Designed Ankyrin Repeat Protein (DARPin) Neutralizers of TcdB from Clostridium difficile Ribotype 027

We report the engineering and characterization of designed ankyrin proteins as potent neutralizers of TcdB toxin secreted by a hypervirulent ribotype 027 strain of Clostridium difficile. We further show that although TcdB toxins from both ribotype 027 and VPI 10461 interact efficiently with TcdB receptors CSPG4 and Pvrl3, TcdB027 lacks significant ability to bind the only known physiologically relevant TcdB receptor, Frizzled 1/2/7.

antibiotics such as metronidazole and vancomycin (6,7). However, disturbing trends of increased morbidity and mortality due to relapse of C. difficile-infected patients after antibiotic treatment have since emerged (8)(9)(10)(11)(12). These trends correlated with the emergence of the "hypervirulent" BI/NAP1/027 strains of C. difficile (11,13,14) (C. difficile 027), which at one point were responsible for ϳ1/3 of the CDI in the United States (15). Infection with C. difficile 027 is associated with more-severe disease and a higher death rate (15). The exact reason for the increased virulence of C. difficile 027 remains enigmatic, although many factors such as antibiotic resistance, sporulation ability, and toxin production have been proposed to contribute to its virulence (12,(16)(17)(18)(19).
The pathology of CDI is primarily due to the action of two bacterial secreted exotoxins, toxin A (TcdA) and toxin B (TcdB) (20), that target small GTPases within the host cells, leading to disruption of tight junctions, loss of colonic epithelial barrier function, and bloody diarrhea (21). Administration of spores from nontoxigenic C. difficile strain M3 was found to significantly reduce CDI recurrence (22), highlighting a pivotal role of the toxins in CDI pathology. Vaccines against C. difficile toxins are currently being actively pursued and have enjoyed some preliminary successes (23)(24)(25)(26)(27). However, since CDI most often afflicts elderly hospitalized patients, the efficacy of vaccine in this unique population may be less than ideal (28,29). The TcdB-neutralizing monoclonal antibody (MAb) bezlotoxumab was found to reduce the CDI recurrence rate from 28% to 16% in a phase III clinical trial (30) and was approved by the FDA in 2016 (31). Curiously, despite its toxin neutralization ability, bezlotoxumab did not improve the initial cure rate of CDI in patients (31) and is not approved by the FDA as a treatment for CDI. Bezlotoxumab neutralizes TcdB by directly blocking its carbohydrate binding pocket and thus preventing its attachment to the colonic mucosal cells (32). Although bezlotoxumab exhibits potent neutralization activity against TcdB from a broad range of C. difficile strains, its potency is significantly (ϳ185-fold) weaker against toxin from C. difficile 027 than against that from laboratory strain VPI 10463 (33).
In this study, we performed phage panning and functional screening to identify a panel of new DARPins with significantly improved TcdB UK1 neutralization activity. The best DARPin, D16, neutralized TcdB UK1 with an EC 50 of 0.5 nM, making it Ͼ66-fold more potent than bezlotoxumab (EC 50 of ϳ33 nM) in vitro. Importantly, D16 also potently neutralizes TcdB VPI (EC 50 of 5 nM) and TcdB M68 (EC 50 of 1.6 nM). Competitive ELISAs showed that all our anti-TcdB UK1 DARPins block the toxin interaction with CSPG4. Since DARPin U3 from our previous study impedes TcdB interaction with FZD1/2/7 receptor, which binds an epitope distinct from that of CSPG4, we constructed multiple-dimer DARPins composed of D16 and U3 with different linker sizes and topologies with a view to creating synergistic blocking of toxin-receptor interactions. These dimers were expected to exhibit enhanced neutralization activity through the avidity effect. All the dimer DARPins exhibited (10-fold-to-20-fold) enhanced neutralization potency against TcdB VPI and TcdB M68 , pointing to their potential as new antitoxin biologics for treating CDI and/or preventing its recurrence.
Intriguingly, none of the constructed dimer DARPins showed enhanced neutralization activity against TcdB UK1 . Subsequent ELISAs revealed that, unlike TcdB VPI , which binds strongly to both purified ectodomains of CSPG4 and FZD2, TcdB UK1 lacks significant ability to interact with FZD2 despite strong ability to associate with CSPG4. Consistent with this result, TcdB UK1 was found to be minimally toxic to Caco-2 colon epithelium cells, which express multiple Frizzled proteins, including FZD2 and FZD7 (38,39), but lack CSPG4 (38,40). A closer analysis of the crystal structure of TcdB Frizzled binding domain (FBD) (TcdB-FBD; PDB code: 6C0B) revealed multiple differences between TcdB UK1 and TcdB VPI at the FBD binding interface which likely abolish or significantly weaken the TcdB UK1 -FZD1/2/7 interaction.

