Identification of Fibroblast Growth Factor Receptor 3 (FGFR3) as a Protein Receptor for Botulinum Neurotoxin Serotype A (BoNT/A)

Botulinum neurotoxin serotype A (BoNT/A) causes transient muscle paralysis by entering motor nerve terminals (MNTs) where it cleaves the SNARE protein Synaptosomal-associated protein 25 (SNAP25206) to yield SNAP25197. Cleavage of SNAP25 results in blockage of synaptic vesicle fusion and inhibition of the release of acetylcholine. The specific uptake of BoNT/A into pre-synaptic nerve terminals is a tightly controlled multistep process, involving a combination of high and low affinity receptors. Interestingly, the C-terminal binding domain region of BoNT/A, HC/A, is homologous to fibroblast growth factors (FGFs), making it a possible ligand for Fibroblast Growth Factor Receptors (FGFRs). Here we present data supporting the identification of Fibroblast Growth Factor Receptor 3 (FGFR3) as a high affinity receptor for BoNT/A in neuronal cells. HC/A binds with high affinity to the two extra-cellular loops of FGFR3 and acts similar to an agonist ligand for FGFR3, resulting in phosphorylation of the receptor. Native ligands for FGFR3; FGF1, FGF2, and FGF9 compete for binding to FGFR3 and block BoNT/A cellular uptake. These findings show that FGFR3 plays a pivotal role in the specific uptake of BoNT/A across the cell membrane being part of a larger receptor complex involving ganglioside- and protein-protein interactions.


Introduction
Botulinum neurotoxin serotype A (BoNT/A) is produced by Clostridium botulinum and is a member of the Clostridial neurotoxin family that includes BoNT/A-G and Tetanus neurotoxin (TeNT). BoNT/A causes transient muscle paralysis by entering motor nerve terminals (MNTs) where it cleaves nine amino acids from the C-terminus of the soluble N-ethylmaleimide-sensitive factor attachment receptor (SNARE) protein SNAP25 (SNAP25 206 ) to yield SNAP25 197 [1]. Intact SNAP25 is required for neurotransmitter release and cleavage of SNAP25 disrupts exocytosis, which blocks neurotransmitter release [2][3][4][5]. BoNT/A has become a useful pharmacological and biological tool. Because of its high potency and specificity for pre-synaptic nerve terminals, BoNT/A at picomolar concentrations, is used to treat a wide range of neuromuscular disorders [6][7][8], pain disorders including migraine [9], and excessive sweating [10]. The key to the exceptional specificity of BoNT/A is believed to be the mechanism of uptake across the presynaptic membrane of neurons that involves a combination of low and high affinity interactions known as the double receptor model [11][12][13].
The low affinity receptor for BoNT/A is the ganglioside GT1b with a binding pocket within the C-terminal portion of the receptor binding domain [12,14,15]. According to the APR receptor model [13], an array of presynaptic receptors (APRs), clustered in microdomains at the presynaptic membrane, are responsible for specific uptake of neurotoxins, including BoNT/A. It is the binding to high density ganglioside GT1b that mediates the initial binding step and via a low affinity interaction concentrates BoNT/A on the cell surface. GT1b has been shown to bind BoNT/A with a K D ,200 nM in vitro [16]. Once anchored in the membrane, lateral movements within the plasma membrane facilitate intermolecular interaction of BoNT/A with additional lower density but higher affinity protein receptors, including the three isoforms of Synaptic Vesicle (SV) glycoprotein 2, SV2A (ENSG00000159164), B (ENSG00000185518) and C (ENSG00000122012) that are exposed on the outer plasma membrane after fusion of synaptic vesicles to the presynaptic membrane [17][18][19][20][21][22]. BoNT/A specifically recognizes the fourth luminal domain (LD4) of SV2 [17,18]. The specific sequence in the BoNT/A binding domain that interacts with SV2 has not been identified [23]. Glycosylated SV2A, B, and C have also been identified as receptors for BoNT/F [22,24] and glycosylated SV2A and B have been identified as receptors for BoNT/E [20]. BoNT/ D was reported to enter neurons via two ganglioside binding sites, one site at a position previously identified in BoNT/A, B, E, F, and G, and the other site resembling the second gangliosidebinding pocket of TeNT [25]. Recently, BoNT/D has also been shown to use SV2 (all three isoforms) to enter hippocampal neurons, but BoNT/D bound SV2 via a mechanism distinct from BoNT/A and E [26]. Surprisingly, SV2A and SV2B have also been reported to mediate binding and entry of TeNT into central neurons [27].
Analysis of the first crystallographic structure of BoNT/A revealed a ganglioside binding site with structural homology to that within TeNT [28,29] (Figure 1A). A potential protein binding site was also identified with structural homology to basic fibroblast growth factor (FGFb or FGF2; ENSG00000138685), agglutinin, and the toxin abrin. The report of a potential FGF2 protein binding site within BoNT/A served as the basis for the identification of FGFR3 (Fibroblast Growth Factor Receptor 3, ENSG00000068078) as a receptor for BoNT/A. FGFR3 [30,31] is one of four receptor-tyrosine kinases (FGFR1-4) that act as receptors for FGFs. FGFRs are composed of an extra-cellular ligand-binding domain consisting of three immunoglobulin-like loops (L1-L3), a transmembrane domain, and a split cytoplasmic tyrosine kinase domain. FGFRs are activated by dimerization induced by ligand binding that enables the cytoplasmic kinase domains to transphosphorylate one another at specific tyrosine residues [32][33][34][35]. FGFR1-3, but not FGFR4, exist in three different splice variants that differ in the C-terminal half of L3 and determine their individual ligand affinity and specificity. The splice variants are referred to as ''a'', ''b'', and ''c'' [36][37][38][39]. The ''b'' and ''c'' variants are expressed on the cell surface, while the ''a'' splice variant, which lacks the transmembrane domain [40], becomes a secreted extra-cellular FGF-binding protein [41]. Among the 22 known FGFs, FGF1 (ENSG00000113578), members of the subfamilies FGF8 (FGF8, 17,18) and FGF9 (FGF9; ENSG00000102678, 16,20) have been shown to function as ligands for both FGFR3b and c but with different levels of affinity. FGF2, and the subfamilies FGF4 (FGF4, 5,6) and FGF19 (FGF19, 21,23) have been shown to function as ligands for FGFR3c [39,42]. Moreover, FGFs bind their receptors in the presence of one or more low affinity co-factors including heparin sulfate (HS), gangliosides, neuropilin-1, Klotho, and anosmin that function to modulate receptor activity [43][44][45][46][47][48].
The studies presented in this manuscript identify FGFR3 expressed in motor neurons at MNTs as a functional protein receptor for BoNT/A. The C-terminal binding domain of BoNT/ A, H C /A, binds to the second and third extra-cellular ligand binding domain of FGFR3 and results in the phosphorylation of FGFR3. It is demonstrated that cellular uptake of BoNT/A is dependent on the level of FGFR3 expression. Native ligands for FGFR3; FGF1, FGF2, and FGF9 compete for binding to FGFR3 and block BoNT/A uptake in a cell-based assay. Moreover, peptides derived from the FGFR3 subtype b and c extra-cellular domain block BoNT/A uptake in neuronal cells. Both FGFR3 subtype b and c bind to rH C /A, but FGFR3b has the highest affinity with a K D ,15 nM in vitro. These data suggest that FGFR3 is a potential high affinity component of a receptor complex for BoNT/A on the presynaptic membrane.

