Structural Features of Clostridium botulinum Neurotoxin Subtype A2 Cell Binding Domain

Botulinum neurotoxins (BoNT) are a group of clostridial toxins that cause the potentially fatal neuroparalytic disease botulism. Although highly toxic, BoNTs are utilized as therapeutics to treat a range of neuromuscular conditions. Several serotypes (BoNT/A-/G, /X) have been identified with vastly differing toxicological profiles. Each serotype can be further sub-categorised into subtypes due to subtle variations in their protein sequence. These minor changes have been attributed to differences in both the duration of action and potency for BoNT/A subtypes. BoNTs are composed of three domains—a cell-binding domain, a translocation domain, and a catalytic domain. In this paper, we present the crystal structures of the botulinum neurotoxin A2 cell binding domain, both alone and in complex with its receptor ganglioside GD1a at 1.63 and 2.10 Å, respectively. The analysis of these structures reveals a potential redox-dependent Lys-O-Cys bridge close to the ganglioside binding site and a hinge motion between the HCN and HCC subdomains. Furthermore, we make a detailed comparison with the previously reported HC/A2:SV2C structure for a comprehensive structural analysis of HC/A2 receptor binding.


Introduction
Botulinum neurotoxins (BoNTs), although highly toxic, are now routinely used as therapeutics with over 100 medicinal applications [1]. They function by cleaving soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins, halting the release of acetylcholine at the neuromuscular junction (NMJ) resulting in flaccid paralysis [2]. This makes them exceptional candidates for the treatment of a range of neuromuscular disorders, and advancements in biotechnology have resulted in the further expansion of their therapeutic potential [3,4]. Several immunologically distinct BoNT serotypes produced by Clostridium botulinum (BoNT/A-/G, /X) have been identified [5], with a growing number of subtypes (e.g., BoNT/A1, /A2, and /A3) that arise due to subtle variations in amino acid sequences [6]. Furthermore, BoNT-like molecules have also been identified in non-clostridial species, such as Weissella oryzae [7], Enterococcus faecium [8,9], and Paraclostridium bifermentans [10]. The substantial number and varying toxicological profiles of BoNT serotypes and subtypes provide opportunities of developing fine-tuned medicines for specific applications.

Structure of H C /A2
The structure of H C /A2 was determined to a resolution of 1.63 Å (Table 1) by molecular replacement. It has an N-terminal β-jelly roll fold and a C-terminal β-trefoil ( Figure 1A) that is consistent with other H C /A subtypes [25][26][27]. The quality of the electron density map is good throughout except for a small loop region (Arg 1269-Phe 1277). This loop (which is conserved across all BoNT/A subtypes) precedes the ganglioside binding site (GBS) and appears to be disordered for other H C /A subtypes [28,29]. This is likely due to the inherent flexibility of this loop so that it can accommodate ganglioside binding. The structure of H C /A2:GD1a complex was solved by molecular replacement to 2.1 Å (Table 1) with two molecules (A and B) in the asymmetric unit ( Figure 1B). Clear positive difference map electron density was observed at the GBS of molecule B that could be readily modelled as GD1a, whereas for molecule A, the GBS is inaccessible to GD1a due to crystal packing. Therefore, the H C /A2:GD1a asymmetric unit contains both the GD1a-bound and unbound states of H C /A2. A total of 5/6 monosaccharides are clearly defined by the electron density ( Figure 1C). GD1a forms a total of 10 hydrogen bonding interactions to H C /A2, two of which are water-mediated ( Figure 2). This binding mode is identical to what was observed in H C /A3, with the exception of Trp 1266, which binds Sia 6 in H C /A2:GD1a, but is unmodelled in H C /A3:GD1a [28].
The superimposition of molecule 'A' and 'B' (for C α atoms) yields an RMSD of 0.47 Å, indicating that the overall structure of the molecule does not change upon binding GD1a. However, the loop spanning residues 1269-1277 appears to widen upon GD1a binding-the distance between the C α atoms of residues 1269 and 1277 is~4.5 Å greater in molecule B than that in molecule A ( Figure 1D). This change in loop positioning is accompanied by the rotation of Phe 1278 towards the GBS, a feature that has been observed previously in the H C /A3:GD1a, H C /A4:GD1a, and H C /A5:GM1b structures [26,28]. This residue, along with Phe 1117 and Phe 1252, forms a hydrophobic pocket occupied by Sia 5 ( Figure 1E).

