Exposure of Neutralizing Epitopes in the Carboxyl-terminal Domain of TcdB Is Altered by a Proximal Hypervariable Region*

Background: TcdB from 027 strains is not as sensitive to antibody neutralization as TcdB from 012 strains, although neutralizing epitopes are present on both forms. Results: Sequence alterations in strain-specific TcdB influence intramolecular protein-protein interactions and exposure of neutralizing epitopes. Conclusion: Our work uncovers a mechanism used by TcdB to shield neutralizing epitopes. Significance: This study demonstrates how strain-specific TcdB can avoid antibody targeting. The sequence, activity, and antigenicity of TcdB varies between different strains of Clostridium difficile. As a result, ribotype-specific forms of TcdB exhibit different toxicities and are not strongly cross-neutralized. Using a combination of biochemical and immunological approaches, we compared two important variants of TcdB (TcdB012 and TcdB027) to identify the mechanisms through which sequence differences alter epitopes and activity of the toxin. These analyses led to the discovery of a critical variation in the 1753–1851 (B2′) region of TcdB, which affects the exposure of neutralizing epitopes in the toxin. Sequence comparisons found that the B2′ region exhibits only 77% identity and is the most variable sequence between the two forms of TcdB. A combination of biochemical, analytical, and mutagenesis experiments revealed that the B2′ region promotes protein-protein interactions. These interactions appear to shield neutralizing epitopes that would otherwise be exposed in the toxin, an event found to be less prominent in TcdB012 due to sequence differences in the 1773–1780 and 1791–1798 regions of the B2′ domain. When the carboxyl-terminal domains of TcdB012 and TcdB027 are swapped, neutralization experiments suggest that the amino terminus of TcdB interacts with the B2′ region and impacts the exposure of neutralizing epitopes in the carboxyl terminus. Collectively, these data suggest that variations in the B2′ region affect protein-protein interactions within TcdB and that these interactions influence the exposure of neutralizing epitopes.

entry into the cytosol of cells (1). TcdB is expressed by diseasecausing strains of C. difficile, is essential to C. difficile infection (CDI), and is capable of recapitulating many aspects of CDI (2)(3)(4)(5)(6). Studies focused on TcdB have provided insight into C. difficile pathogenesis, and findings suggest that blocking this toxin could ameliorate disease.
TcdB (2366 residues) enters cells by receptor-mediated endocytosis and translocates into the cytosol following acidification of the endosome (7)(8)(9)(10). TcdB undergoes acid pH-induced conformational changes within the endosome and, using a mechanism reminiscent of diphtheria toxin, inserts into the membrane with its pore-forming domain (residues 1035-1107) (11). The steps following insertion into the membrane and translocation into the cytosol have not been described, but the glucosyltransferase domain (GTD; residues 1-543) is known to be released into the cytosol following autoprocessing by the proximal cysteine protease domain (CPD; residues 554 -767) (12,13). The GTD and CPD appear to be reasonable drug targets, but recent reports indicate that TcdB mutants lacking these activities remain cytotoxic (14 -16). This raises the possibility that the toxin has other, yet undefined, intracellular activities, and two recent reports found that TcdB modulates the NADPH oxidase complex and causes pyknosis in the absence of glucosyltransferase activity (15,17). Thus, vaccination and other strategies to block earlier steps in cellular intoxication, such as cell binding, may be needed to prevent all of the known and unknown intracellular activities of TcdB.
Development of TcdB-targeted vaccines or therapeutic antibodies for treating most cases of CDI faces an important obstacle; strain type-specific forms of TcdB vary in sequence, toxicity, and antigenicity (18 -22). Earlier findings from our group suggested that conformational variations may influence the differences in toxicity between the two forms of TcdB. For example, experiments found that TcdB produced by the hypervirulent C. difficile 027 ribotype (TcdB 027 ) undergoes unfolding and exposure of hydrophobic domains at a higher acidic pH than TcdB produced by the less virulent C. difficile ribotype (termed TcdB 012 herein and representative of several histor-ical ribotypes, such as 001, 003, 012, and 087) (22). Studies have also shown TcdB 027 , but not TcdB 012 , adopts a structure where the CPD is occluded from external labeling by a substrate-based fluorescent probe (21). This conformational difference corresponds with more efficient autoprocessing by TcdB 027 . As a result, TcdB 027 appears to intoxicate cells more efficiently than TcdB 012 (21,22). Corresponding to the sequence differences and possible variations in structure, we also reported that rabbit polyclonal antisera recognizing the carboxyl-terminal domains (residues 1652-2366) of TcdB 027 and TcdB 012 do not strongly cross-neutralize (20). However, the extent to which substantial conformational differences, rather than more subtle sequence variation in specific epitopes, limits cross-neutralization is not known.
In the current study, we show that the carboxyl-terminal domain of TcdB 027 is prone to intermolecular interactions and forms a higher order complex, which precludes the exposure of neutralizing epitopes through a process we termed "epitope cloaking." Cloaking appears to require a distinct 98-residue domain, which exhibits only 77% identity between the two forms of the toxin. These findings support a model in which TcdB 027 exists in a conformation that can both impact toxicity and the exposure of neutralizing epitopes.

