Engineering botulinum neurotoxin to extend therapeutic intervention Sheng Chen and Joseph T. Barbieri1 Department of Microbiology and Molecular Genetics, Medical College of Wisconsin, 8701 Watertown Plank Road, Room 256, Milwaukee, WI 53226 Edited by R. John Collier, Harvard Medical School, Boston, MA, and approved April 22, 2009 (received for review March 20, 2009) Clostridium botulinum neurotoxins (BoNTs) are effective therapeu- glabellar facial lines, and axillary hyperhidrosis (11). BoNT/A tics for a variety of neurological disorders, such as strabismus, efficacy in dystonia and other disorders related to involuntary blepharospam, hemificial spasm, and cervical dystonia, because of skeletal muscle activity, coupled with a satisfactory safety pro- the toxin’s tropism for neurons and specific cleavage of neuronal file, has prompted empirical/off-label use in a variety of secre- soluble N-ethylmaleimide-sensitive fusion protein-attachment pro- tions and pain and cosmetic disorders (12). tein receptors (SNARE) proteins. Modifying BoNT to bind nonneu- The clinical use of BoNTs is limited to targeting inflictions ronal cells has been attempted to extend therapeutic applications. affecting neuromuscular activity (11, 12). Elucidation of the However, prerequisite to develop nonneuronal therapies requires structure-function relationship of BoNTs has enabled the design the retargeting the catalytic activity of BoNTs to nonneuronal of therapies that retarget BoNT to unique neurons and non- SNARE isoforms. Here, we reported the engineering of a BoNT neuronal cells. Replacement of BoNT HCR domain with nerve derivative that cleaves SNAP23, a nonneuronal SNARE protein. growth factor, lectin from Erythrina cristagalli, or epidermal SNAP23 mediates vesicle-plasma membrane fusion processes, in- growth factors enable retargeting of BoNT/A to neuronal or cluding secretion of airway mucus, antibody, insulin, gastric acids, nonneuronal cells such as nociceptive afferents and airway and ions. This mutated BoNT/E light chain LC/E(K224D) showed epithelium cells (13–15). However, the selective cleavage of extended substrate specificity to cleave SNAP23, and the natural neuronal-specific SNARE proteins by BoNT has limited devel- substrate, SNAP25, but not SNAP29 or SNAP47. Upon direct protein opment of therapies in these nonneuronal systems. Prerequisite delivery into cultured human epithelial cells, LC/E(K224D) cleaved to develop therapies requires the retargeting of the catalytic endogenous SNAP23, which inhibited secretion of mucin and IL-8. activity of the BoNTs to nonneuronal SNARE isoforms. Here, These studies show the feasibility of genetically modifying LCs to we extend the substrate specificity of BoNT/E by engineering a target a nonneuronal SNARE protein that extends therapeutic catalytic derivative that cleaves the nonneuronal SNARE pro- potential for treatment of human hypersecretion diseases. tein, SNAP23, as a platform to develop therapies for nonneu- ronal human secretory diseases (16, 17). SNAP23 ͉ SNAP25 ͉ SNARE proteins Results Clostridium botulinum neurotoxins (BoNTs) are the most potent Previous studies identified residues 167–186 as the minimal, protein toxins for humans (1). BoNTs elicit neuronal-specific optimal peptide of SNAP25, a 206 amino acid protein, for LC/E flaccid paralysis by targeting neurons and cleaving neuron- in vitro cleavage (18). SNAP25 (167–186) comprises 2 subsites specific soluble N-ethylmaleimide-sensitive fusion protein- that include a substrate binding ‘‘B’’ region and an active site attachment protein receptors (SNARE) proteins. BoNTs are ‘‘AS’’ region (Fig. 1A). LC/E recognizes the P3 residue to organized into 3 functional domains: an N-terminal zinc-metal- facilitate alignment of the P2 and P1Ј residues of SNAP25. The loprotease light chain (LC), a translocation domain (HCT), and S1Ј pocket of LC/E is formed by F191,T159, and T208 with a C-terminal receptor binding domain (HCR) (1, 2). BoNTs bind hydrophobic interactions between F191 of LC/E and the P1Ј luminal domains of synaptic vesicle proteins, upon the fusion of residue I181 of SNAP25 (19). The basic S2 pocket contains K224, synaptic vesicles with the plasma membrane (3–5). BoNTs are which recognizes the P2 residue, D179, through a predicted salt internalized into endosomes and upon acidification, the LC is bridge. Docking the P2 and P1Ј residues of SNAP25 into the translocated into the cytoplasm, where SNARE proteins are active site pockets of LC/E aligns the scissile bond for cleavage cleaved (1, 2). (18, 19). Mammalian neuronal exocytosis is driven by the formation of Binz and coworkers reported that BoNT/E did not cleave protein complexes between the vesicle SNARE, VAMP2, and human SNAP23 (8), which provided a framework for defining the plasma membrane SNAREs, SNAP25 and syntaxin 1a (6). SNAP isoform specificity of the BoNTs. Many of the residues There are 7 serotypes of BoNTs (termed A–G) that cleave that contributed to LC/E