A camelid single-domain antibody neutralizes botulinum neurotoxin A by blocking host receptor binding

Antibody treatment is currently the only available countermeasure for botulism, a fatal illness caused by flaccid paralysis of muscles due to botulinum neurotoxin (BoNT) intoxication. Among the seven major serotypes of BoNT/A-G, BoNT/A poses the most serious threat to humans because of its high potency and long duration of action. Prior to entering neurons and blocking neurotransmitter release, BoNT/A recognizes motoneurons via a dual-receptor binding process in which it engages both the neuron surface polysialoganglioside (PSG) and synaptic vesicle glycoprotein 2 (SV2). Previously, we identified a potent neutralizing antitoxin against BoNT/A1 termed ciA-C2, derived from a camelid heavy-chain-only antibody (VHH). In this study, we demonstrate that ciA-C2 prevents BoNT/A1 intoxication by inhibiting its binding to neuronal receptor SV2. Furthermore, we determined the crystal structure of ciA-C2 in complex with the receptor-binding domain of BoNT/A1 (HCA1) at 1.68 Å resolution. The structure revealed that ciA-C2 partially occupies the SV2-binding site on HCA1, causing direct interference of HCA1 interaction with both the N-glycan and peptide-moiety of SV2. Interestingly, this neutralization mechanism is similar to that of a monoclonal antibody in clinical trials, despite that ciA-C2 is more than 10-times smaller. Taken together, these results enlighten our understanding of BoNT/A1 interactions with its neuronal receptor, and further demonstrate that inhibiting toxin binding to the host receptor is an efficient countermeasure strategy.

To further investigate the biological activities of ciA-C2 against BoNT/A1 at the motor nerve terminals-the site-of-action for all BoNTs-we carried out an ex vivo mouse phrenic nerve (MPN) hemidiaphragm assay, a method that closely reproduces in vivo respiratory failure caused by BoNT intoxication 14 . As shown in Fig. 1C, pre-incubation with 1 nM of ciA-C2 displayed almost complete neutralization of 1.63 pM BoNT/A1. ciA-F12, which binds to BoNT/A1 with an affinity (K d ~0.24 nM) higher than ciA-C2, showed weak inhibition of BoNT/ A1 (1.63 pM) even at a high concentration up to 100 nM. These findings confirmed that ciA-C2 is a potent antitoxin against BoNT/A1, despite its small size that is less than 10% of a conventional IgG.
ciA-C2 inhibits cell-surface binding of BoNT/A1. In general, neutralization of BoNTs can be achieved by interfering with at least one of the four major steps required for successful nerve cell intoxication. These steps are: (1) BoNT binds to cell surface receptors on motoneurons; (2) BoNT is internalized by receptor-mediated endocytosis; (3) the protease domain of BoNT (light chain, LC) translocates across vesicle membrane to reach the cytosol; and (4) the LC specifically cleaves SNARE proteins 20 . We performed an in vitro cleavage assay using the recombinant LC/A with the recombinant SNAP-25 as a substrate, and found that neither ciA-C2 nor ciA-F12 could protect SNAP-25 from being cleaved by LC/A (Fig. S1). This is consistent with our earlier studies suggesting that ciA-C2 binds to H C A1 10 . Therefore, we suspected that ciA-C2 may interfere with H C A1-mediated receptor binding.
We then carried out two complementary assays to test this hypothesis. First, we examined BoNT/A1 binding to the cultured rat hippocampal neurons, and found that pre-incubation with ciA-C2 almost completely abolished neuronal binding of BoNT/A1, whereas ciA-F12 did not ( Fig. 2A). Second, we conducted a synaptosome binding assay using 35 S radiolabeled H C A1 on freshly prepared rat brain synaptosomes. ciA-C2, but not ciA-F12, drastically reduced the binding of H C A1 to synaptosomes (Fig. 2B). These findings strongly suggest that ciA-C2 prevents BoNT/A1 from attaching to motor nerve terminals by interfering with its receptor binding.