RESULTS
Selection of TcdB-neutralizing monomer DARPins. DARPins were designed based on repeat modules of natural ankyrin proteins and consist of an N-terminal capping repeat (N-cap), three (N3C) internal ankyrin repeats (ARs), and a C-terminal capping repeat (C-cap) (41). In a DARPin library, each internal repeat contains six randomized positions on the flexible surface-exposed loop and one partially randomized position on the hinge region, yielding a total of 18 plus 3 randomized positions in each N3C DARPin. Additional mutations at the framework positions can emerge during repeated PCR amplification. A previously constructed DARPin library comprising a complex of ϳ2 ϫ 10 9 variants was used in phage panning against biotinylated TcdB UK1 . DARPin variants from the 4th round of panning, which showed significantly elevated levels of TcdB UK1 binding ability (see Fig. S1 in the supplemental material), were cloned into the pET28a vector and expressed in BL21(DE3) Escherichia coli cells and underwent functional screening. Among the 760 individual clones screened, 57 clones rescued Vero cell viability from TcdB UK1 toxicity by Ͼ70%. All these clones were sequenced, and 16 unique clones were identified (see Table S1 in the supplemental material). Interestingly, these 16 unique DARPins were found to be composed of only 6 distinct ARs in 4 different configurations (Fig. 1), suggesting that these DARPins likely bind to overlapping epitopes on TcdB UK1 . Four DARPins (one from each unique configuration [D2, D3, D8, and D16]) that lacked any additional framework mutations were selected for further characterization.
In addition to neutralizing TcdB UK1 , all four DARPins also inhibited TcdB VPI and TcdB M68 (Fig. 2A). The best DARPin, D16, exhibited an EC 50 of 0.5 nM against TcdB UK1 , representing a Ͼ66-fold-higher in vitro potency than bezlotoxumab against the same toxin. D16 also exhibited low nanomolar neutralization potency against TcdB VPI (ribotype 087, EC 50 of 5.2 nM) and TcdB M68 (ribotype 017, EC 50 of 1.6 nM), representing lower potencies for these toxins relative to bezlotoxumab. The ability of the DARPins to bind the different toxins generally matches their toxin neutralization potency (Fig. 2B), with D16 being both the strongest binder and the most potent neutralizer of all the toxins.
Mechanism of TcdB UK1 neutralization by monomeric DARPins. There are three known TcdB receptors: chondroitin sulfate proteoglycan 4 (CSPG4), poliovirus receptorlike 3 (PVRL3 or NECTIN3), and members of the Frizzled protein family FZD1/2/7, which share identical sequences in the TcdB-binding region (37,38,42). In our previous study performed with TcdB VPI , most of the identified antitoxin DARPins interfered with the TcdB-CSPG4 interaction based on Cryo-EM structural studies and competitive ELISA results (34). The high frequency of antitoxin DARPin hits that block CSPG4 binding seen in our previous study likely stemmed from the use of Vero cells, which express a high level of CSPG4 (40), in our functional screening. We repeated the competitive ELISA for the four unique DARPins (i.e., D2, D3, D8, and D16). The wells of the ELISA plate were first coated with TcdB UK1 (4 g/ml) overnight at 4°C. The next day, a green fluorescent protein (GFP)-tagged extracellular domain of CSPG4 (CSPG4-EC-GFP) was added in the presence or absence of the different DARPins. The plate was incubated at room temperature for 2 h, and the amounts of bound CSPG4-EC-GFP were detected using anti-GFP antibody. As shown in Fig. 3, all four DARPins significantly reduced the binding signal from CSPG4-EC-GFP, with D16 producing the most signal reduction, indicating that all these monomer DARPins interfere with the TcdB UK1 -CSPG4 interaction. Thus, consistent with our previous finding, inhibition of CSPG4 interaction emerged as a dominant mechanism used by antitoxin DARPins in Vero cell-based functional screening.
