Synaptotagmins I and II act as nerve cell receptors for botulinum neurotoxin G.

Botulinum neurotoxins (BoNTs) induce muscle paralysis by selectively entering cholinergic motoneurons and subsequent specific cleavage of core components of the vesicular fusion machinery. Complex gangliosides are requisite for efficient binding to neuronal cells, but protein receptors are critical for internalization. Recent work evidenced that synaptotagmins I and II can function as protein receptors for BoNT/B (Dong, M., Richards, D. A., Goodnough, M. C., Tepp, W. H., Johnson, E. A., and Chapman, E. R. (2003) J. Cell Biol. 162, 1293-1303). Here, we report the protein receptor for a second BoNT serotype. Like BoNT/B, BoNT/G employs synaptotagmins I and II to enter phrenic nerve cells. Using pull-down assays we show that only BoNT/G, but neither the five remaining BoNTs nor tetanus neurotoxin, interacts with synaptotagmins I and II. In contrast to BoNT/B, interactions with both isoforms are independent of the presence of gangliosides. Peptides derived from the luminal domain of synaptotagmin I and II are capable of blocking the neurotoxicity of BoNT/G in phrenic nerve preparations. Pull-down and neutralization assays further established the membrane-juxtaposed 10 luminal amino acids of synaptotagmins I and II as the critical segment for neurotoxin binding. In addition, we show that the carboxyl-terminal domain of the cell binding fragment of BoNT/B and BoNT/G mediates the interaction with their protein receptor.

Clostridial neurotoxins (CNTs) 1 are extremely potent bacterial toxins. Among them, the seven serologically distinct botulinum neurotoxins (BoNTs, serotypes A-G) cause botulism, whereas the tetanus neurotoxin (TeNT) provokes the disease tetanus. Each neurotoxin is composed of four domains. Their light chains (LCs) act as zinc-dependent endopeptidases and specifically hydrolyze certain proteins of the vesicular fusion machinery, whereupon the Ca 2ϩ -triggered fusion of synaptic vesicles with the presynaptic membrane is disrupted (reviewed in Refs. [1][2][3]. The heavy chains (HCs) are tethered to the LCs via a single disulfide bond and encompass the three remaining domains. The HC serves as the vehicle that delivers the LC to the cytosol of neuronal cells. Therefore, the extreme toxicity has to be largely ascribed to the specific binding of the molecule to nerve terminals at the neuromuscular junction. The aminoterminal segment of the HC, the H N domain, is responsible for translocating the LC from the lumen of an acidic intracellular compartment into the cytosol subsequent to cell binding and receptor-mediated endocytosis. The carboxyl-terminal segment of the HC, the so-called H C -fragment (H C ), comprises two domains, H CN and H CC . The latter was shown in the case of TeNT to suffice for cell binding (4) and internalization (5). TeNT travels retrogradely and eventually arrives at inhibitory interneurons in the spinal cord where it provokes spastic paralysis. So far, the corresponding domain has not been identified for BoNTs, which in contrast act locally at motoneurons and cause flaccid paralysis. No function could yet be attributed to the H CN domain.
It has long been known that polysialogangliosides, i.e. glycosphingolipids that are particularly enriched in the outer leaflet of neuronal cell membranes, are crucial for the binding of CNTs to neurons. TeNT and BoNTs exhibit affinities in the high nM range for isolated polysialogangliosides. The binding of CNTs to neuronal tissue, however, exhibits yet much higher affinities (reviewed in Ref. 6). This discrepancy as well as other findings led to the proposal of the now confirmed two-receptor model (7). This model suggests that the abundant polysialogangliosides trap and accumulate CNTs in the plane of the cell membrane. Here, the neurotoxins wait until achieving contact with their sparsely occurring protein receptor(s), which are assumed to mediate the subsequent specific endocytosis. Synaptotagmin (Syt)-I and Syt-II, two homologous membrane-anchored proteins of synaptic vesicles (8,9) thought to link vesicular fusion to Ca 2ϩ entry (10,11), were demonstrated to interact with BoNT/B in the presence of GT1b (12)(13)(14). Recently, it was conclusively shown that their luminal domain, which becomes temporarily exposed at the synaptic cleft upon fusion of synaptic vesicles, interacts with BoNT/B and mediates the entry of BoNT/B into neurons (15).
