REVIEW ARTICLE Molecular structures and functional relationships in clostridial neurotoxins Subramanyam Swaminathan Department of Biology, Brookhaven National Laboratory, Upton, NY, USA Introduction Clostridium botulinum, an anaerobic bacterium, pro- duces seven antigenically distinct neurotoxins com- monly called botulinum neurotoxins (BoNT, A–G) [1]. These neurotoxins are among the most poisonous known and have an LD 50 of 1–5 ngÆkg )1 weight of humans [2]. Botulinum neurotoxins are closely related to tetanus neurotoxins (TeNT) produced by Clostri- dium tetani. However, their sites of action and pharma- cological effects are different [3,4]. BoNTs cause flaccid paralysis by inhibiting acetycholine release at Keywords botulinum neurotoxin; botulism; catalytic activity; drug discovery; neuroexocytosis; structure–function; substrate–enzyme complex; tetanus; translocation; X-ray crystallography; zinc endopeptidase Correspondence S. Swaminathan, Biology Department, Brookhaven National Laboratory, Upton, NY 11973, USA Fax: 1 631 344 3407 Tel: 1 631 344 3187 E-mail: swami@bnl.gov (Received 28 February 2011, revised 11 May 2011, accepted 17 May 2011) doi:10.1111/j.1742-4658.2011.08183.x The seven serotypes of Clostridium botulinum neurotoxins (A–G) are the deadliest poison known to humans. They share significant sequence homo- logy and hence possess similar structure–function relationships. Botulinum neurotoxins (BoNT) act via a four-step mechanism, viz., binding and inter- nalization to neuronal cells, translocation of the catalytic domain into the cytosol and finally cleavage of one of the three soluble N-ethylmaleimide- sensitive factor attachment protein receptors (SNARE) causing blockage of neurotransmitter release leading to flaccid paralysis. Crystal structures of three holotoxins, BoNT ⁄ A, B and E, are available to date. Although the individual domains are remarkably similar, their domain organization is dif- ferent. These structures have helped in correlating the structural and func- tional domains. This has led to the determination of structures of individual domains and combinations of them. Crystal structures of catalytic domains of all serotypes and several binding domains are now available. The cata- lytic domains are zinc endopeptidases and share significant sequence and structural homology. The active site architecture and the catalytic mecha- nism are similar although the binding mode of individual substrates may be different, dictating substrate specificity and peptide cleavage selectivity. Crystal structures of catalytic domains with substrate peptides provide clues to specificity and selectivity unique to BoNTs. Crystal structures of the receptor domain in complex with ganglioside or the protein receptor have provided information about the binding of botulinum neurotoxin to the neuronal cell. An overview of the structure–function relationship correlating the 3D structures with biochemical and biophysical data and how they can be used for structure-based drug discovery is presented here. Abbreviations BoNT, botulinum neurotoxin; SNAP-25, synaptosome-associated protein 25 kDa; SNARE complex, soluble N-ethylmaleimide-sensitive factor attachment protein receptors; SV, synaptic vesicle; Syt, synaptotagmin; TeNT, tetanus neurotoxin; VAMP, vesicle-associated membrane protein. FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works 4467 the neuromuscular junction, whereas TeNT blocks the release of c-aminobutyric acid and glycine at the inhib- itory neurons of the spinal chord causing spastic paral- ysis. BoNTs are produced as single 150 kDa polypeptide chains and are proteolytically cleaved before release into two chains, a heavy chain (HC) of 100 kDa and a light chain (LC) of 50 kDa, linked by a single disulfide bond [5]. Most of the BoNTs are released as dichains, cleaved by either endogenous or exogenous proteases. In general, dichain BoNTs are more potent than single-chain BoNTs. BoNT ⁄ A, B, E and possibly F are the common source of human infection [6]. BoNT ⁄ C and BoNT ⁄ D are responsible for avian and animal infections [7,8]. BoNTs consist of three functional domains. The HC is made up of two distinct domains, the N-termi- nal (H N ) and C-terminal (H C ) domains, of almost equal molecular mass. H C consists of two subdo- mains, the N-terminal half, H CN and the C-terminal half H CC , each of 25 kDa. BoNTs toxicity is via a four-step process, binding and internalization to neu- ronal cell, translocation of the LC into cytosol and finally the cleavage of one the soluble N-ethylmalei- mide-sensitive factor attachment protein receptor (SNARE) proteins at specific sites [9]. The H C , and especially the H CC subdomain, is responsible for bind- ing to presynaptic neuronal target cells, H N , for translocation of the catalytic domain (LC) into the cytosol. The catalytic domain is a zinc endopeptidase and possesses a conserved zinc-binding HExxH+E motif in all BoNTs. Each BoNT has a specific target in the SNARE complex and cleaves a specific peptide bond. BoNT⁄ A and E cleave synaptosome-associated 25 kDa protein (SNAP-25) at a specific peptide bond. BoNT ⁄ B, D, F and G (and also TeNT) cleave vesicle-associated membrane protein (VAMP), also known as synaptobrevin. BoNT ⁄ C is unique in that it cleaves both SNAP-25 and syntaxin [1]. Large sub- strate peptides and specific scissile bonds are unique to BoNTs. The crystal structures