Dimeric DARPins with enhanced potency against TcdB VPI and TcdB M68 . Fusion of multiple binders to nonoverlapping epitopes has been reported to significantly enhance the overall target-binding affinity via the avidity effect (43,44). Previously, we identified DARPin U3, which interferes with the interaction between TcdB and its receptor FZD1/2/7 (34). Dimer DARPin-DLD4 -consisting of U3 and 1.4E joined by a 3ϫ GGGGS linker exhibited Ͼ100-fold-higher neutralization potency against TcdB VPI than either constituent monomer. Since both D16 and 1.4E interfere with the TcdB-CSPG4 interaction, we reasoned that a dimeric DARPin (Fig. S2B) comprising U3 and D16 joined by the same linker should exhibit stronger toxin neutralization potency than D16 alone.
Indeed, dimeric DARPin U3D16 showed 10-fold-to-20-fold-higher activity toward TcdB VPI and TcdB M68 than D16 alone ( Fig. 4A and B). However, surprisingly, the TcdB UK1 neutralization ability shown by U3D16 was weaker than that seen with D16 alone (Fig. S3A). We subsequently prepared additional dimeric DARPins with a reverse configuration (i.e., D16U3) and/or different linker lengths (4ϫ, 5ϫ, and 6ϫ GGGGS), but to no avail ( Fig. 3 and 5). Furthermore, a mixture containing both D16 and U3 showed activity identical to that seen with D16 alone (Fig. S3C). These results suggest that, although U3 efficiently neutralizes TcdB VPI and TcdB M68 , it is powerless against TcdB UK1 . Subsequent ELISAs confirmed that U3 lacks the ability to bind TcdB UK1 (Fig. 4C). TcdB UK1 lacks significant ability to interact with FZD2. The inability of U3 to bind/neutralize TcdB UK1 was surprising. DARPin U3 neutralizes TcdB by interfering with the interaction between TcdB and the FZD1/2/7 receptor (34). The members of the Frizzled family of receptors are important for Wnt signaling, a key signaling pathway that regulates cell proliferation and self-renewal (45). Unlike CSPG4, which is not present in the colon epithelium but abundant in the subepithelium layer, FZD2 and FZD7 are highly expressed in mouse and human colonic epithelium, making them the most physiologically relevant receptors for TcdB (38).
The lack of binding of U3 to TcdB UK1 prompted us to examine the interaction between TcdB UK1 and FZD2. ELISA plates were coated with TcdB VPI or TcdB UK1 prior to the addition of the extracellular domain of CSPG4 or FZD2. After thorough washing, the amounts of plate-bound CSPG4 and FZD2 were detected using the respective antibod- Serially diluted DARPins were mixed with the appropriate toxins and then added to Vero cells that had been seeded the night before. Cell viability was quantified by the CellTiterGlo assay 72 h later and normalized to naive Vero cells. The error bars represent mean deviations of results from two independent experiments. (C) U3 lacks the ability to bind to TcdB UK1 as determined by ELISA. The ELISA plates were coated with the appropriate toxin and then blocked with BSA prior to the addition of serially diluted DARPin U3. The amounts of plate-bound DARPin were quantified using an anti-c-Myc antibody. The data are representative of results from two independent experiments performed in duplicate. OD 450 , optical density at 450 nm.
ies. As shown in Fig. 5, CSPG4 can bind efficiently to both TcdB VPI and TcdB UK1 , with the TcdB UK1 interaction being the stronger. In contrast, only TcdB VPI was able to significantly associate with FZD2 at the test concentration, indicating that the binding affinity of TcdB UK1 to FZD2 is much weaker than that shown by TcdB VPI . The extracellular domain of PVRL3 appeared to bind TcdB VPI and TcdB UK1 with similar levels of affinity (Fig. 5C).

DISCUSSION
C. difficile infection (CDI) is the most common cause of antibiotic-associated diarrhea and gastroenteritis-associated death in developed countries. The prevalence, mortality, and costs associated with CDI make C. difficile a major threat to public health. The pathology of CDI stems primarily from the two exotoxins secreted by C. difficile bacteria, TcdA and TcdB, of which TcdB is considered the primarily virulence factor in human (46). C. difficile was first reported to cause human disease in 1978 (47,48). In the past, CDI has been routinely treated with supportive therapy and regimens of antibiotics. However, the cure rate has been steadily decreasing over the last decades largely due to the emergence of a hypervirulent (NAP1/BI/027) strain of C. difficile (9)(10)(11).