EXPERIMENTAL PROCEDURES
Production of Native and Recombinant TcdB-Native TcdB was produced by culturing C. difficile (VPI 10463 or NAP1/BI/ 027) with the dialysis method as described previously (22). From these cultures, supernatants were isolated, and TcdA was removed by a thyroglobulin affinity chromatography protocol. After removing TcdA, TcdB was purified using anion exchange (Q-Sepharose) chromatography in 20 mM Tris-HCl, pH 8.0, and 20 mM CaCl 2 . This method yields pure native TcdB, as demonstrated by a single 270 kDa band when analyzed by SDS-PAGE.
Recombinant TcdB was expressed and purified in a Bacillus megaterium system (MoBiTec, Göttingen, Germany) as described previously by others (23). The tcdB 027 gene was amplified from C. difficile genomic DNA and cloned into the B. megaterium expression plasmid (pC-His1622) between the BsrGI and NgoMIV restriction sites. The tcdB 012 gene in pC-His1622 was a gift from B. Lacy. Hybrid forms of TcdB were generated where the carboxyl terminus (amino acids 1668 -2366) of each toxin was swapped. This resulted in TcdB 012 that had the carboxyl terminus from TcdB 027 (TcdB 012 /B2ЈB3 027 ) and TcdB 027 that had the carboxyl terminus from TcdB 012 (TcdB 027 /B2ЈB3 012 ). To perform this switch, a BspEI site was engineered into both the tcdB 012 and tcdB 027 genes in the pC-His1622 plasmid. This BspEI site was created between nucleotides 4996 and 5001 by changing a single nucleotide with the QuikChange II XL site-directed mutagenesis kit (Agilent). This nucleotide substitution did not alter the amino acid sequence of either TcdB 012 or TcdB 027 . After creating this restriction site, both tcdB 012 and tcdB 027 were digested with BsrGI and BspEI, which created a fragment containing the amino terminus of TcdB (amino acids 1-1667). TcdB 027 /B2ЈB3 012 was generated when the segment of DNA containing the amino terminus of TcdB 027 was ligated into pC-His1622-TcdB 012 that had the amino terminus removed by digesting with BsrGI and BspEI. TcdB 012 /B2ЈB3 027 was produced using a similar method. DNA sequencing was used to confirm the construction of TcdB 027 / B2ЈB3 012 and TcdB 012 /B2ЈB3 027 .
To express recombinant TcdB, pC-His1622-TcdB was transformed into B. megaterium following the manufacturer's instructions (MoBiTec, Göttingen, Germany). After selecting for TcdB-positive clones, B. megaterium containing pC-His1622-TcdB was grown at 37°C in LB medium supplemented with 10 g/ml tetracycline. The culture were grown to an A 600 of 0.3 and then induced for 5 h with 0.5% D-xylose. After induction, cells were isolated by centrifugation and resuspended in 20 mM HEPES (pH 8.0), 500 mM NaCl, and 20 mM imidazole. The cells were then lysed, and TcdB was purified by Ni 2ϩ affinity chromatography. After elution, TcdB was bufferexchanged into a storage buffer consisting of 20 mM HEPES (pH 8.0) and 500 mM NaCl. SDS-PAGE analysis of purity demonstrated that recombinant TcdB was ϳ90% pure.
Production of Recombinant GTD, B2ЈB3, and Mutant B2ЈB3-The B2ЈB3-encoding region of the tcdB gene (nucleotides 4961-7111; amino acids 1651-2366) from ribotype 012 or 027 was codon-optimized for Escherichia coli and cloned into pET15b expression plasmid (20). From these clones, PCR was used to amplify the regions encoding the B3 and the 1751-1852 fragment. These PCR products were cloned into the pENTR vector (Invitrogen) and transferred into expression vectors by a recombination reaction using LR Clonase (Invitrogen). The expression vectors used to express these carboxyl-terminal fragments were pET61Dest (EMD Millipore) and pDest17 (Invitrogen). The B2ЈB3 mutants described in the legend to Fig.  1 were generated in pET15b-B2ЈB3 using the QuikChange II XL site-directed mutagenesis kit (Agilent). To express and purify these proteins, E. coli (BL21 star DE3; Invitrogen) were transformed with pET15b-B2ЈB3, pET61Dest-B3, or pDest17-1752-1853. Then cultures of the transformed E. coli were grown to an A 600 of 0.8 and induced with 300 M isopropyl 1-thio-␤-D-galactopyranoside for 16 h at 16°C. These proteins were then purified by Ni 2ϩ affinity chromatography and were ϳ90% pure as assessed by SDS-PAGE.
Recombinant GTD was produced by using PCR to amplify the regions encoding GTD from the tcdB 012 or tcdB 027 gene codon optimized for E. coli. These PCR products were cloned into the pENTR vector (Invitrogen) and transferred into the pDest15 (Invitrogen) expression vectors using LR Clonase (Invitrogen). The pDest15 vector allowed for the expression of the GTD with an amino-terminal GST tag in E. coli. After transforming the pDest15-GTD into E. coli (BL21 star DE3; Invitrogen), cultures were grown to an A 600 of 0.8 and induced with 300 M isopropyl 1-thio-␤-D-galactopyranoside for 16 h at 16°C. From these cultures, the GST fusion protein was isolated via affinity chromatography using glutathione-Superflow agarose (Pierce). SDS-PAGE was used to demonstrate that ϳ90% purity was achieved.
