SV2 Is the Protein Receptor for Botulinum Neurotoxin A

See allHide authors and affiliations

Science  28 Apr 2006:
Vol. 312, Issue 5773, pp. 592-596
DOI: 10.1126/science.1123654


How the widely used botulinum neurotoxin A (BoNT/A) recognizes and enters neurons is poorly understood. We found that BoNT/A enters neurons by binding to the synaptic vesicle protein SV2 (isoforms A, B, and C). Fragments of SV2 that harbor the toxin interaction domain inhibited BoNT/A from binding to neurons. BoNT/A binding to SV2A and SV2B knockout hippocampal neurons was abolished and was restored by expressing SV2A, SV2B, or SV2C. Reduction of SV2 expression in PC12 and Neuro-2a cells also inhibited entry of BoNT/A, which could be restored by expressing SV2 isoforms. Finally, mice that lacked an SV2 isoform (SV2B) displayed reduced sensitivity to BoNT/A. Thus, SV2 acts as the protein receptor for BoNT/A.

Botulinum neurotoxin A (BoNT/A) is one of seven neurotoxins (designated BoNT/A to BoNT/G) produced by the bacterium Clostridium botulinum (1). BoNT/A blocks neurotransmitter release by cleaving synaptosome-associated protein of 25 kD (SNAP-25) (2, 3) within presynaptic nerve terminals. Owing to its long-lasting effects, BoNT/A is used to treat a wide range of medical conditions and has also emerged as a potential biological weapon (4, 5).

It has been proposed that receptors for BoNTs are composed of both gangliosides that bind toxins with low affinity and protein receptors that form high-affinity complexes with toxins (6, 7). Synaptotagmins I and II, two homologous synaptic vesicle proteins, have been shown to function in conjunction with gangliosides to mediate entry of BoNT/B into cells (811). Low-affinity interactions between BoNT/A and gangliosides have also been reported (1215), and depletion of gangliosides in neuroblastoma cells prevented entry of BoNT/A (16). In addition, mice lacking complex gangliosides display reduced sensitivity to BoNT/A (10, 17). Protein components are certainly required for BoNT/A entry into nerve terminals but have not been identified (1, 6, 7).

To identify protein receptors for BoNT/A, we studied the toxin's route of entry. Stimulation of neurotransmitter release is known to reduce the time required for paralysis caused by BoNT/A at the diaphragm (18). We found that depolarization with a high-K+ solution in hemidiaphragm preparations increased toxin binding to neuromuscular junctions (NMJs) by a factor of ∼6 (Fig. 1A), suggesting that synaptic vesicle exocytosis exposes receptors for BoNT/A.

Fig. 1.

BoNT/A enters neurons by way of recycling secretory vesicles and interacts with a luminal domain of the synaptic vesicle membrane protein SV2. (A) Stimulation of mouse hemidiaphragm preparations with high-K+ buffer increases binding of BoNT/A. NMJs were labeled with α-bungarotoxin (α-BTX). The overlay shows an enlarged image of the region indicated by rectangles. Error bars show SEM (P < 0.0001, Student's t test, n = 27 images). Scale bar, 20 μm. ROI, region of interest. (B) Cultured rat hippocampal neurons were exposed to BoNT/A plus Syt IN Ab for 1 min in three indicated buffer conditions. Scale bar, 20 μm. DIC, differential interference contrast. (C) Immunoprecipitation (IP) of Syp and SV2 from rat brain detergent extracts that contained BoNT/A. (D) Coimmunoprecipitation of BoNT/A and SV2 from mouse brain detergent extracts was carried out with or without exogenous gangliosides. (E) The fourth luminal domain (SV2-L4) of all three SV2 isoforms was fused to glutathione S-transferase (GST), immobilized, and used in binding assays with soluble BoNT/A. (F) The critical region for BoNT/A binding was mapped to a short fragment in the SV2C luminal domain with the use of GST fusion protein pull-down assays. TMS, transmembrane segment. (G) Model of SV2 topology. Black circles indicate conserved residues in all SV2 isoforms, gray circles are residues conserved in two SV2 isoforms, and open circles represent nonconserved residues.

