Structure of the Proteus vulgaris HigB-(HigA)2-HigB Toxin-Antitoxin Complex*

Background: Toxin-antitoxin (TA) systems play a crucial role in bacterial survival during stress. Results: Structures of the P. vulgaris HigBA complex reveal novel structural features such as the HigB and HigA interaction and the solvent accessibility of the HigB active site. Conclusion: Antitoxin HigA interacts with toxin HigB in a novel manner. Significance: Our results emphasize that antitoxins are a structurally diverse class of proteins. Bacterial toxin-antitoxin (TA) systems regulate key cellular processes to promote cell survival during periods of stress. During steady-state cell growth, antitoxins typically interact with their cognate toxins to inhibit activity presumably by preventing substrate recognition. We solved two x-ray crystal structures of the Proteus vulgaris tetrameric HigB-(HigA)2-HigB TA complex and found that, unlike most other TA systems, the antitoxin HigA makes minimal interactions with toxin HigB. HigB adopts a RelE family tertiary fold containing a highly conserved concave surface where we predict its active site is located. HigA does not cover the solvent-exposed HigB active site, suggesting that, in general, toxin inhibition is not solely mediated by active site hindrance by its antitoxin. Each HigA monomer contains a helix-turn-helix motif that binds to its own DNA operator to repress transcription during normal cellular growth. This is distinct from antitoxins belonging to other superfamilies that typically only form DNA-binding motifs upon dimerization. We further show that disruption of the HigB-(HigA)2-HigB tetramer to a HigBA heterodimer ablates operator binding. Taken together, our biochemical and structural studies elucidate the novel molecular details of the HigBA TA system.

Toxin-antitoxin (TA) 3 systems are chromosomally or plasmidencoded gene pairs found in free-living bacteria that aid in survival during environmental and chemical stresses (1). TA systems have been implicated in diverse functions such as programmed cell death, growth, and gene regulation, biofilm formation, and persistence during increased antibiotic exposure, but their precise physiological functions are controversial (2)(3)(4)(5)(6)(7)(8). Their roles in persistence, adaptation, and survival mechanisms underscore their great potential as novel antimicrobial targets (9).
Type II TA operons encode both a small antitoxin and toxin protein (8 -12 kDa each) that under normal growth conditions form a tight, nontoxic complex. These complexes transcriptionally autorepress by binding at operator sequences in their promoter region (1). Upon stress, the antitoxin is degraded by proteases, allowing the toxin to target key cellular processes, including replication (DNA gyrase) and translation (free mRNA, ribosome-bound mRNA, or the ribosome itself) (10 -17). Tightly regulating and/or reducing these energetically expensive processes leads to an overall decrease in metabolite consumption and halts cell growth. This bacteriostatic state continues until the stress passes (1).
RelE is one of the best studied ribosome-dependent toxins and functions by degrading mRNAs preferentially at stop codons in the ribosomal A site (11). Recent evidence suggests RelE may recognize additional codons, but the molecular details of this specificity remain unclear (18,19). The additional RelE family member YafQ cleaves at lysine codons, and member YoeB cleaves at both sense and stop codons (14,20,21). The host inhibition of growth B (HigB) protein from Proteus spp. is a RelE family member with a relaxed codon specificity (13,22). HigB preferentially degrades 5Ј-AAA-3Ј codons (lysine), but codons containing only one adenosine are sufficient for degradation by HigB (13).
The Proteus vulgaris HigBA TA system was first discovered on an exogenous plasmid that conferred kanamycin resistance and post-segregational killing at elevated temperatures (23). This plasmid was isolated from a post-operative pyelonephritis, an ascending urinary tract infection (23,24). The higBA gene pair is not found in Escherichia coli K12 but is found chromosomally in pathogens such as Vibrio cholerae, Streptococcus pneumoniae, E. coli CFT073, and E. coli O157:H7 (25).