of holotoxins BoNT ⁄ A, B and E have been determined and have given insight into the function and mechanism of each domain involved in the four-step process [10–12]. BoNTs share significant sequence homology [13] and the structures were expected to be similar. Indeed, individual domains are similar although E differs from A and B in the domain organization. The crystal structures of individual domains and their complexes with substrates or binding partners give information to analyze and understand the structure–function relationships. This review deals with the structure–function relationship of each individual domain as well as the holotoxin. Crystal structures of BoNT ⁄ A and B Crystal structures of BoNT ⁄ A and B have been deter- mined [11,12]. BoNT ⁄ A and B share significant sequence homology (39% identity and 56% similarity) [13] resulting in structural similarity. Because A and B have similar folds, we describe BoNT ⁄ Basitisa higher resolution structure (1.8 A ˚ ). BoNT ⁄ B consists of three distinct structural domains corresponding to catalytic, translocation and binding domains (Fig. 1). The catalytic domain (LC) has an a ⁄ b fold, the trans- location domain (H N ) is mostly helical with two long helices ( 100 A ˚ long) forming a coiled-coil. A large loop, corresponding to residues 481–532 and called the belt region, wraps around the catalytic domain. This region corresponds to the translocation domain in the primary sequence although it is closely associated with the catalytic domain in three dimensions and is an intriguing feature unique to BoNTs. The binding domain consists of two subdomains, H CN and H CC . H CN consists of a 14-stranded b-barrel in a jelly- roll motif, commonly associated with lectin-binding 4411290 852 TDBD S 0441 CD S A B Fig. 1. Clostridium botulinum neurotoxin type B. (A) Linear repre- sentation of BoNT ⁄ B with the individual domains colored as in (B). The interchain disulfide bond is also marked. (B) Ribbon representa- tion of BoNT ⁄ B. BoNT ⁄ A and BoNT ⁄ B are similar in fold and domain organization. The three functional domains, receptor binding (BD), translocation (TD) and catalytic (CD) domains are represented in orange, green and red, respectively. Zinc is shown as cyan ball. The belt region (also in green) wraps around CD. The N- and C-ter- minals are marked. Molecular structures of Clostridial neurotoxins S. Swaminathan 4468 FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works proteins [14]. The H CC domain is mostly made up of loops and b strands with a b-trefoil fold [15]. The three domains are arranged in a linear fashion with the translocation domain in the middle. The binding domain is tilted away from the translocation domain and has only limited interactions with it. The catalytic domain, which is on the other side of the translocation domain, is closely associated with it. In BoNT ⁄ AorB the binding domain and catalytic domains have virtu- ally no contact. The catalytic zinc is located deep inside a wide cavity that is partly covered by the belt region. The cavity in BoNT ⁄ B is wider than that in BoNT ⁄ A. In the following sections, individual domains and their functions, as elicited from the 3D structure and corroborated by biochemical and biophysical results (or vice versa), are discussed. Receptor-binding domain The first step in botulinum toxicity is for the toxin to bind to the presynaptic membrane of the neuronal tar- get cell for uptake into neuronal cell. As early as the 1980s, it was shown that the H C domain is involved in neuronal cell binding [16,17]. Also, neuraminidase-trea- ted cultured cells had reduced affinity for BoNT ⁄ A and bovine chromaffin cells lacking in polysialogan- gliosides became sensitive to BoNT ⁄ A when pretreated with gangliosides [18–20]. Taken together, it was clear that H C domains of BoNTs bind to the neuronal cell via gangliosides. The presynaptic cell surface is rich in gangliosides which first bind to the toxin and then accumulate them on the neuronal surface. Gangliosides are low-affinity but highly abundant lipids with com- plex sugar molecules as head groups. Botulinum neu- rotoxins in general bind to GT1b, GD1b and GD1a which contain charged sialic acids [21–23]. As described earlier, the receptor-binding domain consists of two subdomains (Fig. 2). The N-terminal (H CN ) and the C-terminal (H CC ) domains comprising a jelly-roll motif and a b-trefoil fold, respectively, are connected by a short helix [14,15]. Binding domains of all botulinum and tetanus toxins share a similar fold, even though the sequence identity of the C-terminal domain is low. The variation in sequence is reflected in the length of the connecting loops. The function of the N-terminal domain (H CN ) is not yet understood. Even though it has a carbohydrate-binding fold there is no evidence that it binds to any ganglioside sugar group. However, recent evidence has shown that BoNT ⁄ A H CN interacts weakly with phosphatidylionositol phosphates [24]. Mutations in the C-terminal half of tetanus-binding domain affect ganglioside binding and the 34 residues (1281–1314) at the C-terminus are enough for ganglio- side binding in TeNT [25,26]. Photoaffinity labeling occurred predominantly at His1292 of TeNT and tryp- tophan fluorescence quenching experiments on ganglio- side binding implicated tryptophans at the C-terminus in ganglioside binding [27]. This biochemical evidence