Functional Diversity of Cytotoxic tRNase/Immunity Protein Complexes from Burkholderia pseudomallei *

Contact-dependent growth inhibition (CDI) is a widespread mechanism of inter-bacterial competition. CDI+ bacteria deploy large CdiA effector proteins, which carry variable C-terminal toxin domains (CdiA-CT). CDI+ cells also produce CdiI immunity proteins that specifically neutralize cognate CdiA-CT toxins to prevent auto-inhibition. Here, we present the crystal structure of the CdiA-CT/CdiIE479 toxin/immunity protein complex from Burkholderia pseudomallei isolate E479. The CdiA-CTE479 tRNase domain contains a core α/β-fold that is characteristic of PD(D/E)XK superfamily nucleases. Unexpectedly, the closest structural homolog of CdiA-CTE479 is another CDI toxin domain from B. pseudomallei 1026b. Although unrelated in sequence, the two B. pseudomallei nuclease domains share similar folds and active-site architectures. By contrast, the CdiIE479 and CdiI1026b immunity proteins share no significant sequence or structural homology. CdiA-CTE479 and CdiA-CT1026b are both tRNases; however, each nuclease cleaves tRNA at a distinct position. We used a molecular docking approach to model each toxin bound to tRNA substrate. The resulting models fit into electron density envelopes generated by small-angle x-ray scattering analysis of catalytically inactive toxin domains bound stably to tRNA. CdiA-CTE479 is the third CDI toxin found to have structural homology to the PD(D/E)XK superfamily. We propose that CDI systems exploit the inherent sequence variability and active-site plasticity of PD(D/E)XK nucleases to generate toxin diversity. These findings raise the possibility that many other uncharacterized CDI toxins may belong to the PD(D/E)XK superfamily.


Results
Structure of the CdiA-CT/CdiI E479 Toxin/Immunity Protein Complex-We previously reported that overexpression of the CdiA-CT/CdiI E479 /His 6 complex in E. coli cells leads to tRNA degradation and concomitant growth arrest (16). These observations suggest that the expression construct produces insufficient immunity protein to neutralize tRNase activity. Because CdiA-CT E479 activity precludes protein overproduction, we inactivated the toxin with the D285A mutation to allow high level expression of the toxin/immunity protein complex (16). We purified and crystallized the SeMet-labeled complex and used SAD phasing to produce an initial partial model. The model was subsequently improved using molecular replacement with a native dataset, resulting in a final resolution of 2.0 Å ( Table 1). The final model includes CdiA-CT E479 residues Arg-201-Lys-316 (numbered from Glu-1 of the ELYN peptide motif), CdiI E479 residues Ala-2-Gly-105, and 155 water molecules. The final R work /R free (%) was 19.3/23.7 with 98.8% of dihedral angles in favorable regions and the remaining 1.2% within allowed regions as estimated by Ramachandran plot.
Like other CdiA-CT constructs (17,26,27), the N-terminal region of CdiA-CT E479 (residues Glu-1-Phe-200) is not resolved in the final model. This unresolved region corresponds to the "translocation" domain, which is postulated to mediate CdiA-CT transport across the cytoplasmic membrane of target bacteria (28). The resolved C-terminal domain corresponds to the tRNase domain responsible for growth inhibition activity (16). The CdiA-CT E479 nuclease domain consists of a fivestranded mixed ␤-sheet decorated by four ␣-helices (Fig. 1A). The sheet forms a half ␤-barrel-like structure with helix ␣1 running through its central cavity (Fig. 1A). The C-terminal half of helix ␣2 (␣2b) is bent 90°with respect to the N-terminal portion (␣2a). The CdiI E479 immunity protein consists of a slightly curved three-stranded antiparallel ␤-sheet decorated with four ␣-helices (Fig. 1A). The CdiA-CT/CdiI E479 interface is largely electrostatic with 19 direct salt bridges and hydrogen bonds mediating the interaction ( Fig. 1B and Table 2). Helix ␣2b of the nuclease domain sits in the curve of the CdiI E479 ␤-sheet. Additionally, helices ␣1Ј, ␣4Ј, and the ␤-sheet of CdiI E479 interact with the end of the nuclease ␤-sheet and the extended loop connecting ␣2Ј and ␤3Ј (Fig. 1B). This proteinprotein interface buries 1015 Å 2 of surface area, corresponding to 14% of the nuclease domain and 18% of the immunity protein total surface area. In accord with this extensive interaction surface, biolayer interferometry showed that CdiA-CT E479 and CdiI E479 form a relatively high affinity complex with an apparent equilibrium dissociation constant of 72 Ϯ 23 nM (Fig. 2).