Identification of FGFR3 as a functional receptor for BoNT/A
Analysis of the BoNT/A crystal structure ( Figure 1A) revealed that the H C /A subdomain has structural homology to basic fibroblast growth factor (FGF) [28]. To investigate the interaction of BoNT/A and FGFR, pull-down assays were performed with Neuro-2a and PC-12 cell lines that have been shown to take up BoNT/A with high efficacy after differentiation by serum starvation and trophic factors, and Nerve Growth Factor (NGF) for PC-12 cells [49,50]. Figure S1A shows a representative experiment of how differentiation increases BoNT/A uptake in PC-12 cells. BoNT/A uptake was determined by treatment of cells with BoNT/A, followed by incubation and Western blot analysis of the SNAP25 197 cleavage product. After optimization of differentiation and treatment conditions, differentiated Neuro-2a and PC-12 cells were treated with BoNT/A and EC 50 values of 6065 pM and 47.1613 pM, respectively were determined ( Figure S1B).
A complex containing BoNT/A and its receptor was isolated using three alternative pull-down methods, the results from two of these methods are shown here. First, Sulfo-SBED BoNT/A was used in a biotin transfer experiment to pull down the receptor from intact Neuro-2a cells. Antibodies against FGFR3 detected a band of the correct molecular weight for FGFR3 as part of a 250 kDa complex with BoNT/A. Figure S1C, D, and E demonstrate that a 250 kDa protein complex was isolated and that specific bands for BoNT/A and FGFR3 within this complex could be detected using antibodies specific to BoNT/A or FGFR3. Second, a complex containing both BoNT/A and FGFR3 was isolated from Neuro-2a cells treated with biotin labeled BoNT/A and the cross linking reagent Bis(Sulfosuccinimidyl) suberate (data not shown). Finally, the recombinant binding domain of BoNT/A, His-rH C /A, was used to pull down FGFR3 from differentiated PC-12 cell lysates without the use of cross-linking reagents, demonstrating a strong interaction ( Figure 1B).

Author Summary
Botulinum neurotoxin serotype A (BoNT/A) is one of seven neurotoxins (BoNT/A-G), produced by the bacteria Clostridium botulinum that are both poisons and versatile therapeutics. These toxins enter motor neurons where they prevent the release of acetylcholine at the neuromuscular junction. The specific uptake of BoNT/A across the neuronal cell membrane is dependent on specific receptor interactions. Binding to high density ganglioside GT1b mediates the initial binding step and via a low affinity interaction concentrates BoNT/A on the cell surface. Once anchored in the membrane, lateral movements within the plasma membrane facilitate intermolecular interactions of BoNT/A with additional lower density but higher affinity protein receptors. Here we present data supporting the identification of Fibroblast Growth Factor Receptor 3 (FGFR3) as a high affinity receptor for BoNT/A. We show that BoNT/A binds to FGFR3 with high affinity and functions as an agonist ligand for FGFR3. The identification of this novel receptor for BoNT/A represents an important advance in the understanding of the mechanism of action of BoNT/A, especially on the initial steps of neuronal uptake, and can be the basis for the development of new specific countermeasures and new BoNT/A-based therapeutics.
Having identified FGFR3 as a binding partner for BoNT/A, we investigated the role of FGFR3 as a functional receptor for BoNT/ A. Competition experiments utilizing native ligands for FGFR3; FGF1, FGF2, and FGF9 [39,42], demonstrated that these ligands competed for binding to FGFR3 and blocked BoNT/A uptake in a cell-based assay with differentiated Neuro-2a cells (Figure 1C-D). rH C /A, was used as a positive control and produced a strong blockade of BoNT/A uptake. As a negative control, FGF10 (ENSG00000070193), which is not a ligand for FGFR3, but closely related to the other FGF ligands tested, was used. Pre-incubation with FGF10 did not affect BoNT/A uptake. The experimental data from at least four independent experiments were compiled and fitted to a non-linear exponential decay model; Y = 100*e-CC*log(concentration). The Competition Constant (CC) for each fitted curve was calculated and demonstrated similar competition of the three FGF ligands. The data strongly suggest that BoNT/A utilizes FGFR3 to gain entry into neuronal cells since native FGFR3 ligands blocked its uptake. The hypothesis that BoNT/A acts as an agonist for FGFR3 was further supported by demonstrating that treatment with rH C /A resulted in phosphorylation of FGFR3, achieving similar levels of activation as cells treated with identical concentrations of FGF2 ( Figure 1E-F).

