SNARE tagging allows stepwise assembly of a multimodular medicinal toxin

Abstract

Generation of supramolecular architectures through controlled linking of suitable building blocks can offer new perspectives to medicine and applied technologies. Current linking strategies often rely on chemical methods that have limitations and cannot take full advantage of the recombinant technologies. Here we used SNARE proteins, namely, syntaxin, SNAP25, and synaptobrevin, which form stable tetrahelical complexes that drive fusion of intracellular membranes, as versatile tags for irreversible linking of recombinant and synthetic functional units. We show that SNARE tagging allows stepwise production of a functional modular medicinal toxin, namely, botulinum neurotoxin type A, commonly known as BOTOX. This toxin consists of three structurally independent units: Receptor-binding domain (Rbd), Translocation domain (Td), and the Light chain (Lc), the last being a proteolytic enzyme. Fusing the receptor-binding domain with synaptobrevin SNARE motif allowed delivery of the active part of botulinum neurotoxin (Lc-Td), tagged with SNAP25, into neurons. Our data show that SNARE-tagged toxin was able to cleave its intraneuronal molecular target and to inhibit release of neurotransmitters. The reassembled toxin provides a safer alternative to existing botulinum neurotoxin and may offer wider use of this popular research and medical tool. Finally, SNARE tagging allowed the Rbd portion of the toxin to be used to deliver quantum dots and other fluorescent markers into neurons, showing versatility of this unique tagging and self-assembly technique. Together, these results demonstrate that the SNARE tetrahelical coiled-coil allows controlled linking of various building blocks into multifunctional assemblies.

Molecular biology and the advent of recombinant production of proteins revolutionized science. The use of recombinant polypeptides, functional fragments of proteins, and whole enzymes is now widespread in medicine, diagnostics, nanotechnologies, and consumer bioindustries. Despite the obvious success of recombinant technologies, protein size remains an obstacle to producing the ever-more-sophisticated proteins as single multifunctional units. It is believed that combining multiple functions in supramolecular assemblies, rather than in individual proteins, would allow us to overcome this bottleneck (1, 2). Clearly, achieving such a goal of building nanofactories or nanomachines depends on our ability to link various functional units on demand and with high precision (1, 3). It is surprising that current efforts in this promising field still rely on linking technologies that were invented several decades ago: biotin–streptavidin pairing, antibody–epitope recognition, and chemical linking through amino- or sulfhydryl groups (4). These approaches are often limiting because of the need to chemically modify recombinant proteins or the complexity of antibody-based techniques. The recently-developed “click” chemistry addresses some of these issues but still relies on chemically synthesized organic compounds and, to date, has not achieved linking of recombinant proteins into a proven supramolecular assembly (4). Alternative approaches based on self-assembly of DNA (5) or homo-oligomerizing polypeptides (6) also have their limitations in designing multifunctional recombinant assemblies. Here we explored a possibility of using the Soluble N-ethylmaleimide sensitive factor Attachment protein REceptor (SNARE) complex (7, 8), to achieve irreversible linkage of recombinant polypeptides into a functional unit.

SNARE proteins drive fusion of cellular membranes in every eukaryotic cell by forming a heteromeric, tetrahelical coiled coil (8). The brain-derived SNARE complex consists of three proteins: synaptobrevin, syntaxin, and SNAP25 (9). Whereas syntaxin and synaptobrevin each contribute a single helix, SNAP25 contributes two distinct helices to form the tetrahelical coiled-coil. The four 55-aa-long SNARE motifs assemble in the N- to C-terminal parallel direction using their coiled-coil–characteristic heptad repeats (8). The assembled neuronal SNARE complex is extraordinary in its stability, exhibiting resistance to chaotropic agents, strong detergents, proteases, and elevated temperatures (10). We decided to investigate whether fusing the SNAREs to recombinant proteins would allow controlled building of a supramolecular entity. As an example of a multifunctional molecule, we focused on a botulinum neurotoxin that was deservedly described as a “nanomachine that unites recognition, trafficking, unfolding, translocation, refolding and catalysis” (11). The Botulinum NeuroToxin type A (BoNT/A) has proved to be of great medical importance because of its ability to cause a very long neuromuscular paralysis upon local injections of minute amounts (pM concentration) (12, 13). This toxin is currently produced in its parental Clostridium bacterium, necessitating a complex series of purification steps to obtain the required medical formulation. BoNT/A is translated as a single 150-kDa polypeptide that is then proteolysed by trypsin to yield the so-called Light chain (Lc; 50 kDa) and Heavy chain (100 kDa), the two held together by a disulphide bond (Fig. 1A). The Lc is a metalloprotease that cleaves botulinum neurotoxin's intraneuronal target, namely, SNAP25, causing blockade of neurotransmitter release (14, 15). The Heavy chain assists Lc to enter the intraneuronal environment. The Heavy chain is made of the N-terminal 50-kDa translocation domain (Td) and the C-terminal 50-kDa part. The latter is responsible for binding to neuronal gangliosides and the synaptic vesicle receptor, SV2 (16, 17) and thus is referred to as receptor-binding domain (Rbd). The three main modules (Lc, Td, and Rbd) can be recognized as separate structural units in the X-ray model (18).