TcdB Neutralization Assays-These assays use the ability of TcdB to reduce CHO cell viability as a method to quantify TcdB neutralization. The antiserum used in these assays was produced by immunizing rabbits with B2ЈB3 012. In these experiments, CHO cells (ATCC) were cultured at 37°C in the pres-ence of 6% CO 2 with F12-K containing 10% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin. Prior to these assays, CHO cells were seeded in 96-well plates at a density of 1 ϫ 10 4 cells/well and allowed to attach overnight. In these assays, TcdB was combined with a 1:100 dilution of serum in culture medium and incubated for 30 min before adding this mixture to cells. The toxin-exposed cells were incubated for 24 h, and cell viability was measured by the CCK-8 assay (Sigma), a WST-8 dyebased colorimetric reaction. In the blocking assays, the antiserum was incubated for 30 min with B2ЈB3, and then toxin was added. The incubation was continued for another 30 min, and then the incubation was added to CHO cells to assess neutralization. For examining the potential linear epitope (positions 1773-1780), this blocking assay was performed using a 19amino acid peptide spanning residues 1769 -1787 of TcdB 012 . This peptide was synthesized and purified to 95% by GenScript (Piscataway, NJ).
Cross-linking-The B2ЈB3 or B3 fragment was cross-linked using bis(sulfosuccinimidyl) suberate (BS 3 ), which is a watersoluble homobifunctional cross-linking agent that reacts with primary amine groups. The cross-linking reaction was performed with 35 M purified proteins and 700 M BS 3 in a buffer consisting of 20 mM sodium phosphate (pH 7.2) and 500 mM NaCl. The reaction proceeded for 30 min at 25°C and then was quenched with 40 mM Tris, pH 8.0.
Gel Filtration-To perform gel filtration, 300 g of purified protein was loaded onto a Superose 6 10/300 GL column (GE Healthcare) at a flow rate of 250 l/min. Chromatography was performed in a buffer consisting of 20 mM sodium phosphate (pH 7.2), 500 mM NaCl, and 2 mM DTT. Protein standards (Bio-Rad) were run under the same conditions in order to gauge the retention times of the proteins analyzed in these experiments.
Pull-down-In these experiments, 2 M His-tagged B2ЈB3 was combined with 450 nM native TcdB (no tag) in 50 l of a buffer consisting of 20 mM sodium phosphate (pH 7.2), 137 mM NaCl, and 0.01% Tween 20 (binding buffer). These proteins were incubated for 1 h at 37°C, and then cobalt-coated magnetic beads (Dynabeads; Invitrogen) were added in order to isolate His-tagged proteins. After incubating for 5 min, the magnetic beads were separated from the input and washed three times in binding buffer. The pulled down proteins were eluted in binding buffer with 300 mM imidazole and analyzed by SDS-PAGE. A similar method was utilized to examine binding between His-tagged B2ЈB3 and GST-tagged GTD. Densitometry analysis of Coomassie-stained bands was carried out on digitized images using ImageJ version 1.37 software (Wayne Rasband, National Institutes of Health).
Differential Scanning Fluorimetry-The thermal stability (T m ) of purified B2ЈB3 012 and B2ЈB3 027 was determined by combining these proteins with SYPRO Orange and performing protein melt reactions. SYPRO Orange increases fluorescence emission by binding to hydrophobic regions of proteins that become exposed during temperature-induced unfolding. In these experiments, an Applied Biosystems 7500 real-time PCR system was used to monitor fluorescence emission of the protein/dye mixture as temperature was stepped from 25 to 99°C. These reactions were performed in quadruplicate with 0.5 mg/ml protein in a buffer consisting of 20 mM sodium phos-phate (pH 7.2) and 137 mM NaCl. The T m was determined by plotting the first derivative of fluorescence versus temperature and then locating the temperature at the midpoint of the transition from folded to unfolded.

The Carboxyl-terminal (B2ЈB3) Region of TcdB 027 Forms
Higher Order Complexes- Fig. 1A shows an overview of the primary structure and domain organization of TcdB. A nomenclature consistent with its characteristics as a single polypeptide A-B toxin was adopted in this study. The GTD is labeled as the A domain, and the remainder of the protein is labeled as subdomains of the B fragment, which constitute the putative cellbinding, autoprocessing, and pore-forming domains of the toxin. The overall sequence identity between TcdB 027 and TcdB 012 is 92% (22); however, the level of identity among the different domains of the toxin ranges from 88 to 96%, with the least amount of consensus sequence found in the carboxyl-terminal portion (B2ЈB3) (Fig. 1A). The B2Ј region shows 77% identity, and shorter regions reveal patches of sequence in the B2Ј region with little or no identity. For example, the 1773-1810 region exhibits 17 residue differences (55% identity), including a smaller (residues 1791-1798) region with 6 of 7 residues that differ between the two forms of the toxin (Fig. 1C). Our previous work found that the B2ЈB3 region of TcdB 027 does not generate neutralizing antibodies against either form of the toxin, although the same region of TcdB 012 does induce protection against TcdB 012 but not against TcdB 027 (20). In the same study, solid phase ELISA analysis using synthetic peptides found distinct differences in reactive epitopes between the two forms of TcdB. Interestingly, the sequences of many of these epitopes were identical or nearly identical in the two forms of TcdB yet were only reactive in one form of the toxin and not the other. Because of this, we hypothesized that the B2ЈB3 region of TcdB existed in distinct conformational states that led to significant differences in the exposed epitopes.