The HigB toxin gene and protein are distinguished from those of other RelE family toxins in three ways. First, the higBA operon has an inverted gene structure with the HigB toxin gene preceding its cognate antitoxin (Fig. 1A) (23). This gene arrangement is only seen in the MqsRA and hicAB TA systems (15,25). Second, sequence alignments with other RelE family members indicate that HigB appears to lack conserved catalytic residues required for mRNA recognition and degradation ( Fig.  2A). Third, a single adenosine in the context of a codon is sufficient for degradation by HigB (13). This contrasts with previously proposed strict mRNA sequence requirements for other toxins (11).
We report the structural and biochemical characterization of the novel TA pair HigBA. Remarkably, our structure shows that, unlike most antitoxins, HigA makes relatively few contacts with its toxin partner and does not cover the solvent-accessible HigB active site. This structural arrangement implies a possible novel model of inhibition. We also present biochemical data that demonstrate tetrameric HigBA (henceforth denoted as HigB-(HigA) 2 -HigB to reflect its spatial organization) is required for productive binding to its own DNA operator sequences, validating the functional relevance of our structural data.

EXPERIMENTAL PROCEDURES
Plasmids pET21c-HigBA and pET28a-His 6 HigBA were generous gifts from Dr. Nancy A. Woychik (Rutgers-Robert Wood Johnson Medical School). A C-terminal hexahistidine (His 6 ) tag encoded on the pET21c construct was added to HigA of the pET21c-HigBA construct by removal of the natural HigA stop codon using site-directed mutagenesis to create pET21c-Hig-BAHis 6 . The pET28a-His 6 HigBA(⌬84 -104) plasmid was created by placing a premature stop codon in HigA after the codon 83. All sequences were verified by DNA sequencing (GeneWiz).

Crystallization, X-ray Data Collection, and Structural Determination of HigBA Complexes
HigBA-His 6 (Crystal Form 1)-Crystals of trypsinized selenomethionine-derivatized HigBA-His 6 were grown by sitting drop vapor diffusion in 3-10% PEG 3350, 0.2 M L-proline, and 0.1 M HEPES, pH 7.5, over approximately 2 days at 10°C. Ethylene glycol was used as a cryoprotectant and added in two increments to a final concentration of 30%. Crystals were flashfrozen in liquid nitrogen, and a single anomalous dispersion dataset was collected at the Northeastern-Collaborative Access Team (NE-CAT) 24-IDC beamline at the Advanced Photon Source using 0.979 Å radiation (Table 1). A total of 113,311 reflections were collected, indexed, and reduced to 16,748 unique reflections (unmerged) to a resolution of 2.8 Å with the program HKL2000 (27). Phase determination was carried out using the intrinsic anomalous signals from selenium. A total of 11 heavy atom sites were identified and used for initial phases with the program Autosol of the PHENIX Suite (28). The starting model was initially built by PHENIX Autobuild (28), followed by manual building in Coot (29). During refinement, XYZ coordinates, real space, and B-factors (isotropic) were refined to a final R work /R free of 19.7/23.8. The final model contained two HigB and two HigA molecules per asymmetric unit ( Fig. 1 and supplemental Fig. S1A).
His 6 -HigBA (Crystal Form 2)-Crystals of His 6 -HigBA were grown by sitting drop vapor diffusion in 90 mM sodium acetate, pH 4.6, 180 mM ammonium acetate, 25% PEG 4000, and 4% acetone at 20°C in 1 week. For cryoprotection, dextrose was dissolved in the reservoir solution and added to the crystallization drop in 15% increments up to 30% (w/v) by exchanging the mother liquor. This was followed by 1-2 min of equilibration, flash frozen in liquid nitrogen, and a native dataset was collected at NE-CAT 24-IDE beamline. A total of 172,519 reflections were collected, indexed, and reduced to 31,287 unique reflections with the program XDS (30). The structure was solved to 2.2 Å by molecular replacement using the AutoMR PHENIX program (28) with one HigB and one HigA molecule from the previously solved HigBA complex as a search model (form 1). Three HigB and three HigA molecules were found in the asymmetric unit (supplemental Fig. S1B). A similar PHENIX refinement scheme was used as with form 1 but with the addition of TLS refinement. Manual model building in Coot was performed to a final R work /R free of 17.3/21.1% (29).