Fig. 1.

SNARE tagging allows Rbd-mediated delivery of quantum dots to synaptic endings. (A) Diagram showing structure of native botulinum neurotoxin, composed of the Light chain (Lc), Translocation domain (Td), and Receptor-binding domain (Rbd). (B) Schematic diagram showing SNARE tagging scheme for linking streptavidin-coated quantum dot (Qdot) with Receptor-binding domain of botulinum neurotoxin (Rbd). Biotin(star)-syntaxin peptide (syx, red) allows SNARE tagging of Qdot, whereas synaptobrevin peptide (syb, blue) is fused to Rbd. Addition of SNAP25 (green) allows attachment of Rbd to Qdot. (C) Coomassie-stained SDS gel showing irreversible assembly of biotinylated syntaxin3 peptide, SNAP25, and syb-Rbd into an SDS-resistant complex, biotin-SNARE-Rbd. (D) Qdots carrying Rbd exhibit synaptic binding, as evidenced by coincidence of Qdot fluorescence (green) and immunostaining of synaptic vesicle marker synaptophysin (red) at axonal extensions of cultured hippocampal neurons. Omission of SNAP25 during assembly prevents targeting of Qdots to synaptic terminals.

Results

We fused synaptobrevin to the Rbd (Fig. 1B) and probed whether this fusion protein could deliver quantum dots (Qdots) into neuronal endings. The synaptobrevin–Rbd fusion was able to form the SNARE complex with SNAP25 and a 52-aa syntaxin3 peptide labeled with biotin for binding to streptavidin-coated Qdots (Fig. 1C). We chose to use the syntaxin3 SNARE motif rather than syntaxin1, the conventional neuronal SNARE partner, because the latter has a tendency to homooligomerize (19). The Qdots with prebound biotinylated syntaxin peptide were incubated with synaptobrevin–Rbd in the presence or absence of SNAP25, and the delivery of Qdots into synaptic endings was assessed in cultures of hippocampal neurons obtained from mice. Qdots carrying SNARE-tagged Rbd (+ SNAP25) accumulated at a subset of synaptic contacts as confirmed by staining with the vesicular protein synaptophysin (Fig. 1D). This shows that the targeting part of BoNT/A following recombinant tagging is still capable of recognizing its synaptic receptor and can deliver a large cargo with a SNARE counter tag to neuronal synapses.

Next, we prepared a fusion product consisting of GST (a purification tool), the enzymatic part Lc, translocating part Td, and, finally, C-terminally positioned SNAP25 (Fig. 2A). To mimic the native, trypsin-based nicking between the Lc and Td and to facilitate production of the functional LcTd part during the isolation procedure, we added a thrombin cleavage site between the two domains. The GST-LcTd-SNAP25 fusion was successfully expressed in Escherichia coli and could be purified to homogeneity on glutathione beads. Because the thrombin cleavage site is also present between the purification tag, GST, and LcTd, the addition of thrombin to GST-LcTd-SNAP25 led to elution of the activated LcTd-SNAP25. The two botulinum parts were still held together by the disulphide bond because of the natural oxidation that takes place either inside bacteria or following bacterial lysis. Indeed, when treated with a reducing agent, DTT, this fusion protein separated into two parts showing the functionality of the critical disulphide bond (Fig. 2B). We then tested whether the SNARE tags would allow assembly of the LcTd and Rbd parts into a single entity. Fig. 2C, Left, shows that combining LcTd-SNAP25 with synaptobrevin-Rbd in the presence but not in the absence of the syntaxin peptide led to the emergence, within 60 min, of another molecular entity, LcTd-SNARE-Rbd, as evidenced by the SDS gel. To aid further visualization of neuronal targeting of the reassembled BoNT/A, we used a fluorescein-labeled version of the syntaxin peptide, and the LcTd-SNARE-Rbd indeed could be visualized as a fluorescent protein assembly (Fig. 2C, Right). It is noteworthy that even though the Lc enzyme acts to remove the last nine C-terminal amino acids of SNAP25, our results show that such possible cleavage in the context of LcTd-SNAP25 does not compromise SNARE assembly, in agreement with previous observations (20).

