Botulinum Neurotoxins: Biology, Pharmacology, and Toxicology

1. Binding and Specificity.

After entering the lymphatic and blood circulations, following intestinal absorption or inspiration or injection, the BoNTs rapidly gain access to the perineuronal fluid compartment but do not cross the blood-brain barrier (Simpson, 2013). The known BoNTs bind with high affinity to the presynaptic plasma membrane of skeletal and autonomic cholinergic nerve terminals in numbers estimated to be, for BoNT/A1 or /B1, in the order of hundreds of molecules per square micrometers at the rat neuromuscular junction (NMJ) (Dolly et al., 1984). This restricted tropism is extraordinary, particularly considering that the presynaptic plasma membrane of cholinergic peripheral nerve terminals represents an infinitesimal part of the total surface area of cells and tissues exposed to body fluids. Such neurospecificity and affinity of binding, together with their catalytic activity, is at the basis of the BoNTs toxicity and, at the same time, of their pharmacological and therapeutic use.

This unique nerve terminal binding property of the BoNTs is due to their capacity to interact with two independent receptors of the presynaptic membrane: a polysialoganglioside (PSG) and the glycosylated lumenal domain of a synaptic vesicle protein that mediates the following step of internalization (Montecucco, 1986). Additional receptor(s) might be involved and contribute to the preference for cholinergic terminals (Montecucco et al., 2004; Kammerer and Benoit, 2014). It is also possible that cholinergic nerve terminals express unique N-glycans attached to synaptic vesicle (SV) glycoproteins that contribute to this preferential binding. Fibroblast growth factor and vanilloid receptors were recently reported to bind BoNT/A1 (Jacky et al., 2013; Li and Coffield, 2016), but their functional involvement in toxin binding has to be further validated.

The involvement of PSG in BoNT binding has been extensively investigated and is supported by a large wealth of experimental data (Rummel, 2013), including the demonstration that mice and cell lines devoid of complex PSG are largely resistant to BoNT (Kitamura et al., 1999; Bullens et al., 2002). This fact also contributes to explain why insects that are devoid of PSG are BoNT insensitive, a property that makes them reliable vectors of BoNT spread during outbreaks of animal botulism among birds and fishes (Montecucco and Rasotto, 2015).

The BoNTs-PSG interaction is rather well characterized and reviewed (Rummel, 2013). Briefly, BoNT/A, /B, /CD, /E, /F, and /G (and TeNT) possess within the HC-C subdomain a large cavity defined by the conserved motif sequence E(Q in BoNT/G)…H(K in BoNT/E and G in BoNT/G)…SXWY…G (Fig. 1, pink). The affinity of this PSG-BoNT binding has been estimated and the K_d values are in the 0.1–1 μM range (Rummel, 2013). At variance, BoNT/C, BoNT/D, and the mosaic neurotoxin BoNT/DC bind PSG at a binding site located in a similar position but defined by a different set of amino acids lateral chains and with lower affinity (Strotmeier et al., 2010; Rummel, 2013). Interestingly, BoNT/C, BoNT/D, and BoNT/DC (and TeNT) harbor a second PSG binding site in their HC-C domain, defined by conserved W and Y(F) residues (Strotmeier et al., 2010, 2011), which might bind the carbohydrate portion of a plasma membrane glycoprotein that is endocytosed from the presynaptic membrane. BoNT/DC was recently shown to bind a sialyl residue in a pocket of the HC domain and a cell binding mechanism involving a cooperative contribution of two ganglioside binding sites was proposed (Nuemket et al., 2011).

Together with sphingomyelin, cholesterol, and some proteins, PSG form anionic microdomains in the plasma membrane (Simons and Toomre, 2000; Sonnino et al., 2007; Prinetti et al., 2009). Accordingly, BoNTs bind complex PSGs whose large oligosaccharide head group projects at a distance from the membrane surface and it is flexible and negatively charged. There is evidence that anionic microdomains of the presynaptic membrane including PSG may orient the electric dipole associated to the BoNT molecules whose positive end is, significantly, located around the PSG binding site. This reorienting effect of the membrane on the approaching BoNT molecule would strongly increase its probability of PSG binding, making it a reaction controlled only by diffusion (Fogolari et al., 2009).

Following membrane binding, the BoNTs are internalized. The key observations that neuronal stimulation enhances the toxicity of BoNTs (Hughes and Whaler, 1962; Simpson, 1980, 1985; Black and Dolly, 1986; Keller et al., 2004) and that BoNT/B binds synaptotagmin in cultured cells (Nishiki et al., 1994) led to the suggestion that BoNTs endocytosis is facilitated by a protein receptor consisting of the luminal domain of a synaptic vesicle (SV) protein (Montecucco and Schiavo, 1995). This second binding provides the high affinity necessary to bind the very low amounts of BoNT estimated to be present in the circulation during botulism (10⁻¹³–10⁻¹⁴ M). During stimulation, the increased exoendocytosis rate causes a frequent exposure of the SV lumen, making the vesicle interior transiently available for binding. This also contributes to account for the efficacy of BoNT in treating human syndromes characterized by hyperactive nerve terminals, because the NMJs of these muscles have a higher rate of SVs exoendocytosis, which favors toxin uptake.

Within a few years, the protein receptors of other BoNTs were identified. BoNT/B1, BoNT/G, and BoNT/DC bind segments 39–50 of synaptotagmin I and 47–58 of synaptotagmin II (Nishiki et al., 1994; Dong et al., 2003; Rummel et al., 2007; Peng et al., 2012; Willjes et al., 2013). BoNT/A1 and BoNT/E1 bind different segments of the fourth lumenal loop of SV2, a multispanning integral protein of synaptic vesicles of unknown function (Dong et al., 2006; Mahrhold et al., 2006; Binz and Rummel, 2009; Benoit et al., 2014). Three isoforms of SV2 (A, B, C) are expressed at motor nerve terminals, but SV2C appears to be the one binding BoNT/A1 more efficiently than SV2A or B via its glycosylated fourth luminal domain (Mahrhold et al., 2006; Benoit et al., 2014; Mahrhold et al., 2016), whereas BoNT/E1 binds isoforms A and B, but not C (Dong et al., 2008).