Protein interfaces, surfaces, and assemblies (PISA) program was used to calculate molecular interfaces and oligomeric states (31), and ConSurf was used to map HigB sequence conservation onto the crystal structure (32). Sequence alignments were performed with ClustalW (33), and all figures were generated using PyMOL (34).

Size Exclusion Chromatography (SEC) Assay
One hundred microliters of 75 M protein in SEC buffer were loaded onto a Superdex 75 10/300 column (GE Healthcare). Estimated molecular weights were calculated by comparison with the molecular weight standards (Bio-Rad) (Fig. 5D). Peaks from the SEC chromatogram corresponding to different protein-protein complexes were run on a 15% SDS-polyacrylamide gel for analysis (Fig. 5E).

Electrophoretic Mobility Shift Assay (EMSA)
Assays were performed as described previously (35) but with slight modifications. Double-stranded DNA representing the Phig region was generated by mixing chemically synthesized DNA (IDT), heating to 90°C for 2 min, and slowly cooling to room temperature (Fig. 5A). Protein at a final concentration of 0, 0.25, 0.5, 1, and 2 M was incubated with 10 ng of DNA for 20 min on ice along with 0.5 mg/ml BSA. Free and proteinbound DNA were resolved on a native 8% polyacrylamide gel prepared with Tris borate, pH 8, EDTA buffer (Fig. 5C). The gel was run at 10°C for 1 h, and DNA was stained with SYBR Green dye (Invitrogen) and visualized using a Typhoon Trio (GE Healthcare).

Molecular Modeling HigB on the 70 S Ribosome
HigB was modeled on E. coli RelE bound to the Thermus thermophilus 70 S ribosome ( Fig. 6 and supplemental Fig. S3) (PDB code 3KIQ) (36). The HigB coordinates were optimally superimposed onto RelE using secondary structure matching in Coot (37). Conserved secondary structural motifs of the RNase fold of RelE and HigB aligned with a root mean square deviation (r.m.s.d.) of 2.4 Å (for 63 equivalent ␣-carbon pairs) (supplemental Fig. S3).

RESULTS
Structural Determination of the HigB-(HigA) 2 -HigB Complex-By placing the hexahistidine tag at either the N terminus of HigB or the C terminus of HigA, we were able to solve two different x-ray crystal structures of the HigBA complex ( Fig. 1 and supplemental Fig. S1). Given how small each protein is (the antitoxin is 11.5 kDa and the toxin is 10.7 kDa), we were concerned that the affinity tag may influence potential crystal packing interactions and the overall oligomeric states. However, both crystal structures are entirely consistent, with an overall r.m.s.d. of 0.9 Å for 366 equivalent ␣-carbon pairs with only a single minor difference within loop 5 of HigB (supplemental Fig. S1C) (38).
The HigBA-His 6 complex (form 1) crystallized in the hexagonal space group P3 2 21 with two HigBA heterodimers per asymmetric unit (supplemental Fig. S1A). The initial phases to 2.8 Å were obtained by single anomalous dispersion using selenomethionine-derivatized protein (Table 1). This model was used as an initial search model for the His 6 -HigBA structure (form 2) and was solved using molecular replacement to 2.2 Å ( Table 1). The form 2 complex grew in the hexagonal space group P6 2 with three HigB and three HigA molecules per asymmetric unit (Table 1 and supplemental Fig. S1B). Both forms 1 and 2 contain a HigB-(HigA) 2 -HigB tetramer, although form 2 contains an additional HigBA dimer in the asymmetric unit. A full tetramer is formed by applying two-fold crystallographic symmetry (supplemental Fig. S1B). Thus, the overall subunit compositions of HigB and HigA are identical.