Fig. 2.

SNARE tagging allows stepwise assembly of individual parts of BoNT/A into a single molecular entity. (A) Structure of SNARE-tagged botulinum neurotoxin. Positions of disulphide bond and engineered SNARE tags between LcTd and the Rbd part of BoNT/A are indicated. (B) LcTd, tagged with SNAP25, can be purified and broken into Lc and Td-SNAP25 following treatment with 50 mM DTT. Coomassie-stained SDS gel. (C) LcTd, tagged with SNAP25, can be united with Rbd, tagged with synaptobrevin, upon addition of the syntaxin3 peptide as evidenced by the Coomassie-stained and fluorescently imaged SDS gels.

When the LcTd-SNARE-Rbd was applied to cultured hippocampal neurons, the fluorescent molecule localized to a subset of synaptic contacts along dendritic extensions, suggesting binding to the native target of the BoNT/A (Fig. 3A). Crucially, immunoblotting of proteins from the treated neurons with anti-SNAP25 antibody demonstrated that the intraneuronal SNAP25 had undergone cleavage in the same fashion as when neurons were treated with the native BoNT/A molecule (Fig. 3B). This shows that the enzymatic part was successfully released into neuronal cytosol upon entry of the LcTd-SNARE-Rbd into synaptic vesicles. It is known that the translocation of Lc into the neuronal cytosol takes place only upon acidification of the intravesicular space driven by a proton pump. Because this pump can be inhibited by bafilomycin A1, we tested whether this drug would interfere with LcTd-SNARE-Rbd function. Fig. 3C shows that the addition of LcTd-SNARE-Rbd in the presence of bafilomycin led to blockade of SNAP25 proteolysis, which is in agreement with results previously obtained with native botilinum neurotoxin (21).

Fig. 3.

SNARE-linked botulinum neurotoxin exhibits synaptic localization and cleaves its intrasynaptic target. (A) Fluorescein-labeled LcTd-SNARE-Rbd binds to axonal extensions of hippocampal neurons. Immunostaining with antisynaptophysin antibody highlights presynaptic terminals of cultured hippocampal neurons. (B) Immunoblot showing cleavage of intrasynaptic SNAP25 by assembled neurotoxin in a fashion similar to that of native BoNT/A. (C) Immunoblot showing that cleavage of SNAP25 by LcTd-SNARE-Rbd (1 nM) is blocked by bafilomycin A1, which prevents acidification of recycled synaptic vesicles.

To test the effect of LcTd-SNARE-Rbd on neurotransmitter release, we used a 96-well glutamate release assay that allows simultaneous comparison of multiple factors (22). Fig. 4 A and B shows that LcTd-SNARE-Rbd was able to inhibit calcium- and KCl-dependent release of glutamate from isolated brain nerve endings with dose dependency similar to that of the native BoNT/A. The degree of inhibition of the glutamate release from central synaptic endings is in good agreement with the value obtained previously (23), and suggests that not all central synapses carry the high-affinity SV2C receptor for BoNT/A (16). Importantly, mixing the SNARE-tagged LcTd and Rbd in the absence of the linking syntaxin peptide resulted in inactive molecules, confirming that full SNARE assembly is the key factor in linking recombinant parts into a functional entity (Fig. 4C). Treatment with dithiotreitol of LcTd-SNARE-Rbd inactivated the assembled toxin, indicating the functional importance of the disulphide bond between Lc and Td.

Fig. 4.

