Recombinant botulinum neurotoxin A heavy chain-based delivery vehicles for neuronal cell targeting

Abstract

The long half-life of botulinum neurotoxin serotype A (BoNT/A) in cells poses a challenge in developing post-exposure therapeutics complementary to existing antitoxin strategies. Delivery vehicles consisting of the toxin heavy chain (HC), including the receptor-binding domain and translocation domain, connected to an inhibitory cargo offer a possible solution for rescuing intoxicated neurons in victims paralyzed from botulism. Here, we report the expression and purification of soluble recombinant prototype green fluorescent protein (GFP) cargo proteins fused to the entire BoNT/A-HC (residues 544–1295) in Escherichia coli with up to a 40 amino acid linker inserted between the cargo and BoNT/A-HC vehicle. We show that these GFP-HC fusion proteins are functionally active and readily taken up by cultured neuronal cells as well as by neuronal cells in mouse motor nerve endings.

Introduction

Clostridium botulinum neurotoxins serotypes A through G (BoNT/A-G) are considered the most potent protein toxins for humans (Arnon et al., 2001). The intravenous lethal dose of BoNT/A, which targets peripheral cholinergic neurons, where it selectively cleaves synaptosome-associated protein of 25 kDa (SNAP25) and prevents neurotransmitter release to cause flaccid paralysis, is estimated to be 1–10 ng/kg (Arnon et al., 2001). BoNTs have long half-lives in cells, with the various serotypes causing neuromuscular paralysis for durations ranging from days to months (Foran et al., 2003). Although BoNT/A is listed as a ‘category A-select agent’ by the Centers for Disease Control and Prevention due to its potential use as a bioweapon (Arnon et al., 2001), it is gaining widespread use as a valuable therapeutic agent for various neuronal disorders (Montecucco and Molgo, 2005; Fabbri et al., 2008) and as a cosmetic (BOTOX^®) for anti-wrinkle applications (Carruthers and Carruthers, 2001; Fabbri et al., 2008).

Currently, the only available treatment for botulism is a combination of antitoxin immunoglobulin therapy and long-term respiratory care, which in the event of mass exposure to BoNT would overwhelm current medical infrastructure. After internalization, antibodies or drugs that work in the bloodstream no longer neutralize the toxin, thus there is only a short therapeutic window available for administration of current forms of antitoxin. Antidotes are urgently needed that can reverse the detrimental effects of BoNT particularly once it has been internalized into nerve cells. Clearly there is a keen need for post-exposure therapies against paralysis from botulism. However, one of the major challenges encountered in realizing post-exposure therapies is the selective targeting and delivery of the antitoxin agents into the intoxicated neuron. A logical choice for delivery of anti-botulism agents targeting the toxin molecules that have already entered the neuronal cell would be to utilize the exquisite specificity and potency provided by the normal intoxication mechanism of BoNT that resides within the heavy chain (HC) and includes both binding (BD) and translocation (TD) domains.

The ability of BoNT-HCs to transport large polypeptides (i.e. BoNT-LCs) across cellular membranes, and importantly into the specific cells that are intoxicated, could be exploited for delivery of drugs to cytosolic targets. TeNT conjugates have been used to transport DNA (Knight et al., 1999) and enzymes (Figueiredo et al., 1997; Roux et al., 2005) into neurons. Derivatives of BoNT/A and BoNT/B have been used as delivery vehicles to target compounds specifically to human neuroblastoma cells (Zdanovskaia et al., 2000). Full-length BoNT/D was found to deliver various cargo proteins [dihydrofolate reductase, BoNT/A-LC, luciferase or green fluorescent protein (GFP)] fused at the N-terminus into neurons (Bade et al., 2004). The advantage of using native BoNT entry pathway for delivery is that the cargo will be transported to the site of the target. When Oregon green 488-labeled 10-kDa dextran was conjugated to Cy3-labeled BoNT/A-HC through Cys454, signals from internalized Oregon green 488 and Cy3 were found to be colocalized in vesicles in BoNT/A-sensitive cultured cortical cells, supporting the feasibility of using BoNT/A-HC as a delivery vehicle for intracellular transport of BoNT/A antagonists (Goodnough et al., 2002; Zhang et al., 2009).