High-resolution crystal structure of ciA-C2 in complex with H C A1. In order to understand the molecular details underlying BoNT/A1 neutralization by ciA-C2, we performed co-crystallization and determined the structure of ciA-C2 in complex with H C A1 at 1.68 Å resolution (Table 1). There is one ciA-C2-H C A1 complex in an asymmetric unit (AU) in the crystal with a 1:1 stoichiometry (Fig. 3A). Binding of ciA-C2 does not cause obvious conformational change to H C A1 when compared to H C A1 in its apo state (PDB: 3FUO) (root-mean-square (RMS) deviation of ~0.35 Å over 367 aligned Cα pairs, calculated by Pymol) 21 . ciA-C2 consists of a typical single immunoglobulin domain that contains three complementarity-determining regions (CDRs) and four framework regions (FRs) (Fig. 3A,B). A canonical disulfide bond is formed between residues C22 of FR1 and C102 of FR3.
H C A1 could be further divided into two subdomains: the N-terminal lectin-like domain (H CN A1) and the C-terminal β-trefoil domain (H CC A1) (Fig. 3A). H CC A1 is a cell-surface binding mediator to both PSG and SV2. The role of H CN A1 during toxin uptake is not fully understood, but it is involved in binding of an N-linked glycan on SV2 that strengthens neuron recognition of BoNT/A1 17 . Interestingly, ciA-C2 binds to H C A1 at an area located between H CN A1 and H CC A1. The binding of ciA-C2 bridges the two subdomains, resulting in a buried interface area of 1,150 Å 2 (calculated by PDBePISA v1.51) 22 . Most of the ciA-C2-H C A1 interactions are contributed by residues in the CDR1, FR2, and CDR2 regions of ciA-C2, with few residues in the FR3, CDR3 and FR4 regions ( Fig. 3C and Table S1). Residues G26, G28, W30, and T37 of CDR1 form multiple hydrogen bonds with N1188 and D1289 in H CC A1. A strong cation-π interaction is observed between W30 of ciA-C2 and K1159 of H C A1. T65 of CDR2 and Y66 of FR3 interact with H CN A1 by forming hydrogen bonds with H1064 and T1063, respectively. FR2 of ciA-C2 binds both subdomains of H C A1: E50 forms two hydrogen bonds with T1146 of H CC A1, while L54 forms a hydrogen bond with H1064 of H CN A1. In CDR3 and FR4 of ciA-C2, residue E106 interacts with Y1112 of Figure 2. ciA-C2 prevents BoNT/A1 from binding to neurons. (A) ciA-C2 blocks the binding of BoNT/A1 to neuronal surface. BoNT/A1 (20 nM) was pre-incubated with indicated VHHs (200 nM) in high K + buffer. The mixtures were used to incubate cultured rat hippocampal neurons for 5 min. Neurons were then washed, fixed, and subjected to immunostaining analysis detecting BoNT/A1. SV2 was detected as a marker for pre-synaptic terminals. Scale bar: 20 μm. (B) ciA-C2 blocks synaptosome binding of H C A1. Rat brain synaptosomes were incubated with radiolabeled H C A together with ciA-C2 or ciA-F12 (200 nM). Synaptosomes were then collected by centrifugation, washed, and subjected to SDS-PAGE and autoradiography. Amounts of bound H C A1 were compared with the binding efficiency of the reference H C A1 without VHH.
SCIENTIfIC REPORtS | 7: 7438 | DOI:10.1038/s41598-017-07457-5 H CC A1, while W108 forms a hydrogen bond with E1293 and an aromatic-proline interaction with P1295 at the C-terminus of the toxin. ciA-C2 prevents BoNT/A1 from binding to both the peptide and carbohydrate moieties of SV2C. It is well accepted that all BoNTs bind to polysialogangliosides and most of them further bind specific protein receptors on the neuronal surface in order to achieve high binding affinity and specificity 16,19,23 . Failure of binding to either receptor results in drastic toxicity loss, as demonstrated in many studies 14,16,18,[24][25][26][27] . A structural comparison with the GT1b-bound H C A1 (PDB: 2VU9) 28 clearly shows that ciA-C2 binds H C A1 at a position far from the GT1b-docking site, thus it is unlikely to affect BoNT/A-PSG interaction (Fig. 4A).