FGFR3b Loop 2,3 binds to rH C /A with high affinity
The ligand binding site for FGFR3 has been identified as the second and third extra-cellular loops of FGFR3 ( Figure 2A) [39,51,52]. To further verify the functional role of FGFR3 as a receptor for BoNT/A, we demonstrated that pre-incubation of BoNT/A with a peptide spanning the second and third extracellular loops of FGFR3b (FGFR3b Loop 2,3) inhibited BoNT/A uptake presumably via binding to the receptor binding domain of BoNT/A ( Figure 2C). Inhibition, although to a lesser extent, was also observed using the peptide spanning the luminal domain (LD4) of SV2C, SV2C 529-579 ( Figure 2C). SV2C 529-579 ( Figure 2B) has previously been reported as the minimal peptide region for binding to BoNT/A [17]. As a positive control for inhibition, BoNT/A was pre-incubated with a neutralizing monoclonal antibody directed to the binding domain of BoNT/A, Anti-H C /A. As a negative control, Synaptotagmin II (aa1-20, Syt II 1-20 ), the receptor for BoNT/B [53][54][55][56] was used. The experimental data from at least three independent experiments were compiled and fitted to a non-linear exponential decay model; Y = 100*e-IC*log(concentration). The Inhibition Constant (IC) for each fitted curve was calculated and demonstrated that FGFR3b Loop 2,3 inhibited BoNT/A uptake and was a more effective uptake inhibitor than SV2C 529-579 ( Figure 2D).
Initial experiments designed to explore the combined effect of FGFR3b Loop 2,3 and SV2C 529-579 showed that the peptides bound with good affinity in vitro (data not shown). To address the question as to whether FGFR3 and SV2 interact in neurons, we performed a series of Co-IPs experiments. We tested if an antibody to FGFR3 could pull-down SV2 isoforms from a differentiated Neuro-2a cell lysate, and vice versa, antibodies to SV2 isoforms could pull-down FGFR3. An interaction between FGFR3 and SV2 was detected using the Anti-SV2B (sc-28956) antibody, which recognizes SV2B and, to a lesser extent, SV2A and SV2C. No bands were detected when using antibodies for SV2A or SV2C (data not shown). The result suggests that FGFR3 and SV2B interact in differentiated Neuro-2a cells ( Figure 2E-F).
To characterize the binding of FGFR3 and SV2C to rH C /A, the binding affinity of the two receptor surrogate peptides, FGFR3b Loop 2,3 and SV2C 529-579 to rH C /A was tested in a Surface Plasmon Resonance (SPR) binding assay. FGFR3b Loop 2,3 bound to rH C /A with an average K D = 15.063 nM, n = 4, k a = 1.77E+04 1/Ms, k d = 2.40E-04 1/s ( Figure 3B-C). This is similar to what has been reported earlier upon binding of FGFR2b Loop 2,3 to FGF2, K D = 12.860.3 nM [57]. SV2C 529-579 bound to rH C /A with an average K D = 10566 nM, n = 3, k a = 2.34E+03 1/Ms, k d = 2.47E-04 1/s ( Figure 3A,C). The difference in affinity between FGFR3b Loop 2,3 and SV2C 529-579 is due to a 10 times faster association, k a is estimated to be 10 times higher for FGFR3b Loop 2,3 versus SV2C 529-579 . It can be seen on the curves as a more shallow slope and longer time to equilibrium for SV2C versus FGFR3b Loop 2,3 ( Figure 3A versus 3B).
Having identified rH C /A as an agonist ligand for FGFR3 ( Figure 1E-F) and shown that rH C /A binds to FGFR3b Loop 2,3 in vitro with similar affinity as native ligands for FGFR3, we evaluated if FGFR3 would facilitate uptake of rH C /A and native ligands in a similar fashion. We utilized HEK 293 cells as a model system, because they express FGFR3 but no measurable levels of any of the SV2 isoforms ( Figure S1G). Consequently, uptake of rH C /A via SV2 should be absent in these cells. HEK 293 cells do not express SNAP25 and therefore SNAP25 cleavage could not be used as a measure for BoNT/A uptake. Instead, uptake was measured as an increase in intracellular fluorescence after addition of fluorescently labeled rH C /A or FGF2, a native ligand for FGFR3. The results showed slightly less uptake of rH C /A compared to FGF2, but similar kinetics ( Figure S1H). The slightly higher uptake of FGF2 compared to rH C /A could be due to FGF2 having more receptor targets, since it is a general ligand for FGFRs. These data suggest that FGFR3 can mediate BoNT/A uptake independently of SV2.
To explore the binding sites of FGFR3b Loop 2,3 and SV2C 529-579 on the binding domain of BoNT/A, we performed a series of dual binding experiments using the BIAcore. We tested if the peptides and anti-H C /A, the neutralizing monoclonal antibody previously used in the cell based inhibition assay ( Figure 2C & D), could bind to rH C /A simultaneously. rH C /A was captured by anti-H C /A monoclonal and FGFR3b Loop 2,3 or SV2C 529-579 were flowed across. The results show that binding of anti-H C /A monoclonal blocks binding of FGFR3b Loop 2,3, but not SV2C 529-579 to rH C /A in vitro ( Figure 3F-G), demonstrating that FGFR3 and SV2 bind to different sites on the BoNT/A binding domain. Interestingly, these data also suggest that inhibition of BoNT/A uptake by the neutralizing monoclonal anti-H C /A antibody in the cell based assay is due to blockage of FGFR3 binding.
These results demonstrate that FGFR3b Loop 2,3 and SV2C 529-579 can both inhibit the activity of BoNT/A in a cellbased assay, but FGFR3b Loop 2,3 is a stronger inhibitor than SV2C 529-579 . They show that in an in vitro binding assay, the binding affinity for FGFR3b Loop 2,3 upon binding to rH C /A is higher, due to an estimated 10 times faster association, than the binding affinity for SV2C 529-579 upon binding to rH C /A. The binding affinity for FGFR3b Loop 2,3 to rH C /A, is similar or identical, to the binding affinity for FGFR3b Loop 2,3 upon binding to FGF2 and FGF9, two native ligands for FGFR3. Also, uptake of rH C /A in HEK 293 cells, that express FGFR3, but not SV2, is comparable to uptake of FGF2, supporting a case for uptake of BoNT/A via FGFR3 independent of the presence of SV2. Finally, dual in vitro binding studies using a neutralizing antibody to H C /A, show that the FGFR3 and SV2C peptides bind to rH C /A at different sites, FGFR3 at a site close to or overlapping the binding site for the anti-H C /A, and SV2C in a site distal from the anti-H C /A binding site. Different binding sites for FGFR3 and SV2 would allow a multi-receptor complex to form.

FGFR3 expression levels affect the sensitivity of cells to BoNT/A
Differentiation of neuronal cells increases BoNT/A uptake ( Figure S1A and B). It has been suggested that the increased  sensitivity in differentiated PC-12 cells is due to increased expression of the SNAP25b subtype that is most sensitive to BoNT/A [50]. Since the increased sensitivity could also be a result of increased expression of a receptor for BoNT/A, we studied the expression of FGFR3 as well as SV2A, B, and C before and after differentiation in both Neuro-2a and PC-12 cells. FGFR3 expression levels were similar in both cell lines and the amount of FGFR3 was unchanged after differentiation. Neuro-2a cells expressed mostly SV2C, while PC-12 cells expressed all three SV2 isoforms. Surprisingly, differentiation resulted in decreased expression of SV2 isoforms in both cell lines ( Figure S1F).
Assuming FGFR3 is a functional receptor for BoNT/A, one would expect overexpression of FGFR3 to result in increased binding of BoNT/A on the cell membrane. If receptor binding is a rate-limiting step, this should also result in increased sensitivity to BoNT/A. Experiments to test the sensitivity to BoNT/A were performed under non-depolarizing conditions, where the exposure of SV2 on the cell surface is presumed to be limited. Overexpression of FGFR3 in PC-12 and Neuro-2a cells increased the sensitivity to BoNT/A and produced higher efficacy (increased maximal signal), in a Western blot SNAP25 197 cell-based assay, while overexpression of SV2C did not ( Figure 4A-B).
FGFR3 overexpression also increased binding of transfected cell membranes to rH C /A in a SPR binding assay, while overexpression of SV2C did not ( Figure S2A and C), suggesting that if there is more FGFR3 on the cell surface more BoNT/A will bind, while more SV2C does not increase BoNT/A binding. Human neuroblastoma SH-SY5Y cells were also evaluated in the SPR binding assay because they have low sensitivity to BoNT/A [58] and express very little endogenous SV2C. Even in this situation, there was no effect as a result of over expressing SV2C ( Figure  S2B).
We also demonstrated, utilizing shRNA, that reduced expression of FGFR3 resulted in reduced sensitivity to BoNT/A. A 65% reduction of FGFR3 protein expression resulted in a 5.7-fold decrease in potency and a ,5-fold increase in EC 50 when compared to control cells ( Figure 2C). No change in the protein expression levels of either SV2A, B, or C was detected in those samples. A separate experiment with siRNAs for FGFR3 and SV2C demonstrated that a 4.2-fold reduction in SV2C mRNA resulting in a 2-fold reduction in protein levels did not cause a reduction in BoNT/A uptake ( Figure S2E-F). While a 3-fold reduction of FGFR3 mRNA resulting in a 2-fold reduction in protein levels reduced sensitivity to BoNT/A causing a 3-fold shift in relative potency when compared to control cells ( Figure S2D and F), confirming that, under non-depolarizing conditions, binding to FGFR3 is a rate-limiting step in BoNT/A uptake.
The interaction between FGFR3 and rH C /A was also observed in a photobleaching experiment using the FRET partners AF-488 and TMR. We detected an increase in the fluorescent signal from the AF-488 labeled rH C /A (donor) after photobleaching the TMR labeled FGFR3 (acceptor). The data shows that FGFR3 and rH C /A are proximal enough within PC-12 cells to FRET, suggesting that FGFR3 not only binds BoNT/A on the cell surface, but it is also trafficking with BoNT/A within the cells. There was little change in the fluorescence observed when the experiment was performed with TMR labeled SV2C as the acceptor ( Figure 4E-G).