With these precedents, it is desirable to have a rapid and convenient method for expression and purification of various cargo proteins fused to the entire BoNT-HC, comprising both TD and BD. These fusion proteins could then be screened for their antitoxin activity or other intracellular modulator activity inside neuronal cells. Several experimental and technical barriers in application of this approach need to be overcome to enable replacement of the catalytic domain of the toxin with other cargo molecules. Foremost is the known difficulty in expression and production of stable and soluble BoNT/A-HC fragments containing BD and TD in recombinant systems. Another obstacle is the inefficiency and inconsistency of chemical methods for conjugation of potential inhibitor cargos to a purified HC-based vehicle (Goodnough et al., 2002; Zhang et al., 2009).

Expression and purification of native full-length or truncated BoNTs from Clostridium species is problematic in terms of cost, biosafety requirements and low yields (Byrne and Smith, 2000). Fusion proteins containing only the BD of BoNTs have been expressed in Escherichia coli (LaPenotiere et al., 1995; Baldwin et al., 2008), but these truncated versions lacked the TD that provides a functional endosome to cytosol translocation mechanism reported necessary for intoxication (Koriazova and Montal, 2003; Fischer et al., 2008). Although a fusion protein with the BD alone expresses well in E.coli, obtaining larger sized soluble and stably expressed BoNT-HC with the TD has been considered difficult (Byrne et al., 1998; Zhou and Singh, 2004; Band et al., 2010). It has been speculated that folding or proteolysis may contribute to this difficulty. Indeed, expression of BoNT/B-HC (residues 624–1291) in E.coli resulted in formation of inclusion bodies that required refolding steps during purification (Zhou and Singh, 2004), which leads to questions regarding complete recovery of functionality in the resulting protein. This difficulty has led a number of investigators to resort to alternative, albeit less-convenient expression systems, such as the Pichia pastoris yeast system (Byrne et al., 1998; Bouvier et al., 2003) or Sf9 baculovirus system (Band et al., 2010).

Although we similarly found poor protein production using pET vectors or pTrcHis vectors, a modified vector of pTrcHis, named pGEpi, designed for expression of GFP-fusion proteins, made it possible for production of GFP-BoNT/A(TD–BD). The fusion proteins produced in E. coli using this expression system were soluble and stable throughout our simple purification steps. Moreover, the resulting fusion proteins were expressed well and appeared correctly folded with full functional properties, including binding and entry into neuronal cells and appropriate cellular targeting in nerve muscle preparations isolated from adult mice. Results described herein demonstrate the utility of our expression system for generating functionally viable BoNT-HC-based fusion proteins for delivery of inhibitor cargos or fluorescent markers to neuronal cells.

Material and methods

HIS-select^® nickel affinity gel (P6611), Protease Inhibitor Cocktail (P8849), DNAse I (DN25) and IGEPAL^® CA-630 (I3021) were purchased from Sigma (St. Louis, MO, USA); bovine serum albumin (BSA) standard from Pierce (Rockford, IL, USA); isopropyl-β-d-thio-galactopyranoside (IPTG) and Miller's LB broth from RPI (Mt. Prospect, IL, USA); benzamidine, imidazole, Hyclone bovine growth serum (BGS) from Thermo-Fisher (Waltham, MA, USA); lysozyme, ribonuclease A (RNase), phenylmethylsulfonyl fluoride (PMSF) from Ameresco (Solon, OH, USA); HiTrap ANX and SP columns and PD-10 columns from GE Healthcare (Piscataway, NJ, USA); PCR primers from IDT (Coralville, IA, USA); restriction enzymes from Fermentas (Glen Burnie, MD, USA); fluorescently labeled α-BnTx, Top10 E.coli cells, Zero-Blunt TOPO PCR cloning kit with pCR4-TOPO vector, T4 DNA ligase, Pfx polymerase, dNTPs, tissue culture media and supplies from Invitrogen (Carlsbad, CA, USA); PC12 (CRL-1721) and NG108-15 (HB123-17) cells from ATCC (Manassas, VA, USA); Rosetta(DE3) E.coli cells from EMD/Novagen (Gibbstown, NJ, USA); BoNT/A from Metabiologics Inc. (Madison, WI, USA); mouse anti-troponin I (ab10231) and TRITC-conjugated goat anti-mouse IgG (ab50596) from Abcam; and fluorescently labeled BoNT/A from BBTech (Dartmouth, MA, USA).