In form 1, residues 1-90 (92 total) were built for one HigB molecule, although unambiguous density allowed building of all HigB residues of the second molecule. The C terminus of the fully built HigB is involved in crystal contacts with a neighboring crystallographic symmetry-related molecule, which presumably stabilized this region. The side chain and backbone of HigB residues Lys-57, Asp-59, and Glu-61 have poor electron density in both crystal forms, and two out of the three HigB molecules from form 2 showed little to no C␣ electron density for Lys-57 and Asp-59. Therefore, the backbone was built using neighboring residues as guides for ␣-carbon positions. In form 2, residues 1-90 were built for all three HigB molecules. In both crystal forms, HigA (104 amino acids total) was modeled to either residue 92 or 93 as no interpretable electron density was seen beyond these positions.
Both HigBA structures adopt nearly identical tertiary and quaternary structures (supplemental Fig. S1C). Two HigA molecules form a dimer similar to that observed in previous HigA crystal structures without the toxin (supplemental Fig. S2A) (39). Each HigA interacts with one HigB molecule to form a heterodimer that with additional HigA-HigA interactions completes a dimer of heterodimers (Fig. 1, B and C). Consistent with our structural results, PISA predicts the HigBA complex to exist as a tetramer (31). The structure of the P. vulgaris HigA dimer in the context of the TA complex is very similar to that of HigA alone from E. coli CFT073 (PDB codes 2ICT and 2ICP) (39) and Coxiella burnetti (PDB code 3TRB) with r.m.s.d. of 2.5, 1.6, and 1.6 Å, respectively. This indicates HigA does not undergo large conformational changes upon toxin binding (supplemental Fig. S2A).

vulgaris
HigB with other ribosome-dependent toxins showing residues with 50, 75, or 100% sequence identity as light, medium, or deep purple, respectively. Residues located within the HigB concave surface (purple circles) and E. coli RelE amino acids that recognize and/or degrade mRNA (black circles and triangles, respectively) are indicated. B, HigB toxin structure colored by amino acid conservation among HigB homologs according to the scale shown (1 is least conserved and 9 is the most conserved). Residues located on the concave surface proposed to contain active site residues are shown as sticks. HigB residues that make ionic interactions with HigA are also shown as sticks and colored by conservation. C, E. coli RelE R81A toxin structure (PDB code 4FXI) with residues identified as important for mRNA recognition or cleavage shown as sticks. Water molecules are shown as red spheres and color scheme is the same as in Fig. 1. B, ϳ45°rotated view of A highlighting additional salt bridge and hydrogen bonding interactions. C, hydrophobic interactions formed between HigA and HigB.
HigA Monomer Contains an Intact DNA Binding Domain-The P. vulgaris HigA protein contains a compact five ␣-helical bundle and a disordered C terminus (residues 93/94 -104) (Fig.  1B). All ␣-helices were juxtaposed, and their relative orientation is very similar to members of the xenobiotic response element-helix-turn-helix family (XRE-HTH) of DNA-binding proteins (49). Family members include the P22 C2 and phage 434 proteins, which transcriptionally repress specific genes by binding to their operator regions in the major groove in a sequence-specific manner (50,51).
HigA has a number of unique structural characteristics in addition to the presence of the HTH motif. For example, each HigA monomer contains a defined hydrophobic core unlike other antitoxins that recognize RelE family members. Normally antitoxins only form a hydrophobic core upon self-dimerization and have typically been classified as partially unstructured (41,43,44,52,53). Additionally, most antitoxins that recognize RelE family members form one DNA-binding motif upon dimerization (41, 43, 44, 52, 53). In sharp contrast, each HigA monomer contains a complete DNA-binding motif. Therefore, the HigA dimer contains two DNA-binding motifs that fully extend over the two 9-nucleotide inverted repeats of the hig operator shown to interact with HigA through DNase protection assays (Figs. 1A and 5A and supplemental Fig. S2B) (35). These results imply that a single HigB-(HigA) 2 -HigB tetramer can repress an operator site consisting of two inverted repeat sequences (Figs. 1A and 5A).