FGFR3 is expressed at rat Motor Nerve Terminals
BoNT/A causes transient muscle paralysis through presynaptic blockade of acetylcholine release at the neuromuscular junction. If FGFR3 functions as a receptor for BoNT/A in vivo, then it would be reasonable to presume that FGFR3 should be expressed at the MNTs. The expression pattern of the FGFR3 receptor was examined on cross-sections of rat skeletal muscle to look for potential co-expression with SV2C, SNAP25, and nicotinic acetylcholine receptors (nAChRs). Overall, immuno-reactive (IR) staining for SV2C and SNAP25 were co-expressed exclusively at neuromuscular junctions (NMJs) throughout the muscle ( Figure 5A, A-D). These NMJs were specifically defined by using fluorescently labeled a-bungarotoxin (a-Bgt) nAChRs. In contrast, FGFR3-IR was not only detected at NMJs, but also in extrasynaptic structures, such as myoblasts and blood vessels ( Figure 5A, E). At the NMJs however, the FGFR3 staining pattern corresponded to that of SNAP25 and nAChRs ( Figure 5A, E-H). To verify expression of SV2C and FGFR3 within BoNT/A sensitive NMJs, we treated rat Tibialis Anterior (TA) muscles with BoNT/A and analyzed the staining patterns for SV2C and FGFR3 together with IR-staining for cleaved SNAP25 (SNAP25 197 ). Focusing on individual synapses, we observed overlapping patterns for SV2C-IR and SNAP25 197 -IR that were adjacent to the pattern of post-synaptic nAChR expression ( Figure 5A, I-L). Similarly, the patterns for FGFR3-IR and SNAP25 197 -IR at the NMJ were overlapping and appeared adjacent to the pattern of nAChR expression ( Figure 5A, M-P). Saline-treated rat muscles showed no immuno-staining for SNAP25 197 ( Figure 5B). These qualitative results demonstrate that FGFR3 receptors are present on MNTs and are co-expressed with SV2C and SNAP25.
The second and third extra-cellular loop of FGFR3 is the minimal optimal binding site for H C /A We have shown that FGFR3b Loop 2,3 binds to BoNT/A with low nanomolar affinity. To further identify the binding site for BoNT/A and to test whether the two subtypes of FGFR3, subtype b and c, bound with similar affinities, we constructed eight deletion mutants of FGFR3, containing either FGFR3 Loop 1,2,3 (long or short version), Loop 2,3, or Loop 3 of both subtypes ( Figure 6A). The difference between FGFR3b and FGFR3c lies in the most C-terminal glycosylation sites within the LD4 luminal domain and it is highly glycosylated in vivo [20,71,72]. The red arrows mark the 6 kDa peptide, SV2C 529-579 , which spans part of the LD4 luminal domain of SV2C. (C) Pre-incubation of BoNT/A with FGFR3b Loop 2,3 or SV2C 529-579 inhibit BoNT/A uptake in Neuro-2a cells. BoNT/A, at 1 nM concentration, was incubated for 20 min with increasing concentrations of FGFR3b Loop 2,3 or SV2C 529-579 before treatment of cells. In parallel, BoNT/A was incubated with neutralizing antibodies to H C /A (positive control), or with Syt II 1-20 (negative control). Data are shown as the percentage inhibition compared to inhibition after addition of Anti-H C /A. SNAP25 cleavage, as a measure of BoNT/A uptake, decreased when BoNT/A was pre-incubated with Anti-H C /A, FGFR3b Loop 2,3, or SV2C 529-579 . No inhibition was observed after addition of Syt II [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] . The averages of three or more experiments are included. (D) Data was fitted to a non-linear exponential decay model; Y = 100*e-IC*log(concentration). As a measure for inhibition, the Inhibition Constant (IC) and standard error (SE) for each inhibitor are shown. The IC values are all significantly different from each other, P-value#0.0045. (E-F) Protein Complex Immunoprecipitation (Co-IP) and Western blot (WB). IP's were performed with antibody to FGFR3 (sc-123), SV2A (sc-28955), SV2B (sc-28956), and SV2C (sc-28957). WB's were performed with antibody to FGFR3 (sc-123) (E) and SV2B (sc-28956) (F). FGFR3 and SV2 are glycosylated in vivo and multiple bands are detected on WB. The predicted MW for nonglycosylated protein is 88 kDa for FGFR3 and 77 kDa for SV2B. The positive control, IP and WB with the same antibody, showed clear bands on both blots and the negative control (antibody only, no lysate) showed no bands. An interaction between FGFR3 and SV2 was detected, in both directions, using the Anti-SV2B (sc-28956) antibody that recognizes SV2B and, to a lesser extent, SV2A and SV2C. doi: 10 Figure S3A). All the deletion mutant FGFR3 peptides were able to inhibit BoNT/A uptake in a cell-based inhibition assay, presumably via binding to the receptor binding domain of BoNT/A and preventing binding to cells ( Figure 6B-C). However, in a SPR binding assay a significantly lower affinity was observed for the peptides spanning only Loop 3 compared to the peptides spanning Loop 2,3 or Loop 1,2,3. The association (on-rate) of the peptides spanning only Loop 3 was ,10 times lower than the on-rate of the longer peptides, while the dissociation (off-rate) was similar ( Figure 6E and S3B-C). The similar off-rate can explain why the peptides are able to inhibit equally well BoNT/A binding to the receptor in the cell-based inhibition assay. In the cell-based assay sufficient time (20 min) is available for even slow associating peptides to bind and the ability to inhibit relies more on a slow dissociation. In the SPR binding assay, on the other hand, binding is observed in real time (5-10 min). Based on the lower on-rate of the Loop 3 peptides observed in the SPR binding assay, Loop 2,3, which is the binding region for native FGF ligands, was identified as the minimal optimal binding region for BoNT/A. FGFR3b bound with slightly higher affinity than FGFR3c ( Figure 6D).