Construction of pGEpi vector

A restriction fragment containing the GFP gene from vector pGFPuv (Clontech) was inserted into the KpnI and NotI restriction sites of the pTrcHisC vector (Invitrogen) to yield the pTHC-GFP vector. This plasmid was then used as a template with overlapping primers to construct a synthetic gene encoding GFP with a T7 promoter sequence (TAATACGACTCACTATAGGG) and a lac operator sequence (AATTGTGAGCGGATAACAATT) at the 5′-end and HA-cMyc tag sequence at the 3′-end. This synthetic gene was then introduced back to the pTHC-GFP vector at the EcoRV and HindIII sites. The resulting vector pGEpi (GFP-Epi) contains a truncated lacI_q gene. This vector allows expression of GFP in Top10 cells on LB agar plate without induction.

Construction of BoNT/A-HC gene

A synthetic gene encoding BoNT/A (residues 448–1295) with codons optimized for E. coli, based on the bont/A gene (Genbank accession number ABA29018), was constructed through fragment extension using overlapping PCR primers utilizing standard cloning and PCR procedures, and the final product was cloned into the pCR4-TOPO vector. The resulting construct pCR4-TOPO-BoNT/A(448–1295) was then used as a template to generate the other BoNT/A-HC constructs. The numbering of amino acid residues in all constructs is according to the assigned numbers in the crystal structure PDB 3BTA (Lacy et al., 1998).

Construction of GFP-BD and GFP-HC expression vectors

PCR amplification with pCR4-TOPO-BoNT/A(448–1295) as a template was performed using primers to introduce the C-terminal residues of GFP, a linker (-SGGPG-) containing a SmaI site before residues Thr875 or Leu544, and a HindIII site after the STOP codon of the BoNT/A sequence. The PCR reaction product was digested with SacI and HindIII and inserted into the pGEpi vector to generate pGFP-BoNT/A-BD(875–1295) and pGFP-BoNT/A-HC(544–1295), correspondingly.

Construction of GFP-(DE)n-BoNT/A-HC(544–1295) expression vectors

With the pGFP-BoNT/A-HC(544–1295) vector as a template, PCR amplification was performed using primers to copy HC and introduce a PGDEDEDEDE (DE4) sequence in front of residue Leu544. The SmaI–HindIII fragment from the PCR product was then exchanged back to the template vector to generate the GFP-DE4-BoNT/A-HC(544–1295) vector. With pGFP-DE4-BoNT/A-HC(544–1295) vector as a template, PCR primers were used to copy GFP and mutate the existing SmaI site and add a SmaI after the end of the DE4 sequence. The KpnI–SmaI fragment from the PCR product was then exchanged back into the template vector to generate the GFP-DE8-BoNT/A-HC(544–1295) vector. With the pGFP-DE8-BoNT/A-HC(544–1295) vector as a template, PCR amplification was performed using primers to copy GFP-DE8 and mutate the existing SmaI site and add an SmaI after the end of the DE8 sequence. The KpnI–SmaI fragment from the PCR product was then exchanged back into the template vector to generate the pGFP-DE12-BoNT/A-HC(544–1295) vector. With the pGFP-DE8-BoNT/A-HC(544–1295) vector as a template, PCR amplification was performed using primers to copy DE8-HC(544–1295) and add an NaeI site in front of the DE8 sequence. The NaeI–HindIII fragment from the PCR product was then ligated into the pGFP-DE12-BoNT/A-HC(544–1295) vector to generate the pGFP-DE16-BoNT/A-HC(544–1295) vector.