The binding of BoNT/A1 to the glycosylated SV2C receptor is mediated by protein-protein and protein-glycan interactions that do not appear to influence each other. Of the five putative N-glycosylation sites of SV2C, only the Asn-559 is involved; this N-glycan is highly complex with a tetra antennary structure, which is likely to interact with an extensive area of HC/A1 (Mahrhold et al., 2016). The protein-protein HC/A-SV2C contacts involve mostly the backbones of the two proteins, through the pairing of two solvent-exposed β-strands, one from each partner (Benoit et al., 2014). Asn559 of SV2C is at the center of the BoNT/A1-SV2 area of interaction and its N-linked glycan inserts in a crevice made by one β-strand of the HC-N subdomain (segment 950–954) and two segments of the subdomain HC-C: the β-strand segment 1064–1066 and the C-terminal loop 1289–1292 (Fig. 5A) (Yao et al., 2016). As glycosylation patterns vary during development and among adult individuals (Knezević et al., 2009; Pucic et al., 2010; Lauc et al., 2016), such a feature might be responsible for the different onset and duration of neuroparalysis elicited in human patients by the same dose of injected BoNT/A1 as different amounts of bound toxin are likely to correspond to different numbers of L chains entering the nerve terminal cytosol. In addition, as invertebrate and vertebrate N-glycans are different (Moremen et al., 2012), this may contribute to the lack of sensitivity of invertebrates to BoNTs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Close-up views of the binding interfaces between BoNT/A1 and BoNT/B1 to their synaptic vesicle protein receptors. (A) Two areas of interaction of BoNT/A1 with the synaptic vesicle protein glycosylated-SVC2 (PDB ID 5JLV). One main interaction is mediated by protein-protein contacts through the pairing of the backbones of two solvent-exposed β-strands (black dotted ellipsoid), one from each partner. Essential interactions are mediated by R1156 of BoNT/A1 making a cation-π stacking interaction with P563 of SV2C and also by R1294 of the toxin. The second main interaction is mediated by N559 whose side chain carries a N-glycan modification (shown as sticks), which fits within a crevice formed at the interface between HC-N (purple) and HC-C (green). Amino acids forming the groove are colored in cyan and labeled according to their location (P953, N954, S957, S1062, H1064, and R1065 from HC-N, in purple and T1145, Y1155, D1288, D1289, and G1292 from HC-C, in green). The cartoon also shows essential water molecules (black pellets) and the H bonds (yellow dotted lines), which mediate the interaction of BoNT/A1 with the N-glycan, suggesting the possibility that they serve to adapt the variety of N-glycans that are produced by different kind of neurons and/or by neurons of different individuals and animal species. (B) Interaction among BoNT/B1 and its synaptic vesicle protein receptors Synaptotagmin II (Syt-II) (PDB ID 4KBB). The interface of interaction is at the extreme bottom of BoNT/B and, at variance from BoNT/A1, involves exclusively HC-C (green). Syt-II is unstructured in solution but assumes an helical conformation upon binding to the toxin, forced by the interactions occurring at the level of two hydrophobic pockets. One pocket is formed by HC-C residues P1117, W1178, Y1181, P1194, A1196, P1197, F1204 with Syt-II residues M46, F47, and L50. The second pocket of HC-C is lined by residues K1113, S1116, P1117, V1118, Y1183, E1191, K1192, F1194, and F1204 with Syt-II residues F54, F55, E57, and I58. Only the most significant aminoacids involved in the interaction are shown and labeled. Note the H bond (black dotted line) formed by K1113 and E57, which may also interact electrostatically. The binding sites for the oligosaccharide portion of polysialoganglioside receptor are not shown, but in both BoNT/A1 and /B1 are located within the HC-C subdomain at a distance from the protein receptor binding sites in such a position that they do not interact with them (see text).

The structural basis of BoNT/B1 and of BoNT/DC binding to their synaptotagmin receptor have been determined and are shown in Fig. 5B (Chai et al., 2006; Jin et al., 2006; Berntsson et al., 2013a,b). Such binding is mediated by a pocket present in the HC-C subdomain formed by residues of the segment Pro1114-Phe1202 (BoNT/B1 numbering), which fits the α-helical segment 47–58 of synaptotagmin II located next to the lumenal surface of the SV membrane. Asn24 of synaptotagmin II is predicted to be glycosylated but the potential role of N-glycan linked to this residue on the binding of BoNT/B1 has not yet been tested, although this may be relevant with respect to the therapeutic use of BoNT/B1 in humans, because the pattern of glycosylation in the nervous system varies among individuals. The presence of Leu51 in human synaptotagmin II reduces significantly the binding affinity for BoNT/B1, /DC, and /G with respect to the corresponding Phe54 of mouse synaptotagmin II. This also explains the much higher dosage of BoNT/B1 necessary to achieve the same therapeutic effect of BoNT/A1 (Peng et al., 2012; Strotmeier et al., 2012).

The protein receptors of other BoNTs are not known in such details. BoNT/C might not have a protein receptor, because protease pretreatment does not affect its binding and internalization into cultured cells (Tsukamoto et al., 2005). Conflicting results have been reported for BoNT/D (Kroken et al., 2011; Peng et al., 2011; Rummel, 2013), but based on its similarity with BoNT/E, BoNT/F is expected to bind SV2, and some data support this possibility (Rummel et al., 2009).

2. Internalization into Nerve Terminals.

After intramuscular injection, BoNT/A1 is rapidly taken up and found within the lumen of SV at the neuromuscular junction (Colasante et al., 2013). In cultured primary neurons part of such internalization is clathrin mediated and most of the toxin is detected inside SV (Neale et al., 1999; Couesnon et al., 2009; Harper et al., 2011). Importantly, the average number of BoNT/A molecules per SV were estimated to be one to two (Harper et al., 2011; Colasante et al., 2013), a figure that matches the estimated copy number of SV2 molecules per vesicle (Takamori et al., 2006). The internalization of the other BoNTs has not been visualized by electron microscopy, but it is expected to occur as well via SV because these toxins bind a SV membrane protein.

SV exocytosis is strictly coupled to endocytosis, which can take place in different ways (Saheki and De Camilli, 2012; Cousin, 2015; Jähne et al., 2015; Kononenko and Haucke, 2015; Soykan et al., 2016). Before fusion with the presynaptic membrane at specialized release sites of nerve terminals (active zones), SVs are morphologically similar and, to be functional, they must have a defined proteic and lipidic composition (Takamori et al., 2006; Boyken et al., 2013). However, after fusion with the presynaptic membrane, endocytosed SVs may acquire molecular differences (Wienisch and Klingauf, 2006; Soykan et al., 2016). Linked to this and to the frequency of nerve terminal stimulation, SVs may undergo different forms of retrieval (Saheki and De Camilli, 2012; Jähne et al., 2015; Kononenko and Haucke, 2015; Soykan et al., 2016). In particular a functional SV, right after endocytosis, may be directly recycled and reacidified by the v-ATPase proton pump, which generates an electrochemical gradient driving the accumulation of neurotransmitter via a specific transporter (Parsons, 2000; Omote et al., 2011). Alternatively, some vesicles may fuse with endosomal compartments, where quality-controlled SVs of appropriate composition reform and re-enter the synaptic cycle. Under some conditions, early endosomal compartments can also be generated from the infolding of a large portion of the presynaptic membrane, and then SVs can reform by endosomal budding. Additionaly, it was reported that SV proteins (including synaptotagmins) of exocytosed SV can freely diffuse within the presynaptic boutons, where they are slowly confined in a periactive zone by endocytic adaptors and presynaptic diffusion barriers to be reinternalized by clathrin-dependent or bulk endocytosis mechanisms (Gimber et al., 2015).

Collectively, these different possibilities of SV endocytosis indicate that the trafficking of BoNTs may occur via different pathways, depending at least partially, on their receptors. This is supported by experiments performed with primary cultures of central nervous system neurons, showing a significant variability in the intoxication kinetics of different BoNTs (Keller et al., 2004; Sun et al., 2012). Moreover, the use of the endocytosis inhibitor 4-bromobenzaldehyde N-(2,6-dimethylphenyl)semicarbazone (EGA) (Gillespie et al., 2013) provided indirect evidence that different BoNTs may follow different trafficking routes after the initial endocytosis also in vivo (Azarnia Tehran et al., 2015). This would result in different entry times of the L chain into the cytosol and would explain the “rapid” entry of BoNT/A and /E and the slower one of BoNT/B determined in primary cultures of neurons (Keller et al., 2004; Sun et al., 2012; Colasante et al., 2013).