We then focused on the BoNT/A1-SV2 interactions. Our recent co-crystal structure of H C A1 in complex with the glycosylated SV2C (gSV2C) revealed extensive interactions between them, which are crucial for BoNT/A1's high neuron tropism 17 . Besides the conventional protein-protein interactions, BoNT/A1 recognizes an N-linked glycan on SV2 that is conserved across vertebrates. A structural comparison between the H C A1-ciA-C2 and H C A1-gSV2C complexes showed that ciA-C2 largely occupies the binding-site for the SV2C glycan (Fig. 4A). The side chain of ciA-C2-L54 clashes with the aromatic ring of the second N-acetylglucosamine (NAG), while the main chain of ciA-C2-R57 overlaps with the beta-D-Mannose (BMA). Furthermore, the two hydrogen bonds formed between ciA-C2-E50 and H C A1-T1146 brought the FR2 loop of ciA-C2 up-close to the toxin, resulting in spatial conflicts with two β-sheets of gSV2C including residues T536-T538 and E556-K558 (Fig. 4B). Therefore, ciA-C2 simultaneously inhibits both the protein-and glycan-based interactions between BoNT/A1 and SV2C. These findings were further supported by an in vitro pull-down assay using H C A1 as a prey and bSV2C (non-glycosylated SV2C) and gSV2C as baits. ciA-C2 clearly eliminated H C A1-SV2C interactions regardless of SV2C's glycosylation state (Fig. 4C).
The neutralization mechanism of ciA-C2 is similar to that of a monoclonal antibody in clinical trials. We and other groups have shown that a potent BoNT/A-neutralizing human antibody currently in clinical trials, CR1/CR2, blocks BoNT/A-SV2 recognition 16,17,29 . The crystal structure of BoNT/A1 in complex with CR1-Fab (fragment antigen-binding) revealed two separated binding regions mediated by the variable regions in the heavy chain (V H ) and the light chain (V L ). We found that ciA-C2 almost completely occupies the V L -binding site on H C A1 (Fig. 5A). Surprisingly, even though ciA-C2 is only 1/4 size of CR1-Fab (and <10% of CR1-IgG), it occupies a binding interface as big as that of CR1 (1,126 Å 2 for CR1 vs. 1,150 Å 2 for ciA-C2). Moreover, ciA-C2 binds to both H CN A1 and H CC A1, while CR1 mostly targets H CN A1 (Fig. 5B). While CR1 only shows side-to-side clashes with the peptide portion of SV2C, residue E50 of ciA-C2 directly binds to a key SV2-binding residue T1146 in H C A1 (Fig. 5B), which forms three hydrogen bonds with F557 of human SV2C and two of them are formed via the atom OG1 of T1146 17 . Earlier studies showed that a T1145A-T1146A mutation almost completely blocked the binding between BoNT/A1 and SV2C 15, 17 . However, we suspected that mutating threonine, a strong sheet-forming residue, to an alanine may cause artificial local secondary structure changes on H C A1. Therefore, we designed a new H C A1 mutant, T1145V-T1146V, in which Thr-OG1 was replaced with Val-CG1 to abolish OG1-mediated interactions, while both threonine and valine are strong sheet formers. The wild-type and mutations of H C A1 were then subjected to a pull-down assay employing gSV2C (Fig. 5C) and bSV2C (data not shown) as baits. We found that even a single atom change in T1145 and T1146 was sufficient to abolish the H C A1-SV2C interaction, indicating this area of toxin is extremely important for receptor recognition and highly sensitive to even subtle conformational changes. Both T1145 and T1146 are strictly conserved in all eight known BoNT/A subtypes 30 . Thus, ciA-C2 neutralizes BoNT/A1 by targeting one of the most critical spots on the toxin-receptor interface, which was not observed for CR1/CR2.