established the importance of H C for ganglioside bind- ing. Crystal structures of BoNTs and TeNT with com- plex sugars and GT1b analogs have confirmed this and have mapped the GT1b-binding pocket [12,28,29]. Whereas a single ganglioside-binding site was observed in BoNT ⁄ B, structure determination of TeNT with sugars showed two binding sites (Site 1 and Site 2) for TeNT. Also, when an analog of GT1b was used in cocrystallization, the branched sugar molecule cross- linked two TeNT molecules via the two sites. The crystal structure of BoNT ⁄ B in complex with sialyllactose identified the ganglioside-binding site in BoNT ⁄ B [12]. Sialyllactose is a partial mimic of GT1b and occupies a pocket in the H CC domain (Fig. 3A). The pocket is formed by residues His1240, Ser1259, Trp1261 and Tyr1262 of the conserved motif H CN H CC Fig. 2. The receptor binding domain of BoNT ⁄ B. The H CN domain has lectin binding motif and the H CC domain contains a b-trefoil fold which provides binding pockets for the receptors. S. Swaminathan Molecular structures of Clostridial neurotoxins FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works 4469 H…SxWY…G in BoNTs. Residues Glu1188 and Glu1189 also take part in forming the binding pocket. The sialic acid sits between His1240 and Trp1261. The stacking interaction between Trp1261 and the sialic acid seems to be critical for strong binding. The trisac- charide molecule forms an extensive hydrogen-bonding network with the residues forming the pocket. The res- idues forming this pocket are structurally and sequen- tially similar in BoNT ⁄ A and TeNT suggesting that the ganglioside-binding site will be similar in these tox- ins. This pocket is called Site 1 in this review (also referred to as the lactose-binding site) [21,28]. Later, the crystal structure of TeNT with a GT1b analog (GT1b-b) was determined [29]. Gal–GalNAc moiety of GT1b-b occupies the sialyllactose site of BoNT ⁄ B. However, the other branch – the disiayllac- tose moiety (GD3 part) – binds to an adjacent pocket, called Site 2 in this review (also referred to as the sia- lic-acid-binding site) [21,28]. This pocket is made up of Asp1147, Asp1214, Asn1216, Arg1226 and Tyr1229. Mutational analyses have confirmed that these residues are important for GT1b binding [30]. In the crystal structure, GT1b-b links two molecules via Sites 1 and 2. However, this cross-linking may be an artifact of: (a) crystal packing, and (b) the b2–3 linkage (different from the a2–3 linkage in GT1b) of the disialic acid arm to the central galactose unit. Later mutational analysis and binding studies on TeNT have shown that cross-linking does not take place in solution. The same studies also proved that, whereas there are two Sialyllactose GD3 Sialyllactose Tripeptide Syt IIGT1b AB CD Fig. 3. Binding domain and receptors. (A) The sialyllactose binding site in the H CC domain of BoNT ⁄ B. Only the b-trefoil fold is shown (view almost normal Fig. 2, down trefoil fold). (B) GD3, a part of GT1b, binds at Site 2 in TeNT. The same site is occupied by the GD3 part of the GT1b-b analog. (C) A composite figure of the BoNT ⁄ BH CC domain with sialylllactose and tripeptide (as bound to TeNT) shows the double receptor model. (D) A composite figure of the BoNT ⁄ BH CC domain with GT1b as bound in Site 1 of BoNT ⁄ A and Syt II peptide as in Site 2 of BoNT ⁄ B. These two match with sialyllactose and the tripeptide in (C). Struc- tures are all in a similar orientation. Molecular structures of Clostridial neurotoxins S. Swaminathan 4470 FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works ganglioside-binding pockets in TeNT, both BoNT ⁄ A and B have only one binding site corresponding to Site 1 [30,31]. The second site was later confirmed using the crystal structure of TeNT cocrystallized with the carbohydrate part of GD3 [32] (Fig. 3B). However, the interactions and binding modes are slightly differ- ent because of the b2–3 linkage in GT1b-b. This causes the salt bridge between Arg1226 and the sialic moiety to be different. Whereas the terminal sialic acid is involved in the salt bridge formation in the GT1b-b structure, it is the second sialic acid that is involved in the salt bridge. It is also possible that TeNT binds to GD3, whereas BoNTs do not. Recent structural work on BoNT ⁄ A with a ganglio- side analog has also confirmed the GT1b-binding site, as identified in BoNT ⁄ B [33] (Fig. 3D). Remarkably, the binding mode of GT1b to BoNT ⁄ A is similar to TeNT (at least one branch). Similar to TeNT, it is the Gal–GalNAc moiety that binds in this site. As in TeNT, Gal forms a stacking interaction with the con- served Trp, whereas this interaction is provided by the sialic acid moiety in the BoNT/B structure. This may be because sialyllactose is only a partial mimic of one branch of GT1b. Interestingly, mutational analysis on the conserved Tyr (1262 in BoNT ⁄ B and 1267 in BoNT ⁄ A) shows that Tyr is important for binding and toxicity because when it is mutated to Ala or Phe, the toxicity of BoNT ⁄ A or B is < 2% compared with the wild-type [21,31]. Whereas this tyrosine in BoNT ⁄ B forms two strong hydrogen bonds with sialyllactose, BoNT ⁄ A complex lacks these contacts. Because BoNT ⁄ A and B share high sequence homology in the Site 1 pocket, it is suggested that the GT1b binding mode will be similar to BoNTA [33]. However, there is no structural evidence for this. BoNT ⁄ E has a pocket