CdiA-CT E479 Toxin Domain Is a Member of the PD(D/E)XK Superfamily-We used the DALI server (24) to identify proteins with structural similarity to the CdiA-CT E479 nuclease domain. This search revealed that two other CdiA-CT nuclease domains from B. pseudomallei 1026b and Yersinia pseudotuberculosis YPIII exhibit the greatest similarity with CdiA-CT E479 ( Table 3). The CdiA-CT 1026b and CdiA-CT E479 nuclease domains superimpose with a root-mean-square deviation (r.m.s.d.) of 3.7 Å over 89 of 132 ␣-carbons, corresponding to a DALI Z-score of 7.0. The CdiA-CT YPIII nuclease domain exhibits comparable structural similarity, although its ␤4-␤5 hairpin element is absent from the CdiA-CT E479 nuclease domain (27). Other proteins identified during this search include two closely related XisH endonucleases from cyanobacteria, another CdiA-CT nuclease domain from Escherichia coli EC869, and the AspBHI restriction endonuclease from Azoarcus sp. BH72 ( Table 3). All of these domains share the core structure of the PD(D/E)XK nuclease superfamily, which includes most type II restriction endonucleases and various enzymes involved in DNA recombination and repair. The PD(D/E)XK core domain is a mixed ␤-sheet flanked by two ␣-helices with ␣␤␤␤␣␤ topology. CdiA-CT E479 and CdiA-CT 1026b share this core fold, but the CdiA-CT E479 nuclease domain contains an insertion that forms helix ␣2a (Fig. 3, A and  B). We also note that helix ␣1 from CdiA-CT E479 is significantly shorter than the corresponding helix in the CdiA-CT 1026b structure (Fig. 3, A and B).
Identification of the CdiA-CT E479 Active Site-We previously suggested that Asp-280 and Asp-285 of CdiA-CT E479 may function in catalysis, because mutation of these residues abrogated toxicity (16). However, structural alignment with the nuclease domain from CdiA-CT 1026b indicates that residues Glu-204, Asp-229, Asp-243, and His-275 of CdiA-CT E479 are more likely to catalyze tRNA cleavage (Fig. 3A). These residues overlay well with active-site residues Glu-187, Asp-214, Asp-223, and Lys-242 of CdiA-CT 1026b (Fig. 3C). To test this prediction, we mutated CdiA-CT E479 residues His-275, Asp-243, Asp-229, and Glu-204 to Ala individually and examined the growth inhibition activity of each toxin variant. Induction of wild-type CdiA-CT E479 expression in E. coli cells resulted in immediate growth inhibition (Fig. 4A). By contrast, induction of domains that carry the H275A, D243A, D229A, or E204A mutations had no effect on cell growth (Fig. 4A), consistent with the loss of toxic nuclease activity. We also purified each toxin domain and tested its tRNase activity in vitro. Wild-type CdiA-CT E479 toxin . e R work ϭ ⌺͉F obs Ϫ F calc ͉/⌺F obs . R free was computed identically except where all reflections belong to a test set of 5% randomly selected data. cleaved a large proportion of tRNA molecules as assessed by denaturing PAGE analysis, and this activity was blocked when purified CdiI E479 immunity protein was included in the reaction (Fig. 4B). This tRNase activity was confirmed by Northern blot analysis, which showed cleavage of tRNA Gly molecules in the reactions (Fig. 4B). By contrast, none of the other CdiA-CT E479 mutant variants exhibited detectable RNase activity (Fig. 4B).
To determine whether the mutations adversely affect toxin structure, we tested whether the refolded CdiA-CT E479 proteins still interact with cognate immunity protein using affinity co-purification. Each mutant toxin co-eluted with CdiI E479 -His 6 during Ni 2ϩ -affinity chromatography (Fig. 4C), indicating that the nuclease domains retain their native fold. Taken together with the crystal structure, these findings suggest that CdiA-CT E479 residues His-275, Asp-243, Asp-229, and Glu-204 participate in catalysis. Because CdiI E479 binds directly over this cluster of residues, the immunity protein presumably neutralizes toxin activity by blocking access to tRNA substrates. We previously showed that CdiA-CT E479 cleaves between residues T54 and ⌿55 in the T-loop of tRNA molecules (16). To test whether post-transcriptional modifications at positions 54 and 55 are required for CdiA-CT E479 activity, we examined toxin activity on unmodified E. coli tRNA Asp and tRNA Gln substrates prepared by in vitro transcription. Each substrate was cleaved efficiently by purified nuclease and the activity completely neutralized by CdiI E479 immunity protein (Fig. 4D). Thus, the universal T-loop modifications are not required for CdiA-CT E479 -mediated tRNase activity. Structural Comparison of CdiI E479 and CdiI 1026b Immunity Proteins-Although the CdiA-CT E479 and CdiA-CT 1026b nuclease domains share a common fold and active site, the corresponding immunity proteins appear to be unrelated. Using iterative PSI-BLAST, we were unable to establish a link between CdiI E479 and CdiI 1026b sequences. Moreover, structural superimposition of the two immunity proteins reveals a poor fit between the central ␤-sheets and misalignments of most ␣-helical elements (Fig. 5, A and B). CdiI E479 and CdiI 1026b align with an r.m.s.d. of 3.42 Å over only 42 of 103 ␣-carbons (Z-score of 1.8) indicating low structural similarity. We next used the DALI server to search for proteins with structural similarity to CdiI E479 . The only hit with a Z-score of Ͼ5 was the N-terminal domain of protein NS3 from rice hoja blanca tenuivirus (Table  3). NS3 suppresses RNA interference pathways in host cells by binding to both siRNA and miRNAs (29,30). Although these proteins superimpose with r.m.s.d. of 2.7 Å over 75 of 90 ␣-carbons, the NS3 domain is entirely ␣-helical and lacks the central ␤-sheet found in CdiI E479 .