In the present study, we investigated whether other CNTs than BoNT/B can utilize Syt-I or Syt-II as protein receptors. Our results establish that BoNT/G interacts directly with both Syt-I and Syt-II in a ganglioside-independent fashion. Moreover, neutralization studies employing peptides derived from the luminal domain of Syt-I or Syt-II effectively block the toxicity of BoNT/G in mouse phrenic nerve preparations. Thus, the interaction of BoNT/G with Syt-I or Syt-II is crucial for its entry into motor nerve terminals.

EXPERIMENTAL PROCEDURES
Plasmid Constructions and Recombinant Proteins-Plasmids encoding full-length BoNT/A and BoNT/B as well as the H C -fragments of TeNT, BoNT/A, and BoNT/B were described previously (16,17). Corresponding constructs for the expression of BoNT/C, -D, -E, -F, and -G H C -fragments carrying a carboxyl-terminal StrepTag were generated by using PCR with suitable primers and purified bacteriophage DNA (BoNT/C and -D) or total bacterial DNA (BoNT/E, strain NCTC 11219; BoNT/F, strain NCTC 10281; BoNT/G (Clostridium argentinense)) as template DNA. Plasmids for Escherichia coli expression of carboxylterminal His 6 -tagged H CN and H CC domains of BoNT/A, -B, and -G were constructed by PCR using suitable primers and pQE3 (Qiagen, Hilden, Germany) as parental vector. A plasmid encoding full-length BoNT/G with carboxyl-terminal StrepTag (pBoNTG-wt) was assembled from LC and H N domain-encoding DNA pieces obtained by PCR in the H Cencoding vector. An expression plasmid (pBoNTG-thro) for full-length BoNT/G comprising a thrombin cleavage site between L and H N was constructed in the same manner. Truncated variants of rat Syt-I (1-82, 1-53, 1-43), rat Syt-II (1-90, 1-61, 1-51), and rat Syt-III (1-80) were cloned in pGEX-2T. Nucleotide sequences of all plasmids were verified by DNA sequencing.
Recombinant full-length neurotoxins and the various H C -derived fragments were produced utilizing the E. coli strain M15[pREP4] (Qiagen) during 10 h of induction at RT and purified on Strep-Tactin (IBA GmbH, Göttingen, Germany) or nickel-nitrilotriacetic acid beads (Qiagen) according to the manufacturers' instructions. GST fusion proteins obtained from E. coli TG1 were purified employing glutathione-Sepharose beads. Fractions containing the desired proteins were pooled and dialyzed against Tris/NaCl buffer (20 mM Tris-HCl, 150 mM NaCl, pH 7.2) or, in the case of neurotoxins, directly frozen in liquid nitrogen and kept at Ϫ70°C. GST fusion proteins used for pull-down experiments were prepared in Tris/NaCl buffer supplemented with 0.5% Triton X-100 (Tris/NaCl/Triton buffer).
Syt-I-1-53 and Syt-II-1-61 were also extracted from the respective GST fusion proteins using the thrombin cleavage site present between the GST and Syt sections by applying thrombin (Roche Applied Science) in a 1:100 mass ratio for 2 h at RT. Thrombin was subsequently inactivated using phenylmethylsulphonylfluoride.
GST Pull-down Assays-GST fusion proteins (0.14 nmol each) immobilized to 10 l of GT-Sepharose beads were incubated with fulllength neurotoxins, H C -fragments, H CN or H CC domains (0.06 nmol each) in the absence or presence of a bovine brain ganglioside mixture (18% GM 1 , 55% GD 1a , 10% GT 1b , and 2% other gangliosides; 5 g each; Calbiochem) in a total volume of 200 l of Tris/NaCl/Triton buffer for 90 min at 4°C. Beads were collected by centrifugation and washed three times each with 35 bed volumes of the same buffer. Washed pellet fractions were boiled in SDS sample buffer and analyzed together with supernatant fractions by SDS-PAGE and Coomassie Blue staining.
