Recombinant forms of tetanus toxin engineered for examining and exploiting neuronal trafficking pathways.

Tetanus toxin is a fascinating, multifunctional protein that binds to peripheral neurons, undergoes retrograde transport and trans-synaptic transfer to central inhibitory neurons where it blocks transmitter release, thereby, causing spastic paralysis. As a pre-requisite for exploiting its unique trafficking properties, a novel recombinant single chain was expressed at a high level in Escherichia coli as a soluble, easily purifiable protein. It could be activated with enterokinase to produce a dichain that matched native toxin in terms of proteolytic and neuroinhibitory activities, as well as induction of spastic paralysis in mice. Importantly, nicking was not essential for protease activity. Substitution of Glu(234) by Ala created a protease-deficient atoxic form, which blocked the neuroparalytic action of tetanus toxin in vitro, with equal potency to its heavy chain; but, the mutant proved >30-fold more potent in preventing tetanus in mice. This observation unveils differences between the intoxication processes resulting from retrograde transport of toxin in vivo and its local uptake into peripheral or central nerves in vitro, dispelling a popularly held belief that the heavy chain is the sole determinant for efficient trafficking. Thus, this innocuous mutant may be a useful vehicle, superior to the heavy chain, for drug delivery to central neurons.

Many potential treatments for disorders of the central nervous system are hindered by the impermeability of the bloodbrain barrier to large therapeutic molecules. An alternative entry route entails uptake into peripheral nerve endings followed by retrograde axonal transport to their cell body and trans-synaptic transfer to central neurons. Such fascinating trafficking routes are exploited efficiently by growth factors (1), neurotrophic viruses (2), and tetanus toxin (TeTx 1 (1,3,4)).
Thus, insights could be gained into the fundamental processes of protein and membrane trafficking by investigating the cellular and molecular mechanisms responsible for such movements, for example, of TeTx; also, novel means of preventing tetanus should emerge.
TeTx, a protein from Clostridium tetani composed of a heavy chain (HC) and light chain (LC) linked by a disulfide and non-covalent bonds, causes spastic paralysis by targeted delivery to the spinal cord and lower brain where it blocks transmitter release from inhibitory nerve endings (5). The HC is required for binding to ecto-acceptors on peripheral and central neurons (6,7, and reviewed in Ref. 8), but functional antagonism of whole toxin by recombinant HC (9) has not been demonstrated in vivo. The LC functions independently as a Zn 2ϩdependent protease and cleaves specifically synaptobrevin, a synaptic vesicle protein essential for Ca 2ϩ -elicited neurotransmitter release (10,11). Clearly, this toxin and its chains or derivatives can provide invaluable research tools for elucidating poorly understood intra-and inter-neuronal trafficking pathways/mechanisms. Moreover, TeTx has unique potential as a central nervous system delivery vehicle because of its proven ability to co-transport attached "cargo" such as complexed anti-toxin (12) or conjugated enzymes (13). Because TeTx is one of the most poisonous substances known, ablation of this toxicity is an essential pre-requisite for such exploitation; this can be achieved by mutagenesis of the protease active site in LC but requires subsequent reconstitution with HC, which gives a low yield of dichain (9,14). Due to its ease of preparation (15,16), the C-terminal moiety of the HC, termed H C , has been employed to traffic enzymes into the central nervous system in vivo (15,16) but the transport efficiency was very poor (17,18). Possible reasons for this low efficacy of H C include: (i) reduced binding affinity for central neuronal membranes in vitro compared with HC or TeTx (19,20); (ii) its preferential or predominant interaction with non-productive acceptors on peripheral neurons that do not mediate optimal retrograde traffic; (iii) the absence of the contributions of the N-terminal half of the HC, as well as the inter-chain disulfide, to the internalization of TeTx that results in toxicity (21). In this context, it is notable that motor nerve terminals appear to possess two distinct pools of TeTx acceptor. In addition to the aforementioned productive acceptor that underlies retrograde trafficking leading to the spastic paralysis typical of tetanus, low affinity binding sites exist that apparently release the neurotoxin proximal to its synaptic site of internalization; hence, at high doses TeTx induces flaccid paralysis at neuromuscular junctions in vitro (22) and in vivo (23). These various findings highlight that TeTx, devoid of toxicity, will provide the most effective vehicle for targeted delivery to the central nervous system, being functional at very low concentrations by targeting the high affinity acceptor, which enters the retrograde trafficking pathway and culminates in translocation of LC into the neuronal cytosol. Hence, a new strategy to achieve this goal is described herein.
To circumvent the difficulties of preparing the individual chains of TeTx and achieving adequate reconstitution, TeTx was expressed as a single chain (SC) in Escherichia coli to yield a high level of correctly folded and soluble toxin, thereby, mimicking Clostridium tetani. Attachment of a tag afforded complete purification while incorporation of an enterokinase (EK)-susceptible linker allowed controlled nicking to create the fully active, disulfide-linked dichain. A similarly created protease-inactive, non-toxic variant (TeTx E234A) protected mice from a TeTx challenge, unlike HC; on the other hand, both antagonized the toxin's neuroparalytic action in vitro. Hence, these strategies have provided a unique non-toxic TeTx for future use as a vaccine (24) and an efficient transporter to the central nervous system, as well as for characterizing the functional acceptors and poorly understood but functionally important retrograde trafficking pathways. Moreover, the fully active wild-type will facilitate mutagenesis to delineate the individual roles of each of its multiple domains.