In the initial set of experiments, biochemical differences between B2ЈB3 012 and B2ЈB3 027 were examined. First, the thermal stability of the B2ЈB3 012 and B2ЈB3 027 was measured using differential scanning fluorimetry. These experiments revealed that B2ЈB3 012 had a T m value of 49.9°C, whereas B2ЈB3 027 had a T m value of 44.3°C (Fig. 1B). These differences in the T m values suggested that the two forms of B2ЈB3 have distinct biochemical properties. These biochemical properties were further explored using both cross-linking and gel filtration approaches. As shown in Fig. 2, the B2ЈB3 domains of TcdB 027 and TcdB 012 were treated with the cross-linker BS 3 , and the reaction was examined by SDS-PAGE ( Fig. 2A) or SDS-agarose gel electrophoresis (AGE) (Fig. 2B). As shown by SDS-PAGE analysis, the B2ЈB3 fragments formed higher order complexes in the presence, but not in the absence, of the cross-linker ( Fig. 2A). A comparison of B2ЈB3 027 and B2ЈB3 012 revealed that the multimers formed by B2ЈB3 027 were more extensive than those formed by B2ЈB3 012 (Fig. 2A). Cross-linked B2ЈB3 012 formed bands that migrated as dimers, tetramers, and other multimers, whereas the majority of cross-linked B2ЈB3 027 was too large to enter the gel ( Fig. 2A). Because cross-linked B2ЈB3 027 formed a large complex that could not be resolved by SDS-PAGE, SDS-AGE was used to evaluate the cross-linked form of B2ЈB3 027 . As shown by SDS-AGE, cross-linked B2ЈB3 027 migrated more slowly through this gel than cross-linked B2ЈB3 012 (Fig. 2B). The same cross-linking procedure was applied to the B3 domain to determine whether the B2Ј region was necessary for multimer formation. Results from this experiment revealed that the isolated B3 regions did not form substantial higher order complexes in the presence of cross-linker (Fig. 2). These data suggest that B2ЈB3 027 could exist in a substantially higher order complex than B2ЈB3 012 because of sequence differences in the B2Ј region.
To examine multimer formation more closely, the next set of experiments used gel filtration (GF) to follow multimer formation in solution and examine the effects of mutating candidate residues in the B2Ј region of B2ЈB3 012 . In line with the crosslinking data, neither B3 domain formed a higher order complex (Fig. 3, A and B), and only a small amount of B2ЈB3 012 was able to multimerize (Fig. 3C). However, GF analysis revealed that all of the B2ЈB3 027 eluted in the void volume (Fig. 3D), which is indicative of a high order complex larger than 600 kDa. Experiments were next performed to determine whether specific B2Ј 027 residues involved in multimerization could be identified. To accomplish this, select mutants (summarized in Fig. 1C) were generated wherein residues in B2ЈB3 012 were replaced with the corresponding sequence from B2Ј 027 (Fig. 3, E-H). Although none of the mutants exhibited a GF profile identical to that of B2ЈB3 027 , GF analysis revealed that three of the four mutants (B2Јmut1, B2Јmut2, and B2Јmut5) had either minor or major peaks indicative of complexes larger than observed with B2ЈB3 012 . Additionally, these mutants were analyzed by BS 3 cross-linking, and increased multimerization was also observed in B2Јmut1, B2Јmut2, and B2Јmut5 (Fig. 3J). Finally, candidate residues in B2ЈB3 027 were mutated to alanines (Fig. 1C) and examined to determine whether this decreased multimer formation. As shown by GF, alanine substitutions at the target residues in B2ЈB3 027 reduced the ability of B2ЈB3 027 to form multimers (Fig. 3I). Collectively, these data suggest that amino acids 1773-1780 (B2Јmut1 and B2Јmut6) and 1791-1798 (B2Јmut2) help support multimerization in B2ЈB3 027 .
Identification of Regions That Influence Exposure of Neutralizing Epitopes-The data shown in Figs. 2 and 3 indicate that purified B2ЈB3 027 exists in a higher order complex that is . This depiction demonstrates that the carboxyl terminus of TcdB is the location of much of the sequence variability between TcdB 012 and TcdB 027 . This figure also reveals a large amount of sequence variability within a 99-amino acid span of the B2Ј region, which is the focus on the current studies. B, differential scanning fluorimetry was used to determine the thermal stability (T m ) of B2ЈB3 012 and B2ЈB3 027. The graph depicts increases in relative fluorescence units (RFU) as SYPRO Orange binds to hydrophobic regions of proteins that undergo temperature-induced unfolding. These data from the graph were used to calculate the T m . In the inset, the T m is displayed as the mean and S.D. of four replicates. C, sequence alignment comparing a highly variable region of TcdB 012 and TcdB 027 . This portion of TcdB is the region where the following series of experiments focus and the region targeted for mutagenesis in these studies. The mutants used in these current studies are detailed. dependent on a set of residues in the B2Ј region. We next asked whether these differences in complex formation might impact the exposure of neutralizing epitopes. To identify regions targeted by neutralizing antibodies, we utilized an assay (Fig. 4) in which B2ЈB3 fragments were assessed for their ability to block neutralizing antiserum. Specifically, polyclonal rabbit serum raised against B2ЈB3 012 (␣B2ЈB3 012 ), which we had previously reported to neutralize TcdB 012 , was incubated with the entire B2ЈB3 012 region (residues 1652-2366) or the B3 region (residues 1853-2366) to adsorb antibodies. B2ЈB3 012 but not B3 012 blocked the neutralizing antiserum (Fig. 4A). The region of TcdB important for neutralization with this antiserum was further narrowed down when a purified segment of B2Ј (residues 1753-1852) was found to block toxin neutralization (Fig. 4A). Next, the B2ЈB3 region of TcdB 027 was examined and found to be less effective at blocking neutralizing antibodies than B2ЈB3 012 (Fig. 4A), which corresponds to our previous findings that indicate that B2ЈB3 027 does not generate a strong antibodymediated neutralizing response like B2ЈB3 012 (20).