Mouse Phrenic Nerve Assays-The mouse phrenic nerve assay was set up as described by Habermann et al. (18). Electrical stimulation of the phrenic nerve was performed continuously at a frequency of 1 Hz. Isometric contractions were recorded with a force transducer and analyzed with VitroDat Online software (FMI GmbH, Seeheim, Germany). The time required to reduce the amplitude to 50% of the starting value (paralytic half-time) was determined. Recombinant BoNT/G-thro was applied in triplicate at final concentrations of 2.0, 6.0, 20.0, and 60.0 nM. A concentration-response curve could be described by the power function y ϭ 128.13x Ϫ0.2646 (R 2 ϭ 0.9838). A concentration-response curve described by the power function y ϭ 69.11x Ϫ0.2662 (R 2 ϭ 0.9855) was For inhibition studies scBoNT/A, scBoNT/B, and BoNT/G-thro were mixed with various concentrations of Syt-I-or Syt-II-derived peptides, incubated for 15 min at RT, and subsequently added to the hemidiaphragm preparation. Paralytic half-times measured in the presence of peptides were converted to corresponding concentrations of the respective isolated BoNTs using the equations mentioned above. Resulting toxicities were finally expressed as percentage of inhibition of toxicity. Individual inhibition experiments were conducted three to six times.

BoNT/G Directly Interacts with the Luminal Domain of Syt-I and Syt-II-Previous studies demonstrated that Syt-I and
Syt-II interact with BoNT/B (12)(13)(14) and mediate its internalization (15). In contrast, the entry of BoNT/A and BoNT/E does not depend on the presence of Syt-I and Syt-II (15). To screen whether any of the remaining CNTs utilizes interaction with Syts to enter neuronal cells, we analyzed binding of their H Cfragments to Syt-I and Syt-II in vitro. We conducted GST pull-down experiments employing fusion proteins in which the entire luminal and transmembrane domains of rat Syt-I and Syt-II was fused to GST. As demonstrated in Fig. 1B, GST⅐Syt-I-1-82 and GST⅐Syt-II-1-90 pull down the H C -fragments of BoNT/B and BoNT/G. The finding with BoNT/B agrees with results of Dong and colleagues (15); the finding with BoNT/G constitutes a novel result. Binding of H C B and H C G to GST⅐Syt-II-1-90 is saturable and reaches a 1:1 stoichiometry at saturation (3-fold molar excess of H C ; data not shown). We also detected traces of H C -fragments of BoNT/A and BoNT/E pulled down with GST⅐Syt-I-1-82 and GST⅐Syt-II-1-90 (Fig. 1B, arrowheads). This interaction appears to be unspecific, as H C A is pulled down to a similar extent by GST⅐Syb2 and GST⅐Syt-III-1-80 (Fig. 1C), and probably explains data of a previous report about the binding of BoNT/A and BoNT/E to Syt-I (19).
To substantiate the newly discovered BoNT/G-Syt interaction, we studied the binding of the full-length neurotoxin as well. For this, we constructed expression plasmids producing BoNT/G fused to a carboxyl-terminal StrepTag. One variant (scBoNT/G-wt) comprised the native loop region between the two cystein residues that form the interchain disulfide bridge. In a second variant (BoNT/G-thro), a recognition site for thrombin was inserted in this region at the expense of four of the original amino acids ( Fig. 2A). Both neurotoxin variants could be isolated from E. coli lysates via their affinity tag and exhib-FIG. 1. The H C -fragment of BoNT/G binds to Syt-I and Syt-II. A, schematic drawing of Syt. Syts are comprised of an intraluminal, a transmembrane (TMD), and two C2 domains (C2A, C2B). Syt constructs used in GST pull-down experiments lacked the entire cytoplasmic part. Numbers below specify amino acid positions of Syt-I and Syt-II. SV, synaptic vesicle. B and C, GST fusion proteins (5 g each/reaction) immobilized on glutathione-Sepharose beads were incubated for 90 min at 4°C with CNT H C -fragments (3 g each) in the presence of gangliosides (25 g/ml). Pellet fractions were washed three times, and 20% of the material as well as 2 g of each H C -fragment was subjected to SDS-PAGE. Protein was subsequently visualized by Coomassie Blue staining. Asterisks denote breakdown products of GST fusion proteins. Arrowheads point toward unspecifically bound H C A and H C E.
ited Ͼ85% purity (Fig. 2B). BoNT/G-wt was used for in vitro binding studies as single chain protein. Full-length scBoNT/ G-wt also interacts with Syt-I and Syt-II, whereas no binding was observed for scBoNT/A, underscoring that the pull down of H C A is indeed unspecific (Fig. 3, middle panel).