BoNT/HA and various BoNT/A subtypes respond differently to ciA-C2 due to subtle residue changes. We next examined whether ciA-C2 could also recognize other natural variants of BoNT/A1, as BoNT genes are actively evolved and at least 40 different subtypes of BoNT have been reported 31 . We first analyzed a newly reported mosaic toxin type HA (BoNT/HA, also known as BoNT/FA or BoNT/H), which has a hybrid-like structure including a BoNT/A1-like H C [32][33][34][35][36] . Amino acid sequence alignments showed that the H C of BoNT/HA (H C HA) is ~84% identical to H C A1. The overall structure of H C HA and H C A1 are very similar with a RMS deviation of 0.526 Å over 341 Cα atoms 37 . Some BoNT/A-neutralizing antibodies that target the H C domain were able to neutralize this new toxin 38,39 . For example, a monoclonal antibody RAZ1 showed an almost identical binding affinity to BoNT/A1 and BoNT/HA. However, antibody CR1/CR2 showed a drastic ~540-fold decreased binding affinity for BoNT/HA when compared to BoNT/A1. In earlier studies, we found that residue N954 and  Fig. 3. gSV2C is colored in green while its N559 glycan is shown as pink sticks. GT1b is shown as purple sticks. Close-up views of the interface highlighted in the box is shown in (B), where the structures of H C A1-gSV2C and H C A1-ciA-C2 are shown in the left and the right panel, respectively. T536-T538, E556-N559 and the N559 glycan of gSV2C and the ciA-C2 residues S48-R57 that occupy the same space on H C A1 surface are shown as sticks. (C) ciA-C2 prevents H C A1 from binding to SV2C independent of its glycosylation state. Pull-down assay was performed using H C A1 as prey and the His-tagged bSV2C or gSV2C as baits. After binding, the Ni-NTA resins were washed twice, and the bound proteins were released from the resins and subjected to SDS-PAGE for visualization.
residues S955-S957 of the 3/10-helix of H C A1 form multiple hydrogen bonds with CR1/CR2 (through residues N96 and E97 of CR1), while these polar contacts are missing in H C HA that is very different from H C A1 in this region 37,38 . Furthermore, an important salt bridge between H C A1-R1294 and CR1-D30 is missing in H C HA due to an arginine to serine replacement (H C HA-S1286) 37 . Therefore, subtle residue changes on the toxin could lead to substantially different responses to the available antitoxins.
Sequence analyses showed that most of the ciA-C2-binding residues are conserved in BoNT/HA (Fig. S2) 17 . Nonetheless, ciA-C2 did not bind H C HA based on a pull-down assay and is apparently not able to neutralize BoNT/HA (Fig. 6B). Taking advantage of the structures of H C HA (PDB 5V38) 37 and the ciA-C2-H C A1 complex, we observed a potential clash and a repulsive charge interaction between H C HA-K895 and the CDR2 loop (K60-I63) of ciA-C2, whereas the equivalent H C A1-N905 did not (Fig. 6A). Can this unique amino acid change in BoNT/HA contribute to the inactivation of ciA-C2? We then designed two single-point mutants, H C HA-K895N and H C A1-N905K, where the two amino acids were swapped between the two toxins. The pull down assay demonstrated that ciA-C2 was able to recognize H C HA-K895N, whereas it displayed a decreased binding to H C A1-N905K by ~30% (n = 3), thus confirming our hypothesis (Fig. 6B).
