Multivalency effects of hemagglutinin component of type B botulinum neurotoxin complex on epithelial barrier disruption

Hemagglutinin (HA) is one of the components of botulinum neurotoxin (BoNT) complexes and it promotes the absorption of BoNT through the intestinal epithelium by at least two specific mechanisms: cell surface attachment by carbohydrate binding, and epithelial barrier disruption by E‐cadherin binding. It is known that HA forms a three‐arm structure, in which each of three protomers has three carbohydrate‐binding sites and one E‐cadherin‐binding site. A three‐arm form of HA is considered to bind to these ligands simultaneously. In the present study, we investigated how the multivalency effect of HA influences its barrier‐disrupting activity. We prepared type B full‐length HA (three‐arm form) and mini‐HA, which is a deletion mutant lacking the trimer‐forming domain. Size‐exclusion chromatography analysis showed that mini‐HA exists as dimers (two‐arm form) and monomers (one‐arm form), which are then separated. We examined the multivalency effect of HA on the barrier‐disrupting activity, the E‐cadherin‐binding activity, and the attachment activity to the basolateral cell surface. Our results showed that HA initially attaches to the basal surface of Caco‐2 cells by carbohydrate binding and then moves to the lateral cell surface, where the HA acts to disrupt the epithelial barrier. Our results showed that the multivalency effect of HA enhances the barrier‐disrupting activity in Caco‐2 cells. We found that basal cell surface attachment and binding ability to immobilized E‐cadherin were enhanced by the multivalency effect of HA. These results suggest that at least these two factors induced by the multivalency effect of HA cause the enhancement of the barrier‐disrupting activity.

BoNT, NTNHA, and HA (Fig. 1a). LL-PTC is assumed to be a dimer of L-PTC (5). When assembled into PTC, NTNHA protects BoNT from an acidic pH and proteolytic enzymes in the gastrointestinal tract (6,7). HA facilitates BoNT absorption across the intestinal epithelium by at least two specific mechanisms: attachment to the luminal surface of the intestinal epithelial cells by carbohydrate binding (8)(9)(10)(11)(12)(13)(14), and epithelial barrier disruption by E-cadherin binding (15)(16)(17)(18)(19). We have proposed an HA-mediated three-step mechanism for the intestinal absorption of L-PTC (20). In the first step, L-PTC binds to glycoproteins/glycolipids at the luminal surface of the intestinal epithelial cells, followed by transcytosis. Then, the HA component of L-PTC delivered to the basolateral side of epithelial cells directly binds to E-cadherin, which exclusively resides on the basolateral surface of epithelial cells, leading to epithelial barrier disruption. Finally, the disruption allows PTC to pass through the paracellular route (20).
HA is composed of three subcomponents: HA1, HA2, and HA3 (also known as HA33, HA17, and HA70, respectively) (21). Six HA1, three HA2, and three HA3 are assembled into a triskelion-shaped hetero-dodecamer in which homo-trimeric HA3 forms the core of the complex (Fig. 1a, b) (22)(23)(24)(25). Hence, the triskelion-shaped HA is a trimer of three protomers, composed of two HA1, one HA2, and one HA3, and it has a "three-arm" structure ( Fig. 1a, b). The carbohydrate-binding activity of HA is attributable to HA1 and HA3, which bind to galactose and sialic acid, respectively (9). In addition, HA1 of types C and D also bind to sialic acid (10,26). Triskelion-shaped HAs of types A and B have six galactose-binding sites and three sialic acid-binding sites (23)(24)(25). HAs of types A and B bind to E-cadherin; a region extending over HA2 and the C-terminal domain of HA3 interacts with E-cadherin ectodomain 1-2 (EC1-2) (17)(18)(19). Lee et al. reported that, using ITC, HA (full-length HA, FL-HA) of type A interacts with E-cadherin EC1-2 expressed in Escherichia coli with a K d $2.3 mM and a 1:3 HA-E-cadherin stoichiometry (18). It was also reported that a truncated mutant of HA (termed mini-HA) of type A, which lacks the trimer-forming domain of HA3 and represents "one-arm" of the HA, interacts with EC1-2 with a K d $2.7 mM and a 1:1 HA-Ecadherin stoichiometry (18).
Carbohydrate-binding sites and E-cadherin-binding sites are located within each arm of HA ( Fig. 1c) (18,19,24).  (7,24,25). The figures were generated by PyMOL (38) and UCSF Chimera (39) (b) Schematic models of FL-HA and mini-HA. FL-HA is composed of HA1, HA2, and full-length HA3 (HA3-FL; aa 19-626) and mini-HA is composed of HA1, HA2, and HA3-mini (aa 380-626). N-terminal deletion of HA3-FL was introduced based on N-terminal sequencing of HA proteins produced by Clostridium botulinum (5). (c) Part of FL-HA and mini-HA are represented by surface and ribbon style, respectively. N286 of HA1 (red), R528 (blue) and K607 (black) of HA3 are represented by sphere style and indicated by labeled arrows. Therefore, triskelion-shaped HA is capable of exerting multivalent binding to these ligands. In the present study, we prepared type B FL-HA; three-arm HA, dimers of mini-HA; two-arm HA, and monomers of mini-HA; one-arm HA, and we investigated the multivalency effects of HA on epithelial barrier disruption.