CdiA-CT/CdiI Complexes Have Unique Electrostatic Interfaces-Both toxin/immunity complexes interact primarily via electrostatic interactions and shape complementation; and intriguingly, both immunity proteins use the N-terminal ␣amino group to form a salt bridge with a catalytic Asp residue in the toxin active site ( Table 2) (17). However, the size, charge distribution, shape, and position of these patches differ between the two immunity proteins (Fig. 6). The electrostatic FIGURE 2. Biolayer interferometry of the CdiA-CT/CdiI E479 binding interaction. Immobilized CdiI E479 -His 6 was exposed to varying concentrations (1.5-5 M) of CdiA-CT E479 , and the binding interaction and dissociation monitored a wavelength shift (nm). Representative association and dissociation curves are presented with the overall correlation coefficient (R 2 ) shown for the fit.  surface map shows that CdiI E479 has a positively charged protrusion (formed from ␣1Ј, ␣2Ј, and the ␤-sheet) that is complementary to the negatively charged active-site groove within the CdiA-CT E479 nuclease domain (Fig. 6A). In addition, a small negatively charged patch on CdiI E479 interacts with a positive patch on the cognate toxin. Much like CdiI E479 , the CdiI 1026b immunity protein inserts a positively charged protrusion (formed from ␤3Ј, ␣3Ј, and the connecting loop region) into the negatively charged active-site cleft of its cognate toxin. CdiI 1026b also contains a negatively charged patch, which complements a positive patch adjacent to the Lys-242 active-site residue within CdiA-CT 1026b . Collectively, these electrostatic and shape complementation interactions ensure that each nuclease domain is only neutralized by its cognate immunity protein.
Nuclease Active-site Architecture and tRNA Specificity-CdiA-CT E479 and CdiA-CT 1026b cleave tRNA at distinct positions, and presumably, this specificity is dictated by the shape of the active-site pocket. The CdiA-CT E479 nuclease domain has a slightly larger open pocket with a diameter of 12.6 Å measured from Gln-253 to His-275 (Fig. 6A). By contrast, the active-site cleft of CdiA-CT 1026b is narrower with a diameter of 10.4 Å from Glu-187 to Lys-242 (Fig. 6B). To explore how the nuclease domains interact with substrate, we used Hex 8.0 to dock each toxin onto the structure of E. coli tRNA Cys . We used experimentally determined cleavage sites for each nuclease to guide the docking searches. The models obtained from the simulations show that each nuclease active-site is positioned near the scissile bond in the tRNA substrate (Fig. 7, A and C). The calculated interaction energy for the tRNA/CdiA-CT E479 complex was Ϫ802 kcal/mol, and that of the tRNA/CdiA-CT 1026b complex was Ϫ1,267 kcal/mol. These energies are comparable with those calculated for Hex-assisted docking of each toxin domain with its cognate immunity protein as follows: Ϫ953 kcal/mol for CdiA-CT/CdiI E479 complex and Ϫ1,199 kcal/mol for the CdiA-CT/CdiI 1026b complex. A negative control simulation using CdiI E479 and CysK, which do not interact with one another, yielded a considerably higher energy of Ϫ368 kcal/ mol. Simulations using non-cognate toxin/immunity protein pairs produced models with lower than anticipated energies (Ϫ742 kcal/mol for CdiA-CT E479 /CdiI 1026b and Ϫ626 kcal/mol for CdiA-CT 1026b /CdiI E479 ), but in each instance the solutions did not position the immunity proteins over the nuclease active sites. toxins with proposed active-site residues highlighted in red and conserved residues in blue. Secondary structure elements are colored gold and green for CdiA-CT E479 and CdiA-CT 1026b , respectively. B, superimposition of CdiA-CT E479 and CdiA-CT 1026b nuclease domains. Secondary structure elements that superimpose are color-coded in gold (CdiA-CT E479 ) and green (CdiA-CT 1026b ), and those that do not align rendered in white (CdiA-CT E479 ) and gray (CdiA-CT 1026b ). C, active site of CdiA-CT E479 and CdiA-CT 1026b nuclease domain. Predicted active-site residues are shown in stick representation (nitrogen and oxygen atoms are colored blue and red, respectively).