Previously, using phage display coupled with functional screening, we successfully isolated a panel of DARPins with potent neutralization activity against TcdB from the laboratory strain of C. difficile VPI 10463 (ribotype 087) and the clinical strain M68 (NAP9/CF/017). The goal of this study was to identify strong binders/neutralizers of TcdB from hypervirulent but intractable C. difficile strain UK1 (NAP1/BI/027). The sequence of TcdB UK1 shares 92% and 88% similarity with those of TcdB VPI and TcdB M68 , respectively. In comparison, TcdB VPI shares 93.7% sequence similarity with TcdB M68 . TcdB from a NAP1/BI/027 strain was reported to induce a greater cytopathic effect on a variety of cell types (49) and to exhibit a substantially lower lethal dose and more-extensive brain hemorrhaging in mice than that were seen with the laboratory strain (50).
To investigate these factors, we performed four rounds of panning of phage displaying a randomized library of designed ankyrin repeat proteins (DARPins) against TcdB from the UK1 strain of C. difficile. TcdB UK1 binders that emerged from this approach were subjected to an in vitro potency screen carried out on Vero cells, resulting in the identification of a panel of DARPins with potent neutralization activity against TcdB UK1 . DARPins are synthetic ankyrin repeat proteins composed of three internal ankyrin repeat (AR) domains sandwiched between N-capping and C-capping domains. Interestingly, the top 57 anti-TcdB UK1 DARPins that emerged from this study share the same 6 ARs in 4 different configurations, suggesting that these DARPins likely target regions surrounding a common epitope (Fig. 1). The four DARPins representing each unique repeat configuration and without any framework mutations were further characterized. The most effective DARPin, D16, neutralized TcdB from C. difficile strains UK1, VPI 10463, and M68 with EC 50 values of 0.5 nM, 5.2 nM, and 1.6 nM, respectively. The in vitro potency of D16 toward TcdB UK1 is Ͼ66-fold higher than that of the toxin-neutralizing therapeutic antibody bezlotoxumab (EC 50 of Ͼ33 nM) (Fig. 2).
All four unique anti-TcdB UK1 DARPins from our screen were found to block the interaction of TcdB with the receptor CSPG4 (Fig. 3), much like the antitoxin DARPins. This finding is consistent with results from our previous study in which the vast majority of the isolated anti-TcdB VPI DARPins neutralized the toxin by interfering with the TcdB-CSPG4 interaction (34). There are currently three known receptors for TcdB: CSPG4, Frizzled 1/2/7, and PVRL3. Although Vero cells express high levels of all these receptors (40), the recurrent emergence of CSPG4-interfering DARPins from functional screens using Vero cells points to a dominant role of CSPG4 in mediating TcdB entry in these cells.
Previously, we identified DARPin 1.4E, which inhibited the CSPG4-TcdB VPI interaction but showed neither activity toward nor binding to TcdB UK1 (34). Since both TcdB VPI and TcdB UK1 bind CSPG4 (Fig. 5A), the ability of the antitoxin DARPins reported in this study to neutralize both TcdB UK1 and TcdB VPI , albeit with different potencies, indicates that these DARPins bind epitopes on TcdB that are distinct from those bound by DARPin 1.4E and that these epitopes partially overlap the footprint of CSPG4, which in part overlaps the epitope of DARPin 1.4E. Unfortunately, the data representing the binding interface of CSPG4 and DARPin 1.4E lack sufficient resolution to support detailed mutagenesis studies to elucidate the exact epitopes for these DARPins (34).