H C /A2 Is Primed for Receptor Binding
The conformational changes that occur with H C /A2 upon the binding of GD1a are highlighted in Figure 3. The residues Phe 1252 and Phe 1272 moved towards the GBS and contribute to a hydrophobic patch, while the movement of His 1253 can be attributed to the formation of a hydrogen bond with Gal 4 ( Figure 3B). Compared to a previous structure of H C /A2 in complex with SV2C (PDB: 5MOY [23]), the superimposition of C α atoms gave an RMSD of 0.96 Å, indicating that the structure does not drastically alter conformation upon SV2C binding ( Figure 4A). There are few changes that occur at the residue level across the H C /A2:SV2C interface upon binding ( Figure 4B,C); however, there are noticeable differences in the conformation of two loop regions at residues 1164-1172 and 1225-1236. For the former loop, the differences are due to crystal packing, whereas for the latter loop, the variation is likely due to the inherent flexibility as residues 1228 and 1229 could not be modelled. Similar findings were also observed with another reported structure of H C /A2:SV2C (PDB: 6ES1 [24]) (not shown). Taken together, the analysis of the ganglioside and SV2C binding sites suggest H C /A2 is primed to bind its receptors, requiring minimal conformational change.

HC/A2 Is Primed for Receptor Binding
The conformational changes that occur with HC/A2 upon the binding of GD1a are highlighted in Figure 3. The residues Phe 1252 and Phe 1272 moved towards the GBS and contribute to a hydrophobic patch, while the movement of His 1253 can be attributed to the formation of a hydrogen bond with Gal 4 ( Figure 3B). Compared to a previous structure of HC/A2 in complex with SV2C (PDB: 5MOY [23]), the superimposition of Cα atoms gave an RMSD of 0.96 Å, indicating that the structure does not drastically alter conformation upon SV2C binding ( Figure 4A). There are few changes that occur at the residue level across the HC/A2:SV2C interface upon binding ( Figure 4B,C); however, there are noticeable differences in the conformation of two loop regions at residues 1164-1172 and 1225-1236. For the former loop, the differences are due to crystal packing, whereas for the latter loop, the variation is likely due to the inherent flexibility as residues 1228 and 1229 could not be modelled. Similar findings were also observed with another reported structure of HC/A2:SV2C (PDB: 6ES1 [24]) (not shown). Taken together, the analysis of the ganglioside and SV2C binding sites suggest HC/A2 is primed to bind its receptors, requiring minimal conformational change.

HC/A2 Is Primed for Receptor Binding
The conformational changes that occur with HC/A2 upon the binding of GD1a are highlighted in Figure 3. The residues Phe 1252 and Phe 1272 moved towards the GBS and contribute to a hydrophobic patch, while the movement of His 1253 can be attributed to the formation of a hydrogen bond with Gal 4 ( Figure 3B). Compared to a previous structure of HC/A2 in complex with SV2C (PDB: 5MOY [23]), the superimposition of Cα atoms gave an RMSD of 0.96 Å, indicating that the structure does not drastically alter conformation upon SV2C binding ( Figure 4A). There are few changes that occur at the residue level across the HC/A2:SV2C interface upon binding ( Figure 4B,C); however, there are noticeable differences in the conformation of two loop regions at residues 1164-1172 and 1225-1236. For the former loop, the differences are due to crystal packing, whereas for the latter loop, the variation is likely due to the inherent flexibility as residues 1228 and 1229 could not be modelled. Similar findings were also observed with another reported structure of HC/A2:SV2C (PDB: 6ES1 [24]) (not shown). Taken together, the analysis of the ganglioside and SV2C binding sites suggest HC/A2 is primed to bind its receptors, requiring minimal conformational change.