Small-angle X-ray Scattering (SAXS) Analysis of Toxin/tRNA
Complexes-To test the computational docking models, we sought to produce stable tRNA/toxin complexes for struc-tural analysis using SAXS. We overproduced catalytically inactive toxins that carry N-terminal His 6 tags and purified the proteins by Ni 2ϩ -affinity chromatography. Remarkably, FIGURE 4. CdiA-CT E479 growth inhibition and tRNase activities. A, growth inhibition activity of CdiA-CT E479 variants. The indicated toxins were expressed in E. coli cells from a rhamnose-inducible promoter as described under "Experimental Procedures." Expression was induced at 0 min, and cell growth was monitored by measuring the optical density at 600 nm (OD 600 ). The curve labeled repressed corresponds to un-induced cells carrying the wild-type CdiA-CT E479 construct. The average Ϯ S.E. from three independent biological replicates is presented. B, in vitro nuclease assays. The indicated CdiA-CT E479 variants were purified and incubated with total E. coli RNA. Reactions were run on denaturing 6% polyacrylamide gels and stained with ethidium bromide. C, mutant CdiA-CT E479 domains bind to CdiI E479 immunity protein. Isolated toxin domains were mixed with purified CdiI E479 -His 6 and then subjected to Ni 2ϩ -affinity chromatography. Lanes labeled input show the protein mixtures loaded onto the column; free lanes show proteins that failed to bind the column, and bound indicates proteins eluted from the column with imidazole. Prior work has shown that CdiA-CT E479 does not bind to Ni 2ϩ -NTA-agarose resin (16). D, CdiA-CT E479 cleaves unmodified tRNAs produced by in vitro transcription. E. coli tRNA Gln and tRNA Asp transcripts were incubated with purified CdiA-CT E479 and CdiI E479 , and reactions were analyzed on denaturing 6% polyacrylamide gels stained with ethidium bromide. Experiments in B-D were repeated twice with essentially identical results. Representative data are shown for each experiment. FIGURE 5. Sequence and structure comparison of CdiI E479 and CdiI 1026b immunity proteins. A, structure-based sequence alignment of CdiI E479 (blue) and CdiI 1026b (cyan) with secondary structure elements indicated above and below the sequence alignment. Conserved residues are highlighted in blue. B, superimposition of CdiI E479 and CdiI 1026b structures. Secondary elements that partially or fully superimpose are labeled.
large quantities of endogenous tRNA co-purified with each inactive toxin (Fig. 8A), indicating that the tRNA/toxin complexes are indeed stable. The tRNA/CdiA-CT 1026b complex migrated at ϳ45 kDa on size-exclusion chromatography, indicative of a 1:1 complex with tRNA in solution (Fig. 8B). By contrast, size-exclusion chromatography showed that the tRNA/CdiA-CT E479 complex is ϳ150 kDa (Fig. 8C), suggestive of a higher order complex containing four nuclease domains bound to four tRNA molecules. We used SAXS to generate low resolution electron density envelopes of each nucleoprotein complex. DAMAVER (31) was used to calculate normalized spatial discrepancies (NSD) of 0.979 Ϯ 0.038 for tRNA/CdiA-CT 1026b and 0.959 Ϯ 0.081 for the tRNA/ CdiA-CT E479 complex, with no restorations rejected. These average NSD values imply reasonable stability of the solutions. The Hex-generated models for each tRNA/toxin complex were then fitted into the respective electron density envelopes using Chimera (Fig. 7, B and D) (32). Supcomb (33) was used to calculate the NSD values between the averaged and filtered shape from SAXS and the structural models of 0.9371 for tRNA/CdiA-CT 1026b and 0.9013 for the tRNA/ CdiA-CT E479 complex. These NSD values suggest that the average SAXS envelopes and the structural models agree quite well with each other. The four CdiA-CT E479 nuclease domains form a donut-like structure. Each nuclease domain fits helix ␣3 into the curvature of the ␤-sheet (strands ␤1-␤3) on the adjacent domain. Interdomain contacts also occur between helix ␣4 and ␤5 and the loop connecting strands ␤2 and ␤3. Within the tetramer, the four ␣1 helices are directed toward the center, and the nuclease active sites project outward. The buried surface area of each monomer is consistent with a stable oligomeric state (34). The complex is further stabilized by interactions between tRNA molecules, which pack together with their aminoacyl acceptor stems pointing into the center of the complex. Similar tight packing interactions have been observed in the crystal structure of tRNA Asp (35).