Sequence Variation within B2Ј Alters Exposure of Neutralizing Epitopes-Experiments next examined various B2ЈB3 mutants to determine whether sequences in the B2Ј region influence neutralizing epitopes. As shown in Fig. 4B, when residues from TcdB 027 were substituted in the 1791-1798 (B2Јmut2) and 1805-1810 (B2Јmut3) regions of B2ЈB3 012 , there was no detectable impact on the ability of this fragment to block neutralizing antibodies. In contrast, B2ЈB3 012 lost the ability to block neutralizing antibodies when TcdB 027 residues were substituted into the 1773-1780 (B2Јmut1) region. In line with this, a double mutant (B2Јmut4) possessing mutations at 1791-1798 and 1805-1810 was able to block neutralizing antibodies, whereas a triple mutant (B2Јmut5) containing mutations in all three regions was unable to block neutralizing antibodies. These data suggest that the 1773-1780 region encodes and/or influences the exposure of neutralizing epitopes in the B2ЈB3 region of TcdB 012 .
In our next experiment, we determined whether the linear epitope corresponding to residues 1773-1780 of TcdB 012 could bind neutralizing antibodies. Therefore, we designed a small peptide that contained residues 1773-1780 of TcdB 012 and determined whether this peptide could absorb neutralizing antibodies. As shown in Fig. 4C, we could not detect any loss of activity of the antibody. This result suggested that the 1773-1880 region may be influencing the exposure of neutralizing epitopes in B2ЈB3 027 , perhaps by increasing multimerization, as observed in Fig. 3. To test this possibility, we examined the B2Јmut6 and B2Јmut7 mutants that contained alanine substitutions at the putative critical residues in B2ЈB3 027 . Strikingly, both mutants, containing a string of alanine substitutions, gained the capacity to block neutralizing antiserum (Fig. 4B). These results suggested that the B2ЈB3 027 region encodes neutralizing epitopes, but these are structurally occluded by sequences in the B2Ј subdomain.
Antibody-mediated Neutralization of TcdB Hybrids-The data described above indicated that sequence differences in the B2Ј region of B2ЈB3 012 and B2ЈB3 027 impacted the exposure of neutralizing epitopes. In the next experiment, we confirmed that observation in the context of the holotoxin and determined whether other regions of TcdB outside of B2ЈB3 might also influence the exposure of neutralizing epitopes. For these experiments, DNA segments encoding B2ЈB3 in the two forms of TcdB were swapped to generate hybrids of TcdB 027 and TcdB 012 with the reciprocal B2ЈB3 domains (TcdB 027 /B2ЈB3 012 and TcdB 012 /B2ЈB3 027 ). In initial experiments, all four forms of the toxins had statistically identical viability curves when toxin activity was examined in the presence of serum from unvaccinated rabbits (Fig. 5A). These toxins were then tested for their sensitivity to antibodies raised against the B2ЈB3 012 (␣B2ЈB3 012 ). As shown in Fig. 5B, ␣B2ЈB3 012 was able to neutralize TcdB 012 , whereas TcdB 027 was less susceptible to neutralization by this antiserum, which was observed previously (20). Examination of the two hybrid toxins revealed that TcdB 027 /B2ЈB3 012 was more sensitive to neutralization by ␣B2ЈB3 012 than TcdB 012 /B2ЈB3 027 (Fig. 5B). Unexpectedly, TcdB 027 /B2ЈB3 012 was found to be more sensitive to neutralization by ␣B2ЈB3 012 than TcdB 012 (Fig. 5B). This unexpected observation was confirmed when TcdB 027 /B2ЈB3 012 was also found to be more sensitive than TcdB 012 to neutralization with antibodies against B2ЈB3 027 (data not shown). These data suggest the amino-terminal region of the toxin could be influencing the exposure of epitopes in the B2ЈB3 domain.