Plasmid construction
Genomic DNA was extracted and purified from Clostridium botulinum type B strain Okra. HA1 (aa 7-294) encoding gene was cloned into the KpnI-SalI site of the pET52b(þ) vector (Novagen, Merck Millipore, Madison, WI, USA), and an oligonucleotide encoding a FLAG-tag was inserted at the C-terminus of HA1. HA2 (aa 2-146) encoding gene was cloned into the HindIII-KpnI site of pT7-FLAG-1 vector (Sigma Aldrich, St Louis, MO, USA). The full-length HA3 (HA3-FL, aa 19-626) encoding gene was cloned into the KpnI-SalI site of the pET52b(þ) vector. The truncated mutant of HA3 (termed HA3-mini, aa 380-626) encoding gene was cloned into the NcoI-SalI site of the pET52b(þ) vector, and an oligonucleotide encoding a Strep-tag II tag was inserted at the C-terminus of HA3-mini. Site-directed mutagenesis was carried out using PrimeSTAR Max polymerase (Takara Bio, Shiga, Japan). The inserted regions of these plasmids and the presence of mutation were confirmed by DNA sequencing.

In vitro reconstitution and purification of HAs
The purified HA1, HA2, and HA3 (HA3-FL for FL-HA, HA3-mini for mini-HA) were mixed at a molar ratio of 4:4:1 in PBS and incubated for 3 hr at 37°C. The FL-HA and mini-HA complexes were purified using StrepTrap HP. The mini-HA complex was subjected to further purification on a Superdex 200 Increase 10/300 GL column (GE Healthcare, Uppsala, Sweden) in PBS. Purified proteins were dialyzed against PBS.

Size-exclusion chromatography analysis of HAs
Each purified HA protein (50 mg) was loaded onto a Superdex 200 Increase 10/300 GL column (GE Healthcare) in PBS. Each mini-HA (25 mg) (in a concentration of 2.2 mM as protomers) was incubated at 37°C for the indicated time. Then, the mini-HA was loaded onto a Superdex 200 Increase 10/300 GL column in PBS. Protein concentration of the elution was measured based on absorbance at 280 nm.

Transepithelial electrical resistance assay
Measurement of TER was carried out using Millicell-ERS (Merck Millipore, Billerica, MA, USA), as previously described (15). In brief, Caco-2 cells were grown on 24-well collagen-coated Transwell filters Pull-down assay FL-HAs and mini-HAs (20 pmol protomer of HAs) were immobilized onto Strep-Tactin Superflow agarose gel beads (Novagen, Merck Millipore) by incubation for 1 hr at 4°C in TNCX buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 2 mM CaCl 2 , 0.01% Triton X-100). Concentration of E-cadherin ectodomain protein was adjusted to 100 nM in TNCX buffer. A 200-mL aliquot of E-cadherin solution was subjected to pull-down with beads coupled to FL-HA, a dimer fraction or a monomer fraction of mini-HA for 2 hr at 4°C. The beads were washed three times, and the bound proteins were eluted in SDS-PAGE sample buffer. Samples were separated by 12.5% SDS-PAGE and transferred to a polyvinylidene fluoride (PVDF) membrane (Immobilon-P; Merck Millipore), followed by immunoblotting with rat anti-mouse E-cadherin monoclonal antibody (ECCD-2, 1:500; Takara Bio). Bound antibodies were probed with HRPconjugated goat anti-rat IgG antibody (1:10,000; Jackson ImmunoResearch, West Grove, PA, USA), and the membranes were developed using ECL select (Amersham, GE Healthcare, Buckinghamshire, UK). After immunoblotting, the membranes were stained with CBB to visualize HA proteins.