similar to Site 1 in its H CC [10]. However, in the H…SxWY…G motif, H is replaced by K. There is no structural evidence for GT1b binding to BoNT ⁄ E although the similarity of the pockets suggests that it will be the same as BoNT ⁄ A and B. Crystal structures of the binding domain of BoNT ⁄ F, C, D and G have been reported [8,34–37]. BoNT ⁄ F and BoNT ⁄ G have similar GT1b- binding pockets, except that H in the H…SxWY…G motif is replaced by G in BoNT ⁄ G. Two binding sites have been identified in BoNT ⁄ D and they are similar to Sites 1 and 2 of BoNT ⁄ B (but not exactly the same) [37]. The crystal structure of BoNT ⁄ C in complex with sialic acid identifies one site close to Site 2, although the second possible site has not been identified struc- turally [8]. Neither BoNT ⁄ C nor D possess the H…SxWY…G motif found in other BoNTs, but a Trp is present in a nearby loop called the ganglioside- binding loop (W1258 in C and W1252 in D-SA) [8]. Mutational studies implicate W1258 of this loop in GT1b binding but no structural information is avail- able [38]. The role of this Trp needs further investiga- tion. In summary, like TeNT, BoNT ⁄ D has two GT1b-binding sites. Similarly, BoNT ⁄ C may also have two ganglioside (GT1b ⁄ GD1b) sites. Double receptor model Gangliosides are not the sole receptors for BoNTs because reduction in TeNT binding was observed when rat brain membranes were treated with proteases [39,40]. A similar study with BoNT ⁄ B suggested that proteins may also be involved in BoNT uptake [39–41]. There has also been other biochemical evidence for a second receptor molecule, specifically a membrane pro- tein or a glycosylated protein. In view of this, a double receptor model was proposed [23]. The low-affinity, high-density gangliosides allow BoNTs to concentrate on the surface of the cell and move laterally and bind to a high-affinity, low-density second receptor, a protein. Although most of the BoNTs require GT1b, each BoNT has specific protein receptor(s). Because the binding is a product of the two binding constants, the overall binding is very high and one of the reasons for its nanogram level LD 50 . Specific protein receptors have been identified for BoNT ⁄ A, B, E and G. BoNT ⁄ A uses the three isoforms of synaptic vesicle 2 proteins (SV2A, SV2B and SV2C) [42]. BoNT ⁄ B binds to synaptotagmin (Syt I and Syt II), which also acts as a receptor for BoNT ⁄ G, although with a lower affinity [43–45]. Recently, glycosylated SV2A and B have also been identified as receptors for BoNT ⁄ E [46]. BoNT ⁄ F also requires a second protein receptor and the keratan sulfate moiety of SV2 probably binds to the second receptor site [34], although a later study contradicts this finding [47]. However, BoNT ⁄ C and D are different in that they do not need a second protein receptor for binding to the cell membrane, although this is yet to be confirmed. Instead they may use dual ganglioside binding [8,37], however, the details are still to emerge. Although the biochemical evidence has been gaining ground, structural support is recent. Crystal structures of TeNT H C with a tripeptide (Tyr–Glu–Trp) and BoNT ⁄ B in complex with Syt II peptide support the double receptor model [32,48,49]. TeNT H C with a tripeptide The crystal structure of TeNT H C complexed with GD3 sugar group, disialyllactose, showed that this sugar molecule binds to Site 2 of the binding domain. S. Swaminathan Molecular structures of Clostridial neurotoxins FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works 4471 GD3 is one part of GT1b with two sialic acids and one lactose. The crystal structure of TeNT complexed with a tripeptide Tyr–Glu–Trp (YEW) showed that it binds in exactly the same pocket as GD3. More inter- estingly, the interactions are also similar, Arg1226 forms strong salt bridge with Glu OE1 and OE2, as observed with GD3 [32]. The affinity for the tripeptide (YEW) was also higher than that for the GD3 sugar because when TeNT was cocrystallized with equimolar of YEW and GD3, only YEW was found in the crystal structure. However, the thermal factors were high, indicating spatial disorder. This gave direct support for the double receptor model. The two sites (Sites 1 and 2) might initially be occupied by gangliosides but when the specific protein receptor approaches, the protein receptor might displace the low-affinity ganglioside. Such a possibility has been suggested [30]. It is also possible that the second receptor is a glycosylated pro- tein whose sugar group might bind to this site by dis- placing GT1b. Recently, SV2A and SV2B have been identified as receptor proteins for TeNT [47]. This was the first structural evidence for a double receptor model since GT1b binding to Site 1 (Fig. 3C). BoNT ⁄ B–Syt II structure Recently, crystal structures of either the holotoxin BoNT ⁄ B or its H C domain have been determined in complex with Syt II peptide (part of its luminal domain) [48,49]. The cocrystal structure of BoNT ⁄ B H C with Syt II (8–61) was determined using a fusion protein with a linker connecting the C-terminus of BoNT ⁄ BH C with the N-terminus of the Syt II peptide. However, in the crystal structure, the linker peptide and residues 8–43 of Syt II were not modeled due to poor electron density. The holotoxin and Syt II pep- tide (40–60) were cocrystallized to determine the com- plex structure. In this structure, only the electron density of residues 45–59 