Using cultured rat hippocampal neurons as a model system, we found that BoNT/A was taken up in an activity-dependent manner along with an antibody that recognizes the luminal domain of secretory vesicle protein synaptotagmin I (Syt IN Ab) (11, 19) (Fig. 1B); BoNT/A and the antibody were largely colocalized (fig. S1A). In addition, depolarization resulted in increased cleavage of SNAP-25 by BoNT/A in hippocampal neurons (fig. S1B) and in cultured spinal cord neurons (20). Finally, pretreating neurons with BoNT/B, which blocks synaptic vesicle exocytosis by cleaving Synaptobrevin (Syb) (21), abolished uptake of BoNT/A (fig. S1C). Thus, the BoNT/A receptor is localized to Syb-containing secretory vesicles, which include synaptic vesicles and large dense core vesicles. All subsequent experiments, which used hippocampal neurons and hemidiaphragms, were carried out with the use of high K+ buffer to drive BoNT/A receptors to the cell surface.

We immunoprecipitated known synaptic vesicle proteins from rat brain detergent extracts and assayed the precipitates for bound toxin (22). In this screen, an antibody that recognizes all isoforms of SV2 [pan-SV2 (23, 24)] was able to coimmunoprecipitate BoNT/A (Fig. 1C). An antibody against another abundant integral synaptic vesicle membrane protein—synaptophysin (Syp)—failed to coimmunoprecipitate BoNT/A (Fig. 1C). The addition of exogenous gangliosides increased the level of coimmunoprecipitation of BoNT/A with SV2 (Fig. 1D), particularly at lower toxin concentrations, indicating that gangliosides can promote the formation of stable BoNT/A·SV2 complexes.

SV2 is a conserved integral membrane protein expressed in vertebrates and is localized to synaptic and endocrine secretory vesicles (22). Three highly homologous isoforms have been identified: SV2A, SV2B, and SV2C (22). SV2A and SV2B are widely expressed in the brain, whereas SV2C is more restricted to evolutionarily older brain regions (24, 25). SV2 is a glycoprotein with 12 putative transmembrane domains (Fig. 1G). BoNT/A bound directly to the largest luminal loop (L4) of SV2A, SV2B, and SV2C; BoNT/B and BoNT/E failed to bind (Fig. 1E). The luminal domains of other major synaptic vesicle membrane proteins—including synaptogyrin 1, synaptogyrin 3, SVOP, and Syp—did not bind BoNT/A (fig. S2A). SV2C showed the most robust BoNT/A binding activity (Fig. 1E). A short fragment (amino acids 529 to 566) within the SV2C-L4 loop was able to pull down levels of BoNT/A similar to the full-length L4 loop (Fig. 1, F and G). A sequence alignment between SV2A, SV2B, and SV2C, within this region, is shown in fig. S2B.

Recombinant SV2C-L4 fragments reduced BoNT/A binding to cultured hippocampal neurons but did not affect Syt IN Ab, BoNT/B, or BoNT/E uptake (Fig. 2A and fig. S3C). Preincubation of BoNT/A with SV2C-L4 also reduced BoNT/A binding to NMJs, with no effect on the binding of BoNT/B (Fig. 2, C and D). BoNT/B is an appropriate control because it also enters by binding to a synaptic vesicle protein (8, 11). Binding of BoNT/B was inhibited by adding a peptide (P21) derived from its receptor, synaptotagmin II (11), but this peptide did not affect BoNT/A binding to hippocampal neurons (fig. S2, A and B). Furthermore, using an antibody that only recognizes the cleaved form of SNAP-25, we found that SV2C-L4 inhibited the functional entry of BoNT/A into these neurons, as evidenced by reduced cleavage of SNAP-25 (Fig. 2B).

Fig. 2.

SV2C-L4 inhibits binding and entry of BoNT/A into hippocampal neurons and motor nerve terminals. (A) Hippocampal neurons and (C) mouse hemidiaphragm preparations were exposed to BoNT/A (or BoNT/B) and Syt IN Ab in the presence of either GST alone or SV2C-L4. (B) Cleavage of SNAP-25 by BoNT/A was detected with the use of an antibody (anti–SNAP-25-C) that recognizes only the cleaved form of SNAP-25. (D) SV2C-L4 reduced BoNT/A binding to mouse hemidiaphragm preparations by 65% (P < 0.0001, t test, n = 76), but did not affect binding of BoNT/B (P > 0.05, t test, n = 49). Error bars show SEM.