Cell-ELISA
Caco-2 cells were grown on 24-well collagen-coated Transwell filters until completely confluent. HAs were added to the basolateral side at a final concentration of 51 nM protomer, and the monolayers were incubated in Caco-2 cell culture medium (MEM, 20% FBS) for 40 min at 4°C. After washing with ice-cold PBS, cells on the Transwell filters were fixed with 4% PFA for 30 min at 4°C, followed by blocking with 5% BSA in PBS for 30 min at room temperature. Then, the basolateral sides of the monolayers were incubated with mouse anti-FLAG-tag monoclonal antibody (M2, 1:5000; Sigma Aldrich) in blocking buffer (5% BSA in PBS) for 2 hr at 37°C, followed by HRP-conjugated donkey anti-mouse IgG antibody (1:10,000; Jackson ImmunoResearch) for 1 hr at 37°C. The Transwell filters were developed using ABTS, and the absorbance at 405 nm was read.

Immunofluorescence
Hemagglutinin were added to the basolateral side of Caco-2 cells grown on Transwell filters, and the monolayers were incubated in the Caco-2 cell culture medium for 40 min at 4°C. After fixation with 4% PFA for 30 min at 4°C and permeabilization with 0.5% Triton X-100 in PBS for 5 min at room temperature, the cells were blocked with 5% BSA in PBS for 30 min at room temperature. Then, HA was stained with a mouse anti-FLAG-tag (1:1000) monoclonal antibody, followed by Cy3-conjugated donkey anti-mouse IgG antibody (1:400; Jackson ImmunoResearch). After immunostaining, the Transwell filters were mounted in ProLong antifade reagent (Molecular Probes, ThermoFisher Scientific, Eugene, OR, USA). Images were acquired by confocal microscopy using a CSUX1 confocal scanner unit (Yokogawa Electric, Tokyo, Japan) and IX71 microscope (Olympus, Tokyo, Japan) and processed using Metamorph software (Molecular Devices, Sunnyvale, CA, USA).
The SEC profiles of Alexa-labeled HA complexes were consistent with those of non-labeled HA complexes (data not shown). Basolateral sides of Caco-2 cells grown on collagen-coated 24-well Transwell filters were incubated with 153 nM protomer of Alexa-labeled HA complexes in the Caco-2 cell culture medium for 40 min at 4°C. After a brief wash in the ice-cold Caco-2 cell culture medium, the monolayers were incubated with 153 nM protomer of non-labeled HA complexes in the Caco-2 cell culture medium for the indicated time (10 or 20 min) at 37°C. After pulse or chase labeling, the monolayers were fixed and stained with rat anti-E-cadherin (1:1000; DECMA-1; Sigma-Aldrich) and mouse anti-ZO-1 (ZO1-1A12, 1:250, Zymed; ThermoFisher Scientific, Carlsbad, CA, USA) monoclonal antibodies, followed by Alexa Fluor 488conjugated goat anti-rat IgG antibody (1:400; Molecular Probes, ThermoFisher Scientific, Eugene, OR, USA) and Alexa Fluor 405-conjugated goat anti-mouse IgG antibody (1:400; Molecular Probes, ThermoFisher Scientific, Eugene, OR, USA), respectively. After immunostaining, images were acquired by confocal microscopy.