Interestingly, a BoNT may be taken up even after the inhibition of SV exocytosis by another BoNT in vitro (Keller et al., 1999, 2004; Pellett et al., 2015a) and in vivo (Eleopra et al., 1998; Adler et al., 2001; Meunier et al., 2003; Keller, 2006; Antonucci et al., 2008). Clearly these results are highly dependent on the SNARE isoform cleaved by the BoNT used to inhibit exocytosis and on the extent of cleavage, data that were not reported. Moreover, it is well established that PSG undergoes endocytosis by the ligand-receptor pathway (Sonnino et al., 1992), and therefore it cannot be ruled out that the sole BoNT-PSG interaction is sufficient to sustain neurotoxicity, particularly for those BoNTs that have two ganglioside binding sites. Taken together these results and considerations indicate that more than one way of uptake may be used at nerve terminals by the different BoNTs or by the same BoNT under different conditions.

3. Membrane Translocation.

BoNTs translocate their L chain into the cytosol from an acidic intracellular compartment, and the process can be blocked by different amines and by bafilomycin A1, a specific inhibitor of the v-ATPase (Simpson, 1983; Williamson and Neale, 1994; Keller et al., 2004). However the lumen of the SV inside the nerve terminal is not experimentally accessible, making the study of this process difficult. Therefore, models that bypass the intracellular delivery of the toxin and induce the entry of the L chain of BoNTs from the cell surface across the plasma membrane were devised (Pirazzini et al., 2011; Sun et al., 2011). By using these systems, it was found that the translocation of L takes place between pH 4.5 and 6 (Pirazzini et al., 2011) and that the entire translocation process is rapid (few minutes at 37°C) and strongly temperature dependent (Pirazzini et al., 2013c). Little if any translocation occurs above pH 6, and this makes it unlikely that the L domain translocates from early endosomes, whose internal pH is only slightly acidic. On the other hand, BoNT/A1 has been localized within the SV lumen by immuno electron microscopy in cultured hippocampal neurons (Harper et al., 2011) and within the NMJ (Colasante et al., 2013). An essential component of SV is the electrogenic v-ATPase, which injects protons into the lumen, generating a transmembrane pH gradient ΔpH of 1.4 pH units and an electrical gradient ΔΨ of + 39 mV (Parsons, 2000). By using fluorescent synaptophluorins, the luminal pH of SV was estimated to be ∼5.8 pH units (Miesenbock et al., 1998). Accordingly, the translocation of L into the cytosol can take place from recycling SV or from SV budded from early endosomes. However, it cannot be excluded that other similarly acidic compartments such as late endosomes and lysosomes that could be reached by a BoNT upon endosome maturation, could provide the appropriate environment leading to L chain translocation, although it should be considered that these organelles contain proteases that may degrade the toxin.

Planar lipid bilayer studies have shown that at low pH several BoNTs and tetanus toxin form transmembrane ion channels (reviewed in Montecucco and Schiavo, 1995). Despite the intense effort of several laboratories, the molecular mechanism describing the transformation of the water-soluble BoNT molecule into a transmembrane ion channel that assists the translocation of the ∼440 amino acids of the L metalloprotease domain across the membrane is still missing. However, a remarkable set of patch-clamping experiments (Montal, 2010; Fischer, 2013; Fischer and Montal, 2013) and recent results obtained with the plasma membrane entry model (Pirazzini et al., 2011; Sun et al., 2011) have shed light on the initial molecular events of the process for BoNT/A1 and /B1. We have reviewed and modeled BoNT membrane translocation recently (Pirazzini et al., 2016), and, therefore, only the most relevant aspects are highlighted here.

The starting point is the BoNT molecule bound to the luminal side of the SV membrane via two interactions: with the polysialoganglioside and with the protein receptor. As the pH lowers, some conserved high pKa carboxylates of BoNT become protonated on the face of the molecule containing the interchain disulfide bond (Fig. 1, top), which acquires a net positive charge, as indicated by bioinformatics and mutagenesis analysis; remarkably, the opposite face is devoid of conserved high pK protonatable residues (Pirazzini et al., 2011, 2013b). Consequently, the BoNT molecule, with its positively charged surface, falls onto the anionic surface of the membrane generating a lipid-protein complex. The pH value close to the membrane is estimated to be at least one pH unit lower than in the lumen owing to the Guy-Chapman effect (Nordera et al., 1997), leading to the protonation of carboxylate residues of lower pKa values. This triggers a process of structural change involving the L and H chain together with membrane lipids, leading to the formation of an ion channel. During this process, both the H and L chain enter in contact with the hydrocarbon chain of lipids as determined by photoactive hydrophobic labeling (Montecucco et al., 1989). There is a general consensus that the H chain acts as a sort of transmembrane chaperone for the translocation of the L chain across the membrane (Koriazova and Montal, 2003; Montal, 2010; Fischer, 2013). Different molecular roles of the H chain can be envisaged, and two possibilities with a range of intermediate cases are mentioned here: 1) that of a protein conducting channel that translocates the unfolded L chain, as it occurs in the case of the protective and lethal factors of anthrax (Collier, 2009), 2) the toxin forms a “molten globule”, a protein state that is known to be capable of interacting with the hydrophobic membrane interior (Ptitsyn et al., 1990; van der Goot et al., 1991). Such molten globule would insert into the membrane together with anionic lipids and would deliver the L chain to the other side of the membrane, whereas the H chain would assemble an ion conducting channel. In any case, the belt has to be unfastened to permit the passage of the L chain on the cytosolic face of the membrane (Fischer and Montal, 2007). The understanding of the mechanism of membrane translocation is very important, because it is likely to be common to all BoNTs and may thus present a key target step for the development of pan-inhibitors of the entire family of BoNTs (Fischer et al., 2009; Pirazzini and Rossetto, 2017).

The number of BoNT molecules involved in membrane translocation is another undefined point. The finding that one SV contains one or two molecules of BoNT/A1 leaves open the possibility that one toxin molecule is sufficient to carry out the process and suggests that these toxins may have an in-built membrane translocating system, similarly to the bacterial system SecY (Park and Rapoport, 2012). Diphtheria toxin, a bacterial toxin with a structural architecture similar to that of BoNTs, also has a translocating domain that is mainly α-helical, and the available evidence supports a single molecule process, with few transmembrane helices forming an ion channel (Finkelstein et al., 2000; Gordon and Finkelstein, 2001; Leka et al., 2014). At variance, it was recently reported that a trimer might form the protein conducting transmembrane channel of BoNT/B1 in PC12 cells and BoNT/E1 at physiologic pH (Sun et al., 2012). Clearly, additional experiments are needed to clarify this essential step of the mechanism of action of the BoNTs.