Eight BoNT/A subtypes, termed BoNT/A1-A8, have been identified. The crystal structure of H C A2 in complex with the non-glycosylated SV2C-L4 (PDB 5MOY) was reported recently 40 . Structural comparison between H C A1 and H C A2 suggested that a replacement of H C A1-H1064 by R1064 in H C A2 may cause a decreased binding to ciA-C2, because H C A2-R1064 does not fit well into the H C A1-H1064-binding pocket and it clashes with L54 of ciA-C2 and may also repel a neighboring R57 in ciA-C2 (Fig. 6C). Sequence analyses among BoNT/A1-A8 (Fig. S2) showed that BoNT/A3, A6, and A8 also have a H C A2-like arginine at this position, thus likely weakens their binding to ciA-C2 as well. Furthermore, the replacement of a ciA-C2-interacting P1295 in BoNT/A1 with S1295 in BoNT/A2 (also exists in BoNT/A3, A7, and A8) may further weaken the binding to ciA-C2. Therefore, the structure of H C A2 further confirms the notion that BoNTs could potentially escape from the existing antitoxins via subtle residue changes 33 .

Discussion
BoNTs are large proteins that induce a wide variety of antibodies that recognize them at different sites and may or may not be able to neutralize the toxins. Even the neutralizing antibodies display vastly different potencies, largely dependent on their neutralizing mechanisms and their affinities. Therefore, a comprehensive understanding of the mechanisms for neutralizing antibodies is urgently needed for further therapeutic development. In this study we present the first high-resolution crystal structure of a BoNT/A1-neutralizing VHH, ciA-C2, in complex with the receptor-binding domain of the toxin. The strong neutralizing activity of ciA-C2 against BoNT/A1 was clearly demonstrated by the ex vivo MPN hemidiaphragm assay and cell-based assays using cultured primary neurons. Although ciA-C2 is small in size, it occupies a large, critical surface on H C A1 spanning both H CN A and H CC A subdomains, which is almost as big as that recognized by the human anti-BoNT/A IgG mAb, CR1/CR2. All three CDRs and three out of four FR regions of ciA-C2 interact with H C A1 through a network of hydrogen bonds and a cation-π interaction. Most importantly, ciA-C2 effectively prevents BoNT/A1 from recognizing SV2 by simultaneously blocking toxin binding to the peptide and the carbohydrate moiety of SV2.
Structural analyses of the ciA-C2-H C A1 complex also provide valuable guidance for rational evolution of ciA-C2 to target other BoNT/A subtypes or other closely related toxins such as the newly identified BoNT/HA. Despite the high sequence identity and similar overall structures between H C HA and H C A1, BoNT/HA effectively dampens the binding of ciA-C2 and CR1/CR2 by changing selective antibody-interacting residues. We believe that understanding the structural basis of antibody-toxin recognition will inform us in this host-pathogen arms race. As a proof of concept, we showed that a single point mutation in the toxin (H C HA-K895N), which was designed based on the crystal structure, successfully revived ciA-C2 in term of H C HA binding.
As potential therapeutic agents, BoNT-neutralizing VHHs have some unique features in comparison to the conventional antibodies. First, camelid VHHs have been shown to be poorly immunogenic in animals 41 and share a high degree of sequence identity with human V H , thus are expected to display low immunogenicity in human recipients. In contrast, the currently available antitoxin BAT is a mixture of equine-derived polyclonal IgG antibodies proteolytically processed to F(ab)2 to reduce immunogenicity. The immunogenicity of camelid VHHs could be further reduced by the well-developed humanization strategy 42 . Second, VHHs could be rapidly and economically produced using bacteria, a major advantage over the expensive and complicated IgG production using mammalian systems. Third, VHHs could be administered using injection, inhalation, or oral delivery to function in the gastrointestinal tract 43 , while antibodies typically need an intravenous delivery. For example, an anti-RSV (respiratory syncytial virus) VHH intended for administration through inhalation is currently in phase IIa clinical trials. Recently, a number of neutralizing VHHs against tetanus and Clostridium difficile toxins have also been developed [44][45][46][47] , largely motivated by the unique features of VHHs. VHHs have also been generated against other therapeutic targets, some are already in clinical trials, e.g. the anti-vWF (von Willebrand factor) VHH in phase III clinical trial (Ablynx). With their unique features added to the full antigen binding capacity and high specificity, VHHs are becoming promising components of new anti-BoNT antitoxins.