It was recently shown that the interaction of BoNT/B with Syt-I depends on the presence of gangliosides. Therefore, we checked next whether BoNT/G displays the same binding mode. However, BoNT/G⅐Syt-I complexes already form in the absence of gangliosides and exhibit a similar affinity as compared with conditions with micelle-incorporated gangliosides (Fig. 3, right panel).
BoNT/G Binds to the Membrane Proximal Region of Syt-I and Syt-II-Because the mode of binding differs between BoNT/B and BoNT/G with respect to the requirement of gangliosides, it was interesting to investigate whether BoNT/B and -G associate with the same segment of Syt. Dong et al. (15) showed that the membrane proximal region of Syt-II, i.e. residues 40 -60, comprises the binding site for BoNT/B. We therefore mapped the binding site for BoNT/G by truncating Syt-I and Syt-II. Results presented in Fig. 4B show that H C G binds to GST⅐Syt-II-1-61. No binding is detected, however, to GST⅐Syt-II-1-51, suggesting that BoNT/G shares the interac-tion site with BoNT/B. The binding site for BoNT/B in Syt-I could not be determined via carboxyl-terminal-truncated Syt-I, because this interaction only occurs when Syt-I and gangliosides are concomitantly present in micellar structures (Fig. 4A) (15). On the other hand, we were able to establish that the corresponding membrane proximal segment of Syt-I, i.e. residues 43-53, is crucial for the binding of BoNT/G because H C G is pulled down by GST⅐Syt-I-1-53 but not by GST⅐Syt-I-1-43. This interaction occurs as with full-length BoNT/G-wt (Fig. 3) whether gangliosides are present or not.
The Isolated Luminal Domain of Syt-I and Syt-II Blocks the Toxicity of BoNT/G at the Mouse Phrenic Nerve-To provide evidence for the physiological relevance of the BoNT/G-Syt interaction, we investigated whether the neurotoxicity of BoNT/G could be blocked by preincubating the neurotoxin with peptides derived from the luminal domain of Syt-I and Syt-II. As an assay system we chose the established mouse phrenic nerve toxicity test (18). In the first set of experiments, we inspected whether it succeeded in blocking the effect of BoNT/B. scBoNT/B was applied in a 2-nM concentration, which results in a 50% reduction of the hemidiaphragm muscle contractile force within 59.2 Ϯ 3.6 min (paralytic half-time). Preincubation of scBoNT/B with a 1000-fold molar excess of GST⅐Syt-II-1-61 leads to a 61% inhibition of neurotoxicity (paralytic half-time: 76.0 Ϯ 5.7 min), whereas a 10,000-fold molar excess results in a 95.1% inhibition (paralytic half-time: 132.3 Ϯ 13.2 min; Fig. 5). Removal of the GST portion increases the potency of Syt-II-1-61 10-fold because a peptide concentration of 1.95 M raises the paralytic half-time to 138.5 Ϯ 10.6 min (corresponding to a 95.9% inhibition). In line with results of the binding experiments, GST⅐Syt-II-1-51, a variant devoid of the complete interaction site, has no influence on neurotoxicity. Secondly, GST⅐Syt-I-1-53, as well as the corresponding peptide Syt-I-1-53, does not significantly prolong the paralytic half-time of scBoNT/B because preincubation occurred in the absence of micellar gangliosides (Fig. 5).