Discussion
In this study we identified FGFR3 as a high affinity protein receptor for BoNT/A. Pull-down experiments with neuronal cells resulted in the identification of a protein complex containing BoNT/A and FGFR3. Native ligands for FGFR3; FGF1, FGF2, and FGF9 compete with rH C /A for binding to the receptor and binding of rH C /A results in phosphorylation of FGFR3, demonstrating that BoNT/A acts as an agonist ligand for FGFR3. Since ligand binding and activation of FGFRs are known to result in receptor-mediated endocytosis of both receptor and ligand [59], we propose that binding of BoNT/A to FGFR3 also results in endocytosis and that FGFR3 may mediate BoNT/A uptake in both stimulation independent and stimulation dependent manners. This hypothesis is supported by the fact that depolarization of nerve cells increases uptake (stimulation dependent), while at the same time BoNT/A uptake can take place in resting neurons (stimulation independent) [3,4,[60][61][62][63]. This is also supported by the observation that the uptake, but not the initial binding step, is altered by nerve stimulation [64].
Motor neurons at MNTs take up BoNT/A with high affinity, resulting in inhibition of exocytosis and muscle paralysis. Thus, MNTs should presumably express a BoNT/A receptor(s), and our results clearly demonstrated that both FGFR3 and SV2C are present at MNTs. These data support the hypothesis that FGFR3 functions as a high affinity receptor for BoNT/A uptake, and that most likely, SV2 is only available as a receptor for BoNT/A after depolarization and vesicular exocytosis.
Using a SPR binding assay, we demonstrated that a peptide spanning the second and third extra-cellular loops of FGFR3, FGFR3b Loop 2,3, binds to rH C /A with a K D ,15 nM and that a peptide spanning the luminal domain of SV2C, SV2C 529-579 , binds to rH C /A with a K D ,100 nM in vitro. The observed ,15 nM affinity for binding of rH C /A to FGFR3b Loop 2,3 was similar or identical to the affinity for binding of two native ligands for FGFR3, FGF2 and FGF9 to FGFR3b Loop 2,3 in the same assay. Also, comparable uptake of rH C /A and FGF2 was observed in HEK 293 cells, a cell line that express FGFR3 and not any of the SV2 isoforms, suggesting that FGFR3 can mediate uptake of BoNT/A independently of SV2.
A recent publication [63] clearly supports our findings. The authors observed limited co-localization of SV2C and H C /A or BoNT/A after treatment of spinal cord motor neurons under resting conditions and this co-localization did not significantly increase under depolarizing conditions. Moreover, inhibition of exocytosis by pre-treatment with BoNT/D did not prevent the internalization of H C /A. The authors concluded that BoNT/A may exploit an alternative pathway(s), largely independent of stimulated synaptic endo-exocytosis, to enter neuronal cells in both resting and depolarizing conditions.
Pre-incubation of BoNT/A with the FGFR3 and SV2C peptides before treatment of cells, blocked uptake in neuronal cells, presumably by interacting with the binding domain of Figure 3. FGFR3b Loop 2,3 binds to rH C /A in vitro with high affinity compared to SV2C 529-579 and the affinity is similar to the binding affinity of the native ligands FGF2 and FGF9. (A) BIAcore SPR Binding Affinity measurements of SV2C 529-579 binding to rH C /A. Shown here are sensorgrams from a single experiment, including the binding of increasing concentrations of SV2C 529-579 : 21, 61, 185, 556, 1667, and 5000 nM respectively to a CM5 sensor chip covered with rH C /A (,3000 RUs). The curves were fitted to a 1:1 kinetic-binding model (A + B « AB) with the BIAevaluation 3.0 software and the binding constants; the association constant (k a ), dissociation constant (k d ), and the binding affinity at equilibrium (K D, = K d /K a (at equilibrium)) were estimated. The binding constants for the experiment shown were estimated to; K D = 96.2 nM, k a = 1.29E+03 1/Ms, k d = 1.24E-04 1/s. (B) BIAcore SPR Binding Affinity measurements of FGFR3b Loop 2,3 binding to rH C /A. Shown here are sensorgrams from a single experiment, including the binding of increasing concentrations of FGFR3b Loop 2,3: 12, 37, 111, 333, and 1000 nM respectively to a CM5 sensor chip covered with rH C /A (,3000 RUs). The curves were fitted to a 1:1 kinetic-binding model (A + B « AB) with the BIAevaluation 3.0 software and the binding constants; k a , k d , and K D, = K d /K a (at equilibrium) were estimated. The binding constants for the experiment shown were estimated to; K D = 15.1 nM, k a = 1. . The result show that SV2C 529-579 and anti-H C /A antibody can bind simultaneously to rH C /A, while FGFR3b Loop 2,3 binding is abolished. Background drift from dissociating rH C /A was subtracted. As a negative control, the peptides were flowed across anti-H C /A antibody before capture of rH C /A. As a positive control, the peptides were flowed across a surface with ,1000 RU immobilized rH C /A in parallel. Upon immobilization by amine coupling the protein is randomly oriented, in contrast to what occurs when the protein is captured by antibody, where the protein is uniformly oriented. This means that the number of free binding sites differs despite equal amounts of ligand protein used in the experiment. This can be seen as a larger binding response for SV2C 529-579 flowed across captured rH C /A compared to amine-immobilized rH C /A. These data suggest that the anti-H C /A monoclonal antibody binds to an epitope close to or at the same site as FGFR3b Loop2, 3  BoNT/A and preventing binding to the receptor on cells. In accordance with the observed lower affinity for SV2C 529-579 compared to Loop 2,3 of FGFR3, the FGFR3 peptide produced a stronger blockade than the SV2C peptide. These data suggest that FGFR3 may function as a high affinity receptor for BoNT/A and that SV2C may function as a medium affinity receptor.
SPR experiments demonstrated that the FGFR3b and SV2C peptides bind to different sites on H C /A, FGFR3 in a site overlapping the epitope of a neutralizing monoclonal antibody to H C /A (6B1, provided by Dr. L. Smith, USAMRIID) and SV2C in a site distal from both. So far the binding site for SV2 has not been identified, but it has been suggested that SV2 binds to the Cterminal half of the binding domain, H CC , similar to how Synaptotagmin II binds to BoNT/B [23]. Different binding sites for FGFR3 and SV2 on the binding domain of BoNT/A would allow formation of a multi-receptor complex. Interestingly, Co-IP experiments show that FGFR3 and SV2 can interact in live Neuro-2a cells, suggesting a step-wise binding and/or formation of a multi-receptor BoNT/A complex.
We concur with others in the field that the specificity of BoNT/ A for neuronal cells and specially motor neurons, which is higher than binding to a single receptor can explain, is due to the fact that uptake of BoNT/A is a multi-step process involving at least two crucial steps [13,23]. The first crucial step is binding to gangliosides like GT1b (KD,200 nM) that are abundantly present in the outer leaflet of the plasma membrane of neuronal  Figure S2) . (C) shRNA knockdown of FGFR3. Dose response curves of PC-12 cells stably transfected with Scrambled (negative control) or FGFR3 shRNAs and then treated with 0-30 nM BoNT/A for 6 hours. The curves were plotted as cleaved SNAP25 (SNAP25 197 ) vs. BoNT/A concentration. Cells transfected with shFGFR3 were less sensitive to BoNT/A, the EC 50 was ,5-fold higher, compared to cells transfected with random control shRNA. (D) Table showing the average fold difference in relative potency. The relative potency is calculated by dividing the potency of a test sample with the potency of a reference sample. A relative potency of 1 means that both samples are identical. If the test sample is more potent than the reference the relative potency is smaller than 1. If the sample is less potent than the reference, then the relative potency is higher than one. In this case, cells transfected with FGFR3 shRNA are 5.7-fold less sensitive than cells transfected with scrambled shRNA (See also Figure S2). (E-F) Photobleaching experiment where the recovery of AF-488 labeled rH C /A (donor) fluorescence was measured after photobleaching of either TMR labeled FGFR3 (E) or TMR labeled SV2C (acceptor). (G)