could be modeled. In both structures, the Syt II peptide occupied the same bind- ing pocket, namely, Site 2 which is adjacent to Site 1. The peptide, which is unstructured in the native pro- tein, is induced to form a helix when it binds to BoNT ⁄ B and occupies a hydrophobic pocket. Phe47, Leu50, Phe54, Phe55 and Ile58 of the Syt II peptide are buried into the binding groove and form hydro- phobic and stacking interactions with BoNT ⁄ B residues. Charged residues Glu57 and Lys51 interact with residues of complementary charges in the protein. There are other hydrogen bonds and electrostatic inter- actions enabling the peptide to bind strongly. Taken together, this Syt II-binding site and the sialyllactose- binding site support the proposed double receptor model for BoNT ⁄ B [23]. Comparison of this with the YEW–TeNT H C complex suggests a common binding site for protein receptors (Fig. 3D). The location of YEW in TeNT is analogous to Syt II in BoNT ⁄ B, however, the chemical identities of the interacting resi- dues are different. A similar pocket exists in BoNT ⁄ G and Syt I or II can bind in a similar fashion [44]. It is expected that in BoNT ⁄ A, E and F their protein receptor will bind at Site 2. However, no structural information is yet available. Translocation domain Once toxins bind to membranes, a temperature- and energy-dependent process internalizes them. The neu- rotoxins have to escape from the vesicles into the cyto- sol by crossing the hydrophobic vesicle barrier. This is achieved by decrease in pH to acidic levels, allowing conformational change of the translocation domain and leading to penetration into membrane for channel formation so that the catalytic domain can escape the endosome. The transmembrane region has been pre- dicted in BoNTs. In the crystal structures, this region, 653–673 in BoNT ⁄ B and 650–672 in BoNT ⁄ A, does not take a helical conformation and is at one tip of the translocation domain apposing one of the long helices (Fig. 4). It is speculated that the region will take a helical conformation when the pH becomes Belt region Fig. 4. The translocation domain of BoNT ⁄ B. The belt region which wraps around the catalytic domain loses its hydrophobic interaction when the catalytic domain separates. The predicted transmem- brane region is shown in magenta. Molecular structures of Clostridial neurotoxins S. Swaminathan 4472 FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works acidic. However, structural determination of BoNT ⁄ B at various pH values (as low as 4) did not show any change in this region [50] although this could be because of crystal packing. The most intriguing part of the translocation domain is the N-terminal region (449–545 in BoNT ⁄ A), especially the 492–545 loop wrapping around the catalytic domain and hence called the belt region Although individual catalytic and binding domains have been crystallized and their struc- tures determined, this information is lacking for the translocation domain, partly because it is hydrophobic and forms aggregates making crystallography a chal- lenge. Also, without the support of the catalytic domain, the belt region may not retain its conforma- tion and may fold back up or down. Although there is a lack of structural work, recent biochemical and biophysical studies provide valuable information. It is postulated that the heavy chain acts as a chaperone for the light chain to translocate the catalytic domain [51]. When the pH becomes acidic, the H N domain penetrates the membrane and translo- cates LC from the N- to the C-terminus, during which the channel is occluded by the LC. A reduction of disulfide in the cytosol is required for LC to separate and then to cleave its target. The interchain disulfide bond plays a critical role in translocation and must be intact for translocation but reduced for translocation to be completed [52]. Recently, it has been shown with the LC–H N complex that: (a) the binding domain is not required for translocation, and (b) translocation can take place at neutral pH, unlike with the holotoxin [53–55]. However, the physiological relevance of this is not clear. The role of the belt region is not yet well understood. Recent studies show that the belt plays a role in translocation. In another study, it was sug- gested that lowering of the pH neutralizes the acidic residues in the belt region and nullifies the repulsion between the negative charge on the membrane and the protein [56]. However, this does not explain transloca- tion at neutral pH in the LC–H N complex. The belt region also acts like a pseudosubstrate and inhibits LC protease activity. The substrate occupies the groove vacated by the belt region when the light chain sepa- rates [57,58]. The nature of the channel formed by HC is not understood although a low-resolution electron micro- graph shows that BoNT ⁄ B forms a tetrameric channel [59]. The channel diameter is observed to be 15 A ˚ and is not large enough for the intact catalytic domain to enter and exit. The catalytic domain unfolds, escapes the endosome and refolds in the cytosol. In summary, structural work on the translocation domain is sparse and more is needed to understand this process well. Catalytic domain This is the most studied domain in BoNTs, both struc- turally and biochemically. The catalytic domain of BoNT is a zinc endopeptidase similar to thermolysin [11,12,60]. Crystal structures of the