RESULTS
Preparation of one-, two-, and three-arm HA Full-length HAs (FL-HAs) and mini-HAs of type B were reconstituted in vitro from purified HA subcomponents expressed in E. coli (Fig. 2, Fig. S1a). We constructed the mini-HA of type B, which consists of HA1, HA2, and an HA3 mutant-deleted N-terminal 379 amino acid (aa) residues, in the same way as mini-HA of type A reported by Lee et al. (24). Reconstituted mini-HA showed two major peaks by SEC (Fig. 2a). Among all HA subcomponents, only HA3-mini showed the two similar peaks (Fig. S1a). It is considered that the dimerization of mini-HA is as a result of a weak association between HA3-minis. After SEC separation of the two major peaks of mini-HA, the re-run profiles of each fraction on SEC showed single major peaks individually at 10.5 mL (dimer, corresponding to two-arm HA, predicted mass: 232 kDa) and 12.5 mL (monomer, corresponding to onearm HA, predicted mass: 116 kDa) (Fig. 2b). Although the height of each minor peak of mini-HAs was slightly increased after incubation at 37°C for 24 hr in PBS, each major peak was mostly retained (Fig. S1b, c). FL-HA was predominantly eluted at 9.0 mL (corresponding to threearm HA, predicted mass: 474 kDa) in SEC (Fig. 2b).

Fractionated mini-HAs disrupt the epithelial barrier of Caco-2 cells
In polarized epithelial cells, junctional complexes such as TJ, E-cadherin-based AJ, and desmosomes are oriented in an apical-to-basal arrangement (27). Epithelial TJ seal adjacent epithelial cells and regulate paracellular permeability (28). Thus, macromolecules, e.g. HA and PTC, do not gain direct access to E-cadherin from the apical side of the epithelial cell, suggesting that a transcytosis step to the basolateral side is required for apically applied HA to interact with E-cadherin (15). To assess the direct barrier-disrupting activities of HAs, we measured the TER of Caco-2 cell monolayers following basolateral treatment with HAs (Fig. 3). FL-HA and dimer and monomer fractions of mini-HA disrupted the epithelial barrier of Caco-2 cells (Fig. 3). The barrierdisrupting activities of dimer and monomer fractions of mini-HA were decreased to approximately one-half and one-fifth of that of FL-HA, respectively (Fig. 3). The TER value of Caco-2 cells with FL-HA was decreased to 84% at 2 hr post-addition, and that with the higher concentration of dimer (102 nM as protomer, 2Â dimer) and monomer (255 nM, 5Â monomer) fractions of mini-HA was decreased to 54 and 69%, respectively (Fig. 3). However, the TER values after 6 hr  post-addition with mini-HAs (2Â dimer, 41$30%; 5Â monomer, 46$41%) was higher than that of FL-HA (37$17%) (Fig. 3).

Binding of HAs to E-cadherin
To assess the interaction between HAs and the Ecadherin extracellular domain, we carried out pull-down assays and sandwich ELISAs using recombinant Ecadherin EC1-5-His, which was expressed in HEK293 cells and purified (Fig. 4). In the pull-down assay, Ecadherin EC1-5 comparably bound to Strep-tag II tagged FL-HA WT and mini-HA immobilized on Strep-Tactin beads (Fig. 4a). In the sandwich ELISA, E-cadherin EC1-5, which contains a hexa-histidine tag at the C-terminus, was immobilized to anti-His tag antibody-coated plates, and then HA were added to these plates. As a result, the E-cadherin-binding abilities of dimer and monomer fractions of mini-HA were relatively weak, and 65% and 10% of that of FL-HA WT, respectively (Fig. 4b). It has been reported that HA3 Lys607 is located at the interface with E-cadherin (Fig. 1c), and its alanine substitution (K607A, KA) impairs the E-cadherin-binding and barrier-disrupting activity (18,19). FL-HA harboring a KA mutation (FL-HA KA ) did not interact with E-cadherin EC1-5 in these experiments (Fig. 4).