SNARE-linked botulinum neurotoxin inhibits neurotransmitter release from brain nerve endings and in neuromuscular junctions. (A and B) Fluorometric measurements of glutamate release from isolated rat brain synaptic endings (synaptosomes) indicate similar degree of inhibition between LcTd-SNARE-Rbd and BoNT/A. Real-time glutamate release graph (A) and dose-dependence graph (B, assessed after 15-min stimulation with 35 mM KCl and 2 mM CaCl₂) were obtained following 1-h incubation of synaptosomes with toxins. (C) Individual SNARE-tagged neurotoxin parts do not block glutamate release following 1-h incubation with synaptosomes, as assessed after 15-min stimulation. (D) LcTd-SNARE-Rbd blocks spontaneous synaptic activity (MEPPs) in mouse flexor digitorum brevis preparations. Degree of inhibition is further enhanced when exocytosis is stimulated with 30 mM KCl during toxin application (activity dependent). Data were recorded for five myofibers for each condition in eight muscle preparations (shown with SEM). (E) Example of endplate potentials following application of 15 mM KCl in control and LcTd-SNARE-Rbd–treated synapses.

LcTd-SNARE-Rbd also potently inhibited spontaneous exocytosis [miniature end-plate potentials (MEPPs)] at the neuromuscular junctions of mouse paw muscle (flexor digitorum brevis). The frequency of MEPPs in nerve terminals that were intoxicated with LcTd-SNARE-Rbd under resting conditions decreased to ∼30% of that in control samples treated with LcTd-SNAP25 and synaptobrevin-Rbd only (Fig. 4D). To determine whether the degree of inhibition depended on synaptic activity, some muscle preparations were stimulated with elevated KCl during incubation with the toxins. After recovery from stimulation, spontaneous synaptic activity in muscle preparations treated with LcTd-SNARE-Rbd ceased almost completely compared with respective control (activity dependent, Fig. 4D). This block of spontaneous exocytosis never developed in the absence of the linking syntaxin peptide. Our data indicate that the uptake and action of LcTd-SNARE-Rbd depends on the exocytotic activity of nerve terminals. In the diaphragm muscle preparations, LcTd-SNARE-Rbd significantly depressed exocytosis evoked by elevated concentrations of KCl, leading to a threefold decrease in the frequency of Ca-dependent evoked end-plate potentials (0.36 ± 0.04 s vs. 0.12 ± 0.01 s for treated and untreated, respectively; Fig. 4E).

Finally, we measured performance of LcTd-SNARE-Rbd in biological assays. We applied several concentrations of the assembled neurotoxin on isolated mouse hemidiaphragm preparations and tested the paralytic response of phrenic nerves. The time required to decrease the amplitude to 50% of the starting value (paralytic half-time) was determined. Fig. 5A shows that LcTd-SNARE-Rbd paralyzed the diaphragm muscle at subnanomolar concentrations (190 pM) within 72 min. Finally, we applied the LcTd-SNARE-Rbd to brain slices and probed its ability to block a long-term physiological function. Daily (circadian) pacemaking in the suprachiasmatic nucleus (SCN), the brain's principal circadian clock, is dependent on interneuronal signaling, and treatment of organotypic SCN slices with BoNT/A has been shown to compromise circadian rhythms (24). Application of LcTd-SNARE-Rbd (6 nM) to SCN slices caused a quick (within 24 h) and persistent (more than 5 d) damping of the circadian gene expression rhythm, as monitored by bioluminescence (Fig. 5B). The amplitude of the circadian oscillation declined by 81% ± 2.5% (vs. the 30% ± 12.6% decline in vehicle-treated slices, arising from declining luciferin substrate availability). Moreover the precision of the circadian cycle was significantly reduced (relative amplitude error: pretreatment, 0.03 ± 0.01; posttreatment, 0.12 ± 0.02). Thus, LcTd-SNARE-Rbd efficiently down-regulated circuit-dependent functions of central SCN neurons in culture.

Fig. 5.

SNARE-linked botulinum neurotoxin inhibits neuromuscular and CNS functions. (A) Graph showing dose-dependent inhibition of isometric contractions of mouse diaphragm by LcTd-SNARE-Rbd. Error bars represent SD; n = 3. (B) Representative trace of circadian bioluminescence rhythms of suprachiasmatic nucleus (SCN) slice treated with 6 nM LcTd-SNARE-Rbd. Note rapid and sustained suppression of amplitude of circadian oscillation following application of the assembled toxin (arrow).