With a view to creating higher-potency antitoxin molecules, the dimeric DARPin U3D16 was created by fusing monomeric D16 to DARPin U3, which was earlier found to disrupt the interaction of TcdB from C. difficile VPI with the Frizzled 1/2/7 receptor. U3D16 exhibits 10-fold-to-20-fold-enhanced neutralization potency against TcdB VPI and TcdB M68 relative to the D16 monomer, likely through an avidity effect (Fig. 4). However, unexpectedly, all the tested dimeric DARPins composed of U3 and D6 (which included variations in configuration and linker length) not only did not show enhanced activity but showed ϳ10-fold-reduced activity (see Fig. S3 in the supplemental material). U3 targets an adjacent epitope on TcdB and blocks its interaction with FZD1/2/7 (34). Further studies showed that neither U3 nor FZD2 bound TcdB UK1 efficiently ( Fig. 4C and  5). To corroborate this finding, we compared the levels of toxicity of TcdB UK1 and TcdB VPI in Caco-2 cells and Vero cells. The Caco-2 cells are derived from the colon epithelium and lack detectable CSPG4 expression (38). On the other hand, Vero cells, derived from the kidney epithelium cells of an African green monkey, express all three known TcdB receptors (CSPG4, FZD2/7, and PVRL3) (40). At the same molar concentration, TcdB UK1 is more toxic to Vero cells than TcdB VPI but is far less toxic to Caco-2 cells than TcdB VPI (Fig. 6), indicating that TcdB UK1 lacks significant ability to enter cells via the FZD1/2/7 receptor. Our ELISA results indicated that, while TcdB UK1 and TcdB VPI bound PVRL3 with similar affinities, TcdB UK1 associated more strongly with CSPG4 than TcdB VPI and lacked significant ability to bind to FZD1/2/7 (Fig. 5). The higher affinity of TcdB UK1 than TcdB VPI for CSPG4 is likely responsible for its greater toxicity in Vero cells, whereas the weaker affinity of TcdB UK1 for FZD1/2/7 may explain its reduced toxicity in Caco-2 cells.
Sequence alignment of TcdB UK1 with TcdB VPI revealed six residue differences at the FZD1/2/7 binding interface (Fig. S4). Among these, we posit that four differences are most likely responsible for the weaker affinity between TcdB UK1 and FZD1/2/7, namely, E1468K, D1501N, Y1509C, and F1597S. The E1468 and Y1509 residues in TcdB VPI form charge interactions/hydrogen bonds with Q83 and H74, respectively, in FZD2 (Fig. 7A). The presence of a positively charged Lys in position 1468 of TcdB UK1 instead of a negatively charged glutamic acid likely abolishes this hydrogen bond interaction. The same applies to the presence of a relatively small cysteine residue in position 1509 of TcdB UK1 in place of tyrosine. Residue D1501 in TcdB UK1 was previously shown to be critical for binding to FZD2, as the D1501A mutation abolished the ability of TcdB VPI to interact with FZD2 (51). In TcdB UK1 , position 1501 is occupied by polar residue Asn (Fig. 7B), which reduces the original two hydrogen bond interactions with K127 in FZD2 to only one. Finally, the F1597S substitution may reduce the hydrophobic interaction between Phe and the nearby F130 on FZD2. In fact, an earlier study found that the F1597G mutation in TcdB VPI abolished its ability to associate with FZD2 in a pulldown assay (51). Collectively, the data indicate that there appears to be a strong structural basis for the lack of interaction between TcdB UK1 and FZD1/2/7.
The weakened ability of TcdB UK1 to use FZD1/2/7 for cell entry is surprising, as FZD7 is abundant in the colon epithelium. On the other hand, CSPG4 is absent from the colon epithelium but is predominantly expressed in multinucleated intestinal subepithelial myofibroblasts (52). Independently, the same phenomenon was recently reported by Chung et al., who noted a lack of interaction between TcdB R20291 (with the same amino acid sequence as TcdB UK1 ) and FZD2-Fc by Western blotting (53), and by Lopez-Urena et al., whose results demonstrated that the uptake of TcdB VPI by HeLa cells (expressing both CSPG4 and Frizzled receptors [38]) was not blocked by TcdB NPI (with the same protein sequence as TcdB UK1 ) (57).
The explanation for the reduced activity of the dimeric DARPin composed of U3 and D16 toward TcdB UK1 compared to D16 is not immediately clear. It may due to a nonspecific interaction between U3 and D16 which partially obscures the target binding interface on D16. The presence of a binding partner of U3 on TcdB VPI and TcdB M68 draws U3 away from D16, enabling the two DARPins to simultaneously bind the toxin and enhancing the neutralization potency.
In summary, we identified a panel of DARPins with potent neutralization potency toward TcdB from the hypervirulent UK1 strain of C. difficile (NAP1/BI/027). These DARPins neutralize TcdB by blocking its interaction with the receptor CSPG4. We further showed that TcdB UK1 does not strongly associate with FZD1/2/7, bringing into question the significance of this receptor in CDI caused by ribotype 027 hypervirulent strains of C. difficile.