Protein expression and purification

Seed culture, each with 50 ml LB broth (ampicillin 100 mg/l) in 125 ml baffled flask, was inoculated with a single colony picked from an ampicillin-LB agar plate and incubated overnight at 30°C on a shaker. Eight or nine production flasks, each with 500 ml LB broth and ampicillin (ampicillin 100 mg/l) in 1 l baffled flasks, were inoculated with 1 ml of overnight seed culture and incubated on a shaker at 30°C until an OD₆₀₀ of 0.6. The bacterial cultures were induced with IPTG (100 mg/l) and transferred to a shaker at 18°C. Bacteria cells were harvested 16–20 h after induction by centrifugation at 3000 × g for 15 min. The pellets were resuspended in 200 ml of lysis buffer (PBS, pH 7.4, 1% IGEPAL, 200 µl Inhibitor Cocktail, 200 mg benzamidine, 0.2 Kunitz-unit DNase I, 4 mg RNase A, 60 mg lysozyme, 60 mg PMSF and 100 µl β-mercaptoethanol). This cell suspension was chilled on ice and subjected to ultrasonication for three to five 1-min cycles, with 1-min intervals cooling on ice. The resulting cell lysate was centrifuged at 20 000 × g for 2 h. The supernatant was passed twice through a HIS-Select Ni-affinity column (10 ml bed volume in a 50-ml column). The matrix was washed with 10 bed volumes of Tris buffer (50 mM Tris–HCl, pH 7.5 and 2.5% glycerol) and eluted with 10 bed volumes of Tris buffer containing imidazole (100 mM) and then loaded onto a HiTrap SP cation exchange column (5 ml bed volume). The proteins were separated on the HiTrap column by FPLC using a NaCl gradient in Tris buffer. Fractions showing green fluorescence were analyzed by using 10% SDS–PAGE gel. Additional HiTrap ANX anion and SP cation exchange columns were used for further purification. The final protein was then desalted by using a PD-10 column, and quantified by densitometry of a Coomassie-stained SDS–PAGE gel using BSA as a standard. The desalted protein was flash frozen with liquid nitrogen and stored at −20°C until use.

Cell culture and fluorescence microscopy

PC12 cells and NG108-15 cells were cultured and maintained at 37°C and 5% CO₂ in DMEM medium supplemented with 10% heat-inactivated BGS, 100 units/ml penicillin G and 100 µg/l streptomycin, pH 7.4. For microscopy experiments, cells were seeded onto cover slips (20–30% cell density) placed in 6-well plates in medium with 2% BGS and incubated overnight. Cells were treated with proteins at the indicated concentrations and treatment times. The cells were washed three times with 1× PBS and fixed with 3.7% formaldehyde for 30 min, followed by washing three times with 1× PBS before mounting on slides. Protein localization was visualized by fluorescence microscopy using Olympus IX70 inverted microscope equipped with an Olympus DP70 digital camera. Images were acquired with a setting of ISO800, 2-s exposure and 4080 × 3072 pixels.

Confocal imaging of nerve-muscle preparations

Triangularis sterni nerve-muscle preparations dissected from Swiss Webster mice were pinned to a Sylgard-lined Plexiglas chamber and bathed in HEPES-Ringer solution. For evaluation of neurotoxin uptake, preparations were bathed in 10 μg/ml of GFP-BoNT/A-HC(544–1295), GFP-BoNT/A-BD(875–1295) or BoNT/A labeled with Alexa-647 for 90 min at 22°C and protein uptake was initiated in response to HEPES-Ringer solution containing 40 mM KCl (osmolarity adjusted by reducing NaCl to 100 mM). The preparations were washed three times in normal HEPES-Ringer solution. Postsynaptic acetylcholine receptors were then labeled by exposing the fixed tissue to 1 ng/ml of α-bungarotoxin (α-BnTx) labeled with Alexa-647 or Alexa-488 at 4°C for 6 h. The muscle endplate region was cut out and mounted in vectashield on a slide and kept frozen at −20°C prior to imaging with a NIKON LSM-410 confocal microscope equipped with argon and HeNe lasers. Images were saved and represented as TIFF files.

Animals

The Institutional Animal Care and Use Committee at UMDNJ-NJMS approved all animal procedures. Adult Swiss Webster mice were anesthetized with an intraperitoneally injected mixture of ketamine (100 mg/kg) and xylazine (9 mg/kg) prior to surgical exposure of the peroneal nerve's entry into the extensor digitorum longus (EDL) muscle. BoNT/A alone (3 μl of a 1 ng/ml stock solution), GFP-BoNT/A-HC(544–1295) (n= 3; 3 µl 10 mg/ml), GFP-BoNT/A-BD(875–1295) (n= 3; 3 µl 10 mg/ml) or control HEPES buffer (3 μl, Ctrl) was injected bilaterally into the space surrounding the innervation site of the EDL with a 26-gauge Hamilton syringe. The skin incision was then surgically closed using aseptic technique. The toe-spread reflex was evaluated 24 h after injection of BoNT/A, GFP-BoNT/A-HC(544–1295), GFP-BoNT/A-BD(875–1295) or control.