catalytic domains (LCs) of all BoNT serotypes and TeNT are available and they share similar fold [61–71]. The fold is a compact globule consisting of a mixture of a helices and b sheets. The characteristic zinc-binding motif, HExxH+H, is in the middle of the primary sequence of LC. The active site zinc is bound deep inside a large open cavity that has a high negative electrostatic potential (Fig. 5). The zinc ion is coordinated by two histidines and one glutamate. The fourth coordination is provided by a water molecule which acts as a nucle- ophile. The nucleophilic water molecule forms a strong hydrogen bond with the first Glu in the zinc-binding motif which acts as a base for the catalytic action. Remarkably, the active sites of all BoNTs share a simi- lar architecture and significant sequence conservation. The conserved residues within 10 A ˚ of zinc form iden- tical contacts. In BoNT ⁄ E, the zinc ion coordinates with His211, His215, Glu250 and the nucleophilic water [61]. His211 forms a hydrogen bond with the conserved Glu335, which in turn forms hydrogen bond with the conserved Arg347. The nucleophilic water forms a hydrogen bond with the conserved Tyr350. His215 forms a hydrogen bond with the conserved Fig. 5. Electrostatic potential surface of the catalytic domain of BoNT ⁄ B. Zinc is in a deep cavity which is highly electronegative. Zinc is shown as gray sphere. S. Swaminathan Molecular structures of Clostridial neurotoxins FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works 4473 Glu249, which in turn forms a hydrogen bond with the conserved His218. These interactions are conserved in all BoNTs, including the nucleophilic water to Glu212 (Fig. 6). Mutational analyses on these con- served residues have confirmed that they affect the cat- alytic activity, some more than the others. Most importantly, mutation of Tyr350 and Glu212 resulted in undetectable catalytic activity in the BoNT ⁄ E light chain. Mutations of Arg347, Glu335, and Glu249 dras- tically reduce the K cat compared with wild-type. This is equally true for BoNT ⁄ A, and by extension others also [62,72–74]. The interactions in the vicinity of the zinc ion are exactly the same, resulting in a common cata- lytic mechanism. However, unlike other zinc endopep- tidases, substrates for BoNTs are large polypeptides and hence have numerous contacts with the enzyme which are unique for each serotype. This results in BoNTs exhibiting high specificity for the substrate and scissile bond selection. Although the active site is con- served, these interactions away from the active site are different and dictate the specificity of the substrate. For example, BoNT ⁄ A and C specifically cleave adja- cent peptide bonds (Gln197–Arg198 and Arg198– Ala199, respectively) of the same substrate SNAP-25. This is true for BoNT ⁄ F and BoNT ⁄ D as well, which specifically cleave adjacent peptide bonds of VAMP. This unique specific bond selection is achieved by the interactions remote from the active site enabling the specific scissile bond to be positioned for cleavage [1]. More structural work is needed to better understand this scissile bond selection. Role of zinc in BoNTs The role of zinc in proteins could be either structural, functional or both. A catalytic zinc is normally coordi- nated by three amino acids and one water, whereas a structural zinc is coordinated by four amino acids [75,76]. In BoNT, the zinc is coordinated by three amino acids and a water molecule. However, it was thought that its role could be structural from tertiary structural studies [77]. But structural work has categor- ically proved that removal of zinc does not change the conformation and that its role is functional since the catalytic activity is lost on zinc removal [50,62,78,79]. Enzyme–substrate complex Most useful information about the enzyme–substrate interactions and the catalytic mechanism is obtained from the crystal structures of enzyme–substrate com- plexes. However, because the substrate is cleaved on binding to the enzyme a strategy has to be adopted for forming the complex without cleavage; either an inactive mutant or an uncleavable mutant substrate is used to form the complex [57,67,80–82]. An inactive double-mutant (E224Q, Y366F) of BoNT ⁄ A was cocrystallized with a SNAP-25 peptide (141–204) for Glu249 Glu250 2.79 2.75 2.11 3.23 2.05 2.07 2.11 2.85 Glu212 His211 2.59 Glu335 3.04 Arg347 Tyr350 His215 His218 Fig. 6. The interactions at the active site in the vicinity of zinc. The active site of BoNT ⁄ E is shown. These interactions between the conserved residues are con- served across all serotypes. Molecular structures of Clostridial neurotoxins S. Swaminathan 4474 FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works structure determination [81]. This crystal structure gave detailed information about the interactions between the enzyme and the substrate. Because a mutant was used, reasonable information about the interactions near the active site could not be obtained (Fig. 7). However, the interactions away from the active site were mapped detailing the exosites that define the specificity of scissile bond. The crystal struc- ture of BoNT ⁄ A in complex with an hexapeptide, SNAP peptide 197 QRATKM 202 , containing the scissile bond clearly demarcated the active site interactions [82]. Taken together, these two structures faithfully map out the interactions between the enzyme and the