BoNT/G proved to be far less active at the mouse phrenic nerve than BoNT/B and BoNT/A, probably because of a lower affinity to synaptosomal membranes 2 as well as to a lower catalytic activity of its LC versus other synaptobrevin-hydrolyzing CNTs (20). A final concentration of 105 nM scBoNT/G-wt resulted in a paralytic half-time of 120 min (data not shown). Bath concentrations of up to 1 mM GST⅐Syt would have been required for inhibition studies. Attempts were undertaken to acquire nicked, i.e. proteolytically activated, BoNT/G because nicked CNTs are generally far more potent. However, activation of recombinant scBoNT/G-wt with trypsin leads to an inadvertent hydrolysis of peptide bonds within the HC (Fig.  2B). To circumvent this problem, we utilized BoNT/G-thro, which is to a great extent specifically activated by E. coli proteases during the purification procedure. BoNT/G-thro proves to be about 80-fold more active than scBoNT/G-wt. A final bath concentration of 20 nM results in a paralytic halftime of 55.8 Ϯ 3.8 min and was consequently used for neutralization studies. In contrast to BoNT/B, both GST⅐Syt-II-1-61 and GST⅐Syt-I-1-53 are able to efficiently diminish the neurotoxicity of BoNT/G-thro on preincubation with a 1000-fold molar excess by 66.2% and 75.1% (according to 74.4 Ϯ 6.5 min, 80.7 Ϯ 2.5 min paralytic half-time), respectively. Syt-II-1-61 and Syt-I-1-53 applied at 100-fold molar excess decrease the neurotoxicity by 37.9% (63.3 Ϯ 5.9 min) and 66.4% (74.5 Ϯ 4.9 min), respectively, and thus nearly approximate the efficacy of their 10-fold higher concentrated GST-tagged variants. In addition, preincubation of BoNT/G-thro with an equimolar mix-2 A. Rummel and T. Binz, unpublished results. min, respectively, that do not significantly differ from the untreated control. The results with the latter truncated Syt variants also demonstrate that inhibition through GST⅐Syts is specific and does not interfere with any other physiological process in the assay system. As a further control experiment, we assessed whether the toxicity of scBoNT/A is effected by preincu- FIG. 3. The binding of full-length BoNT/G to Syt-I does not depend upon the presence of gangliosides. Binding assays were conducted as outlined in the legend to Fig. 1. However, binding assays were conducted partly in the absence of gangliosides. Asterisks denote breakdown and byproducts of GST fusion proteins. Arrowheads indicate truncated Strep-tagged BoNT/G fragments that also bound specifically to GST⅐Syt-II-1-90. These fragments are difficult to detect in GST⅐Syt-I-1-82, because they co-migrate with a byproduct of this fusion protein. Notably, BoNT/B does not, but BoNT/G does interact with Syt-I in the absence of gangliosides.  open columns), and BoNT/G-thro (filled columns) were preincubated for 15 min at RT with various amounts of truncated GST⅐Syt-I, GST⅐Syt-II, or isolated Syt peptides as indicated. Mixtures were then added to electrically stimulated mice hemidiaphragm preparations, and isometric contractions were recorded and analyzed. The time required to decrease the amplitude to 50% of the starting value (paralytic half-time) was determined as well as percent inhibition versus the respective untreated neurotoxin (for details see "Experimental Procedures"). bation with GST⅐Syt, because entry of this neurotoxin has previously been shown to be independent of an interaction with either Syt (15). As presented in Fig. 5, even a 19.5-M final concentration of GST⅐Syt-II-1-61 has no detectable effect on the toxicity of 0.22 nM scBoNT/A. This finding further validates the specificity of the assay system. Together, these results prove that Syt-I and Syt-II indeed do mediate the entry of BoNT/G into peripheral nerve cells.
The H CC Domain of BoNT/B and BoNT/G Mediates the Interaction with Syt-It is known that the H CC domain of the TeNT and BoNT H C -fragments is responsible for ganglioside binding (17,21,22). No function could so far be allocated to H CN . To assign the Syt binding site to either of the two domains, we expressed each as His 6 -tagged proteins in E. coli and conducted pull-down experiments. H CN A and H CC A served as control peptides in this assay. Fig. 6 clearly demonstrates that the H CC , but not the H CN , domain of BoNT/B and -G interacts with GST⅐Syt-II-1-61. So, in addition to ganglioside binding, the H CC domain of BoNT/B and BoNT/G mediates the interaction with the protein receptor as well. DISCUSSION BoNTs, the causative agents of botulism, disrupt the neurotransmission of cholinergic nerves at the neuromuscular junction. More than one decade ago, the underlying molecular basis for the inhibition of neurotransmitter release was deciphered for all BoNTs and TeNT, which turned out to be the proteolysis of one of three intracellular soluble NSF attachment protein receptor proteins by their catalytic domains. On the other hand, receptors that mediate the productive uptake of CNTs into nerve terminals have so far only been unequivocally identified for BoNT/B (12,15).