The B2Ј Region Facilitates Binding between B2ЈB3 027 and Full-length TcdB-The results above suggest that protein-protein interactions between domains of TcdB may be influencing the overall folding and exposure of neutralizing epitopes. Based on the multimerization data, we hypothesized that the B2Ј region of TcdB may serve as an important site for mediating protein-protein contacts between domains of TcdB. To test this possibility, we first determined whether we could detect binding between B2ЈB3 and the holotoxin. Therefore, we performed pull-down experiments using B2ЈB3 027 or B2ЈB3 012 immobilized to cobalt-coated magnetic beads via their His tags. In these experiments, the B2ЈB3 fragment was mixed with TcdB 027 or TcdB 012 , and the pellets from the pull-down were then examined by SDS-PAGE. As shown in Fig. 6A, when the pull-down was performed using B2ЈB3 027 as the bait, both TcdB 027 and TcdB 012 were detected on the gel. In contrast, using B2ЈB3 012 as the bait did not result in the pull-down of either form of the holotoxin (Fig. 6A). When the B3 027 domain alone was used in the pull-down, full-length TcdB could not be detected on the gel (Fig. 6B), which indicates that the B2Ј region is necessary for protein-protein interactions between B2ЈB3 027 and full-length toxin. The residues in B2Ј capable of interacting with fulllength TcdB were narrowed down using the mutants described in Fig. 1C. As shown in Fig. 6, C and D, B2ЈB3 012 gained the ability to bind TcdB when amino acids 1773-1780 (B2Јmut1) or 1791-1798 (B2Јmut2) were replaced with the corresponding sequence from B2ЈB3 027. In contrast, B2ЈB3 012 did not gain the ability to bind TcdB when amino acids 1805-1810 were changed to the B2ЈB3 027 sequence. These data demonstrate that the B2Ј region of the B2ЈB3 027 is important for binding to full-length TcdB and suggest that B2Ј might be involved in mediating interdomain contacts.  Because the amino-terminal region of TcdB appears to influence the exposure of epitopes in the B2ЈB3 region (Fig. 5), we hypothesized that protein-protein interactions between the B2ЈB3 region and the holotoxin could be impacted by alterations to the amino-terminal region. To begin to explore the influence of the amino terminus on binding between B2ЈB3 and the toxin, we first determined whether B2ЈB3 027 could bind the GTD, which corresponds to amino acids 1-543 of TcdB (Fig.  1A). As shown by the pull-down experiments in Fig. 7, A-C, B2ЈB3 027 was not able to bind either GTD 012 or GTD 027 . As another method for investigating the influence of the amino terminus on binding between the B2ЈB3 region and the toxin, we took advantage of the autoprocessing characteristic of TcdB. Upon induction of autoprocessing, the CPD cleaves TcdB between residues 543 and 544, thus separating the GTD (residues 1-543) from the remainder of the toxin (residues 544 -2366). Therefore, if the amino-terminal region modulates interactions between B2ЈB3 and the holotoxin, differences should be detected in the level of B2ЈB3 027 binding to the processed and unprocessed forms of the toxin. Hence, we carried out in vitro autoprocessing under conditions that yielded a mixture of uncleaved and cleaved toxin, with a majority of the toxin remaining in the uncleaved form (Fig. 7D). A pull-down was then performed between B2ЈB3 027 and partially processed TcdB. As shown in Fig. 7D, B2ЈB3 027 preferentially pulled down processed TcdB (residues 544 -2366) rather than unprocessed TcdB (residues 1-2366) (Fig. 7D). A similar result was observed for both TcdB 012 and TcdB 027 . These results indicate that B2ЈB3 027 binds a region located between 544 and 2366 and suggest FIGURE 6. Analysis of interactions between full-length TcdB and the B2B3 domain. Pull-down experiments were utilized to examine binding between full-length TcdB and B2ЈB3. These experiments were carried out with TcdB purified from C. difficile and recombinant His-tagged B2ЈB3 purified from E. coli. Incubations were done for 1 h at 37°C with 450 nM TcdB and 2 M B2ЈB3. His-tagged B2ЈB3 was pulled down with cobalt-coated magnetic beads, and binding between B2ЈB3 and TcdB was visualized by SDS-PAGE and Coomassie staining. A, SDS-PAGE demonstrating binding between B2ЈB3 and TcdB from different ribotypes. B, SDS-PAGE depicting lack of binding between B3 027 and TcdB 027 . C, SDS-PAGE examining binding between TcdB 027 and mutant forms of B2ЈB3 described in the legend to Fig. 1C. D, the bar graph represents the densitometry analysis of three Coomassie-stained gels from the data in A-C. Results are presented as mean (n ϭ 3) Ϯ S.D. (error bars). Asterisks indicate significance when compared with B2ЈB3 012 . *, p Ͻ 0.01. that the GTD (residues 1-543) is obstructing the site on TcdB that B2ЈB3 027 binds.

DISCUSSION
The emergence of new strains of C. difficile presents an ongoing challenge to clinicians. Strains such as the 027 ribotype have predominated in several geographical regions during the past decade and correlate with increases in the mortality rate of CDI (24 -26). Not only does this change in C. difficile epidemiology complicate treatment, but patients are more likely to relapse if infected by the 027 ribotype (27). Furthermore, our previous work has shown that antibodies capable of neutralizing TcdB from historical strains do not equally neutralize TcdB from 027 strains (20). As part of the previous study, we also mapped reactive epitopes in the B2ЈB3 regions of the two forms of TcdB. Curiously, despite sharing common sequences, antibodies recognizing epitopes in B2ЈB3 012 did not react with the identical sequences in B2ЈB3 027 . This led us to speculate that both sequence variations and conformational differences influenced antibody reactivity and altered sensitivity to neutralization. In the current study, we sought to identify key sequence differences in TcdB 027 and TcdB 012 that account for variations in neutralization by antibodies against TcdB. The data indicate that sequence differences may not only alter specific neutralizing epitopes but also confer structural variations that cloak epitopes that would otherwise be subject to antibody targeting.