4. Interchain Disulfide Reduction.

The importance of the interchain disulfide for the toxicity of clostridial neurotoxins is demonstrated by the fact that reduced toxins are inactive (Schiavo et al., 1990; de Paiva et al., 1993; Fischer and Montal, 2007). At the cellular level, Fischer and Montal (2007) demonstrated that the premature reduction of this S-S bond, at any stage before its exposure to the cytosol, aborts the L chain translocation. Also in the plasma membrane translocation model, the L chain has to be disulfide linked to the H chain to cross the membrane (Pirazzini et al., 2011). The detachment of the L chain from the cytosolic face of the membrane by reduction releases its metalloprotease activity in the cytosol (Fischer and Montal, 2007).

The lumen of most intracellular organelles is oxidizing and low pH prevents the reduction of disulfide bonds. At variance, the cytosol has a higher redox potential owing to the presence of several reducing molecules (NADH, NADPH, reduced glutathione, cysteine, etc.). The maintenance of an appropriate redox balance is particularly important for the activity of key proteins (Arner and Holmgren, 2000; Meyer et al., 2009; Hanschmann et al., 2013). The NADPH-thioredoxin reductase (TrxR)-thioredoxin (Trx) system is a major redox system of the cell that reduces protein disulfides. Its involvement in the BoNTs and TeNT entry into neurons was first identified using a pharmacologic approach (Pirazzini et al., 2013a), and then TrxR and Trx were shown to be extrinsic proteins of the cytosolic side of the SV membrane (Pirazzini et al., 2014), where translocation of the L chain is expected to occur. Several inhibitors of the TrxR-Trx redox couple prevent the display of the SNARE specific metalloprotease activity of the L chain of all serotypes of BoNTs, from A to G in cultured neurons. More importantly, these inhibitors largely prevent the BoNT-induced paralysis in mice in vivo, regardless of the serotype involved (Zanetti et al., 2015). The reduction of the single interchain disulfide bond is therefore a general and fundamental step of the BoNT [and TeNT (Pirazzini et al., 2013a; Zuverink et al., 2015)] mechanism of nerve terminal intoxication. As such it has to be considered a step per se in the sequence of passages leading from BoNT nerve terminal binding to neuroparalysis. More recently it was found that the chaperone Hsp90 is also present on the cytosolic face of SV and that it cooperates with TrxR and Trx in the entry of a folded and active L chain in the cytosol (Azarnia Tehran et al., 2017).

5. SNARE Protein Cleavage.

SNARE proteins form a large superfamily comprising many isoforms of VAMP (Rossi et al., 2004) and syntaxins but relatively fewer isoforms of SNAP. Specific isoforms of SNARE proteins selectively interact and form heterotrimeric coiled-coil complexes (SNARE complexes), which mediate most intracellular events of vesicle–target membrane fusion (Jahn and Scheller, 2006). The discovery that TeNT and BoNTs are metalloproteases (Schiavo et al., 1992b,c) and that TeNT and BoNT/B1 cleave the SV protein VAMP/synaptobrevin (Schiavo et al., 1992a) opened the way to the the rapid identification of the other SNARE proteins targets of the other BoNTs (Blasi et al., 1993a,b; Schiavo et al., 1993a,b,c). The cleavage by TeNT and BoNTs of proteins that were previously identified by molecular biology and biochemical methods (Elferink et al., 1989; Oyler et al., 1989; Bennett et al., 1992) explained the molecular basis of tetanus and botulism. At the same time, these findings provided the strongest evidence of the SNARE protein involvement in neurotransmitter release and, more generally, that the three SNAREs are the core proteins of the nanomachine that mediates vesicle fusion to target membrane.

The BoNT proteolytic activity is highly specific and directed toward unique peptide bonds within the sequence of the respective SNARE substrates (summarized in Tables 1, 2, and 3) (Binz, 2013; Blasi et al., 1993a; Pantano and Montecucco, 2014; Rossetto et al., 2014). All BoNTs, except the BoNT/As, cleave large portions of the cytosolic domains of their respective substrates (Fig. 3), preventing the formation of a stable SNARE complex and consequently of exocytosis (Hayashi et al., 1994; Sutton et al., 1998). At variance, BoNT/As generate a truncated SNAP-25, which is still capable of forming a stable SNARE complex (Hayashi et al., 1994; Otto et al., 1995) and has a life time within the nerve comparable to that of intact SNAP-25 (Foran et al., 2003). Remarkable work showed that BoNT/A truncated SNAP-25 (SNAP-25_1–197) by itself inhibits exocytosis (Huang et al., 1998), and results of several experiments indicate that the BoNT/A cleavage of a 10–15% fraction of total SNAP-25 both within the NMJ (Raciborska et al., 1998) and spinal cord neurons (Keller and Neale, 2001; Keller et al., 2004; Montecucco et al., 2005) is sufficient to cause paralysis. These results led to the suggestion that SNAP-25_1–197 acts as a dominant negative factor in the function of a multimeric radial super-SNARE complex because the C-terminal segment is necessary for protein-protein interactions underpinning the formation of a radial SNARE super-complex (Montecucco et al., 2005; Pantano and Montecucco, 2014). Electron microscopy data indicate that multimeric SNARE supercomplexes exist in the CNS (Rickman et al., 2005), and indirect evidence for the existence of an octameric neuroexocytosis radial nanomachine have been obtained (Megighian et al., 2013). A variety of experimental approaches have been used to estimate the number of SNARE complexes forming the nanomachine that mediates vesicle-target membrane fusion and a range of figures have been produced (reviewed in Pantano and Montecucco, 2014). Recent improvement in single particle cryo electron microscopy (Cheng, 2015) may soon reveal the structure of the nanomachine that mediates neuroexocytosis, allowing one to understand the molecular consequences of the cleavage of SNAP-25 by BoNT/A and BoNT/C. However, at the present stage it cannot be excluded that SNAP-25 exists in different pools within the nerve terminal and that a subpool consisting of 10–15% of total SNAP-25 is the one involved in neuroexocytosis. It is also possible that the C-terminal segment of SNAP-25 inserts in the lipid bilayer playing an essential role in membrane fusion.

The specificity of L chains for the SNARE proteins relies on extended and multiple protein-protein interactions with their specific substrate, which include the cleavage site and exosites (Rossetto et al., 1994; Pellizzari et al., 1996; Breidenbach and Brunger, 2004; Jin et al., 2007; Agarwal et al., 2009; Brunger and Rummel, 2009). Such extended enzyme-substrate interaction results in highly specific recognition and explains why no additional protein substrates of the BoNTs have yet been found. The selectivity of BoNTs is well shown by two examples that are relevant to the evolutionary biology of animal botulism. The replacement of a Gln with a Val at the P1′ position of the VAMP cleavage site by TeNT makes rats and chicken resistant to tetanus (Patarnello et al., 1993). BoNT/D is the most potent toxin on mice (lethal dose < 0.1 ng/kg) but poorly toxic for humans (Eleopra et al., 2013) and rats because one of their VAMP1 exosites is not conserved (Peng et al., 2014).

Different events of vesicle trafficking are mediated by different VAMP isoforms, and this fact has been highlighted by the recent identification of the VAMP isoforms involved in evoked and asynchronous release (Kavalali, 2015). To favor the use of the BoNTs as a simple knockout of specific SNARE isoforms and study their involvement in neurophysiological events, Tables 1, 2, and 3 list the known mouse, rat, and human isoforms of the three SNARE proteins and predict their susceptibility to the 10 neurotoxins whose cleavage sites are known.