SV2A and SV2B single-knockout mice and SV2A and SV2B double-knockout mice have been generated (26, 27). Although mice lacking SV2A display severe seizures and die within 2 to 3 weeks of birth (26), cultured hippocampal neurons from SV2A and SV2B double-knockout mice develop normal synaptic structures and are capable of releasing neurotransmitters (26, 28). Because hippocampal neurons express SV2A and SV2B, but not SV2C (24, 25) (fig. S4, A and B), neurons from SV2A and SV2B double-knockout mice serve as an ideal loss-of-function model to study the potential role of SV2 as the receptor for BoNT/A.

We tested the function of SV2B by comparing BoNT/A binding with neurons from SV2B knockout mice (SV2B–/–) and their wild-type littermates. Neurons were simultaneously exposed to BoNT/A and BoNT/B. SV2B knockout neurons showed reduced binding of BoNT/A as compared with that of wild-type neurons (28% reduction) (Fig. 3A). We next asked whether the remaining BoNT/A binding activity was mediated by SV2A. SV2A+/–SV2B–/– mice were bred with each other, and therefore all of the progeny were SV2B–/–, but they had varying levels of SV2A: SV2A+/+, SV2A+/–, and SV2A–/– (Fig. 3B). Binding of BoNT/A to SV2A+/–SV2B–/– neurons was reduced by 53% as compared with that of SV2A+/+SV2B–/– neurons (Fig. 3C). BoNT/A did not bind to SV2A and SV2B double-knockout neurons (Fig. 3, B and C). Binding of BoNT/B to neurons of each genotype remained the same (Fig. 3, A to C).

Fig. 3.

SV2 expression is required for BoNT/A binding and entry into cells. (A) Hippocampal neurons from SV2B knockout (SV2B–/–) mice displayed reduced binding of BoNT/A (28% reduction, P < 0.0001, t test, n = 18) compared with neurons from wild-type (WT) littermates. Binding of BoNT/B remained the same (P > 0.05, t test, n = 22). Error bars show SD. (B) Hippocampal neurons from littermates with the indicated genotypes were exposed to BoNT/A and BoNT/B. Triple immunostaining was performed. (C) SV2A+/–SV2B–/– neurons displayed a 53% reduction of BoNT/A binding activity compared with SV2A+/+SV2B–/– (P < 0.0001, t test, n = 11); BoNT/A did not bind to SV2A–/–SV2B–/– double-knockout neurons. The binding of BoNT/B remained the same for all genotypes (P > 0.05, t test, n = 11). Error bars show SD. (D) Rat SV2A was expressed in SV2A–/–SV2B–/– double-knockout hippocampal neurons. BoNT/A selectively bound to cells that expressed SV2A. The overlay shows an enlarged image of the region indicated by a rectangle. (The color of the SV2 signal was changed from cyan to green to visualize colocalization with red BoNT/A signals.) (E) (Top) A short hairpin–mediated RNA (shRNA) construct was transfected into PC12 cells to establish stable SV2A knockdown cell lines. The indicated cell lines were exposed to BoNT/A, and cell lysates were analyzed by Western blot; the arrow indicates the cleaved form of SNAP-25. Syp served as a loading control. (Bottom) An SV2A knockdown PC12 cell line (number 16) was transiently transfected with rat SV2B or SV2C. Cleavage of SNAP-25 by BoNT/A was assayed as described for the top panel. Actin served as a loading control. (F) Neuro-2a cells were transfected with either a pool of synthesized SV2C small interfering RNA (siRNA) that target the mouse SV2C coding sequence or with a pool of control siRNA that had random sequences. SV2C siRNA inhibited SV2C expression in transfected cells, and these cells failed to take up BoNT/A.

We carried out rescue experiments by transfecting SV2A, SV2B, or SV2C into SV2A and SV2B double-knockout neurons, using a vector that expresses both SV2 and green fluorescent protein (GFP) under the control of separate promoters. Binding of BoNT/A was observed only in neurons that expressed SV2A, SV2B, or SV2C (Fig. 3D and fig. S6A). Enlarged, overlaid images showed a high degree of colocalization between SV2 and BoNT/A at synapses of transfected neurons (Fig. 3D and fig. S6A, overlay).