The experiments in this study focused on the carboxyl-terminal region of TcdB with a particular emphasis on the B2Ј sequence. This region is of particular interest because recent work by Zhang et al. (28) showed that the 1756 -1852 region was required for TcdB toxicity. Moreover, a comparison of the B2Ј regions showed only 77% identity between the two forms of TcdB, which is notably less than the 92% identity found across the full length of the two toxins. Thus, given the sequence differences between the two forms of the toxin and the demonstrated importance of this region in intoxication, we reasoned that this portion of the toxin could encode critical neutralizing epitopes that differed between TcdB 027 and TcdB 012 . In Fig. 4, an antibody-blocking assay was used to map critical neutralization regions to the B2Ј and pinpoint sequence variations in the B2Ј that alter neutralizing epitopes. Interestingly, results from the antibody blocking assay indicated that both B2ЈB3 012 and B2ЈB3 027 have neutralizing epitopes, but B2ЈB3 027 exists in a conformation that cloaks these epitopes. To explain the mechanism of epitope cloaking, biochemical analyses of the B2ЈB3 regions were conducted that revealed that these regions exist in a multimeric state. Using a chemical cross-linking approach as well as GF analysis, we found that the B2Ј region was necessary for the multimerization of B2ЈB3 because this activity was not observed when only the B3 fragment was examined (Figs. 2 and  3). These data also indicated that B2ЈB3 027 multimerized more readily than B2ЈB3 012 , and increased multimerization could be conferred to B2ЈB3 012 by substituting amino acids encoded by the 027 form of the toxin, which correlates with the antibodyblocking assay. FIGURE 7. Binding between full-length TcdB and B2B3 027 is influenced by the amino terminus of TcdB. A-C, pull-down experiments were used to examine binding between GTD and B2ЈB3. These experiments were carried out with purified His-tagged B2ЈB3 and purified GST-tagged GTD. Incubations of 450 nM GTD with 2 M B2ЈB3 were performed for 1 h at 37°C. His-tagged B2ЈB3 was pulled down with cobalt-coated magnetic beads, and interactions between B2ЈB3 and GTD were assessed using SDS-PAGE and Coomassie staining. D, IP6 processing was used to separate the amino terminus from TcdB and determine whether removal of the amino terminus influences binding between TcdB and B2ЈB3 027 . To activate TcdB processing, 450 nM TcdB was incubated with 100 M IP6 for 1 h at 37°C. This was followed by the addition of 2 M B2ЈB3 027 and another incubation for 1 h at 37°C. His-tagged B2ЈB3 027 was pulled down with cobalt-coated magnetic beads, and binding between B2ЈB3 027 and TcdB was visualized by SDS-PAGE and Coomassie staining. The gel labeled Input illustrates the ratio of processed to unprocessed TcdB that does not pull down with B2ЈB3 027 . The gel labeled Pull down shows the ratio of processed to unprocessed TcdB that binds B2ЈB3 027 .
The ability of the B2ЈB3 region to multimerize provides important clues about the function of this toxin and the mechanism of epitope cloaking. The multimerization of the B2ЈB3 suggests that full-length toxin may form multimers, but data from our group suggest that the full-length toxin exists as a monomer at neutral pH (data not shown). However, this does not exclude the possibility that TcdB may form multimers under certain buffer conditions that the toxin encounters during the cellular entry process. Another interpretation of the multimerization data suggests that B2ЈB3 multimerization may be related to intramolecular interactions that exist within monomeric TcdB. Thus, the multimerization assay may be detecting an intramolecular interaction that appears as multimerization when the analysis is performed on the B2ЈB3 fragment alone. This latter possibility is supported by experimental data in Figs. 5 and 6.
In Fig. 6, we determined whether the molecular contacts that confer multimerization had functionality in the full-length toxin. In these pull-down experiments, we demonstrated that the B2ЈB3 027 fragment could bind full-length toxin, whereas B2ЈB3 012 could not. As with the multimerization data, the B2Ј region was essential for this activity, and binding between B2ЈB3 012 and TcdB could be conferred to B2ЈB3 012 by substituting amino acids encoded by the 027 form of the toxin. Next, we further explored the idea of intramolecular interactions cloaking neutralizing epitopes within the full-length toxin. In Fig. 5, the B2ЈB3 regions of TcdB 012 and TcdB 027 were swapped, yielding TcdB hybrids (TcdB 012 /B2ЈB3 027 and TcdB 027 /B2ЈB3 012 ). These experiments suggested that the amino-terminal portion of the toxin influences the exposure of neutralizing epitopes. For instance, when B2ЈB3 012 is combined with the amino terminus of TcdB 027 , the B2ЈB3 domain appears to become more accessible to the neutralizing antibody (Fig.  5B). This suggests that the amino-terminal region interacts with the carboxyl-terminal portion of the toxin, thereby changing the exposure of neutralizing epitopes. This possibility was further examined by taking advantage of the autoprocessing activity of TcdB. Following autoprocessing, which dissociates residues 1-543 from the 544 -2366 region, B2ЈB3 027 was found to bind the larger processed fragment (residues 544 -2366) better than the intact holotoxin (residues 1-2366) (Fig. 7D). This observation suggests that the amino terminus of TcdB is obscuring regions that can bind the B2ЈB3 region, thus supporting the idea that the amino-terminal region of TcdB is influencing the accessibility of neutralizing epitopes in the B2ЈB3 region of TcdB. Interestingly, the cryo-EM structure of TcdB suggests contacts between the GTD and a region proximal to the B3 domain of TcdB, and this contact is lost following autoprocessing (29). Further studies that map these contact regions may provide important information on how these two distal portions of the toxin interact.