1. Reversibility of the Neuroparalysis Induced by Botulinum Neurotoxins.

Compared with other bacterial and plant toxins, which kill cells, a very remarkable aspect of the neuroparalysis caused by the BoNTs is its complete reversibility, i.e., a botulism patient survives even a major intoxication provided that respiration is mechanically supported and supportive care is provided (Johnson and Montecucco, 2008). There are several reports of botulism patients that have survived botulism after many months in an emergency room. This is due to the fact that BoNTs are neither cytotoxic nor they cause any axonal degeneration, although toxicity in vitro has been reported for BoNT/C and by the combined use of BoNTs targeting different SNAREs (Williamson and Neale, 1998; Berliocchi et al., 2005; Zhao et al., 2010; Peng et al., 2013; Arsenault et al., 2014). However, in evaluating the latter, it should be considered that: 1) BoNTs were used at concentrations much higher than those causing botulism, 2) isolated neurons in primary cultures are more sensitive to any form of stress than neurons in situ where they are involved in multiple structural and physiologic interactions with neighbor cells and extracellular matrix, 3) no neurotoxicity of any kind was detected in human volunteers injected with a “therapeutic dose” of BoNTs (Eleopra et al., 1997, 2002, 2004). More importantly, the very extensive and long-term experience with BoNT/A1 and BoNT/B1 as therapeutics has provided no indications of neuronal damage after repeated treatments extended over many years (Naumann and Jankovic, 2004; Naumann et al., 2006; Ramirez-Castaneda and Jankovic, 2014). Therefore, the presently available data provide no evidence for any neurodegeneratve action of the BoNT tested so far when used at concentrations sufficient to paralyze the NMJ. This does not exclude the possibility that cytotoxic toxins might be found among the large number of novel, yet untested, BoNTs.

The duration of the paralysis varies extensively depending on: 1) type of BoNT, 2) dose, 3) animal species, 4) mode of administration, and 5) type of nerve terminal. Since BoNT duration of action is of paramount importance because it determines the duration of hospitalization of botulism patients and the duration of the therapeutic effects, this aspect of BoNTs biology and pharmacology will be discussed in some detail below.

The order of duration of action in mice and humans is: BoNT/A ∼ BoNT/C > BoNT/B ∼ BoNT/D, BoNT/F, and /G > BoNT/E (Foran et al., 2003; Eleopra et al., 2004; Keller, 2006; Morbiato et al., 2007) with the exception of BoNT/D that is poorly active in humans but very potent in mice (Eleopra et al., 2013). The duration of action of BoNTs is about three times longer in humans than in mice (i.e., BoNT/A 3–4 months versus 1 month, respectively), and skeletal muscles recover about three times faster than autonomic cholinergic nerve terminals (in humans 3–4 months versus ∼ 1 year for BoNT/A1).

It has been estimated that as few as 1,000 molecules of BoNT are sufficient to inactivate nerve transmission in a muscle (Hanig and Lamanna, 1979), rendering the L chain half-life difficult to study in vivo with the currently available biochemical methods. As a consequence, this investigation has been tackled mainly in vitro, by transfecting L chains tagged with protein reporters (Fernandez-Salas et al., 2004a,b; Tsai et al., 2010) and mainly on BoNT/A1 and BoNT/E1 L chains, because they display respectively the longest and the shortest persistence in vivo. Within the limits of the methods used, it appears that BoNT/A1 L chain has a longer lifetime than that of BoNT/E1 because it escapes the action of the cell degradation systems (Tsai et al., 2010). In fact, BoNT/E L chain is ubiquitinated and targeted to the ubiquitin-proteasome system, whereas BoNT/A1 L chain appeared refractory to this degradation pathway. This resistance may come from the ability of the L chain of BoNT/A1 to recruit deubiquitinases, specialized enzymes that remove polyubiquitin chains, thus sparing proteins from ubiquitin-proteasome system degradation (Shoemaker and Oyler, 2013). Ubiquitination of BoNT/B1 was also reported and associated with a decreased proteolysis of VAMP (Shi et al., 2009). Other factors that may come into play are the localization of the BoNT/A L chain on the cytosolic face of the membrane, the presence of a di-leucine motif at its C terminus, and the recruitment of septins (Fernandez-Salas et al., 2004a,b; Wang et al., 2011; Vagin et al., 2014).

Recent evaluation of BoNT/As L chain persistence in cultures of neuronal cells indicate that it can remain active for over a year as detected by cleavage of SNAP-25 (Whitemarsh et al., 2014); the functionality of the ubiquitin proteasome system in these neurons was not tested. Such a long lifetime should not surprise, because it has been reported that purified BoNT/A1 remains unaltered for 4 years at room temperature (Frevert, 2009). These findings support the generally accepted view that duration of the paralysis induced by the BoNTs reflects the lifetime of the L chain of a BoNT inside the neuronal cytosol (Shoemaker and Oyler, 2013; Pantano and Montecucco, 2014). However, the situation is different in the cases of the BoNT/As, the serotype predominantly used in human therapy (see next section), and of BoNT/C. Using animal ex vivo preparations or primary cultures of spinal cord neurons no correlation was found among the amount of SNAP-25 cleaved by BoNT/A and the extent of neuroparalysis, i.e., the cleavage of 10–15% of SNAP-25 was found to be sufficient to cause a complete blockade of neurotransmitter release (Raciborska et al., 1998; Keller et al., 2004), and 2–3% cleavage was sufficient to silence miniature postsynaptic currents (Beske et al., 2015). On the contrary, a good correlation was found in the case of BoNT/E, which removes a large part of the C-terminal half of SNAP-25, whereas BoNT/A and BoNT/C removes only nine and eight residues from the C terminus, respectively (Fig. 3). As mentioned above, the BoNT/A truncated SNAP-25 (SNAP-25_1–197), but not the BoNT/E truncated SNAP-25 (fragment_1–180), is still capable of forming the SNARE complex. Other seminal findings were that SNAP-25_1-197 inhibits insulin secretion from insulinoma cells as BoNT/A does (Huang et al., 1998) and that a single mutation within the C terminus of SNAP-25 was able to impair exocytosis in chromaffin cells (Criado et al., 1999). The simplest explanation of these results is that there are different pools of SNAP-25 within nerve terminals and that only a 10–15% of total SNAP-25 is actively involved in neurotransmitter release and is accessible to the proteolytic action of BoNT/A and BoNT/C. However, this rationale does not explain the inhibitory activity of SNAP-25_1–197, and the fact that all SNAP-25 molecules are accessible to BoNT/E. SNAP-25_1–197 may persist for long periods of times after generation by BoNT/A (Keller et al., 1999; Foran et al., 2003; Meunier et al., 2003). Taken together, these data are consistent with the suggestion that SNAP-25_1–197 may contribute significantly to the long duration of action of BoNT/A by acting as a dominant negative component of a neuroexocytosis nanomachine, consisting of several SNARE complexes arranged radially whose assembly/function is critically dependent on an intact SNAP-25 C-terminus (Megighian et al., 2013; Pantano and Montecucco, 2014). Also the long duration of action of BoNT/C in human and mice is likely to be due to an inhibitory action of a long-living SNAP-25_1–198 fragment.