In vivo injections of GFP-BoNT/A-HC(544–1295) and GFP-BoNT/A-BD(875–1295)

Adult Swiss Webster mice were anesthetized by an intraperitoneal injection of a mixture of ketamine (100 mg/kg) and xylazine (9 mg/kg) prior to surgical exposure of the peroneal nerve's entry into the EDL muscle. GFP-BoNT/A-HC(544–1295) (n= 3, 3 µl 2.8 mg/ml) or GFP-BoNT/A-BD(875–1295) (n= 3, 3 µl 0.55 mg/ml) was injected bilaterally into the space surrounding the innervation site of the EDL using a 26-gauge Hamilton syringe. The skin incision was then surgically closed with an aseptic technique. The injection protocol was repeated on two consecutive days. On Day 3, EDL nerve muscle was removed by dissection, and the preparations were mounted on a sylgard-coated chamber, fixed and permeabilized as previously described (Thyagarajan et al., 2009). The skeletal muscle was stained for troponin using anti-troponin I primary antibody (1:1000 dilution in PBS) at 4°C for 6 h. Postsynaptic acetylcholine receptors were then labeled by exposing the fixed tissue to 1 ng/ml of α-BnTx labeled with Alexa-647 at 4°C for 6 h. Preparations were washed thrice with PBS for 20 min each and then visualized using an Olympus BX61WI upright fluorescence microscope, equipped with a photometrics CCD camera and DG4 Illumination system. Images were analyzed using Metamorph software and prepared using Adobe Photoshop.

Results

Purification of GFP-BoNTA-BD(875–1295)

GFP-BoNT/A-BD(875–1295) was purified in good yields (40–50 mg/l) from Rosetta(DE3) E.coli cells carrying the vector pGFP-BoNT/A-BD(875–1295) induced with IPTG at 18°C for 20 h. Shown in Fig. 1 is the analysis of a typical protein preparation, where the desired protein could be obtained in reasonable purity after Ni²⁺-chelation chromatography followed by SP cation exchange chromatography. The GFP-BoNT/A-BD(875–1295) protein could also be obtained from culture induced at 30°C, but with slightly lower yields (data not shown). This protein precipitates at concentrations higher than ∼5 mg/ml, if allowed to stand on ice. Some insoluble precipitate was evident after thawing if the protein was frozen slowly. Irreversible precipitation could be avoided if kept at concentrations below 5 mg/ml in PBS with 10% glycerol and the sample was flash frozen. Smaller-binding domains, such as GFP-BoNT/A-BD(1091–1295), remained soluble even at high concentrations (data not shown).

Fig. 1.

Purification of GFP-BoNT/A-BD(875–1295). Shown are Coomassie blue-stained 10% SDS–PAGE gels of protein fractions during various steps in the purification of GFP-BoNT/A-BD(875–1295). Left panel: Ni²⁺-chelation column flow through (F), wash (W) and SP column elution fractions 15–24. Right panel: pooled SP column fractions (16–20) after desalting, lane 1–1.25 µg, lane 2–2.5 µg.

Purification of GFP-BoNT/A-HC(544–1295) and GFP-(DE)n-BoNT/A-HC(544–1295)

The gene encoding GFP-BoNT/A-HC(544–1295) was originally cloned into a pTrcHis vector, but no intact fusion protein could be isolated using this vector, despite the appearance of green fluorescence in cell pellets and cell free extract (data not shown). Attempts to use high expression vectors, such as pET21, pET28 or pET33, or vectors for expressing proteins with solubility-enhancing tags, such as pGEX (GST fusion) or pMal (maltose fusion), were not successful. Our pGEpi vector derived from pTrcHis with a truncation in lacI_q and an addition of a T7-promotor was initially intended as an expression vector for convenient preparation of GFP fusion proteins. For instance, it was used for expression of GFP-SNAP25 protein as a BoNT/A substrate for gel-shift assay (Pires-Alves et al., 2009).