MATERIALS AND METHODS
Protein expression and purification. TcdB UK1 was recombinantly expressed in Bacillus megaterium cells and purified via the use of a nickel-nitrilotriacetic acid (Ni-NTA) column essentially as described previously (54). The fractions containing TcdB were combined and concentrated and subjected to buffer exchange using phosphate-buffered saline (PBS) (10-fold dilution of 10ϫ PBS [Fisher catalog no. BP3991]) and ultrafiltration units (Amicon; molecular-weight cutoff [MWCO], 100 kDa). Protein purity was confirmed using SDS-PAGE. The concentration of purified protein was determined by calculating the absorbance at 280 nm with a theoretical extinction coefficient of 293,620 M Ϫ1 cm Ϫ1 . We typically obtain ϳ2 mg of purified toxin per liter of culture. The purified protein was stored at Ϫ20°C in 50% glycerol. TcdB VPI , TcdB M68 , CSPG4-EC-GFP, and bezlotoxumab were recombinantly expressed and purified as described previously (34,54).
Phage panning and functional screening. An in-house N3C DARPin library with a diversity level of ϳ10 9 was used in the phage panning essentially as described previously (34,55). Purified TcdB UK1 was biotinylated via the use of EZ-Link-Sulfo NHS-LC [succinimidyl 6-(biotinamido)hexanoate] biotin (Pierce) and used as the target protein in four rounds of sequential phage panning. A significant level of TcdB UK1 binding enrichment was observed after 4 rounds of selection using phage ELISA (55), indicative of successful phage panning.
DARPin variants from the fourth-round phage library were cloned into pET28a vector (containing an N-terminal His tag and a Myc tag; see Fig. S2A in the supplemental material) via the use of BamHI and HindIII restriction sites. A total of 760 individual E. coli BL21(DE3) clones were picked and grown in v-bottom 96-well plates (200 l/well) in Luria broth (LB) supplemented with kanamycin (50 g/ml) at 37°C with shaking for ϳ18 h. Cells were harvested by centrifugation (1,048 ϫ g for 10 min at 4°C). Each cell pellet was resuspended in 200 l lysis buffer (PBS supplemented with 1 mM CaCl 2 , 0.5 mM EDTA, and 200 g/ml lysozyme) and incubated at 37°C for 30 min. These cells then underwent 2 cycles of freeze-thaw between Ϫ80°C and 37°C and centrifugation at 1,048 ϫ g for 20 min at 4°C. The soluble cell lysates (2 l/well) were added to Vero cells that had been seeded the previous day at 1,500 cells/well together with TcdB UK1 (1 pg/ml final concentration) in 200 l complete growth medium (Dulbecco's modified Eagle's medium [DMEM] supplemented with 10% fetal bovine serum, 1ϫ nonessential amino acids, and 1ϫ antibiotic antimycotic (Life Technologies catalog no. 15240062). The plates were incubated at 37°C and 5% CO 2 for 72 h, and the viability of these Vero cells was quantified by the use of CellTiter-Glo reagent (Promega) and normalized to that of naive Vero cells.
In vitro toxicity and neutralization assay. To compare the levels of toxicity of the different toxins (i.e., TcdB UK1 , TcdB VPI , and TcdB M68 ), purified toxins were serially diluted and added to Vero cells or Caco-2 cells that had been seeded in 96-well plates the day before at 1,500 cells/well. The plates were incubated at 37°C and 5% CO 2 , and the viability of these cells was quantified 72 h later using CellTiter-Glo reagent. The minimum concentrations of the toxin that led to Ͻ20% Vero cell viability were determined to be 1 pg/ml for TcdB UK1 , 5 pg/ml for TcdB VPI , and 5 pg/ml for TcdB M68 .
To determine the neutralization potencies of the different DARPins, serially diluted DARPins were added to Vero cells (seeded at 1,500 cells/well the day before) together with the minimum dose of the appropriate toxin that led to Ͻ20% cell viability.
Modeling study. A homology model of TcdB UK1 Frizzled binding domain (FBD; amino acids 1284 to 1803) was constructed using SWISS-MODEL and was overlaid onto the FBD of TcdB VPI (PDB code: 6C0B). The complex was visualized using Visual Molecular Dynamics (VMD) (56).
Data availability. The sequences of our engineered proteins are presented in Table S1.

ACKNOWLEDGMENTS
The funding for this work was graciously provided by the National Institutes of Health through grants R21AI126025, DP2OD024146, and R21AI137803 (to Z.P., R.S., and Z.C.).