Hinge Motion between the H CN and H CC Subdomain
The superimposition of H C /A2:GD1a and H C /A2:SV2C structures with the H C /A2 structure revealed a slight misalignment across the entirety of the C α trace (RMSD values of 0.79 Å and 0.99 Å, respectively). The superimposition of just the C-terminal subdomains (H CC ) showed a nearly identical alignment, with the N-terminal subdomain (H CN ) rotated out of position relative to each other ( Figure 5). Using DynDom (which estimates domain motions in proteins) [30], the H CN /A2 subdomain appears to rotate 3.6 • in one direction when bound to GD1a ( Figure 5A), but 6.7 • in the other direction when bound to SV2C ( Figure 5B). This hinge-like motion is not believed to facilitate SV2C receptor binding because it does not alter the position of the key binding residues. However, considering that the H CC subdomain is responsible for anchoring BoNT to the NMJ cell membrane, it is possible that the hinge between H CC and H CN may aid in the orientation of the H N and LC towards the membrane in preparation for translocation, which is consistent with what has been proposed previously [31].

Hinge Motion between the HCN and HCC Subdomain
The superimposition of HC/A2:GD1a and HC/A2:SV2C structures with the HC/A2 structure revealed a slight misalignment across the entirety of the Cα trace (RMSD values of 0.79 Å and 0.99 Å, respectively). The superimposition of just the C-terminal subdomains (HCC) showed a nearly identical alignment, with the N-terminal subdomain (HCN) rotated out of position relative to each other ( Figure 5). Using DynDom (which estimates domain motions in proteins) [30], the HCN/A2 subdomain appears to rotate 3.6° in one direction when bound to GD1a ( Figure 5A), but 6.7° in the other direction when bound to SV2C ( Figure 5B). This hinge-like motion is not believed to facilitate SV2C receptor binding because it does not alter the position of the key binding residues. However, considering that the HCC subdomain is responsible for anchoring BoNT to the NMJ cell membrane, it is possible that the hinge between HCC and HCN may aid in the orientation of the HN and LC towards the membrane in preparation for translocation, which is consistent with what has been proposed previously [31].

Lys 1236-X-Cys 1280 Bridge near the GBS
In both the HC/A2 and HC/A2:GD1a structures, clear electron density was observed between Lys 1236 and Cys 1280 ( Figure 6A,B) indicating the presence of an unusual covalent bridge between the two residues. A similar observation was made in the structure of HC/A5 involving equivalent residues [27]. The Lys-X-Cys bridge is a recently reported interaction that may be a widespread phenomenon in many protein structures [32,33]. There are two possible bridging atoms where X is either an O or a C (as a methylene group, CH2). The formation of an -O-bridge occurs via the spontaneous oxidation of the cystine sulfhydryl group in the absence of reducing agents, whereas the formation of a -CH2bridge has been suggested to occur by the reaction of a Lys with CO2 or CH2O [32,34].