HigA Mediates the Formation of the HigB-(HigA) 2 -HigB Complex-HigA dimerizes to form a dimer of heterodimers (Fig. 1, B and C). These HigA dimers interact in a two-fold symmetrical manner mainly stabilized by hydrophobic interactions (Fig. 4, B and C). HigA ␣5 packs against ␣5Ј of the partner HigA molecule in an antiparallel fashion (Fig. 4C). Loop 6Ј packs against ␣4 of its partner HigA and caps the junction formed by ␣1, ␣2, and ␣4 of the adjacent HigA molecule (Fig.  4, C and D). This 1,240 Å 2 interface is mediated primarily via hydrophobic amino acids (Ile-54, Leu-68, Leu-76, Leu-79, Ile-83, Ile-88, and Tyr-91) from both molecules (Fig. 4C). For comparison, the HigB-HigB interface is 280 Å 2 (Fig. 1B). Thus, the HigA-HigA interaction plays a major role in driving the formation of the tetrameric HigB-(HigA) 2 -HigB complex. HigA Does Not Mask the HigB Active Site-The HigB active site is likely located at a concave surface where a high density of conserved residues reside, including the proposed catalytic HigB amino acid His-92 (Fig. 2, A and B) (13). RelE amino acids that contact and cleave mRNA cluster in a similar concave surface (Fig. 2C) (36). This surface is located ϳ20 Å distal opposite to the HigA-HigB interface (Fig. 1B). Additionally, the active site is solvent-accessible, and this suggests that simple active site steric occlusion by HigA is not the mechanism of HigB inactivation.
The higBA operator sequence used in the EMSAs includes two endogenous operator sites, both of which in turn contain two inverted repeats (Fig. 5A). Wild-type HigB-(HigA) 2 -HigB binds to its own operator DNA with increasingly higher oligomeric states (Fig. 5C, lanes 2-5). This indicates more than one HigB-(HigA) 2 -HigB complex interacts with its promoter. However, we found that HigBA(⌬84 -104) is unable to bind to this same DNA promoter (Fig. 5C, lanes 7-10). Considering that each HTH motif of HigA is left intact in this mutant, it is surprising that HigBA(⌬84 -104) is unable to bind DNA. Two possibilities for this result exist. The first is that the HigA mutation caused destabilization of the protein, and little to no soluble HigA is produced. The second possibility is that removal of the C terminus of HigA disrupts the HigA-HigA dimerization interface resulting in a HigBA heterodimer.
The SEC results show wild-type HigB-(HigA) 2 -HigB elutes as the expected tetrameric complex of 56 kDa, whereas purified HigBA(⌬84 -104) elutes with an apparent molecular mass of 23 kDa (Fig. 5D). This is approximately the molecular mass of a dimer of HigA or HigB or HigBA heterodimer. To differentiate between these options, we analyzed the fractions of each peak with SDS-PAGE (Fig. 5E). The 56-kDa peak of the wild-type HigB-(HigA) 2 -HigB complex shows a large band at ϳ10 kDa, which is most likely both His 6 -HigB (13.0 kDa) and HigA (11.5 kDa) (Fig. 5E, lane 1). These bands could be separated by treatment with thrombin to release the N-terminal His 6 tag and the linker region of HigB; this allows the identification of both HigB and HigA (Fig. 5E, lane 2). The HigBA(⌬84 -104) complex that runs at ϳ23 kDa shows two distinct bands on the SDS-polyacrylamide gel (Fig. 5E, lane 3). Upon thrombin treatment to release the N-terminal His 6 tag and the linker region of HigB, we again see the appearance of tag-free HigB (Fig. 5E, lane 4). Therefore, HigBA(⌬84 -104) is indeed a heterodimer of HigB and truncated HigA(⌬84 -104). Taken together, these results demonstrate that the oligomeric state of the HigBA complex, specifically a HigB-(HigA) 2 -HigB tetrameric state, is required for productive DNA interaction. This is despite each HigA monomer containing a full HTH motif.