2. Reversibility of Botulinum Neurotoxin Action in Vitro and in Vivo.

The comparison of the very long duration of neurotransmitter release inhibition induced by local BoNT/A1 and BoNT/B1 injection at the human autonomic cholinergic nerve terminals (even more than a year) (Naumann et al., 2013b), and of the similar one found for BoNT/A in stem cell-derived neurons in culture (Pellett et al., 2015c), with the shorter duration of action of BoNT/A1 in skeletal nerve terminals (3/4 months in humans), indicates that other factors come into play in vivo to determine duration and that at least part of these factors are external to the nerve terminal itself. Indeed it is long known that the BoNT poisoned NMJ undergoes a profound remodelling (Duchen, 1971). Novel nerve terminals sprout from their unmyelinated motor axon terminal and, to a lesser extent, from the first node of Ranvier (Juzans et al., 1996; Meunier et al., 2002). These nerve sprouts follow projections that emerge from perisynaptic Schwann cells, which multiply and migrate from the original NMJ to other sites of the sarcolemma soon after inactivation of the motor axon terminal. New contacts with muscle fibers are formed. Also the postsynaptic component undergoes critical changes with spreading of ACh receptors over the sarcolemma, without formation of the organized clusters characteristic of the mature NMJ (Sanes and Lichtman, 1999). The new synapses, although immature, can sustain vesicle recycling (de Paiva et al., 1999; Meunier et al., 2002), but are poorly efficient in ACh release (Rogozhin et al., 2008), providing a limited contribution to the recovery of the neurotransmission from nerve to the muscle fiber. Once a certain level of functionality is reestablished at the original site, terminal and nodal sprouts are pruned, and the newly formed synaptic specializations are eliminated (Johnson and Montecucco, 2008). Apparently, the neuroparalysis due to injection of BoNT/A followed by recovery of function can take place many times, one after the other, without loss of NMJ function, and this is at the basis of the repeated injections of this neurotoxins in human therapy (Borodic, 2007).

The duration of BoNTs activity has a paramount significance with respect to their therapeutic use and to the length of recovery of botulism patients in intensive care units and ensuing rehabilitation. On the therapeutic side, long lasting BoNTs require fewer injections and lower doses, limiting the possibility of immunization. On the other hand, a potent but short lasting BoNT could be desirable in some pathologies such as disjointed bone fractures, where a short duration might be more useful in ameliorating the course and outcome of the illness. Accordingly, there is a growing area of research that aims at changing binding specificity, affinity, and duration of BoNT action to obtain tailor-made therapeutic agents and more sophisticated tools to be used in cell biology studies (Chen and Barbieri, 2009; Ferrari et al., 2011; Masuyer et al., 2015; Sikorra et al., 2016).

1. Dystonias.

Dystonias are a heterogeneous group of disorders characterized by sustained involuntary muscle contractions, frequently causing repetitive twisting movements, abnormal postures, and pain (Albanese and Lalli, 2009, 2012). Localized injections of BoNT provide a transient symptomatic relief in primary and nonprimary focal dystonia syndromes, as demonstrated by several randomized controlled studies and by a large number of uncontrolled studies (Hallett et al., 2013; Guidubaldi et al., 2014).

In general, the therapeutic benefits of BoNT in dystonias derive from a decreased acetylcholine release from alpha motoneurons at the NMJ and the ensuing relaxation of the affected skeletal muscle. However, substantial evidence suggests that BoNT injected peripherally may also influence indirectly CNS function (Gracies, 2004). In fact, the BoNT inhibition of intrafusal muscle fibers by blocking also the gamma motoneurons, reduces muscle spindle afferent input to the CNS and may contribute to the therapeutic effects of BoNT in focal dystonias (Giladi, 1997; Hallett, 2000; Rosales and Dressler, 2010). Moreover there is considerable evidence that BoNTs, particularly at high doses, can reach higher structures in the brain. This is an area that deserves further investigation as a detailed understanding of direct central effects of BoNTs will provide valuable information for present and future therapeutic uses of BoNTs, particularly in the case of lower limb spasticity, which requires the use of high doses (Caleo and Schiavo, 2009; Mazzocchio and Caleo, 2015).

BoNT injection is considered the treatment of choice for most focal and segmental dystonias and in hemifacial spasm. The efficacy of BoNT for the treatment of these movement disorders, which include blepharospasm, cervical dystonia, hemifacial spasm, oromandibular dystonia, focal limb dystonias, and laryngeal dystonia has been recently reassessed by a thorough meta-analysis study, which included a total of 42 clinical trials (Hallett et al., 2013; Del Sorbo and Albanese, 2015).

Blepharospasm is a form of dystonia of the periocular muscles that produces forced eyelid closure, sometimes leading to functional blindness. FDA approved BoNT/A1 for the treatment of blepharospasm in 1989. Typically, 30–60 U of Botox or Xeomin, 60–180 U of Dysport or 1200–3600 U of Myobloc/Neurobloc are applied to the orbicularis oculi, procerus and corugator supercilii muscles bilaterally. Several studies demonstrated the efficacy and safety of BoNT therapy for this indication (Hallett et al., 2013). Response rates of around 90% make blepharospasm one of BoNT’s most successful indications. Adverse effects are usually mild and always transient and include local hematoma, ptosis, and diplopia (Dressler, 2012).

In cervical dystonia, also referred to as “spasmodic torticollis”, where dystonia of the neck and shoulder girdle muscles produces impairment of head control and pain, there is adequate evidence of BoNT efficacy, based on data from eight randomized, controlled, Class I clinical trials (Hallett et al., 2013). In clinical practice, the average total dose injected in patients with cervical dystonia is 100–300 U Botox or Xeomin, 400–800 U Dysport, or 10,000–20,000 U Myobloc/Neurobloc. These doses can vary considerably as the recommended range has to be adjusted depending on the individual patient’s features and sensitivity to the toxin. BoNT/A1 is the preferred treatment, whereas BoNT/B1 is recommended for A-resistant patients, because injections are generally more painful, possibly owing to the acidic pH of its solution, efficacy is shorter and immunogenicity higher (Bentivoglio et al., 2015).

Focal limb dystonia refers to dystonia affecting one arm or leg. Primary dystonia of the upper extremity commonly begins during selective, usually highly skilled, and repetitive motor tasks and are also referred as occupational or attitudinal dystonia. Typical upper limb dystonias include musician’s cramps and writer’s cramps where BoNT/A1 has been reported to be effective (Karp, 2004). However, due to the functional complexity, the usually high performance levels required and the narrow therapeutic window in the forearm muscles, functional outcome of treatments is sometime disappointing compared with those of blepharospasm or cervical dystonia. This also derives from the difficulty in obtaining the requested quality of voluntary movement without some weakness. BoNT application under electromyographic or ultrasound guidance is recommended as it improves the outcome.