Discussion
The results presented here demonstrate that the CdiA-CT E479 nuclease domain is a member of the PD(D/E)XK nuclease superfamily. Together with the previously characterized CdiA-CT 1026b and CdiA-CT o11 EC869 nuclease domains (17,27), there are at least three CDI toxin classes that share the PD(D/E)XK core fold. The amino acid sequences of these toxins are distinct (15-18% pairwise sequence identity) and show no apparent relationship to one another through iterative PSI-BLAST analyses. However, structural superimposition of the CdiA-CT E479 , CdiA-CT 1026b , and CdiA-CT o11 EC869 toxins reveals significant similarities. The PD(D/E)XK fold consists of a central four-stranded mixed ␤-sheet flanked by two ␣-helices with a characteristic ␣ 1 ␤ 1 ␤ 2 ␤ 3 ␣ 2 ␤ 4 topology. The core structure serves as a scaffold to arrange catalytic residues. The canonical PD(D/E)XK active site found in type II restriction endonucleases is built from a conserved Asp residue at the N terminus of ␤2 and the (D/E)XK sub-motif within ␤3 of the core (23,36). However, there are several variations in the active-site configuration, with catalytic residues migrating to other secondary structure elements during evolution (23,37,38). For the CdiA-CT E479 nuclease domain, Asp-229 and Asp-243 occupy canonical positions within ␤2 and ␤3, but Glu-204 and His-275 are contributed by ␣1 and ␣2, respectively. This arrangement is very similar to the active site of CdiA-CT 1026b and the type IIS restriction endonuclease BspD6I (17,39). The DNase domains of CdiA-CT o11 EC869 and CdiA-CT YPIII have yet another configuration that was first described for EcoO109I (17,27,40,41). In these latter enzymes, Glu of the (D/E)XK sub-motif has migrated from ␤3 to helix ␣1 to produce an alternative E-PDXXK motif. For the CdiA-CT o11 EC869 class of toxins, the active sites use an E(F/Y)DSXK sequence motif (17,27). Interestingly, these latter DNases contain an additional ␤-hairpin inserted between ␣2 and ␤4 of the PD(D/E)XK core. This ␤-hairpin constitutes much of the binding interface with the cognate immunity protein, and its sequence varies between family members (27). Analogous insertions into the PD(D/E)XK core have been detected in other superfamily members (23), again underscoring the flexibility of the core fold.
Most PD(D/E)XK enzymes are phosphodiesterases involved in DNA restriction, transposon excision, recombination, and repair. By contrast, there are relatively few family members with RNase activity. EndA/Sen15 tRNA splicing endonucleases were the first PD(D/E)XK enzymes to be implicated in RNA metabolism (23,42). More recently, Rai1 has been reported to act as a phosphodiesterase to remove 5Ј-cap structures from eukaryotic mRNAs (43). Our findings show that the PD(D/ E)XK fold has been adopted to produce RNases with novel specificities. The CdiA-CT E479 nuclease domain cleaves tRNA between residues T54 and ⌿55 of the conserved T⌿C-loop (16). Positions 54 and 55 are modified to thymidine and pseudouracil in eubacterial tRNAs, but these universal post-tran-scriptional modifications are not required for CdiA-CT E479 activity in vitro. CdiA-CT 1026b is a novel RNase that cleaves near the 3Ј-end of tRNA (16,17). The computational docking  studies reported here represent the first steps toward a detailed understanding of tRNA-binding specificity. Docking of tRNA onto the CdiA-CT 1026b domain provides a reasonable model for toxin binding to the aminoacyl-acceptor stem. The interaction between CdiA-CT E479 and substrate appears to be more complicated, and it is unclear why CdiA-CT E479 /tRNA complexes oligomerize in solution. It should be noted that CdiA-CT E479 in the absence of tRNA also forms a tetramer (data not shown). Although the models are still vague, it is tempting to speculate that the additional helix ␣2a within CdiA-CT E479 contributes to T-loop binding specificity. Helix ␣2a forms a ridge along the lower edge of the putative tRNA-binding surface. Residues Phe-260 and Phe-261, which form a prominent hydrophobic patch adjacent to the nuclease-active site, may participate in substrate binding by stacking onto nucleobases. An elucidation of specific contacts must await high resolution structural studies of toxin/substrate complexes. Given that inactive versions of each nuclease domain bind to tRNA with high affinity, it should be possible to generate specific nucleoprotein complexes for high resolution crystallography.