DISCUSSION
TA systems commonly contain at least two operator regions with two imperfect inverted repeats comprising a single operator site (Fig. 5A). Antitoxin proteins belonging to the ribbonhelix-helix, AbrB, and PhD/YefM superfamilies require dimerization to form a single DNA binding domain that recognizes an inverted repeat (Figs. 1A and 5A). Direct binding of either an antitoxin dimer or a toxin-antitoxin complex confers transcriptional autorepression. The strength of the repression correlates to differences in binding affinities of either antitoxin dimers or TA complexes.
The crystal structures of the HigB-(HigA) 2 -HigB complex presented here reveal the TA complex is a tetramer containing two, rather than one, DNA-binding motifs (Fig. 1, B and C). Our biochemical results indicate that a HigB-(HigA) 2 -HigB tetrameric complex is essential for DNA binding (Fig. 5C). The loss of DNA binding upon disruption of HigA dimerization (thus forming a HigBA heterodimer) may result from a diminished interaction surface, culminating in reduced binding. Both the HigA dimer and HigB-(HigA) 2 -HigB tetramer provide the same surface area for the inverted repeats to interact with, which is halved in the context of the HigBA dimer. Moreover, another possible reason for the HigBA dimer ablating DNA binding is that the HTH motifs, in the context of the HigB-(HigA) 2 -HigB tetramer, are tethered or rigidly held in place because they are part of the HigA globular domain formed upon HigA dimerization (Fig. 1, B and C). In this manner, the precise structural arrangement may be functioning as a molecular ruler for specific recognition of both inverted repeats as seen in other HTH-containing proteins, such as Fis (58). In summary, both the DNA interaction area formed by the HigA dimer and the spatial organization of the HTH motifs coordinate to recognize DNA and render the area impenetrable to RNA polymerase.
Transcriptional autorepression by TA complexes has been proposed to occur by regulation of the overall molar ratio of toxins and antitoxins as shown in vivo for RelEB (57). By varying the molar ratio, different stoichiometric complexes form, which may function to repress transcription to different magnitudes or cause complete derepression. In vitro biochemical experiments for the RelEB, PhD-Doc, and CcdAB TA systems demonstrate that once a saturated TA-DNA complex is formed, increasing the amount of free toxin protein destabilizes the DNA-TA interaction and probably allows for derepression of transcription (56,57,59,60). These studies led to a model referred to as conditional cooperativity (56,57). This model can help explain why toxins function as either anti-or corepressors depending upon environmental changes that require bacteria to respond and regulate metabolic processes quickly.
Despite these studies, the mechanism by which different TA complexes repress transcription is still not entirely clear. Structural and modeling studies of the RelEB and the PhD-Doc TA complexes suggest two tetrameric TA complexes sterically clash at single operator sites, although in the case of RelEB, modeling studies indicated that trimeric complexes can coexist (44,56). However, the proposed trimeric complexes of RelEB and PhD-Doc have not been observed structurally. Therefore, it is not obvious how the diverse oligomeric states from structural studies fit with these models (43,44,56).
In contrast, structures of TA complexes such as N. gonorrhoeae FitAB and Rickettsia felis VapBC bound to two inverted repeats suggest that higher oligomeric states can simultaneously bind to two inverted repeats without steric clashes (55,61). Because the spacing between inverted repeats may play a role in which oligomeric complexes fit, it is interesting that the fitAB promoter contains 12 bp between inverted repeats, although the vapBC promoter has only two. So, in this context, both short and long spacings between inverted repeats give rise to a higher oligomeric state binding to DNA. Modeling of our HigB-(HigA) 2 -HigB structure on the structure of phage 434 bound to a double-stranded 20-nucleotide DNA (PDB code 1RPE) indicates that it is possible for both HTH motifs of the HigB-(HigA) 2 -HigB tetramer to interact with one complete operator site (two inverted repeats) without steric clashes (supplemental Fig. S2B).