C-terminal region (amino acids 141–204) of the sub- strate and provide invaluable information for design- ing substrate-based inhibitors. Although the overall conformation of the two enzyme–substrate peptide complexes is very similar (RMSD  1A ˚ for 400 Ca atoms), loops 200, 250 and 370 vary significantly. This conformational change may be because of either the recognition of a-exosites in the complex with a larger peptide or an artifact of crystal packing. In the structure with the hexapeptide, loops 200, 250 and 370 pack together tightly, whereas in the structure with a larger peptide, loop 200 moved away. This also points to the induced fit when the lar- ger substrate peptide is used. In the hexapeptide (QRATKM)–BoNT⁄ A complex, the carbonyl oxygens of P1 (Gln197) and P1¢ (Arg198) form strong hydrogen bonds with the side chains of Tyr366 and Arg363, respectively. The amino nitrogen of P1 displaces the nucleophilic water and coordinates with zinc. Also, P1¢ (Arg198) forms a salt bridge with Asp370 of the enzyme (Fig. 8). These interactions dem- onstrate the critical role played by these residues in addition to the zinc-coordinating residues, and explain the mutational analyses [62,72]. Based on this, a cata- lytic mechanism has been proposed (see Fig. 6 in Ref. [67]). This is supported by mutagenesis studies on BoNTs. Conserved Tyr and Arg help to position, ori- ent and stabilize the substrate for cleavage. Glu224 acts as a general base by absorbing a proton from the nucleophilic water. The nucleophilic water attacks the carbonyl carbon of the scissile bond, which forms a tetrahedral transition intermediate. The zinc ion and Tyr might stabilize this intermediate transition state. The shuttling of protons with the help of Glu224 assists the subsequent formation of a stable leaving amino group. This model is consistent with that pro- posed for BoNT ⁄ B, E and F [57,62,83] and will hold good for all BoNTs. The proposed noncanonical self protease activity could be due to the high concentra- tion of protein ⁄ substrate and low pH used in crystalli- zation and may not be physiologically relevant [68]. Crystal structures of BoNT ⁄ F in complex with two VAMP peptides, VAMP 22–58 ⁄ Gln58D-cysteine and VAMP 27–58 ⁄ Gln58D-cysteine, use an active enzyme with uncleavable substrate inhibitor peptides with K i  1nm [57]. These crystal structures mapped out the interactions between BoNT ⁄ F and the VAMP sub- strate. Three exosites were identified which may govern substrate specificity. Interestingly, conformational changes involving rotamer positions of side chains of enzyme residues were observed when the substrate binds. These changes, which are due to induced fitting when a complex is formed, either open up the site for substrate to enter or reorient to make better contact with the substrate. The movement of loop 370 observed in BoNT ⁄ A is not seen in this structure. Bio- chemical and mutational studies confirmed that BoNT ⁄ F recognizes VAMP via unique exosites. This structure established that Arg133, Glu164 and Arg171 are important residues determining substrate specific- ity. The biochemical and structural results agree well [57,84]. Extending substrate beyond the C-terminal of the inhibitor peptide improved hydrolysis, suggesting additional interactions of the region 59–65 [85]. How- ever, the substrate inhibitor in the BoNT ⁄ F complex structure stops at P1 and does not provide any information in this regard (Fig. 9). Fig. 7. SNAP-25 peptide bound to BoNT ⁄ A catalytic domain. The enzyme is shown in light blue, SNAP-25 in green and the zinc ion in magenta as a sphere. Coordinates were taken from PDB 1XTG. S. Swaminathan Molecular structures of Clostridial neurotoxins FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works 4475 Although VAMP is small compared with SNAP-25 used in the BoNT ⁄ A complex, it showed distinct exosite interactions. The orientation in which the substrate binds to the enzyme is the same in both and the active site interactions are conserved. However, VAMP is dif- ferently positioned compared with SNAP-25. The three major exosites in BoNT ⁄ F are completely different from those in BoNT ⁄ A. By comparison with SNAP-25 in the complex, it is concluded that the substrate takes the place of the belt region (Fig. 10). Again as in BoNT ⁄ A, the unstructured substrate takes a helical conformation induced by contact with the enzyme. Exosite 1 forms a short helix and its hydrophobic side chains point towards the hydrophobic core of the enzyme. Surpris- ingly, the V1 SNARE motif ( 38 QVDEVVDIMR 47 )is not a helix, but the adjacent region (N-terminal side) is helical. In summary, the enzyme–complex structures help in understanding the interactions between the two, leading to clues for drug design against botulism. Whether the exosites, alone or in conjunction with the active site, can be used as targets for drug design should be explored. Structure-based drug discovery for botulism Even though botulinum neurotoxins are potential bio- warfare agents and a public health hazard, effective drugs are yet to be developed, especially for post intoxication with botulinum toxins. Antibody thera- peutics is emerging, but more than one antibody may be needed to contain the effect of a single serotype and there are limitations [86]. An equine antitoxin is also available for post-exposure therapeutics. Small molecules can be used effectively to treat