Lys 1236-X-Cys 1280 Bridge near the GBS
In both the H C /A2 and H C /A2:GD1a structures, clear electron density was observed between Lys 1236 and Cys 1280 ( Figure 6A,B) indicating the presence of an unusual covalent bridge between the two residues. A similar observation was made in the structure of H C /A5 involving equivalent residues [27]. The Lys-X-Cys bridge is a recently reported interaction that may be a widespread phenomenon in many protein structures [32,33]. There are two possible bridging atoms where X is either an O or a C (as a methylene group, CH 2 ). The formation of an -O-bridge occurs via the spontaneous oxidation of the cystine sulfhydryl group in the absence of reducing agents, whereas the formation of a -CH 2 -bridge has been suggested to occur by the reaction of a Lys with CO 2 or CH 2 O [32,34].
Both possibilities were modelled into the electron density of the H C /A2 structure and refined ( Figure 7A,B). However, it was not possible to determine the identity of the bridging atom; crystallographically, both were equally possible ( Figure 7A-C). This illustrates the difficulty in determining the precise nature of the bridging atom, and there appears to be a divided debate on this topic [35]. Both possibilities were modelled into the electron density of the HC/A2 structure and refined ( Figure 7A,B). However, it was not possible to determine the identity of the bridging atom; crystallographically, both were equally possible ( Figure 7A-C). This illustrates the difficulty in determining the precise nature of the bridging atom, and there appears to be a divided debate on this topic [35].
On the other hand, the previously reported crystal structures of HC/A2 bound to SV2C showed either the formation of a disulphide bond between Cys 1280 and Cys 1235 instead ( Figure 6C), or no interaction between the two residues ( Figure 6D). Interestingly, both of these structures (PDB: 6ES1 and 5MOY, respectively) were determined from crystals grown under reducing conditions, in contrast to the present structure of HC/A2 where crystals were grown under non-reducing conditions. This indicates that the formation of the Lys-X-Cys bridge may be dependent on the redox environment. This type of redoxdependency has been observed previously in the transaldolase enzyme from the Neisseria gonorrhoeae bacterium, where a Lys-O-Cys bridge serves as an allosteric redox switch [36]. On the other hand, the previously reported crystal structures of H C /A2 bound to SV2C showed either the formation of a disulphide bond between Cys 1280 and Cys 1235 instead ( Figure 6C), or no interaction between the two residues ( Figure 6D). Interestingly, both of these structures (PDB: 6ES1 and 5MOY, respectively) were determined from crystals grown under reducing conditions, in contrast to the present structure of H C /A2 where crystals were grown under non-reducing conditions. This indicates that the formation of the Lys-X-Cys bridge may be dependent on the redox environment. This type of redoxdependency has been observed previously in the transaldolase enzyme from the Neisseria gonorrhoeae bacterium, where a Lys-O-Cys bridge serves as an allosteric redox switch [36].
Further, a recent report based on a systematic study on the presence of Lys-Cys bridges in protein structures revealed that oxygen is the most likely bridging atom [33]; therefore, we modelled this atom in the final deposited coordinates.
Although the biological relevance of this Lys 1236-O-Cys 1280 bridge in the present H C /A2 structure is currently unknown, it is interesting to note that both these equivalent residues are conserved in all subtypes of BoNT/A, and that they are situated in a dynamic region of the protein close to the GBS. Cys 1280 is located close to Phe 1277, whose sidechain is known to flip orientation upon ganglioside binding, and Lys 1236 is positioned within a β-hairpin (residues 1220-1240) that appears to possess a flexible loop based on the lack of electron density for residues 1224-1236 ( Figure 8). Further investigation will be required to confirm the identity and biological function (if any) of this Lys-O-Cys bridge. Further, a recent report based on a systematic study on the presence of Lys bridges in protein structures revealed that oxygen is the most likely bridging atom therefore, we modelled this atom in the final deposited coordinates.
Although the biological relevance of this Lys 1236-O-Cys 1280 bridge in the pre HC/A2 structure is currently unknown, it is interesting to note that both these equiv residues are conserved in all subtypes of BoNT/A, and that they are situated in a dyn region of the protein close to the GBS. Cys 1280 is located close to Phe 1277, w sidechain is known to flip orientation upon ganglioside binding, and Lys 1236 is tioned within a β-hairpin (residues 1220-1240) that appears to possess a flexible based on the lack of electron density for residues 1224-1236 ( Figure 8). Further inves tion will be required to confirm the identity and biological function (if any) of this Ly Cys bridge.