cells. This initial step increases the local concentration of BoNT/A
and allows it to diffuse in the plane of the membrane to bind its protein receptor(s) [15,16], similar to what has been observed for heparin sulfate and FGF2 [65]. BoNT/A that is diffusing within microdomains of the plasma membrane will be presented to FGFR3 and/or SV2, bind to the receptor, and undergo endocytosis representing a second crucial step.
There are several lines of evidence suggesting that the initial binding to gangliosides is critical to specifically accumulate BoNT/ A on the membrane of neuronal cells. For example, It has been shown [66] that, in the absence of GT1b, Neuro-2a cells are insensitive to BoNT/A and that knockout mice defective in the production of polysialogangliosides show reduced sensitivity to BoNT/A and BoNT/B [20,26,67]. Moreover, a mutant version of H C /A, W1266L & Y1267S that does not bind to GT1b, does not extend paralysis time caused by BoNT/A in murine phrenic nerve-hemidiaphragm preparations demonstrating an impaired ability to bind to neuronal cells [68]. Here we propose that only after BoNT/A is anchored at the neuronal membrane the second crucial step, binding to FGFR3 and/or SV2, can occur. This explains how BoNT/A can specifically enter motor neurons by recognizing FGFR3, a receptor also expressed by non-neuronal cells that lack gangliosides in their membranes. As evidence for a second crucial step, we demonstrate that, in Neuro-2a cells, if either FGFR3 or SV2C binding is blocked, BoNT/A uptake is impaired.
BoNT/A uptake is affected by the cellular levels of FGFR3 expression. We demonstrated that overexpression of FGFR3 increased binding of membrane extracts to rH C /A as well as BoNT/A uptake in three different neuronal cell lines, while downregulation of FGFR3 reduced uptake of BoNT/A. In contrast, no changes in BoNT/A uptake were observed when increasing or decreasing the expression of SV2C, suggesting that FGFR3, but not SV2C represents a rate-limiting step in BoNT/A uptake under resting conditions. This is consistent with our finding that FGFR3, but not SV2C, co-localized with BoNT/A in un-stimulated PC-12 cells.
By testing eight deletion mutant peptides of the FGFR3b and c extra-cellular domain in the SPR binding assay, we identified the extra-cellular Loop 2,3 of FGFR3 as the minimal optimal binding site for rH C /A. Native ligands for FGFR3 also bind to Loop 2,3 [39,51,52] and these data demonstrate that the binding site for rH C /A overlaps the binding site for native ligands of FGFR3. The affinity measurements also demonstrated that FGFR3b bound with slightly higher affinity than FGFR3c, the K D for the subtype b peptides was ,15 nM, while the K D for the subtype c peptides was ,25 nM. The FGFR3c subtype is the subtype expressed in the nervous system, while expression of the FGFR3b subtype is restricted to epithelial structures [69,70]. It is therefore more likely that BoNT/A utilizes the FGFR3c subtype in vivo to gain access into neuronal cells.
In conclusion, this paper presents evidence for FGFR3 as a high affinity receptor for BoNT/A, potentially being part of a larger receptor complex involving sugar-and protein-protein interactions. FGFR3 is present in the target motor neurons. Overexpression of FGFR3 in several neuronal cells increases efficacy and sensitivity to BoNT/A while decreased FGFR3 expression renders the cells less sensitive. BoNT/A binds to FGFR3 at the same extra-cellular region and with the same affinity as native ligands for FGFR3 and functions as an agonist ligand inducing FGFR3 phosphorylation. Moreover, BoNT/A uptake can be blocked by native FGFR3 ligands or by peptide fragments containing the extra-cellular region of FGFR3. Together, these results expand our knowledge of BoNT/ A uptake in neuronal cells and present a potential new pathway mediating BoNT/A entry and trafficking into neurons under both resting and depolarizing conditions.   Figure S3)). (B-C) Pre-incubation of BoNT/A with eight deletion mutant FGFR3 peptides, corresponding to the extra-cellular loops of FGFR3b/c (A) inhibit BoNT/A uptake in Neuro-2a cells. BoNT/A at 1 nM was incubated for 20 min with increasing concentrations of individual FGFR3 peptides before treatment of cells. In parallel, BoNT/A was pre-incubated with antibodies to H C /A (positive control), or Syt II 1-20 (negative control). Data are shown as percentage BoNT/A uptake relative to the uptake after pre-incubation with the negative control Syt II [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] . SNAP25 cleavage, as a measure of BoNT/A uptake, was decreased when BoNT/A was pre-incubated with Anti-H C /A or with any of the eight deletion mutant FGFR3 peptides. The averages of three or more experiments were included. The data was fitted to a non-linear exponential decay model; Y = 100*e-IC*log(concentration). As a measure for inhibition, the Inhibition Constant (IC) and standard error (SE) for each inhibitor are shown. The IC value for Syt II 1-20 was significantly lower than all the other IC values, p#0.0096 and p#0.0003 respectively. The IC value for Anti-H C /A was significantly higher than all the other IC values, p#0.0125 and p#0.0068 respectively. The IC value for FGFR3b Loop 2,3 was significantly higher than the IC values for the other FGFR3b peptides, p-value#0.0489. There was no significant difference in IC value among the FGFR3c peptides. (D) Results from testing the eight FGFR3 deletion mutant peptides in a BIAcore SPR binding assay. Increasing concentrations, 0-4000 nM of each of the deletion mutant FGFR3 peptides were flowed across a CM5 sensor chip covered with rH C /A. The curves were fitted to a 1:1 kinetic-binding model For differentiation, PC-12, Neuro-2a, and SH-SY5Y cells were plated in 96-well plates at 5610 4 cells/well in 100 ml differentiation media for three days.

His tag pull-down assay
The assay was performed according to the protocol from Pierce (21277; Pierce). 150 mg of rH C /A (26 mg/ml stock conc.) was used as ''Bait'' protein and 500 ml of differentiated PC-12 cell lysate (from 1.5610 6 cells) was used as ''Prey'' protein. As negative controls, samples without either ''Bait'' or ''Prey'' protein were run in parallel. The eluted samples were analyzed by SDS-PAGE and Western blot analysis.
Pull-down using Sulfo-SBED (Biotin transfer) 10 mg BoNT/A was reacted with 2 mM Sulfo-SBED (33073; Thermo Scientific) (solubilized at 125 mM in DMSO) in 0.1 ml PBS for 2 hours. The reaction was stopped by addition of 0.1 ml 0.4 M Tris. As a control 10 mg of BSA was also reacted with 2 mM Sulfo-SBED. Sulfo-SBED BoNT/A and BSA were added to 1610 8 Neuro-2a cells and mixed by rotisserie at 4uC for 4 hours. The reagent was photoactivated for 15 minutes with a UV light source. The cells were washed 4 times with cold TBS and then lysed by incubation for 2 hours in T-X-100 lysis buffer (50 mM Tris, 150 mM NaCl, 1% Triton X-100, 10 mM EDTA, pH 7.2). The biotinylated proteins were precipitated using Monomeric Avidin (Thermo Scientific), washed 4 times in the TX100 lysis buffer and then analyzed by SDS-PAGE and Western Blot Analysis. blocking agent. Secondary antibody was anti-rabbit IgG H+L HRP conjugate (81-6120; Invitrogen), anti-rabbit IgG veriBlot for IP secondary antibody (HRP) (ab131366, Abcam) (used for IP only), and anti-mouse IgG H+L HRP conjugate (62-6520; Invitrogen) diluted 1:5000 in TBS-T plus 2% blocking agent. Membranes were developed using ECL Plus Western Blotting System Detection Reagents (RPN2132; GE Healthcare). The Chemiluminescence was captured using a Typhoon 9140 (GE Healthcare) set to the following parameters: 455 nm excitation laser and detector set to all wavelengths below 520 nm emissions. The intensity of the gel bands were calculated using Image Quant software TL V2005 (GE Healthcare). The data was analyzed using PLA and SigmaPlot v 10.0 (Systat Software Inc.). Intensity values were plotted against concentration of BoNT/A in log scale and fitted to a 4-parameter logistics function (Y = Y 0 +a/[1+(X/X 0 ) b ]) without constraints. Based on the fitted curves the EC 50 values, corresponding to ''X 0 '', were determined.