In addition to neurons, BoNT/A can enter cell lines, including PC12 (11) and Neuro-2a cells (16). These cell lines provide homogeneous model systems to study SV2 function in which the expression of SV2 isoforms can be readily altered. PC12 cells express only SV2A (25) (fig. S4A) and Neuro-2a cells express only SV2C (fig. S4A). We established three stable SV2A knockdown PC12 cell lines (Fig. 3E). All three knockdown cell lines were relatively resistant to BoNT/A entry, as evidenced by reduced cleavage of SNAP-25 (Fig. 3E). Transient transfection of SV2B or SV2C in an SV2A knockdown cell line (number 16) rescued BoNT/A entry (Fig. 3E). Similarly, knockdown of SV2C in Neuro-2a cells inhibited entry of BoNT/A (Fig. 3F). This defect was rescued by expressing rat SV2A or SV2B GFP fusion proteins in these cells (fig. S6B). An enlarged confocal section of transfected cells revealed a high degree of colocalization between SV2A-GFP and internalized BoNT/A (fig. S6C).

Disruption of neurotransmitter release at diaphragm motor nerve terminals is the cause of death from BoNT poisoning (29). All three SV2 isoforms were detected by immunostaining at diaphragm motor nerve terminals (fig. S4C). Because SV2C knockout mice are currently not available and SV2A and SV2B double-knockout mice have severe seizures, we studied binding of BoNT/A to diaphragm preparations from SV2A+/– SV2B–/– mice (26). These mice are healthy and have the least amount of SV2 expression available (fig. S4, D and E). Binding of BoNT/A to NMJs of these mice was reduced by 72% as compared with that of wild-type controls (Fig. 4, A and B).

Fig. 4.

SV2 knockout mice have reduced BoNT/A binding activity at diaphragm motor nerve terminals and are less sensitive to BoNT/A. (A and B) Binding of BoNT/A to mouse hemidiaphragms, prepared from wild-type (WT) and SV2A+/–SV2B–/– mice, was assayed as described in Fig. 1A. SV2A+/–SV2B–/– mice exhibit reduced BoNT/A binding activity compared with that of wild-type mice (72% reduction, P < 0.001, t test, n = 47), whereas binding of BoNT/B was the same (P > 0.05, t test, n = 20). Error bars show SEM. (C) The same amount of BoNT/A was injected into SV2B–/– mice and their wild-type littermates. SV2B–/– mice live longer than wild-type mice (43 min versus 33 min, P < 0.05, paired t test), corresponding to a 60% loss in toxicity in the SV2B knockout mice (30). LD50, mean lethal dose.

Finally, we carried out whole-animal studies to determine whether changes in SV2 expression in mice altered their susceptibility to BoNT/A. To minimize potential defects in synaptic transmission in vivo, we compared the BoNT/A sensitivity of SV2B knockout mice to their wild-type littermates, because SV2B knockout mice are phenotypically normal (26). Sensitivity to BoNT/A was assessed with an established rapid assay, in which large doses of toxin are injected intravenously (10, 11, 30). The survival time of each mouse was determined and converted to intraperitoneal toxicity (30). SV2B knockout mice survived significantly longer than wild-type littermates in this assay. The effective toxin concentration in SV2B knockout mice was reduced by 60% as compared with that of wild-type mice (Fig. 4C). The remaining toxicity in SV2B–/– mice was probably mediated by SV2A and SV2C, which are not altered, and by gangliosides, which serve as low-affinity receptors.

By using the secretory vesicle protein SV2 as its protein receptor, BoNT/A attacks active neurons with high selectivity because active neurons expose more receptors during exocytosis. Because BoNT/A blocks vesicle exocytosis, theoretically, a successful BoNT/A entry event will shut down the subsequent exposure of more receptors, allowing toxin molecules to enter active synapses that have yet to be poisoned. It has been shown that BoNT/B uses synaptotagmins I and II to enter cells (8, 9, 11). BoNT/G has also been reported to bind synaptotagmins I and II (31). Thus, secretory vesicle recycling pathways may have been exploited by several BoNTs, thereby contributing to the efficiency of this class of toxins.

Supporting Online Material

Materials and Methods

Figs. S1 to S6


References and Notes

View Abstract

Navigate This Article