We have now reported crystal structures for four different CDI toxin classes. As described above, three of these toxins are nucleases of the PD(D/E)XK superfamily. The other toxin, CdiA-CT ECL from Enterobacter cloacae ATCC 13047, is an Ntox21 family member and adopts a fold common to barnase, endonuclease poly(U)-specific, colicin E5/D, and RelE (BECR) toxins (14,26). Sequence analyses by Aravind and co-workers (14,15) indicate that CDI systems encode several other toxin families with distinct protein folds and activities. However, most CdiA-CT sequences do not have Pfam designations nor predictions for their biochemical activities (3). Given that the CdiA-CT E479 , CdiA-CT 1026b , and CdiA-CT YPIII toxins were not identified as PD(D/E)XK nucleases by prior computational surveys, it remains possible that other uncharacterized CDI toxins also belong to the superfamily. Because of extreme sequence variability and catalytic residue migration, PD(D/ E)XK enzymes are notoriously difficult to identify through computational approaches (23,25,38). This problem is compounded by insertions and circular permutations of the core structure (23). Aravind and co-workers (14) recently predicted five new restriction endonuclease-like domains (Tox-REase-2, -3, -5, -7, and -9) that are associated with prokaryotic competition systems. Only the Tox-REase-7 family (Pfam PF15649) is found in CdiA effectors, and these CDI toxins appear to be limited to Pseudomonas and Acinetobacter species. These systems are under considerable positive selection to diversify, presumably due to the competitive advantage obtained with novel toxins. Similar pressures are postulated to drive the impressive diversity of restriction endonucleases, which is the result of the complex interplay between bacteria and their phages (44). Thus, it is not surprising that the versatile PD(D/E)XK core structure has been adopted by CDI and other prokaryotic competition systems.

Experimental Procedures
Plasmid Constructions-Plasmids used in this study are listed in Table 4. Constructs for the overproduction of CdiA-CT/ CdiI 1026b -His 6 (pCH7590), CdiA-CT(D285A)/CdiI E479 -His 6 (pCH8288), and wild-type CdiA-CT/CdiI E479 -His 6 (pCH7770) complexes have been described previously (16). Active-site mutations were made in the CdiA-CT E479 nuclease domain using mega-primer PCR. Plasmid pCH7770 was amplified with primer E479-cdiI-Spe-rev (  TGC GGC TGC TGC GG). The resulting products were used as mega-primers in subsequent reactions with forward primer E479-Nco-for (5Ј-CGG CCA TGG CAT CGA ACG TCG AGC TTT AC). The final products were digested with NcoI and SpeI and then ligated to plasmid pET21 to generate mutant CdiA-CT/CdiI E479 -His 6 expression constructs. These plasmids were used as templates to amplify cdiA-CT E479 coding sequences with primers E479-Nco-for and E479-CT-Xho-rev (5Ј-GCC ACT CGA GCC TTA CTT GAT CAG AAT AAT C). The products were digested with NcoI and XhoI and then ligated to plasmid pSCRhaB2 (45) to generate L-rhamnose-inducible expression constructs to monitor growth inhibition activities. Plasmid pCH8479 was amplified with oligonucleotides 1026b-Spe-for (5Ј-ATA ACT AGT GCA TCG AAC GTC GAG C) and 1026b-CT-Xho-rev (5Ј-AAT CTC GAG TTA ATT CCC CTT TGG), and the resulting fragment was ligated into plasmid pSH21 to generate a construct that overproduces inactive His 6 -CdiA-CT(D214A) 1026b . The cdiA-CT(D285A) E479 coding sequence was amplified from pCH8427 with primers E479-CT-NdeI-H6for (5Ј-GAT CAT ATG ATG GGG GCA AGC TCA GGT AGT AAT ATC) and E479-CT-EcoRI-rev (5Ј-GAT GAA TTC TCA CTT GAT CAG AAT AAT CTT CGC CTG CAG TTT). The product was digested with NdeI/EcoRI and ligated to pET28b. The D243A mutation was introduced via site-directed mutagenesis using primers E479-CT-D243A-for (5Ј-CGG TAA GAC CGT GGC GTT TAT GCT TAC GCC-3Ј) and E479-CT-D243A-rev (5Ј-GGC GTA AGC ATA AAC GCC ACG GTC TTA CCG-3Ј) to produce an expression construct that overproduces His 6 -CdiA-CT E479 carrying the D243A and D285A mutations. Protein Overexpression and Purification-CdiA-CT E479 (from residue Gly-157, numbered from Glu-1 of the ELYN motif) was co-expressed with CdiI E479 -His 6 and overproduced in E. coli BL21 (DE3) cells grown aerobically at 37°C in LB medium supplemented with 50 g/ml ampicillin. Protein expression was induced by addition of isopropyl ␤-D-thiogalactoside to 1 mM final concentration once the culture reached an absorbance at 600 nm (A 600 ) of ϳ0.8. Induced cells were incubated for 4 h and then harvested by centrifugation at 5,100 ϫ g for 20 min. The cell pellet was resuspended in 20 mM sodium phosphate (pH 7.0), 200 mM NaCl supplemented with 10 mg/ml lysozyme, and 1 mM phenylmethylsulfonyl fluoride (PMSF), and the cells were broken by sonication. The lysate was clarified by centrifugation at 14,000 ϫ g for 30 min, and the soluble fraction passed through a 0.22-m filter before loading onto a Ni 2ϩ -charged Hi-trap column (GE Healthcare). The column was washed with 20 mM sodium phosphate (pH 7.0), 200 mM NaCl, 15 mM imidazole, and the CdiA-CT/CdiI E479 -His 6 complex eluted with a linear gradient of 15-250 mM imidazole. The purified complex was concentrated with a 10-kDa centrifugal concentrator and then run on a Superdex 200 size-exclusion column equilibrated in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl. SeMet-labeled proteins were overproduced in E. coli BL21 (DE3) cells grown in M9 minimal medium supplemented with L-leucine, L-isoleucine, and L-valine at 50 mg/liter; L-phenylalanine, L-lysine, and L-threonine at 100 mg/liter; and SeMet at 75 mg/liter as described (46). The SeMet-labeled CdiA-CT/CdiI E479 -His 6 complex was purified as described above.