Methods
Study approval. All procedures using rat hippocampal neurons were conducted in accordance with the guidelines approved by the Institute Animal Care and Use Committee (IACUC) at Boston Children's Hospital (#3030). For the mouse phrenic nerve (MPN) hemidiaphragm assay and the rat brain synaptosome binding assay, animals were sacrificed by trained personnel before dissection of organs according to §4 Abs. 3 (killing of animals for scientific purposes, German animal protection law (TSchG)). Animal number is reported yearly to the animal welfare officer of the Central Animal Laboratory and to the local authority, Veterinäramt Hannover.
Plasmid construction. H C A1 (residues N872-L1296 of BoNT/A1) was cloned into expression vector pQE30 with an N-terminal His 6 -tag followed by a PreScission protease cleavage site. H C HA (residues E860-L1288 of BoNT/HA) was cloned into pGEX-6P-1 vector with an N-terminal glutathione S-transferase (GST) and a PreScission protease cleavage site. All point mutations for H C A1 and H C HA were generated with QuikChange site-directed mutagenesis (Agilent). Human SV2C-Loop4 (residues V473-T567) was cloned into two different vectors: pET28a vector for E. coli expression (bSV2C) and pcDNA vector for mammalian cell expression (gSV2C), as previously described 17 . SUMO (Saccharomyces cerevisiae S288c Smt3p) was introduced to the N-terminus of bSV2C to facilitate protein expression and folding. The VHHs ciA-C2 and ciA-F12 were cloned into the pET-32 vector downstream of the thioredoxin (Trx) tag and an enterokinase cleavage site.
Protein expression and purification. H C A1, H C HA, bSV2C and the two VHHs were expressed in E.
coli strain BL21-Star (DE3) (Invitrogen). Bacteria were cultured at 37 °C in LB medium containing appropriate selecting antibiotics. The temperature was set to 18 °C when OD 600 reached 0.4. For induction, IPTG (isopropyl-b-D-thiogalactopyranoside) with a final concentration of 0.2 mM was added to the culture when OD 600 reached 0.7. Protein expression was continued at 18 °C for 16 hours after induction. The cells were harvested by centrifugation.
Bacteria expressed proteins were first purified using Ni-NTA (Qiagen) affinity resins in a buffer containing 50 mM Tris-HCl, pH 8.0, 400 mM NaCl, and 40 mM imidazole. The bound His-tag proteins were eluted with a high-imidazole buffer (50 mM Tris-HCl, pH 8.0, 400 mM NaCl, and 300 mM imidazole) and then dialyzed at 4 °C against a buffer containing 20 mM HEPES-NaOH, pH 7.5, and 150 mM NaCl. The His-tags of H C A1 and H C HA were cleaved by PreScission protease. The Trx-tags of the VHHs were cleaved by enterokinase.
Tag-cleaved H C A1 and H C HA were further purified by MonoS ion-exchange chromatography (GE Healthcare) in a buffer containing 50 mM MES, pH 6.0, and eluted with a NaCl gradient. The peak fractions were then subjected to Superdex 200 size-exclusion chromatography (SEC, GE Healthcare) in a buffer containing 20 mM sodium phosphate, pH 6.0, and 50 mM NaCl. bSV2C and the VHHs were purified using Superdex 200 SEC in a buffer containing 20 mM HEPES-NaOH, pH 7.5, and 150 mM NaCl.
gSV2C was expressed and secreted from HEK 293 cells and purified directly from cell culture medium using Ni-NTA resins. gSV2C was eluted from the resins with a high concentration of imidazole and dialyzed against a buffer containing 50 mM Tris-HCl, pH 8.0, and 400 mM NaCl. gSV2C was then subjected to Superdex 200 SEC in a buffer containing 20 mM HEPES-NaOH, pH 7.5, and 150 mM NaCl.
