutants. The mariner transposon was introduced into E. coli CH7175 cells through conjugation with E. coli MFDpir donor cells carrying plasmid pSC189. Donor cells were grown to mid-log phase in LB medium supplemented with Amp and 30 M diaminopimelic acid. Donors and recipients were mixed and incubated on LB agar for 4 h at 37°C. Cell mixtures were harvested with a sterile swab, diluted into 1ϫ M9 medium, and plated onto Kansupplemented LB agar to select transposon mutants. The following day, the mutant pool (~100,000 Kan r colonies) was harvested into 1 ml of 1ϫ M9 medium. Mutant pools were cocultured with E. coli EPI100 cells carrying plasmid pCH13604 in LB medium overnight at 37°C, and target bacteria were selected on Kan-supplemented LB agar. The surviving target cell colonies were harvested into 1ϫ M9 medium and subjected to two additional cycles of coculture selection. Four CDI r clones were randomly selected from each of the three independently prepared transposon mutant pools. Transposon insertion sites were identified by rescue cloning. Chromosomal DNA from each CDI r mutant was digested with NspI overnight at 37°C. The restriction endonuclease was inactivated at 65°C for 20 min, and reaction mixtures were supplemented with ATP and T4 DNA ligase for overnight incubation at 16°C. The ligated DNA was electroporated into E. coli DH5␣ pir ϩ cells. Plasmids were isolated from resulting Kan-resistant transformants, and transposon insertion junctions were identified by DNA sequencing using primer marinerrev-seq (5=-CAA GCT TGT CAT CGT CAT CC).