HigA is not the only antitoxin that contains an HTH motif. MqsA and HipB antitoxins also possess HTH motifs, but there are key structural and functional differences among the three (62,63). First is the location of the HTH motif relative to the toxin (supplemental Fig. S2C). The HipA toxin binds to HipB at a location orthogonal to its HTH motif, although MqsA has a separate N-terminal toxin neutralization domain and a C-terminal DNA binding domain (supplemental Fig. S2C). The HipA toxin is also a much larger protein and is not homologous to either MqsR or HigB (63). Second, HipA contacts two HipB antitoxins, whereas HigB and MqsR only contact one. Finally, an important difference is that the MqsA antitoxin alone, and not the MqsRA TA complex, binds DNA (64). Thus, in this example, the toxin does not appear to act as a corepressor. In summary, although all three have an HTH motif in common, they show substantial functional and structural differences underscoring antitoxin plasticity.
HigB is a ribosome-dependent RNase that cleaves codons containing at least a single adenosine located, most likely, in the A site of the ribosome (13). Our HigB structure reveals a solvent-exposed concave surface containing highly conserved residues (Fig. 2B). Several lines of evidence suggest this HigB concave surface is its active site. Microbial RNases such as RNase T1 or RNase Sa contain similar concave surfaces, and ribosome-dependent toxins have been proposed to degrade mRNA in an analogous manner (36,65,66). An x-ray crystal structure of RelE bound to the 70 S ribosome shows the same concave surface interacts with mRNA (36). Additionally, mutagenesis experiments of other toxins also implicate the same concave surface residues as important for function (36,41,43,62). In summary, HigB appears to use the same tertiary fold and surface to recognize ribosome-bound mRNA.
A hallmark of toxin inactivation is a direct interaction in which the antitoxin wraps around the toxin much like a pincer (Fig. 4E) (43,44,52,56,61,62,(67)(68)(69). This toxin-antitoxin interaction ablates activity of the toxin, and although the precise mechanism is unknown, it has been proposed to occur via antitoxin masking of the toxin active site. Our structure reveals that the antitoxin HigA does not wrap around and mask the active site of HigB. Instead, only two regions of contact are made, both of which are distant from the active site ( Figs. 1 and  3). The MqsRA and the HipBA TA pairs also do not wrap around their cognate toxins but interact in a manner and location distinct from HigB-(HigA) 2 -HigB ( Fig. 4F and supplemental Fig. S2C) (62,63).
Comparison of toxin active sites in toxin alone, toxin-antitoxin, and toxin bound to the ribosome structures reveals there are only minor structural rearrangements of the toxin (36,41,53). To further explore the inhibitory mechanism of HigB by HigA, we superimposed the HigB-(HigA) 2 -HigB complex on the structure of RelE bound to mRNA on 70 S and found that HigB-(HigA) 2 -HigB cannot be accommodated (Fig. 6) (36). A steric clash exists between HigB-(HigA) 2 -HigB and ribosomal protein S12 and 16 S rRNA helix 18 (h18). The position of a large portion of the N terminus of HigA (␣1-4) overlaps with the entire h18 (Fig. 6, clash 1). The second clash site is with the C terminus of HigAЈ that overlays with S12 residues 108 -113 and the tetraloop of h18 (Fig. 6, clash 2). Thus, we propose binding of HigA sterically inhibits HigB from interacting with mRNA in the A site of the ribosome.
Taken together, our results expand the molecular understanding of how diverse antitoxins counteract the activity of toxin proteins. As Blower et al. (70) described, the tertiary structures of toxins are a static scaffold that may contain myriad possible active site residues that dictate substrate specificity. This appears to be consistent with what is known about ribosome-dependent toxins. However, the antitoxin structure and interaction with its cognate toxin varies and can be structurally divergent depending upon its DNA binding domain and the structural features of the antitoxin and toxin interface. This antitoxin structural plasticity underscores the expansive nature of TA-mediated bacterial survival mechanisms.