Pull-down assay. Pull-down assay was performed using Ni-NTA resins in a buffer containing 50 mM Tris-HCl, pH 8.0, 400 mM NaCl, 10 mM imidazole, and 0.1% Tween-20. bSV2C-SUMO, gSV2C, or ciA-C2 were used as baits while H C A1 or H C HA variants as preys. Protein baits were pre-incubated with Ni-NTA resins at 4 °C for 1 hour, and the unbound proteins were washed away. The resins were equally divided into small aliquots with each has ~5 μg of bound protein baits, and then H C A1 or H C HA variants (~30 μg) were added. For experiment shown in Fig. 4C, ciA-C2 (~18 μg) was used at 2-fold molar excess over H C A1. The pull-down assay was carried out at 4 °C for 1.5 hours. The resins were washed twice before the addition of SDS sample buffer (with 300 mM imidazole), and analyzed by SDS-PAGE. The experiment was performed in triplicates and a representative gel photo was shown. Protein band intensity was quantified by densitometry.
Crystallization. Initial crystallization screens for the H C A1-ciA-C2 complex were carried out using a Phoenix crystallization robot (Art Robbins Instrument) with high-throughput crystallization screening kits (Hampton Research and Qiagen). Extensive manual optimization was then performed. Single crystals were grown at 20 °C by the hanging-drop vapor-diffusion method in a 1:1 (v/v) ratio of protein and the reservoir containing 100 mM sodium acetate, pH 5.0, and 18% Polyethylene glycol (PEG) 6,000. The crystals were cryoprotected in the original mother liquor supplemented with 20% (v/v) glycerol and flash-frozen in liquid nitrogen.
Data collection and structure determination. X-ray diffraction studies were performed at the Advanced Photon Source (APS) and the Stanford Synchrotron Radiation Lightsource (SSRL). The best data set were collected at APS and processed with XDS 48 . The structure of the H C A1-ciA-C2 complex was determined with Phaser 49 by molecular replacement using the structure of H C A1 (PDB: 3FUO) and ciA-F12 (PDB: 3V0A) as search models. The structural modeling and refinement were carried out iteratively using COOT 50 and Refmac from the CCP4 suite 51 . All refinement progress was monitored with the R free value with a 5% randomly selected test set 52 . The structure was validated through the MolProbity web server 53 . Data collection and structural refinement statistics are listed in Table 1 Antibodies and toxins. Mouse monoclonal antibodies for Syb (Cl69.1), SNAP-25 (Cl71.2), SV2 (pan-SV2) were generously provided by E. Chapman (Madison, WI) and are available from Synaptic Systems (Goettingen, Germany). BoNT/A1 and rabbit polyclonal antibodies for BoNT/A1 were generously provided by E. Johnson (Madison, WI).
Primary neuron culture, immunostaining, and immunoblot analysis. Dissociated rat hippocampal neurons were prepared from E18-19 embryos as described previously 54 . Briefly, dissected hippocampi were dissociated with papain following manufacture instructions (Worthington Biochemical, NJ). Cells were plated on poly-D-lysine coated glass coverslips and cultured in Neurobasal medium supplemented with B-27 (2%) and Glutamax (Invitrogen). Experiments were carried out using mature neurons (between 12-18 days in vitro). The high K + buffer contains (mM: NaCl 87, KCl 56, KH 2 PO 4 1.5, Na 2 HPO 4 8, MgCl 2 0.5, CaCl 2 1). Immunostaining was carried out by fixing cells with 4% paraformaldehyde, permeabilized with 0.3% Triton in PBS solution, and incubated with indicated primary antibodies for 1 h at room temperature, followed by incubation with secondary antibodies for 1 h at room temperature. Images were collected using a confocal microscope (Leica TCS SP5; 40× oil objective). Immunoblot analysis was carried out by lysing cultured neurons with RIPA buffer (50 mM Tris-HCl, 1% NP40, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS) plus a protease inhibitor cocktail (Sigma-Aldrich). Lysates were centrifuged for 10 min at maximum speed using a microcentrifuge at 4 °C. Supernatants were subjected to SDS-PAG and immunoblot analysis using the enhanced chemiluminescence (ECL) method (Pierce).