botulinum poisoning both before and after exposure and research in this direction is expanding fast. Botulinum neuro- toxin could be deactivated by targeting any one of the three major steps in its toxicity pathway, binding, translocation and catalytic activity. This can be done His227 His223 2.97 2.12 2.11 2.10 2.07 2.00 Glu262 Tyr366 2.37 2.75 3.26 Gln197 Glu224 Glu164 Phe163 Gln162 Ile161 Phe194 2.90 3.05 Lys201 Thr200 Rg Ala199 2.95 3.32 2.81 Arg198 3.04 63 2.58 P370 Met202 Fig. 8. BoNT ⁄ A and substrate peptide inter- actions. Interactions between the main chain atoms with the conserved residues are maintained across serotypes. The car- bonyl oxygen of P1 and P1¢ hydrogen bond with the conserved Tyr and Arg (Tyr366 and Arg363 in BoNT ⁄ A). These interactions help to position the substrate and stabilize the intermediate transition state. Molecular structures of Clostridial neurotoxins S. Swaminathan 4476 FEBS Journal 278 (2011) 4467–4485 Journal compilation ª 2011 FEBS. No claim to original US government works [...]... use of BoNTs in therapeutics to other ailments Molecular structures of Clostridial neurotoxins molecule more globular (Fig 12) In BoNT ⁄ A and B there are only limited interactions between the translocation and binding domains and no interaction between the binding and catalytic domains In BoNT ⁄ E, all three domains have contacts with one another The belt region in E is similar to that in A or B, but... holotoxin structure contains one extra water molecule and is reminiscent of what has been observed in recombinant BoNT ⁄ A catalytic domain structure [78] Surprisingly, although the individual domains are similar, the domain organization in BoNT ⁄ E is different from in A or B Whereas the translocation domain in A or B is flanked by catalytic and binding domains on either side in a linear fashion, in BoNT... However, a low-resolution image obtained with electron microscopy using nicked, dichain BoNT ⁄ E shows a somewhat similar arrangement [107] So the difference is not because of the single chain molecule The position of the binding domain of E could be obtained by rotating the binding domain of A or B about the linker region connecting the translocation and binding domains This linker region (858–870, BoNT ⁄... not been continued Recently, interest in the binding domain has gained momentum because the binding sites (both gangliosides and protein receptors) have been identified [21] These sites will be good targets for developing small molecules to prevent the toxin binding to neuronal membranes Because these two sites are independent, adjacent and nonoverlapping, two molecules connected by a linker to block... the receptor binding fragment Hc of tetanus neurotoxin Nat Struct Biol 4, 788–792 Molecular structures of Clostridial neurotoxins 15 Murzin AG, Lesk AM & Chothia C (1992) betaTrefoil fold Patterns of structure and sequence in the Kunitz inhibitors interleukins-1beta and 1alpha and fibroblast growth factors J Mol Biol 223, 531–543 16 Simpson LL (1984) The binding fragment from tetanus toxin antagonizes... neurotoxin light chain Biochemistry 44, 7450–7457 72 Binz T, Bade S, Rummel A, Kollewe A & Alves J (2002) Arg362 and Tyr365 of the botulinum neurotoxin type A light chain are involved in transition state stabilization Biochemistry 41, 1717–1723 73 Li L, Binz T, Niemann H & Singh BR (2000) Probing the mechanistic role of glutamate residues in the zincbinding motif of type A botulinum neurotoxin light chain...S Swaminathan Molecular structures of Clostridial neurotoxins by designing small molecules to block the catalytic site, binding site or the channel formed by translocation domain The following section discusses how structural information could be exploited to achieve this goal Binding domain as target Fig 9 VAMP-based inhibitor bound to BoNT ⁄ F catalytic domain The inhibitor stops at... are the binding substances in neural cells for tetanus and botulinum toxins in mice Biochim Biophys Acta 1441, 1–3 23 Montecucco C (1986) How do tetanus and botulinum toxins bind to neuronal membranes? Trends Biochem Sci 11, 314–317 24 Muraro L, Tosatto S, Motterlini L, Rossetto O & Montecucco C (2009) The N-terminal half of the receptor domain of botulinum neurotoxin A binds to microdomains of the... and function of tetanus and botulinum neurotoxins Q Rev Biophys 28, 423–472 6 Sobel JD (2005) Botulism Clin Infect Dis 41, 1167– 1173 7 Oguma K, Fujinaga Y & Inoue K (1995) Structure and function of Clostridium botulinum toxins Microbiol Immunol 39, 161–168 8 Karalewitz AP, Kroken AR, Fu Z, Baldwin MR, Kim JJ & Barbieri JT (2010) Identification of a unique ganglioside binding loop within botulinum neurotoxins. .. translocation domain and allows the binding domain to traverse to the other side However, in BoNT ⁄ E this region is a loop (830–845) and allows the binding Crystal structure of BoNT ⁄ E The crystal structure of the holotoxin BoNT ⁄ E (unnicked, single chain) has provided information on its novel domain organization [10] BoNT ⁄ E has 39.8% and 37.2% identity with BoNT ⁄ A and BoNT ⁄ B, respectively In view of . binding domain of E could be obtained by rotating the binding domain of A or B about the lin- ker region connecting the translocation and binding domains SNAP-25 in green and the zinc ion in magenta as a sphere. Coordinates were taken from PDB 1XTG. S. Swaminathan Molecular structures of Clostridial neurotoxins FEBS