Attachments of HA to the basolateral surface of Caco-2 cells
We examined the abilities of HA to attach to the basolateral cell surface by cell-ELISA using Caco-2 cell monolayers (Fig. 5a). FL-HA WT attached to the basolateral cell surface (Fig. 5a). To identify how HA attaches to the basolateral surface of Caco-2 cells, we carried out these experiments using functional mutants of FL-HA. The galactose-binding site of HA1 and the sialic acid-binding site of HA3 have already been identified in type A, B, and C by co-crystallization with saccharides ( Fig. 1c) (24,(29)(30)(31). It has also been reported that alanine substitutions of HA1 Asn286 (N286A, NA) and HA3 Arg528 (R528A, RA) abolish each carbohydrate binding (19,24,29). FL-HA harboring an NA mutation (FL-HA NA ) rarely attached to the basolateral cell surface (Fig. 5a). The basolateral cell surface attachment of FL-HA harboring an RA mutation (FL-HA RA ) was 70% of that of FL-HA WT, and that of FL-HA KA was comparable with that of FL-HA WT (Fig. 5a). These results indicate that HA attached to the basolateral surface of Caco-2 cells by binding mainly to galactose. The basolateral cell surface attachment of dimer and monomer fractions of mini-HA were 20% and 7% that of FL-HA WT, respectively (Fig. 5a). Next, HAs that attached to the basal cell surface were visualized using immunofluorescence (Fig. 5b). FL-HA WT, FL-HA RA , and FL-HA KA attached to the basal surface of Caco-2 cells homogeneously (Fig. 5b). The FL-HA NA attached to the cell surface was rarely observed (Fig. 5b). The attachment of the dimer fraction of mini-HA to the basal cell surface was weaker than FL-HA WT, and the monomer fraction of mini-HA showed very weak attachment (Fig. 5b).

Pulse-chase analysis
The z-stacked images showed that basolaterally applied HAs localized only to the basal cell surface, and not to the lateral cell surface at 4°C (Fig. 6a). It has been reported that basolaterally applied FL-HA WT localized to the basal and lateral surfaces of Caco-2 cells when incubated continuously at 37°C (17,19). Therefore, we postulate that HA attaching to the basal cell surface moves to the lateral cell surface. To examine this hypothesis, we carried out pulse-chase experiments (Fig. 6). After chasing for 10$20 min at 37°C, some of the FL-HA WT that attached to the basal cell surface moved to the lateral cell surface (Fig. 6b,c). In a similar way, the dimer fraction of mini-HA moved to the lateral cell surface, albeit at a lower level compared with FL-HA WT. The monomer fraction of mini-HA moved to the lateral cell surface only slightly (Fig. 6c). FL-HA KA did not localize to the lateral cell surface, and some of the FL-HA KA were endocytosed into the cytoplasm (Fig. 6b,c). E-cadherin showed major localization throughout the lateral cell surface and infrequent localization to the basal cell surface (Fig. 6a,b).