Mouse phrenic nerve (MPN) hemidiaphragm assay. The MPN assay was performed as described previously 14,55 . To limit the consumption of mice, the left and right phrenic nerve hemidiaphragms were excised from female mice of strain RjHan:NMRI (18-25 g, Janvier, St Berthevin Cedex, France) and placed in an organ bath containing 4 mL of Earle's Balanced Salt Solution. The pH was adjusted to 7.4, and oxygen saturation was achieved by gassing with 95% O 2 and 5% CO 2 . The phrenic nerve was continuously electro-stimulated at a frequency of 1 Hz via two ring electrodes. The pulse duration was 0.1 ms and the current was 25 mA to achieve maximal contraction amplitudes. Isometric contractions were recorded with a force transducer (Scaime, Annemasse, France) and the software VitroDat (Föhr Medical Instruments GmbH (FMI), Seeheim, Germany). The resting tension of the hemidiaphragm was approximately 10 mN. In each experiment, the preparation was first allowed to equilibrate for 15 min under control conditions. Then, the buffer was changed to 4 mL of Earle's Balanced Salt Solution supplemented with 0.1% BSA and the BoNT/A1-VHH-containing solution. To determine neutralization by the VHH, a constant BoNT/A1 concentration of 1.63 pM yielding a paralysis time t ½ of 53.7 min in the absence of VHH was incubated for 15 min at 37 °C with increasing concentrations of VHH. The previously reported calibration curve determined for BoNT/A1 (y (BoNT/A1; 0.408/0.81/1.63 pM) = −16.062Ln(x) + 60.648, R 2 = 0.99) 56 was used to calculate the concentration of non-neutralized BoNT/A1. Difference of applied and non-neutralized BoNT/A1 calculates the amount of neutralization.
Binding of H C A1 to synaptosomes. Rat brain synaptosomes were freshly prepared and H C fragment of BoNT/A1 ( 35 S-H C A1) was radiolabeled by in vitro transcription/translation as previously described 14 . 35 S-H C A1 was bound in the presence of 200 nM nM VHH to synaptosomes in a total volume of 100 µL of physiological buffer (140 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 20 mM HEPES-NaOH, 10 mM glucose, 0.5% BSA, pH 7.4) with the final synaptosomal protein adjusted to a concentration of 10 mg/ml for 2 h at 4 °C. Controls were performed with samples lacking synaptosomes and/or VHH. After incubation, synaptosomes were collected by centrifugation (5,000 × g; 5 min) and unbound material in the supernatant fraction was discarded. The pellet fractions were washed two times each with 100 µL of physiological buffer and incubated for 20 min at 37 °C in SDS sample buffer and subsequently subjected to SDS-PAGE and autoradiography. Bound 35 S-H C A1 was quantified with the Advanced Image Data Analyzer (AIDA 2.11) software (Raytest, Straubenhardt, Germany) and calculated after subtraction of the value obtained for control sample in the absence of synaptosomes as the percentage of the total 35 S-H C A1 in the assay.

SNAP-25 endopeptidase assay.
For SNAP-25 endopeptidase assays, 10 nM BoNT/A1 was pre-incubated for 15 min at 37 °C with 30 nM VHH and subsequently added to 3 µM of rSNAP-26H6 in a total volume of 120 µL in toxin assay buffer (20 mM HEPES-KOH, pH 7.2, 150 mM K-Glu, 10 mM DTT, and 0.2 mM ZnCl 2 ) at 37 °C for 1 h. Reactions were stopped by mixing with 4-fold Laemmli buffer. Samples were boiled for 2 min and subjected to 15% SDS-PAGE. Proteins were visualized by staining with Coomassie blue. Data Availability. Coordinates and structure factors for the H C A1-ciA-C2 complex have been deposited in the Protein Data Bank under accession code 5L21.