recognition of SNAP25 were con- specific residues on 1 of 3 SNARE proteins: serotypes B, D, F, served in human SNAP23, except T173/A179,D179/K185,M182/T188, and G cleave VAMP-2, serotypes A and E cleave SNAP25, and and E183/D189, respectively (Fig. 1B). T173 in SNAP25 played only serotype C cleaves SNAP25 and syntaxin 1a (1). Thus, neuronal a limited contribution for LC/E substrate recognition (19) and specificity is based upon BoNT binding to neurons and cleaving only main chain interactions of M182-D186 contributed to LC/E neuronal isoforms of the SNARE proteins. For example, substrate recognition. Thus, the T173/A179,M182/T188, and E183/ BoNT/A cleaves human SNAP25, but not the human nonneu- D189 differences between SNAP25 and SNAP23 did not appear ronal isoform SNAP23 (7, 8). The nonneuronal SNARE iso- to contribute to inability of LC/E to cleave SNAP23. In contrast, forms are involved in divergent cellular processes, including fusion reactions in cell growth, membrane repair, cytokinesis, and synaptic transmission (reviewed in 9). Author contributions: S.C. and J.T.B. designed research; S.C. performed research; S.C. and The reversible nature of muscle function after BoNT intoxi- J.T.B. analyzed data; and S.C. and J.T.B. wrote the paper. cation that replace toxin-affected nerves with new nerves (10) The authors declare no conflict of interest. has turned the BoNT from a deadly agent to therapies for This article is a PNAS Direct Submission. neuromuscular conditions. As early as 1989, BoNT/A was ap- 1To whom correspondence should be addressed. E-mail: [email protected]. proved by the FDA to treat strabismus, blepharospam, and This article contains supporting information online at www.pnas.org/cgi/content/full/ hemificial spasm and then for cervical dystonia, cosmetic use, 0903111106/DCSupplemental. 9180–9184 ͉ PNAS ͉ June 9, 2009 ͉ vol. 106 ͉ no. 23 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0903111106 Downloaded by guest on October 1, 2021 Fig. 2. Cleavage of SNAP23 by LC/E(K224D). (A) Five micromolar SNAP23 was incubated with indicated amounts of LC/E(K224D) and subjected to SDS/PAGE [stained gel is shown in insert, SNAP23 (152–211) is designated (SN23 (152– 211)] and the cleavage product SNAP23 (152–186) is designated *. % SNAP23 cleavage was determined by densitometry. (B) Kinetic constant for LC/E to cleave SNAP25 and LC/E(K224D) to cleave SNAP23. E(K224D) did not cleave SNAP23(I187D) (Fig. 3B) supported that LC/E(K224D) cleaved human SNAP23 between residues BIOCHEMISTRY 186R-I187. SNAP25 isoforms include SNAP25a, SNAP25b, SNAP23a, SNAP23b, SNAP29, and SNAP47 (22, 23). SNAP23 and SNAP25 mediate synaptic membrane fusion in nonneuronal and neuronal cells, respectively, whereas SNAP29 and SNAP47 have Fig. 1. K185 of human SNAP23 contributes to the substrate recognition by not been implicated in membrane fusion events. SNAP29 was BoNT/E. (A) Substrate recognition by LC/E. Two subsites in SNAP25 contribute shown to inhibit SNARE disassembly and was implicated in to substrate binding ‘‘B’’ (Km) and catalysis ‘‘AS’’ (kcat), where the P3, P2, and synaptic transmission (24). Whereas the function of SNAP47 is P1Ј residues contribute to recognition by LC/E. (B) Sequence alignment of not clear, SNAP47 can substitute for SNAP25 in SNARE human SNAP25 (SN25) and human SNAP23 (SN23). (C)(Upper) modeled complex formation and proteoliposome fusion. The substrate complex structure of LC/E-SNAP25 predict the recognition of P site residues of 224 SNAP25 by LC/E. (Lower) modeled complex structure of LC/E(K224D)-SNAP23 specificity of LC/E(K D) on SNAP25 isoforms including predict the recognition of P site residues of SNAP23 by LC/E(K224D). Models SNAP23a, SNAP25b, SNAP29, and SNAP47 were tested. were generated by SWISS-MODEL, using LC/E crystal structure (PDB:3d3x), and SNAP23b and SNAP25a were not tested because the a-b iso- images were generated in PyMol. forms of SNAP23 and SNAP25 were identical at the LC/E the P2 residue of SNAP25, D179, is recognized by the basic S2 pocket of LC/E via the basic residue, K224, which contributes to LC/E substrate recognition (Fig. 1C, upper panel). This allowed the hypothesis that ‘‘the salt bridge between K224 of LC/E and D179 of SNAP25 contributes the ability of LC/E to cleave SNAP25 and that charge repulsion between K224 of LC/E and the P2 residues of SNAP23, K185, contributes to the inability of LC/E to cleave SNAP23.’’ To test this hypothesis, a point mutation, K224D, was introduced into LC/E and tested for the ability to cleave human SNAP23 (Fig. 1C, lower panel). 224 LC/E(K D) cleaved human SNAP23 with a Km of approxi- Ϫ1 mately 3 ␮M and kcat of approximately 17 S (Fig. 2), within 2-fold of the Km and 5-fold of the kcat of LC/E for the cleavage of human SNAP25. The specific activity for the cleavage of SNAP23 by LC/E(K224D) was similar to the cleavage of VAMP-2 by the B serotype of BoNT and approximately 10-fold faster for the cleavage of VAMP-2 by tetanus toxin (20, 21).