Our next approach was to use the pGEpi vector for expression of GFP-BoNT/A-HC(544–1295) with 4 repeats of Asp-Glu (DE4) inserted between GFP and the HC for enhanced solubility. Similar poly-charged solubility-enhancing tags have been described before to improve protein solubility (Zhang et al., 2004). The partially purified GFP-DE4-BoNT/A-HC(544–1295) fusion protein was obtained after Ni²⁺-chelation and ANX anion exchange chromatography (Fig. 2A). It was necessary to use multiple ANX column steps to achieve complete purification of GFP-BoNT/A-DE4-HC(544–1295) (see Supplementary data, Fig. S1, left panel). The rationale for inserting additional DE-repeats to improve solubility did not fully materialize, since the longer DE-repeats as in GFP-BoNT/A-DE16-HC(544–1295) caused enhanced degradation (see Supplementary data, Fig. S2) and did not appear to contribute to enhanced production of the fusion proteins. Although the DE-repeats did help the purification of proteins by anion exchange chromatography, it was not possible to separate full-length protein from degraded fragments with ANX column alone.

Fig. 2.

Purification of GFP-DE4-BoNT/A-HC(544–1295). Shown are Coomassie blue-stained 10% SDS–PAGE gels of protein fractions during various steps in the purification of GFP-DE4-BoNT/A-HC(544–1295). (A) Cell-free extract (C), Ni²⁺-chelation column flow through (HF), wash (HW), 100 mM imidazole elution (H) and ANX column elution peak fraction (AN). (B) Cell-free extract (C), Ni²⁺-chelation column flow through (HF), 100 mM imidazole elution (H), SP column flow through (SF) and SP column elution peak fraction (SP).

Inclusion of an SP-column step in the purification scheme eliminated the need for a DE linker. Although the GFP-BoNT/A-DE4-HC(544–1295) protein has an overall pI value of 6.49, the concentration of positively charged residues in the binding domain (pI = 8.65) allowed binding of the fusion protein to an SP cation exchange column (Supplementary data, Fig. S1, right panel). Replacing the SP column step in the overall purification scheme allowed for significantly enhanced purification of GFP-DE4-BoNT/A-HC(544–1295) (Fig. 2B). Indeed, inclusion of additional ANX-column and SP-column steps after the initial SP-column step further optimized this method, such that the fusion protein without DE repeats, GFP-BoNT/A-HC(544–1295) (Fig. 3A), as well as fusions with higher number of DE repeats (DE4 and DE8), were obtained as soluble proteins in reasonable purity and with practical yields (Fig. 3B). All preparations have been repeated over 10 times, with yields of 4.8 mg (DE0), 4.9 mg (DE4), 3.9 mg (DE8), 0.5 mg (DE12) and 0.4 mg (DE16) from 4.0 to 4.5 l of E. coli culture.

Fig. 3.

Purification of GFP-BoNT/A-HC(544–1295) and DE fusion proteins. Shown are Coomassie blue-stained 10% SDS–PAGE gels of protein fractions during various steps in the purification. (A) GFP-BoNT/A-HC(544–1295), cell-free extract (C), Ni²⁺-chelation column flow through (HF), wash (HW), 100 mM imidazole elution (H), pooled elution fractions from first SP column (SP1), pooled elution fractions from ANX column (AN) and pooled and desalted elution fractions from the second SP column (SP2). (B) Similarly purified fusion proteins after desalting, lane 1, GFP-BoNT/A-BD(875–1295); lane 2, GFP-BoNT/A-HC(544–1295); lane 3, GFP-DE4-BoNT/A-HC(544–1295); lane 4, GFP-DE8-BoNT/A-HC(544–1295); lane 5, GFP-DE12-BoNT/A-HC(544–1295); lane 6, GFP-DE16-BoNT/A-HC(544–1295).