Crystallization and Structure Determination-CdiA-CT/ CdiI E479 was crystallized by hanging drop-vapor diffusion against a 1-ml reservoir of crystallization buffer (0.1 M HEPES (pH 7.0), 20 mM MgCl 2 , 30% (w/v) polyacrylic acid). Polyacrylic acid (Sigma) with average molecular mass of 5,100 Da was prepared as a 50% (w/v) solution, filtered, and used to facilitate crystallization. Hanging drops were prepared from a 1:1 (v/v) mixture of protein solution (20 mg/ml) and crystallization buffer supplemented with 20 g/ml chymotrypsin. Crystals were soaked in cryo-protectant solution containing 1:1 mixture of 40% (v/v) glycerol and crystallization buffer and then collected by flash freezing. A native dataset was acquired at 70 K at 0.97591 Å on beamline 7-1 at Stanford Synchrotron Radiation Lightsource. Data were processed using HKL2000 (47), resulting in a 99.24% complete dataset to 2.0 Å resolution. The CdiA-CT/CdiI E479 complex crystallized in space group P22 1 2 1 with two complexes per asymmetric unit and unit cell dimensions of 54.5 ϫ 73.3 ϫ 110.0 Å. Diffraction data were initially indexed and scaled to P222 1 ; however, the best solution obtained by molecular replacement using Phaser in the PHENIX suite was P22 1 2 1 . The SeMet-labeled complex was crystallized using 0.1 M MES (pH 6.5), 0.01 M ZnCl 2 , 15% (w/v) PEG-6000 as the buffer. SeMet-labeled CdiA-CT/CdiI E479 crystallized in space group I4 with three complexes per asymmetric unit and unit cell dimensions of 117.2 ϫ 117.2 ϫ 111.6 Å. A SAD was collected at 70 K at 0.97591 Å on beamline 7-1 at Stanford Synchrotron Radiation Lightsource. Data were processed using HKL2000, yielding a 99.9% complete dataset to 3.3 Å resolution. We used Autosol in the PHENIX suite (48) to detect 18 selenium atoms with a figure of merit of 0.408 and overall score of 36.2 Ϯ 14.2. Autosol also built a partial model consisting of 416 residues with an R work /R free (%) of 38.4/44.9. This model showed little secondary structure except for ␣1 of CdiA-CT E479 and ␣2Ј and ␣3Ј of CdiI E479 . Molecular replacement was carried out with Phaser in the PHENIX suite using the partial model together with higher resolution data from native crystals. The Phaser-generated model was then subjected to Autobuild and phenix.refine (48). The final model includes residues Arg-201-Lys-316 of CdiA-CT E479 and residues Ala-2-Gly-105 of CdiI E479 with a final R work /R free (%) 19.3/23.7. The Ramachandran plot shows 98.8% in the favorable allowed regions and the other 1.2% in the allowed regions. Data collection and refinement statistics are presented in Table 1. Intermolecular hydrogen bonds and salt bridges were determined using PDBePISA (49).
Toxin Immunity Protein Binding Kinetics-The apparent equilibrium dissociation constant for the CdiA-CT/CdiI E479 complex was determined by biolayer interferometry using a BLitz instrument (ForteBio) (50). CdiA-CT E479 was separated from CdiI E479 -His 6 by Ni 2ϩ -affinity chromatography under denaturing conditions. The isolated toxin was refolded by dialysis and run on a Superdex S200 size-exclusion column. CdiI E479 -His 6 was purified by Ni 2ϩ -affinity chromatography as described previously (16). CdiI E479 -His 6 was loaded onto a Ni 2ϩ -NTAcoated biosensor in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl at 25°C. Sensor-bound immunity protein was then incubated with 1.5-5 M CdiA-CT E479 toxin for 180 s. The sensor was then washed with buffer, and toxin dissociation was monitored over 180 s. Curve fitting was run following reference subtraction using the BLitz Pro Software to calculate dissociation constants. Local fit analyses were performed for individual association-dissociation curves, followed by averaging to obtain the final apparent K d value and standard deviation.
Growth Inhibition Assays-E. coli X90 cells harboring rhamnose-inducible CdiA-CT E479 expression plasmids (Table 4) were grown to mid-log phase in LB media supplemented with 100 g/ml trimethoprim. Cells were then diluted to A 600 ϭ 0.05 in fresh LB media supplemented with 100 g/ml trimethoprim and either 0.4% D-glucose to repress or 0.4% L-rhamnose to induce CdiA-CT E479 expression. Cultures were incubated with shaking at 37°C, and cell growth was monitored by measuring the A 600 every 30 min.