The lower extremity is commonly affected in subjects with genetic dystonia (e.g., DYT-1 Dystonia), whereas upper limb and foot dystonia can be either idiopathic, in the context of a generalized dystonia, or symptomatic, as in Parkinson’s disease or in juvenile cerebral palsy. Successful treatments with BoNT have been reported but no controlled trials are available (Schneider et al., 2006). Higher doses of BoNT may need to be injected in the limb than in hand dystonia because motor control is less refined.

Hemifacial spasm describes synchronous unilateral contractions presumably caused by vascular compression of the facial nerve and is generally treated with lower doses than blepharospasm (Bentivoglio et al., 2009). Although experience with BoNT in hemifacial spasm treatment mainly originates from open-label trials, there is no doubt on its efficacy and safety (Guidubaldi et al., 2014). Notwithstanding the limited number of studies available, it appears that other nondystonic disorders including tics and tremors also benefit from BoNT therapy (Lotia and Jankovic, 2016).

2. Spasticity.

Spasticity describes the combination of a central paresis together with various forms of muscle hyperactivity, including dystonia, rigidity, and spasms often associated with pain. Most frequent etiologies include cerebral stroke, multiple sclerosis, traumatic brain injury, spinal cord injury, and infantile cerebral palsy. The goal of spasticity treatment is to reduce motor overactivity to improve movement without worsening weakness. BoNT treatment of spasticity can reduce muscle tone, improve function, facilitate nursing, and prevent contractures and decubitus. However, successful spasticity management requires a multiprofessional task force where all medical and surgical treatments need to be combined with physical interventions. Rehabilitation treatment of poststroke patients (i.e., muscle stretching exercises, eventually combined with neuromuscular electrical stimulation), when begun immediately after injection enhances the efficacy of BoNT therapy (Ward, 2002). However, recent systematic reviews have concluded that BoNT/A1 injections have to be considered as the pharmacological treatment of choice in focal spasticity to improve limb position, functional ability and to reduce pain (Esquenazi et al., 2013; Simpson et al., 2016), whereas limited data are available on the efficacy of BoNT/B1 in spasticity. BoNT therapeutic effects are better established for spasticity in the upper rather than lower limb (Esquenazi et al., 2013). It usually involves large BoNT doses: typically 300–500 U Botox or Xeomin or 600–1000 U Dysport for arm or leg spasticity. Recent studies reported treatment of poststroke patients with up to 800 U of Botox, 1800 Dysport, and up to 1200 U of Xeomin without signs of systemic toxicity (Dressler et al., 2015; Santamato et al., 2015). A higher dilution results in larger injection volumes and higher degree of paralysis (Kutschenko et al., 2016), probably because of larger spreading to neuromuscular junctions remote from injection site (Gracies et al., 2009).

3. Autonomic Disorders.

Botulinum neurotoxins inhibit neuroexocytosis from cholinergic nerve terminals of the sympathetic and parasympathetic autonomic nervous systems. The use of BoNT within the broad category of autonomic indications includes hypersecretory disorders, such as hyperhidrosis, sialorrhea, and chronic rhinorrhea, and smooth muscle hyperactivities in urologic and gastrointestinal disorders. Owing to an apparently higher affinity for cholinergic autonomic nerve endings, BoNT/B1 is as efficacious as BoNT/A1 in the therapy of secretory disorders both in terms of extent of inhibition and duration of the beneficial effects and at lower relative doses than those required in skeletal muscles (Bentivoglio et al., 2015).

Hyperhidrosis is medically benign but may be a socially devastating condition characterized as it is by excessive sweating, which may be occurring focally within the axillary region or it may extend to palms and soles. A formalized assessment of the American Academy of Neurology confirmed that BoNT therapy is safe and effective for axillary hyperhidrosis (Naumann et al., 2013b); total doses per axilla of 50–100 U Botox or Xeomin, 100–200 U Dysport or 2500 U NeuroBloc/Myobloc may be used (Bentivoglio et al., 2015).

Hypersalivation (or sialorrhea) refers to the presence of excessive saliva in the mouth, which may cause drooling with consequent severe embarrassment for the affected people. In several cases hypersalivation is secondary to other pathologic conditions including Parkinsonian syndromes, motoneuron diseases (amyotrophic lateral sclerosis) and cerebral palsy and is caused by impaired swallowing of saliva. The total therapeutic dose injected into the parotid glands for sialorrhea is tipically 2500 U for NeuroBloc/Myobloc, whereas BoNT/A1 ranges from 50 U for Botox to 75–450 U for Dysport (Naumann et al., 2013b).

Allergic rhinitis is a common disorder, one that is not disabling by itself, but it imposes a substantial burden in medical costs and indirect costs due to loss of productivity (Simoens and Laekeman, 2009). BoNT/A1 may be a good alternative for the treatment of rhinitis patients who are unresponsive to other treatment methods. Some restrictions such as painful injections and needing of specialized experience for adminitration prevent its wide use (Ozcan and Ismi 2016). BoNT for the treatment of allergic rhinitis is supported by a level B recommendation (probably effective) (Naumann et al., 2013b).

4. Urologic Pathologic Conditions.

The therapeutic applications of BoNT in urology include detrusor sphincter dyssynergia, lower urinary tract symptoms due to benign prostatic hyperplasia, and detrusor overactivity (both neurogenic and idiopathic). These conditions are common among patients with spinal cord lesions of different types (multiple sclerosis, spinal cord tumors, other spinal cord diseases, traumatic spinal cord injuries). The mechanism(s) underlying the therapeutic actions of BoNT in urologic disorders is not fully understood but a complex inhibitory action on receptors, neuropeptides, and neurotransmitters involved in the pathophysiology of the overactive bladder has been proposed (Chancellor et al., 2008). In the bladder, BoNT is thought to act primarily by inhibiting acetylcholine release from parasympathetic nerve endings inducing detrusor muscle relaxation (Schurch et al., 2000; Fowler et al., 2008). In addition, basic and clinical evidence suggests that BoNT may have sensory inhibitory effects unrelated to its actions on acetylcholine release (Apostolidis et al., 2005). Indeed, BoNT/A1 reduces the release of glutamate, of substance P, and of calcitonin gene-related peptide from the peripheral terminals of afferent bladder neurons (Duggan et al., 2002; Rapp et al., 2006) and of neurotrophins from the urothelium (Chancellor et al., 2008). The BoNT inhibition of secretion of these neuropeptides and ATP as well as cyclooxygenase-2 products, which are inflammatory response mediators released from nociceptive sensory endings in response to noxious stimuli (Khera et al., 2004; Chuang et al., 2008), would explain the therapeutic benefit of BoNT injection in painful bladder conditions (Smith et al., 2004; Wang et al., 2016).

BoNT/A1 injection for overactive bladder treatment is also associated with a significant decrease of purinergic receptors P2X3 and capsaicin receptors TRPV1, two receptors involved in nociception in the suburothelial nerve fibers (Apostolidis et al., 2005; Liu et al., 2014). These changes may reflect a direct activity of BoNT on the afferent innervation of the bladder and/or a secondary effect on the efferent innervation of the detrusor. Moreover, evidence of BoNT/A retrograde transport to the CNS after bladder injection in rats has been recently provided (Papagiannopoulou et al., 2016). Further studies are needed to elucidate this transport and its significance with respect to possible changes induced by BoNTA1 in the CNS after bladder injection.