Conclusions
The crystal structures of HC/A2 alone and in complex with GD1a reveal a total of 9 residues that form 10 hydrogen bonding interactions with the sugar moiety, accompanied by a conformational change of a loop (residues 1269-1277) located near the GBS. Furthermore, structural comparison with HC/A2 bound to its protein receptor, SV2C, revealed

Conclusions
The crystal structures of H C /A2 alone and in complex with GD1a reveal a total of 9 residues that form 10 hydrogen bonding interactions with the sugar moiety, accompanied by a conformational change of a loop (residues 1269-1277) located near the GBS. Furthermore, structural comparison with H C /A2 bound to its protein receptor, SV2C, revealed features not previously reported in the literature. The H CN and H CC subdomains appear to rotate about a common hinge position depending on which receptor molecule H C /A2 binds-ganglioside or SV2C. This motion may be involved in orienting the translocation domain towards the cell surface following dual-receptor-initiated endocytosis. We also note the presence of a Lys 1236-O-Cys 1280 bridge in the two crystal structures of H C /A2 presented here that is located on a loop near the GBS. The biological significance of both the hinge and Lys-O-Cys bridge is unknown and requires further investigation. This information might be valuable in the bioengineering and manufacture of BoNT/A subtypes for enhanced therapeutic applications.

Expression and Purification of H C /A2
The pJ401 vector containing H C /A2 (BoNT/A2 residues 871-1296) was transformed into One Shot BL21 (DE3) Star competent cells (Thermo Fisher Scientific, Loughborough, UK), as previously described [37]. Cultures were grown in TB at 37 • C until an OD 600 of 0.6, and protein expression was induced with 1 mM IPTG at 16 • C for at least 16 h. Cells were lysed in 50 mM Tris pH 7.4, 20 mM imidazole, and 0.5 M NaCl. H C /A2 was captured by Ni 2+ affinity chromatography and eluted with 0.5 M imidazole in 50 mM Tris pH 7.4, 0.5 M NaCl via a gradient elution. H C /A2 was further purified by gel filtration using a superdex 200 column into a final buffer of 50 mM Tris pH 7.4, and 150 mM NaCl. The purified protein was flash frozen in liquid nitrogen for storage at −20 • C until required for crystallisation.

X-ray Crystallography
Crystals of H C /A2 and H C /A2:GD1a were grown using the sitting drop vapour diffusion method at concentrations of 17 mg/mL and 8 mg/mL, respectively. For the latter, H C /A2 was incubated with 5 mM GD1a for at least 1 h prior to setting up crystallisation screens. H C /A2 crystals grew at 16 • C in 0.1 M Sodium acetate, pH 4.5, 22% v/v PEG smear broad (4.55% PEG 400, 4.55% PEG 500 MME, 4.55% PEG 600, 4.55% PEG 1000, 4.55% PEG 2000, 4.55% PEG 3350, 4.55% PEG 4000, 4.55% PEG 5000, 4.55% PEG 6000, 4.55% PEG 8000, and 4.55% PEG 10,000), whereas H C /A2:GD1a crystals grew at 16 • C in 0.2 M lithium citate tribasic tetrahydrate, 20% w/v PEG 3350. Crystals were mounted into a cryoloop and flash frozen in liquid nitrogen. Diffraction data were collected on I04 beamline at Diamond Light Source (Oxon, UK). A total of 7200 images were collected at 0.1 • oscillation with exposure times of 0.01 s, for both H C /A2 and H C /A2:GD1a crystals. Data processing was carried out in DIALS [38] and both structures were determined by molecular replacement using PHASER [39] as part of the CCP4 package [40]. The H C /A2:SV2C (PDB: 5MOY) structure (excluding the coordinates for SV2C) was used as a search model for H C /A2 [23], and the refined H C /A2 structure was subsequently used as a search model for H C /A2:GD1a. Both structures were refined using REFMAC [41] and Phenix [42], with modelling performed in COOT [43]. The structures were validated using Molprobity [44] and PDB validation [45]. Figures were produced using CCP4mg [46].

Informed Consent Statement: Not applicable.
Data Availability Statement: The atomic coordinates and structure factors of H C /A2 and H C /A2:GD1a were deposited in the protein data bank under accession codes 7Z5T and 7Z5S, respectively.