In Vitro tRNase Assays-Purified CdiA-CT/CdiI E479 -His 6 complexes were denatured in binding buffer supplemented with 6 M guanidine-HCl and CdiA-CT E479 isolated from the void volume during Ni 2ϩ -affinity chromatography (1). Toxins were refolded by dialysis against binding buffer, and all purified proteins were quantified by absorbance at 280 nm. Total RNA was isolated from E. coli X90 cells as described (51) and used as a substrate for in vitro nuclease assays. E. coli RNA (5 g) was incubated with CdiA-CT E479 variants (5 M) in reaction buffer (20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10 mM MgCl 2 , 0.2 mg/ml BSA) for 1 h at 37°C. Where indicated, CdiI E479 was included at 17.5 M final concentration. tRNA Asp and tRNA Gln substrates were generated by in vitro transcription using RNA polymerase from bacteriophage T7 RNA. The tRNA Asp template was prepared with oligonucleotides 5Ј-tRNA-Asp (5Ј-AAT TCC TGC AGT AAT ACG ACT CAC TAT AGG AGC GGT AGT TCA GTC GGT TAG AAT ACC TG) and 3Ј-tRNA-Asp (5Ј-TGG CGG AAC GGA CGG GAC TCG AAC CCG CGA CCC CCT GCG TGA CAG GCA GGT ATT CTA AC), and the tRNA Gln template with oligonucleotides 5Ј-tRNA-Gln (5Ј-AAT TCC TGC AGT AAT ACG ACT CAC TAT AGG GGG TAT AGG GGG TAT CGC CAA GCG GTA AGG CAC CGG) and 3Ј-tRNA-Gln (5Ј-TGG CTG GGG TAC GAG GAT TCG AAC CTC GGA ATG CCG GAA TCA GAA TCC GGT GCC TT). Annealed oligonucleotides were end-filled with Klenow fragment of DNA polymerase I. Templates were incubated with T7 RNA polymerase, 2 mM NTP, 10 mM dithiothreitol, and 10 mM MgCl 2 for 3 h at 37°C. Template DNA was removed with RNase-free DNase I, and the transcripts purified with the Direct-zol RNA MiniPrep kit (Genesee Scientific). Nuclease reactions were analyzed by denaturing electrophoresis on 50% urea, 6% polyacrylamide gels in 1ϫ Tris borate-EDTA (TBE) buffer. Gels were stained with ethidium bromide or transferred to nylon membrane for Northern blot hybridization with 5Ј-radiolabeled oligonucleotide glyV probe (5Ј-CTT GGC AAG GTC GTG CT) as described (16,51).
Molecular Docking-Hex 8.0 (52-54) was used to dock CdiA-CT nuclease domains onto tRNA to generate models of enzyme/substrate complexes. CdiA-CT 1026b and CdiA-CT E479 nuclease domains were docked onto the structure of E. coli tRNA Cys (PDB code 1B23) (55). The active site of each nuclease was positioned adjacent to the known scissile bond and the origin set to sample multiple orientations in search of the low energy interactions. Positive control docking simulations were performed using cognate and non-cognate toxin/immunity proteins pairs from B. pseudomallei E479 and 1026b. As a negative control, CdiI E479 and CysK from Salmonella typhimurium LT2 (PDB code 1OAS (56)) were docked onto one another.
SAXS-Inactive CdiA-CT(D243A/D285A) E479 and CdiA-CT(D214A) 1026b (17) toxins carrying N-terminal His 6 epitope tags were purified by Ni 2ϩ -affinity chromatography under nondenaturing conditions. Under these conditions, endogenous tRNA co-purifies with the inactive nuclease domains. Toxin/ tRNA complexes were exchanged into 20 mM sodium phosphate (pH 7.4), 150 mM NaCl using a Superdex S200 size-exclusion column and diluted to several concentrations ranging from 0.5 to 5 mg/ml for SAXS analysis. SAXS data were collected on SIBYLS beamline 12.3.1 at the Advanced Light Source using a Pilatus3 2 M detector with exposure times of 0.5, 1, 2, and 4 s. Buffer subtracted data were analyzed using PRIMUS (31), following modification with GNOM (57); P(r) output files with d max of 91 and 190 for tRNA/CdiA-CT 1026b and tRNA/CdiA-CT E479 complexes (respectively) were used to generate electron density envelopes via GASBOR (58). Density envelopes (12 per tRNA/toxin complex) were averaged using DAMAVER (31), and docking solutions were fitted into the final envelopes using Chimera (32) and Crysol (59). SAXS parameters and statistics are provided in Table 5 according to (60). Crysol outputs, together with Guinier, Kratky and P(r) plots, are presented in Fig. 9.