The extent to which the variations in TcdB will influence interactions with antibodies has not been explored in detail and could have important consequences as disease-causing strains of C. difficile continue to emerge. In addition to ribotype 027, ribotype 078 has recently surfaced as a highly virulent strain in humans. Examining the sequence of TcdB from ribotype 078, we discovered that the B2Ј region more closely resembles the 012 ribotype (98% identity) than the 027 ribotype (78% identity). When we inspect residues that were identified in this study as critical, we find that residues 1773-1780 from TcdB 078 exactly match TcdB 012 . However, examination of residues 1791-1798 shows that only 6 of the 8 amino acids agree with TcdB 012 . Interestingly, one of the residues in this region of TcdB 078 matches TcdB 027 (Ser-1797), and one of the residues is unique to TcdB 078 (Ala-1798). This unique residue is a serine in TcdB 012 and a threonine in TcdB 027 . These subtle sequence changes in TcdB 078 could impact the structure of TcdB as well as interactions with antibodies. Additionally, these changes could provide clues to the future of emergent stains of C. difficile.
These variations in TcdB could also influence vaccination and therapeutic antibody efficacy. A study by Marozsan et al. (30) found that the humanized monoclonal antibodies CDA1 and CDB1, targeting TcdA and TcdB, were substantially less effective against toxin from the 027 ribotype compared with other ribotypes, such as the 012 ribotype. CDB1 is also referred to as bezlotoxumab, which is a TcdB-specific humanized monoclonal antibody currently in phase 3 clinical trials (31). Recently, Orth et al. (32) determined the crystal structure of bezlotoxumab bound to epitopes in the carboxyl-terminal region of TcdB. The sequence of the target antigen for the crystallization studies was derived from TcdB 012 , and using a combination of approaches, the authors identified regions of the toxin that were blocked by bezlotoxumab. Within the regions found to be blocked by bezlotoxumab, we noted that only two residues (Asn-1912 and Glu-2033) differed between TcdB 027 and TcdB 012 . In TcdB 027 , Asn-1912 is replaced with aspartic acid, and Glu-2033 is replaced with an alanine. Both of these changes could impact charge interactions, but they do not seem to be dramatic in the context of the entire sequence. Moreover, of the 22 TcdB residues situated within 4.0 Å of contact with the antibody complementarity-determining region, only four differed between the two forms of the toxin. Of those residues, none appear to show intimate interactions with residues in the complementarity-determining region. Based on this assessment, we can reasonably conclude that bezlotoxumab should interact with target epitopes both in TcdB 027 and TcdB 012 , yet the neutralization experiments suggest that TcdB 027 is less sensitive to the antibody (30). More recently, Hernandez et al. (33) performed a comprehensive analysis of bezlotoxumab reactivity with TcdB from several ribotypes and found that subtle sequence variations do indeed affect the EC 50 , but this can be overcome by reasonable increases in the amounts of antibody applied to the toxin. The extent to which intramolecular or intermolecular interactions shield these and other specific neutralizing epitopes has yet to be determined.
Identifying the sequence of epitopes cloaked by the 1752-1851 region is the subject of ongoing studies in our laboratory; however, evidence in the literature supports the idea that candidate epitopes reside between residues 1853 and 2066. Leuzzi et al. (34) found that a TcdB fragment covering the 1853-2366 region of the toxin generated a protective antibody response, whereas smaller fragments (2056 -2366 and 2157-2366) did not induce a protective response, despite stimulating high antibody titers. This suggests that the 1853-2056 region could con-tain neutralizing epitopes, which correlates with the effectiveness of bezlotoxumab mentioned above. In a previous study, we mapped reactive epitopes in B2ЈB3 for both forms of the toxin (20). TcdB 012 has four unique epitopes, and TcdB 027 has only one unique epitope in this region. Moreover, the toxins appear to share at least five common epitopes in this region of the protein. It will be of great interest to determine whether the two forms of TcdB share critical neutralizing epitopes, which are more effectively blocked in TcdB 027 .
A high resolution three-dimensional structure has not been reported for either TcdB 012 or TcdB 027 , making it difficult to know the precise structural differences between the two forms of the toxin. Moreover, the crystal structure itself might not reveal intermolecular interactions that influence exposure of epitopes in the B2ЈB3 region of the toxin. However, the findings that structural differences influence exposure of epitopes with shared sequences are bolstered by two earlier studies from our group. Previously, we found that TcdB 027 underwent acid-induced conformational changes at a higher pH than TcdB 012 (22). In a second study analyzing the autoproteolytic activity of TcdB, we found that the CPD of TcdB 012 was more sensitive to labeling with an exogenous probe than the CPD from TcdB 027 (21). Taken together, these data consistently support the idea that TcdB 027 and TcdB 012 exist in distinct conformations that affect their biochemical activities and antigenicity.