In this study, we have identified the cellular receptor for the second of the eight CNTs, BoNT/G, which like BoNT/B is guided into neurons through its specific interaction with Syt-I or Syt-II. Two lines of evidence support this suggestion. First, by means of GST pull-down experiments we were able to show that BoNT/G interacts directly with the luminal domain of Syt-I and Syt-II. Second, peptides derived from the luminal domain of Syt-I and Syt-II are capable of blocking the entry of BoNT/G into motoneurons that innervate the mouse diaphragm. This only occurs when these peptides contain the membrane anchor-juxtaposed luminal 10 amino acids of Syt. The identified segment of Syt-I and Syt-II becomes transiently exposed on the membrane surface only when synaptic vesicles fuse with the presynaptic membrane at synapses. This is in line with the well documented finding that nerve stimulation generally accelerates the uptake of BoNTs and concomitant poisoning of nerve terminals (23). In a similar manner, TeNT was previously also shown to be taken up via recycling of synaptic vesicles (24). Together, these data suggest that CNTs in general enter nerve terminals via this route and may associate with segments of resident synaptic vesicle proteins that are exposed to the luminal side.
In contrast to what was reported for BoNT/B (15), the binding of BoNT/G to Syt-I is apparently independent of the presence of gangliosides. The most obvious explanation for this discrepancy is that a low affinity of BoNT/B for Syt-I does not allow detection by GST pull-down experiments. Here, only the simultaneous interaction with Syt-I and a ganglioside molecule ultimately guarantees high affinity binding. In BoNT/G, sequence variation in its Syt binding fold may account for stronger binding to Syt-I. Alternative explanations like the generation of a high affinity binding site for Syt-I because of ganglioside-induced structural changes is unlikely to occur, because significant structural changes were not observed in several crystals of BoNT/B and its complexes with sialyllactose (21) or doxorubicin (25). Furthermore, gangliosides incorporated in Triton micelles did not support binding of BoNT/B to a truncated Syt-I variant that lacked the membrane anchor domain, indicating strict requirement of both Syt-I and ganglioside within the context of membranes.
Our study illustrates that the H CC domains of BoNT/B and BoNT/G are responsible for the binding to their protein receptor. This interaction is requisite for productive uptake into neurons. Indirect evidence for an employment of H CC in this process has recently been suggested for TeNT as well (5). Thus, the H CC domains likely function as a closed cell entry module in all CNTs, whereas the function of the H CN domain still awaits elucidation.
The novel finding of the present study that BoNT/G, just like BoNT/B, utilizes Syt-I and Syt-II for its entry into nerve terminals is consistent with their degree of sequence conservation. The similarity score for aligning their H C -fragments is 42.4% and is only exceeded by the score of 58.0% for H C E (strain NCTC11219)/H C F (Clostridium baratii) (ClustalW software). Almost all other pairwise alignments result in less than 30% similarity. We have further shown that the H CC domain of BoNT/B and -G mediates the binding to Syt-I and Syt-II. With respect to sequence alignments of H CC , H CC B/H CC G actually reaches by far the highest similarity score (39.5%), whereas those for many other pairs drop down to less than 20%. This agrees with the notion that none of the remaining six CNTs binds to Syt-I or Syt-II and the observation that BoNT/B and -G are incapable of competing binding of BoNT/A and E to synaptosomal membranes. 2 These data add further support to the premise that all other CNTs associate with different protein receptors to become endocytosed. Interestingly, TeNT, which is sorted into the retrograde axonal transport route upon endocytosis, was demonstrated to bind simultaneously to two separate carbohydrate structures (16,26), one of which could be part of the recently discovered glycosylated 15-kDa protein receptor (27). Therefore, the use of a separate protein receptor that ensures exit from the lysosomal transport route of BoNTs is plausible explicitly for TeNT.