Immunoprecipitation of phosphorylated membrane proteins
Differentiated Neuro-2a cells transfected with FGFR3 were treated with 0.5 nM or 50 nM FGF2 (233-FB; R&D Systems) or FGFR3 Is a Functional Protein Receptor for BoNT/A rH C /A (26 mg/ml stock concentration) for 10 minutes. Membrane extracts were prepared and 100 mg of total protein was incubated with 40 ml of a 50% slurry of anti-phosphotyrosine conjugated beads (16-101; Millipore) for 24 hours at 4uC. Samples were washed 4 times with MEB buffer (50 mM Tris pH 7.5, 150 mM NaCl) containing phosphatase inhibitor cocktail 1 and 2 (P2850 and P5726; Sigma) and complete protease inhibitor cocktail (11 873 580 001; Roche) and analyzed by SDS-PAGE and Western blot analysis, using antibody against FGFR3.

Differentiated cells (see Cell Lines and Growth Conditions)
were treated with 0-30 nM BoNT/A (0.41 mg/ml stock concentration) in differentiation media for 6 or 24 hours followed by overnight or two day incubation in toxin-free media. Cell lysates were analyzed by SDS-PAGE and Western blot analysis, using antibody against cleaved SNAP25, anti-SNAP25 197 (Allergan). The toxin containing media was then removed and replaced with fresh media followed by overnight incubation. Cells lysates were analyzed by SDS-PAGE and Western blot analysis, using antibody against cleaved SNAP25, Anti-SNAP25 197 . Blots were quantified and the amount of SNAP25 197 produced at each concentration, as a measure of BoNT/A uptake, was used to calculate the percent competition/inhibition. The amount of BoNT/A uptake for each competitor/inhibitor was compared to the amount of BoNT/A uptake after pre-treatment/pre-incubation with the negative control. Each experiment was conducted at least three independent times and each dose was tested in triplicates in each individual experiment, the percent average for each of the three or more independent experiments were used to generate inhibition curves. The curves were fitted to a non-linear exponential model; Y = 1006e 2 b*log(concentration) , where ''b'' was defined as either the competition constant (CC) or the inhibition constant (IC).

Protein Complex Immunoprecipitation (Co-IP)
Differentiated Neuro-2a cells were washed with PBS and then lysed by incubation at 4uC for 30 minutes in lysis buffer containing 20 mM Tris, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, and 1 mM EGTA pH 7.2 plus complete protease inhibitor cocktail (11 873 580 001; Roche). The supernatant was collected by centrifugation and the total protein concentration was measured using Bradford Reagent (500-0205; Bio-Rad). The Co-IP reaction was performed by mixing 1 mg of cell lysate with 10 mg antibody in a total volume of 1 ml. The reaction was incubated at 4uC overnight. As negative controls, a sample without antibody (lysate only) and a sample without lysate (antibody only) were prepared in parallel. Then 100 ml Protein A/G Magnetic Beads (88802; Thermo Scientific) was added and the reaction was incubated at 4uC for 1 hour. Three times the beads were sedimented using a dynamag-2 magnet (Invitrogen; 12321D) and washed with PBS. Finally, the beads were re-suspended in 26 SDS-PAGE running buffer and Western Blot was performed. Cell lysate was run in parallel.

Labeling of rH C /A and FGF2
Purified rH C /A (0.4 mg) was dialyzed against 4 L of 50 mM HEPES pH 7.0-7.2 150 mM NaCl in a 0.5 ml dialysis unit (Harvard Apparatus) for 16 hours at 4uC using a 25 kDa cut-off cellulose acetate membrane. FGF2 (0.2 mg) (233-FB/CF; R&D Systems) was re-suspended in 0.5 ml of 50 mM HEPES pH 7.2, 150 mM NaCl. rH C /A was labeled with either Alexa Fluor 488 C5-maleimide (A10254; Invitrogen) (cell imaging and photobleaching) or Alexa Fluor 633 C5-maleimide (A20342; Invitrogen) (rH C /A uptake) and FGF2 was labeled with Alexa Fluor 633 C5maleimide (10:1 molar ratio free label to protein) overnight at 4uC in the dark. To remove free label, the proteins were either dialyzed again, using the conditions listed above, or ran on a PD-10 desalting column (17-0851-01; GE Healthcare). The column was equilibrated with 50 mM HEPES pH 7.2 150 mM NaCl. The concentrations of rH C /A and FGF2 were determined by measuring the UV absorbance at 280 nm using a Beckman Coulter DU 800.

Uptake of rH C /A and FGF2 in HEK 293 cells
Cells were plated at 20,000 cells per well in a Cell Carrier microplate (Perkin Elmer) in culture media and allowed to attach overnight. Cells were treated with 1 mg/ml Cell Tracker Green CMFDA (C2925; Invitrogen) and Hoechst 33342 (H10295; Invitrogen) for 30 minutes prior to adding fluorescently labeled rH C /A or FGF2. After removing the staining media, 0.1 ml of 0-25 nM of Alexa Fluor 633-rH C /A or -FGF2 was added. Fluorescence was measured using the Operetta High Content Imaging System (Perkin Elmer), set to the following parameters: 206 WD objective, 9 fields, non-confocal, 15% excitation, Blue (Ex 380-410/Em 430-460), Green (Ex 460-490 nm/Em 500-550 nm) and Far Red (Ex 630-645 nm/Em 660-900 nm). Uptake, measured as increasing amounts of Far Red signal in the cells, was monitored for 15 hours, with 30 minutes time points. The results were analyzed using Harmony 3.1 software.
Cell imaging and photobleaching PC-12 cells transfected with Halo tagged FGFR3 or SV2C were plated on Collagen IV coated glass bottom dishes (P35GCOL-0-10-C; MatTek) and differentiated for 3 days. The cells were incubated with 5 mM Halotag TMR ligand (G8251; Promega) for 15 minutes and washed 3 times for 10 minutes with fresh media. Cells were then treated with 1 mM Alexa Fluor 488 labeled rH C / A. After 2 hours incubation, the cells were fixed using 5% paraformaldehyde. Cells were imaged using a LSM710 confocal microscope and analyzed using ZEN 2009 software (Carl Zeiss INT, Germany). The Alexa Fluor 488 label and TMR star labels were imaged using the following respective settings: excitation 488 nm/emission 500-510 nm, excitation 561 nm/emission 595-620 nm. The TMR was photobleached by exciting the fluor with the 561 nm laser 100 times for 0.1 s with 5% laser power. The amount of fluorescence intensity of the donor fluor (AF-488) was measured before and after photobleaching of the acceptor (TMR).