Most studies have reported injection doses of 100–300 U of Botox, with 100 U being a typical starting dose for idiopathic detrusor overactivity and higher doses for those patients with insufficient response to an initial injection or those with neurogenic detrusor overactivity (Gomelsky and Dmochowski, 2014; Cheng et al., 2016). The injected volume is a key factor in the spread and action of BoNT/A in the bladder, with a wider diffusion of the toxin when higher volumes are injected (Coelho et al., 2012). In an experimental animal model of severe bladder pain, intrathecal injection of Botox has a strong analgesic effect, and this route of administration should be further explored for “intractable” forms of pain (Coelho et al., 2014).

5. Pain.

Despite current availability of a large number of analgesic drugs, management of chronic pain is still a challenge for clinicians. Over the past decade, BoNT treatment has shown efficacy in a large spectrum of human pain disorders, and the potentially relevant use of BoNT/A1 has been reported as efficacious in some form of chronic pain refractory to other treatments. The established ability to produce neurotoxins by recombinant protein expression (Bade et al., 2004; Band et al., 2010; Masuyer et al., 2014, 2015) and to modify them to create engineered neurotoxins with enhanced specificity for nociceptive nerve terminals will expand the therapeutic utility of BoNT for pain treatment. In this context, an innovative class of biopharmaceuticals obtained by replacing the native binding domain with another protein to redirect the light chain to a different nerve or cell has been recently proposed (Chaddock et al., 2004; Pickett, 2010; Masuyer et al., 2014).

The analgesic effects of BoNT/A1 were first reported in 1985 in a pilot study of BoNT/A treatment of cervical dystonia, characterized by abnormal, involuntary neck and shoulder muscle contractions and often resulting in significant, disabling musculoskeletal pain (Tsui et al., 1985). Early in its use as a therapeutic agent, BoNT/A1 was observed to provide pain relief in spastic and nonspastic muscle conditions in humans such as myofacial pain syndromes, temporomandibular disorder and bruxism, low back pain, prostatic pain, tension type and migraine headache, different types of neurophatic pain syndromes and many others (Wheeler and Smith, 2013). However, no effect was reported in a model of pain perception in human skin (Schulte-Mattler et al., 2007).

The association between BoNT/A1 and pain relief was originally thought to correlate only to its effect on muscle overcontraction or contractures. However, it is particularly interesting that the analgesia provided by BoNT injection occurs before muscle paralysis and outlasts any muscle weakness. It is now well documented that the analgesic effects of BoNT/A1 are related not only to its paralytic effect, but also to an effect on the nociceptor system (Wheeler and Smith, 2013). The widely reported antinociceptive effect of BoNT/A1 would be primarily mediated by the blockade of neuropeptides and inflammatory mediators release, and by the inhibition of plasma membrane exposure of pain sensors at peripheral level (as described previously for bladder pain). Indeed, in cultured sensory neurons, the TNF-α induced surface trafficking of TRPV1 and TRPA1 channels is mediated by SNAP25, VAMP-1 and syntaxin-1 and is inhibited by the serotypes of BoNTs that selectively cleave their respective SNAREs (Meng et al., 2014, 2016). This BoNT/A1 inhibitory effect would reduce peripheral sensitization and afferent input to the spinal cord, thereby dampening indirectly central sensitization. However, a growing number of experimental and clinical data on BoNT/A1 antinociceptive action cannot be explained adequately only by the inhibition of the release of peripheral neurotransmitter/inflammatory mediators. It is envisaged that the central effect may be directly involved, via retro and anterograde axonal transport along the branches of nociceptive neurons (Matak and Lackovic, 2014; Luvisetto et al., 2015; Mazzocchio and Caleo, 2015; Pellett et al., 2015d). Several studies reported toxin bilateral actions in distant regions after unilateral injection of BoNT/A1, suggesting that peripherally administered BoNT/A1 reaches the CNS region via axonal transport to target neurotransmission of pain sensory circuits. Enzymatic activity of BoNT/A1 has been immunohistochemically detected in spinal cord or brain stem areas, to which sensory and motoneurons in BoNT/A1-injected peripheral regions are projected (Matak and Lackovic, 2014; Luvisetto et al., 2015). Based on these and other evidence it was suggested that the BoNT/A action on pain is dominantly a central effect (Matak and Lackovic, 2014). Moreover, recently, in a rat model of neuropathic pain, BoNT/A was shown to influence microglial activation and to restore neuroimmune balance by reducing the levels of pronociceptive factors (IL-1 β and IL-18). In addition, it appears to increase the level of antinociceptive factors (IL-10 and IL-1RA) in the spinal cord and in the dorsal root ganglia (Zychowska et al., 2016). These findings suggest that neuroimmunological changes are also involved in BoNT/A-mediated analgesia.

Within the pain medicine field, only chronic migraine is an approved FDA indication for BoNT/A1. All other areas are currently considered off label, although for several of them, the available literature already strongly suggests efficacy. Concerning the evidence-based efficacy, the lack of standardized guidelines for BoNT application and dosage for pain management, and the lack of appropriate definition of study primary outcomes, in addition to the small sample size and limited number of randomized controlled clinical trials, have led to contradictory or negative findings in some forms of pain (Jabbari and Machado, 2011). Only some examples of chronic pain conditions for which prospective and controlled data are available on BoNT efficacy are presented in this paragraph.

6. Other Applications.

The therapeutic possibilities of BoNTs are manifold and certainly not yet exhausted. Controlled studies for many indications are frequently lacking because of the limited incidence of some disorders and symptoms.

7. Cosmetic Uses.

The present major use of BoNT/A1 in humans is related to a variety of cosmetic indications (Carruthers et al., 2016). Recently published statistics from the International Society of Aesthetic Plastic Surgery show that BoNT/A1 injection is the most popular of all cosmetic procedures worldwide (Frevert, 2015). Current estimates suggest that approximately half of the medical production of BoNT/A1 is used in aesthetic medicine. The serendipitous observation that BoNT/A1 smoothed facial lines when used therapeutically (Carruthers and Carruthers, 1992) led to study the toxin effect on glabellar lines, which are perceived as a sign of aging and of negative emotions. The repetitive contraction and activity of the procerus and the corrugator supercilii muscles involved in facial expression is mainly responsible for glabellar lines and can be attenuated by BoNT/A1 injection into three to five sites of these muscles (Gendler and Nagler, 2015). A recent meta-analysis showed that a 20 U total dose of BoNT/A1 is remarkably effective and safe for the treatment of glabellar lines (Guo et al., 2015). BoNT/A1 formulations have been FDA approved for glabellar lines (Vistabel, Allergan; Azzalure, Dysport, and Bocouture, Merz), and very recently Bocouture in Europe has been approved also for the treatment of horizontal frown lines and lateral periorbital lines. However, BoNT/A1 can be used to reduce all wrinkles in all skin areas of the face and the neck as long as they are based on increased muscle tone. Best results are obtained when BoNT is combined with filler substances to correct nondynamic aspects of the skin lines (Dressler, 2012). Repeated injections have not been reported to cause permanent damages to the superficial muscles involved and this is in line with the findings obtained in the treatment of dystonic muscles (please see next paragraph).