Discussion

Here we have introduced a unique method that allows nonchemical linking of recombinantly produced proteins. This technique could be used to aid production of functional molecules when (i) expression of a single large protein is problematic, (ii) there are safety issues due to protein toxicity, (iii) one wishes to explore combinatorial profiling of multidomain assemblies, and (iv) nonchemical, quick, oriented attachment of recombinant proteins to surfaces, nanoparticles (25), and imaging reagents is required. We chose to test the SNARE-based linking on botulinum neurotoxin because of its highly toxic nature and its importance in treatment of more than 100 medical conditions, including dystonias, gastrointestinal disorders, facial spasms, strabismus, cerebral palsy, stuttering, hyperlacrymation, hyperhidrosis, spasms of the inferior constrictor of the pharynx, spastic bladder, and migraine (13, 26).

Our results show that SNARE tagging allows a stepwise assembly of BoNT/A (27). The heterotetrameric SNARE bundle offers a number of linking configurations and here we decided to tag the LcTd part with SNAP25 leaving us with two shorter SNARE tags for versatile addition of functional units (e.g., for future chemical synthesis). Tagging of the receptor-binding part Rbd with the synaptobrevin SNARE motif allowed us to reassemble BoNT/A simply by addition of a fluorescent/biotinylated syntaxin peptide. The assembly reaction is complete within 1 h at 20 °C, giving rise to a supramolecular assembly that cannot be broken even by the harsh detergent SDS. Several functional tests used in this study demonstrated that the SNARE-tagged toxin parts are functional. First, the Rbd could be attached via SNARE assembly to either Qdots or fluorescently labeled syntaxin peptide and then can be used to visualize synaptic contacts on neuronal extensions. When tested on central neurons and brain-derived nerve endings, LcTd-SNARE-Rbd displayed activity similar to that of native neurotoxin, showing that SNARE-tagging of LcTd does not compromise translocation and proteolytic activities of corresponding toxin parts. Although the efficiency of the LcTd-SNARE-Rbd in blocking neuromuscular junctions was less than that of the native BoNT/A (16), this result can be explained by either reduced ability of the structurally extended toxin to reach distant active zones within long neuromuscular junctions or by a large volume of the presynaptic ending at the muscle leading to compromised efficiency. In contrast, the suppression by LcTd-SNARE-Rbd of circadian pacemaking in the central neurons of the suprachiasmatic nucleus was comparable, in terms of extent and time-course, to that seen with WT toxin (24). The greater efficiency of the LcTd-SNARE-Rbd in silencing interneuronal communication compared with neuromuscular transmission potentially can be used in the future, for example, to block sensory pathways involved in pain transduction without causing local neuromuscular paralysis.

Specific advantages of the fragment-linking approach in the case of BoNT/A are listed here. First, it is now possible to express an active form of a multimodular medicinal toxin in common bacteria in a safe way; in fact, our use of a “locking” peptide allows an additional safety feature. Few establishments have the capability to make and to use native or mutated neurotoxin for research at the scale required; therefore the ability to reassemble the functional toxin from innocuous parts will go some way in addressing this requirement. It is also possible to use the SNARE-tagged Rbd part to deliver imaging agents and future therapeutics into neurons by tagging them with SNARE counterparts (28). The SNARE tagging of the LcTd part, on the other hand, will allow a retargeting of the active portion of BoNT/A to specific neuroendocrine cells (29). Here one can target neuropeptide or growth factor receptors by making corresponding SNARE-tagged ligands. Such SNARE tagging would allow a convenient combinatorial mixing of various functional units with the aim of finding the most beneficial combination(s) to silence specific subsets of neurons (30).

Although we used BoNT/A as an example of a sophisticated nanomachine, it is clear that SNARE tagging can be used in building, in a highly controlled manner, other multimodular medicinal toxins (e.g., diphtheria toxin and ricin) to target specific cells, including cancer cells. Generation of well-defined, functional supramolecular architectures of nanometer size through controlled linking of suitable building blocks is believed to offer new perspectives for many fields (3). Our results indicate that SNARE tagging allows a unique approach for stepwise construction of various supramolecular assemblies for nanobiotechnology and medicine. The relatively short SNARE motifs allow combination of both recombinantly produced polypeptides and chemically synthesized molecules, as evidenced by incorporation of biotin and fluorescein upon assembly of the botulinum neurotoxin. The greatest advantage of the SNARE coiled coil is its heterotetrameric nature, which could allow linking of up to eight distinct functionalities (i.e., using both ends of all four helices). This potential has yet to be exploited in future medicine and applied technologies.

Methods

Plasmids, protein purification and assembly reactions, neuronal imaging, immunoblotting, measurements of neurotransmitter release, hemidiaphragm paralytic assay, and bioluminescent circadian recordings are described in SI Methods.