Uptake of GFP-BoNTA-HC(544–1295) and GFP-BoNTA-BD(875–1295) into NG108-15 cells

To demonstrate that the recombinant GFP-toxin fusion proteins are folded correctly, we tested for internalization of the proteins into cultured neuronal cells using fluorescence microscopy. NG108-15 neuronal cells readily took up both GFP-BoNT/A-HC(544–1295) and GFP-BoNT/A-BD(875–1295) proteins, with as little as 0.25 µM protein visible after 0.5 h (data not shown). As shown in Fig. 4, the GFP fluorescence was clearly visible after 1 h in vesicles as punctated spots, corresponding to accumulation of GFP-fusion proteins in endosomes or perinuclear compartments. Similar results were observed for PC12 cells (data not shown).

Fig. 4.

Binding and entering of GFP-BoNT/A-BD(875–1295) and GFP-BoNT/A-HC(544–1295) into cultured neuronal cells. Shown are fluorescence micrographs of NG108-15 cells treated for 1 h with either 0.25 µM GFP-BoNT/A-HC(544–1295) (top panels) or 0.62 µM GFP-BoNT/A-BD(875–1295) (bottom panels).

Uptake of GFP-BoNT/A-HC(544–1295) and GFP-BoNT/A-BD(875–1295) into motor nerve terminals of mouse neuromuscular junctions

To further demonstrate functional activity of the GFP-toxin fusion proteins, we tested for specific delivery of the proteins to in vivo target sites, namely mouse motor nerve terminals. As shown in Fig. 5 using EDL nerve-muscle preparations perfused ex vivo with GFP-BoNT/A-HC(544–1295) and GFP-BoNT/A-BD(875–1295), both delivery vehicles localized to the synaptic motor nerve terminals similar to full-length BoNT/A labeled with Alexa-647. This localization was distinct from the α-BnTx-labeled postsynaptic acetylcholine receptors, and was specific for the neuronal cells since GFP label was found exclusively in the nerve cells and not muscle cells. To demonstrate the specificity of targeting to neuronal cells in vivo, the delivery vehicles were also injected into mice at the entry of the peroneal nerve into the EDL muscle and 3 days post-exposure, EDL nerve-muscle tissue was removed, and the preparations stained and visualized by fluorescence microscopy. The results shown in Fig. 6 confirmed the localization of the GFP-containing delivery vehicles at the nerve end plates adjacent to, but distinct from α-BnTx (Fig. 6A–C and E–G) and the underlying skeletal muscle tissue, which was labeled with antibodies against troponin I (Fig. 6D and H).

Fig. 5.

Forty millimolar KCl-stimulated uptake of GFP-BoNT/A-HC(544–1295), GFP-BoNT/A-BD(875–1295) and BoNT/A into mouse motor nerve endings. Left panels: fluorescence micrographs illustrating representative targeting of Triangularis sterni endplate acetylcholine receptors with α-BnTx labeled with Alexa-647 (A and B) or Alexa-488 (C). Middle panels: fluorescence micrographs illustrating representative targeting of corresponding motor nerve endings with GFP-BoNT/A-HC(544–1295) (A), GFP-BoNT/A-BD(875–1295) (B) or BoNT/A labeled with Alexa-647 (C). Right panels: overlap of fluorescence micrographs from left and middle panels.

Fig. 6.

In vivo uptake of GFP-BoNT/A-HC(544–1295) and GFP-BoNT/A-BD(875–1295) into mouse motor nerve endings. Shown are fluorescence micrographs illustrating representative targeting of mouse motor nerve endings 3 days after injection into the peroneal nerve's entry into the EDL muscle with GFP-BoNT/A-BD(875–1295) or GFP-BoNT/A-HC(544–1295). (A, E) Targeting of Triangularis sterni endplate acetylcholine receptors with Alexa-647-labeled α-BnTx. (B, F) Targeting of corresponding motor nerve endings with GFP-BoNT/A-BD(875–1295) (B) or GFP-BoNT/A-HC(544–1295) (F). (C) Overlay of A and B. (G) Overlay of E and F. (D, H) Corresponding EDL nerve-muscle preparations stained for skeletal muscle using anti-troponin I antibodies.

Mouse toe-spread reflex toxicity assay

To determine the toxicity of the GFP-toxin fusion proteins, we compared GFP-BoNT/A-HC(544–1295) and GFP-BoNT/A-BD(875–1295) with full-length BoNT/A in the mouse toe-spread assay (Thyagarajan et al., 2009). In vivo injection of 10million times more GFP-toxin fusion protein than the amount of full-length toxin known to block the toe-spread reflex, showed no detectable inhibition of neuromuscular activity (Supplementary data, Fig. S4).