DISCUSSION
Our results showed that the greater number of arms HA has, the more effectively it disrupts the epithelial barrier of human intestinal Caco-2 cell monolayers; that is, the barrier-disrupting activity of HA is markedly increased by the multivalency effect of HA. Our results also showed that the three-arm form of HA is not essential for epithelial barrier disruption. The monomer fraction of mini-HA slightly compromised the integrity of the epithelial barrier, but significantly compromised it at a higher concentration (Fig. 3). These results imply that one-arm HA is capable of disrupting the epithelial barrier. However, the possibility cannot be excluded that one-arm HA has no activity to disrupt the epithelial barrier, and a small amount of two-arm HA existing in the monomer fraction causes this activity; there might exist dimers of mini-HA in the monomer fraction for some reason; e.g. contamination of dimers in the fractionation procedure, and/or dimerization under the experimental conditions. The barrier-disrupting activities of mini-HAs are weakened at 4-6 hr postaddition, possibly through the instability of mini-HAs because of the truncating mutation.
Consistent with the results of Lee et al. (18), the binding ability to un-immobilized E-cadherin was comparable between FL-HA and mini-HAs (Fig. 4a). However, the binding ability of HA to immobilized E-cadherin was markedly enhanced by the multivalency effect (Fig. 4b). Assuming that the immobilized E-cadherin exists as a cluster on the plates, these results suggest that the multivalency effect of HA might cause the effective interaction with E-cadherin that resides on the lateral cell surface and exists as a cluster. We previously found that the barrier-disrupting activity of FL-HA NA was one-fifth that of FL-HA WT, and the activity of FL-HA RA was slightly less than that of FL-HA WT, when these HAs were applied from the basolateral side, even though they showed the same binding ability to E-cadherin in the pull-down assay (19). Hence, it is considered that the carbohydrate-binding activity of HA increases the opportunities to meet E-cadherin by promoting basolateral cell surface attachment, leading to more effective barrier disruption (19). Our present results show that HA attaches to the basal surface of Caco-2 cells by binding mainly to the terminal galactose residue on glycoproteins/ glycolipids (Figs 5, 6a). It is well-known that the interaction between carbohydrates and lectins could be markedly enhanced by the multivalency effects of lectins (32)(33)(34). In a similar way, the intermediate L-PTC of type D, which has a smaller number of HA1-HA2 complexes than mature L-PTC, attaches to the surface of rat intestinal epithelial IEC-6 cells by carbohydrate binding weaker than mature L-PTC (35). Consistently, our results showed that the attachment of HA to the basolateral surface of Caco-2 cells is markedly enhanced by the multivalency effect of HA (Fig. 5).
Pulse-chase experiments showed that basolaterally applied HAs initially attach only to the basal surface of Caco-2 cells, and then the bound HAs move to the lateral cell surface (Fig. 6). FL-HA KA , which does not interact with E-cadherin, was able to attach to the basal cell surface as well as FL-HA WT (Figs 5, 6a). However, Fig. 6. Pulse-chase analysis with HA-Alexa568. (a-c) After Caco-2 cells grown in the Transwell chamber were incubated with 153 nM protomer of Alexa568-labeled HAs (red) from the basolateral side for 40 min at 4°C (a), the cells were further incubated with 153 nM protomer of non-labeled HAs from the same side at 37°C for 10 min (b) or 20 min (c). Monolayers were stained with DECMA-1 antibody against E-cadherin (green) and ZO1-1A12 antibody against ZO-1 (blue). Images were acquired in the x-z plane (a, b) and x-y plane at the middle of the cell height (c). Scale bars: 5 mm (a, b) or 10 mm (c).
FL-HA KA did not move to the lateral cell surface after chasing (Fig. 6b, c). These results suggest that this movement of HA is dictated by E-cadherin. It is known that cadherin molecules flow within the lateral plasma membrane in a basal-to-apical direction and this flow is called "cadherin flow" (36,37). Therefore, it is assumed that HA binds to E-cadherin at the basal cell surface and then rides on cadherin flow. This movement of HA was increased by the multivalency effect of HA (Fig. 6c). We previously reported that the amount of the carbohydrate-binding defective mutant FL-HA (FL-HA NA/RA , FL-HA harboring both NA and RA) located at the lateral cell surface is much lower than that of FL-HA WT, when these HAs are applied basolaterally and the monolayers are incubated continuously at 37°C (19). These observations indicate that the movement of HA from the basal to the lateral cell surface is promoted by basal cell surface attachment by carbohydrate residues, which consequently increases the opportunities to meet Ecadherin, and this attachment step is enhanced by the multivalency effect of HA.
Recently, we reported that L-PTC of type A enters the host through M cells by glycosylphosphatidylinositol (GPI)-anchored glycoprotein-mediated transcytosis. Binding to these glycoproteins is responsible for HA in a carbohydrate-binding-dependent way (13). Therefore, it is assumed that the multivalency effect of HA enhances the attachment of L-PTC to the luminal surface of M-cells and increases the efficiency of transcytosis across the intestinal lumen.
In conclusion, we showed that basolaterally applied HA initially binds to glycoproteins/glycolipids only at the basal cell surface and then moves to the lateral cell surface, where HA acts to disrupt the epithelial barrier. Our results show that the multivalency effect of HA increases the epithelial barrier-disrupting activity, and this phenomenon is considered to be induced by at least two mechanisms: enhancing the cell surface attachment by carbohydrate residues at the basal cell surface, and enhancing the binding avidity to clustered E-cadherin that resides on the lateral cell surface.