Immunostaining
Sprague-Dawley rats (200-250 g; Charles River) were injected with 10 units of BOTOX (Allergan) into the tibialis anterior (TA) muscle of the right hind limb. Animals receiving injections of 0.9% saline into their TA muscle served as controls. Rats were sacrificed 3 days following injections and their TA muscles were harvested. Muscles were embedded in OCT compound, frozen in liquid nitrogen and stored at 280uC. Prior to staining, muscles were cross-sectioned (10 mm) using a cryostat (Leica), mounted onto microscope slides and stored at 220uC until use. Frozen, slidemounted muscle sections were thawed to room temperature and immediately fixed with 2% paraformaldehyde for 10 min. Sections were blocked with 5% normal serum in PBS, pH 7.4 for 60 minutes and then incubated with primary antibodies for 3 hours at room temperature: rabbit anti-SV2C (1:400, sc-28957; Santa Cruz Biotechnology), rabbit anti-FGFR3 (1:200, sc-123; Santa Cruz Biotechnology), mouse anti-SNAP25 (1:200, SMI-81, Covance), and mouse anti-SNAP25 197 (1:200, Allergan). Muscle nicotinic acetylcholine receptors (nAChR) were labeled with a-bungarotoxin (a-Bgt) Alexa-Fluor 647 conjugate (1:500, Invitrogen). Sections were then washed and incubated with secondary antibodies for 30 minutes at room temperature. Following a final wash, slides were coverslipped and analyzed. Images were acquired using a Zeiss LSM-710 confocal microscope (Carl Zeiss INT).

Surface Plasmon Resonance (SPR) binding analysis
Experiments were performed on a BIAcore 3000 instrument (GE Healthcare). Ligands, rH C /A, anti-H C /A 6B1 (Provided by Dr. L. Smith, USAMRIID), FGF2, or FGF9 (233-FB; and 273-F9; R&D Systems), were immobilized on a CM5 chip (BR-1003-99, GE Healthcare) using an amine coupling kit (BR-1000-50, GE Healthcare). Analytes, either SV2C 529-579 (JPT Peptide Technologies, dissolved in 100% DMSO), FGFR3b/c deletion mutant peptides ( Figure 6A), rH C /A, or membrane extracts were injected over the ligand surfaces at concentrations ranging from 0-5000 nM, or for the membrane extractions at 5 mg/ml. The flow rate was set to 20 or 30 ml/min. Running buffer: HBS-EP buffer (BR-1006-91, GE Healthcare). The surfaces were re-generated by two 1-min injections at 30 ml/min of 10 mM Glycine, pH 1.5 (rH C /A) or 1-min injections at 30 ml/min of either; 10 mM Glycine, pH 1.5 and 0.125% SDS (FGFR3), 10 mM Glycine, pH 1.5 and 20 mM CHAPS (Membrane extracts), or 10 mM NaOH (SV2C). The sensorgram curves were evaluated using the BIAevaluation 3.0 software. The curves were fitted to a 1:1 Langmuir binding model (A + B « AB, where A is the analyte and B is the ligand immobilized on the sensor surface). Based on the fitted curves the association constant, k a , the dissociation constant, k d , and the equilibrium constant, K D (K D = k d /k a ) were determined. The FGFR3b/c peptide curves were also visually compared using the ''normalization'' wizard in the BIAevaluation 4.1 software.

T-test
To assess the significance of the differences from the BoNT/A cellbased competition/inhibition assays and the SPR Binding Analysis assays t-tests were performed using online Graphpad software; www.graphpad.com/quickcalcs/ttest1.cfm?Format = SEM (Graph-Pad Software Inc).  Figure S1 Isolation of a BoNT/A-FGFR3 protein complex and BoNT/A sensitivity and expression of SV2 and FGFR3 before and after differentiation. (A) Undifferentiated and differentiated PC-12 cells were treated with BoNT/A for 6 hours followed by overnight incubation in toxin-free media. Western Blot of PC-12 cell lysates using antibodies specific to SNAP25 197 . Western blot band intensities were plotted against BoNT/A concentration in log scale using SigmaPlot v 10.0 and the EC 50 values were calculated by fitting the curves to a 4parameter logistics (4PL) function. PC-12 cells were more sensitive to BoNT/A after differentiation, observed as an increase in both potency and efficacy. The EC 50 before differentiation was 800620 pM, the EC 50  Western blot under reducing conditions resulting in separation of BoNT/A and its receptor. The blot was probed with antibody against BoNT/A heavy chain (D) or FGFR3 (E) demonstrating that both BoNT/A and FGFR3 were present in the complex. The negative control with Sulfo-SBED BSA was run in parallel (Control). (F) Expression of FGFR3 and SV2 in Neuro-2a and PC-12 cells, before and after 3 days differentiation. Lysates of differentiated and non-differentiated cells were analyzed by Western blot using antibodies to FGFR3 and SV2A, B and C. The lysates, 6 mg per lane for FGFR3 and SV2A and B, and 3 mg per lane for SV2C, were run in parallel to allow direct comparison of intensities. According to the manufacturer the antibody to FGFR3 (Ab133644) recognizes a single band for FGFR3 just below 100 kDa. The antibodies to SV2 recognize multiple diffuse bands for SV2, partly because SV2 proteins are highly glycosylated in vivo [19]. The predicted size for each of them, based on their protein sequence, ,80 kDa, is shown and an arrow is pointing to the band size expected to be SV2 for each lane. Interestingly, the band size for SV2 is higher in PC-12 than in Neuro-2a cells, suggesting different levels of glycosylation. As a loading control the blots we re-blotted using an antibody to Syntaxin. According to the manufacturer, the antibody for Syntaxin, sc-12736 recognizes a single band just above 34 kDa. Expression of FGFR3 was unchanged after differentiation, while the expression of SV2A, B and C was decreased in both cell lines. (G) Expression of FGFR3 and SV2 isoforms in HEK 293 cells. Lysates of HEK 293 cells (6 mg per lane) were analyzed by Western blot using antibodies to FGFR3 and SV2A, B and C. As a positive control, lysates from PC-12 cells were run in parallel. HEK 293 cells express FGFR3, but little or no SV2. (H) Uptake of rH C /A in HEK 293 cells. Uptake of 0-25 nM Alexa Fluor 633 labeled rH C /A (Hb) was measured, as an increase in intracellular fluorescence over time, using the Operetta High Content Imaging System. For comparison, internalization of Alexa Fluor 633 labeled FGF2 (Ha), a native ligand for FGFR3, was measured in parallel. The results show that rH C /A binds to and enters HEK 293 cells and that uptake of rH C /A is comparable to uptake of FGF2, even if slightly lower. (TIF) Figure S2 Overexpression of FGFR3, but not SV2C, increases binding to rH C /A. Reduced expression of FGFR3 but not SV2C decreased sensitivity to BoNT/A. 4000 nM of each of the FGFR3 peptides were flowed across a CM5 sensor chip covered with rH C /A. The curves were fitted to a 1:1 kinetic-binding model (A + B « AB) with the BIAevaluation 3.0 software. The table shows the calculated k a (1/Ms), k d (1/s), and K D (k d /k a (M)) for all eight FGFR3 deletion mutant peptides upon binding to rH C /A. To assess the significance of the differences t-tests were performed, comparing each peptide to the longest peptide of the same subtype and comparing the subtype b peptides that were not significantly different from the longest FGFR3b peptide to the subtype c peptides that were not significantly different from the longest FGFR3c peptide. P-values are shown in the table. Unmarked p-values indicate that the difference is not significant. FGFR3b peptides bound slightly better than subtype c peptides. For both subtypes, the peptides that contained only Loop 3 bound with significantly lower affinity compared to the peptides that spanned Loop 2,3 or Loop 1,2,3. The average K D for FGFR3c Loop 2,3 fell in the middle, being about 3 times higher than the average K D for FGFR3c Loop 1,2,3. There was no significant difference in binding of FGFR3b Loop 2,3 and FGFR3b Loop 1,2,3, showing that Loop 2,3 of FGFR3 is the minimal optimal binding region for rH C