Discussion

The BD of BoNT/A (residues 872–1295), also known as the C-terminal fragment (H_C) of BoNT-HC (residues 544–1295), and the two subdomains of H_C, H_C-N (872–1094) and H_C-C (1095–1295), have been prepared as recombinant proteins in E.coli by a number of laboratories (Rummel et al., 2004; Tavallaie et al., 2004; Baldwin et al., 2005, 2008; Sharma et al., 2006). Although the H_C-C subdomain is responsible for protein receptor-binding (Rummel et al., 2004), the presence of both H_C-N and H_C-C in an intact H_C is required for protective immunity (Tavallaie et al., 2004), suggesting that the entire H_C domain (BD) is required for entry into cells. We have successfully prepared in good yield a recombinant fusion protein in E.coli that contains the entire BD domain with GFP as a cargo. This soluble GFP-BD fusion protein is functionally active and can readily bind and enter cultured neuronal cells as well as motor nerve endings ex vivo and in vivo.

The N-terminus of the heavy chain (H_N), including the TD (residues 544–871) and the belt region (residues 453–543), is believed to be important for delivering the catalytic domain of BoNT through the membrane of the endosome into the cytosol, thereby allowing access to its target substrate (Schiavo et al., 2000; Montal, 2009). Therefore, inclusion of the TD in a delivery vehicle design should enhance the capability of the vehicle to deliver cargo to the cytosol over that of the BD alone. However, achieving stable and soluble recombinant HC proteins consisting of both TD and BD has been challenging. For example, the HC of BoNT/B (residues 624–1291) was expressed in E.coli as inclusion bodies that required solubilization using 6M guanidinium chloride (Zhou and Singh, 2004). Fusion proteins of BoNT/A-HC with cargo peptides or GFP have also been expressed as soluble proteins in a baculoviral system (Band et al., 2010), but this could pose limitations for scale-up.

We have developed an expression vector system that enabled production of soluble GFP-BoNT/A-HC fusion proteins from E.coli, which could be subsequently purified by convenient Ni²⁺-chelation, anion and cation exchange chromatography. Up to 5 mg of purified protein could be obtained from 4 to 5 l of LB broth. This method could readily be scaled up for higher production. Compared with the BD domain alone, the yield for the HC fusion protein is about 10 times less, consistent with previous reports of difficulty due to folding and degradation problems. It was possible to insert up to 20 amino acids between the cargo and HC domains, as in the DE4 and DE8 constructs, with similar yields. However, in our hands, the yield was less with introduction of 30 or 40 amino acids between the cargo and HC, as in DE12 or DE16 constructs, respectively (Supplementary data, Fig. S2). With addition of the TD, our current HC fusion proteins with GFP as cargo were soluble and also functional in binding and entering neuronal cells and nerve terminals in neuromuscular junction preparations ex vivo and in vivo.

In summary, we have now incorporated both the binding and translocation functional features of the native toxin into our working delivery vehicle. We have demonstrated that our GFP-HC fusion protein was able to bind and enter neuronal cells in vitro and in vivo. Our current design will allow us to optimize the translocation and release mechanism for delivering cargo into the nerve cell cytosol by testing various candidate sequences in the linker region between the cargo and HC. Using full-length toxin as a vehicle for delivering cargo poses biosafety concerns for medical applications in terms of residual toxic activity even with removal of active site residues. On the basis of our observations, our fusion proteins with BD or with both BD and TD did not exhibit any neuromuscular inhibitory activity even at concentrations 10-million-fold higher than native toxin. Thus, our experimental approach promises development of a safer inhibitor-delivery platform for blocking BoNT/A activity within target cells.

Supplementary data

Supplementary data are available at PEDS online.

Funding

This research was supported through funds from the Great Lakes RCE NIH/NIAID award U54-AI057153 (to B.A.W.), NIH/NIAID award U01-AI075502 (to B.A.W.), and from the Defense Threat Reduction Agency award HDTRA1-07-C-0031-CBT-BAA (to J.J.M.) and Kirby Foundation Funds (to J.J.M.).