MATERIALS AND METHODS
Restriction endonucleases, DNA purification kits, and E. coli were from Promega. Talon resin and a site-directed mutagenesis kit were bought from CLONTECH and Stratagene, respectively. pTrcHisA vector and enterokinase were purchased from Invitrogen. A monoclonal antibody specific for the poly-histidine tag was obtained from Qiagen. Cell culture media, reagents, and glass beads were purchased from Sigma-Aldrich. Horse serum was supplied by Life Technologies, Inc. and primers by MWG-Biotech.
Preparation of SC TeTx Constructs-Every construct had unique restriction sites inserted for the purpose of validation. The pTrcHisA vector was modified using the QuikChange site-directed mutagenesis kit by the insertion of two restriction sites (SalI (GTC GAC) and HindIII (AAG CTT)) into its multiple cloning site (MCS), ahead of the EK consensus sequence. This was achieved using two primers (a; 5Ј-G ACT GGT GGA CAG CAA GTC GAC CGG AAG CTT TAC GAC GAT GAC G-3Ј and b, 5Ј-C GTC ATC GTC GTA AAG CTT CCG GTC GAC TTG CTG TCC ACC AGT C-3Ј). Because this insertion introduced a second HindIII site into the MCS, the original HindIII site was replaced, by SmaI, using the same method except with another two primers (a, 5Ј-TAC CAT ATG GGA ATT CCC GGG TTG GCT GTT TTG GCG-3Ј; b, 5Ј-CGC CAA AAC AGC CAA CCC GGG AAT TCC CAT ATG GTA-3Ј). The modified pTrcHisA was confirmed by agarose electrophoresis after digestion with SalI, SmaI, or HindIII enzymes.
The LC DNA of either wild-type (WT) or E234A mutant was isolated from the pMAL-LC and pMAL-E234A plasmids (14), respectively, by SalI and HindIII digestion and subcloned, separately, into the modified pTrcHisA pre-cut by the same pair of enzymes; this created pTrcHisA-LC and pTrcHisA-LC (E234A) plasmids. The stop codon at the end of each LC DNA was removed using the above method with the following primers: a, 5Ј-AAT AGA ACT GCA GGA GAA AAG CTT TAC GAC GAT GAC-3Ј; b, 5Ј-GTC ATC GTC GTA AAG CTT TTC TCC TGC AGT TCT ATT-3Ј.
The HC gene was removed from the pMAL-HC plasmid (9) by BamHI digestion and subcloned, in the correct reading frame, into the BamHI pre-cut pTrcHisA-LC and pTrcHisA-E234A plasmids, producing pTrcHisA-SC TeTx WT and pTrcHisA-SC TeTx E234A constructs. These two SC TeTx plasmids were separately transformed into competent E. coli cells (JM109 strain) by a heat-shock method (25). The positive colonies were screened on Luria-Bertani agar plates containing 100 g/ml ampicillin, and the isolated constructs were analyzed by DNA digestion, electrophoresis, and DNA sequencing, as detailed previously (14).
Expression and Purification of SC TeTx and HC-All the work with the SC TeTx WT was carried out under containment level 3, according to an approved strict safety protocol. E. coli transformed with pTrcHisA-SC TeTx WT or E234A were used to inoculate 10 ml of Luria-Bertani (LB) medium, and grown overnight with agitation (200 rpm) at 37°C. An aliquot (2 ml) of each culture was transferred to 200 ml of fresh LB medium with the addition of ampicillin (100 g/ml); the cells were grown further at 37°C with agitation to reach A 600 nm ϭ 0.5 (about 2.5 h). Isopropyl-␤-D-thiogalactoside (IPTG) was then added to a final concentration of 1 mM, and the induction was allowed to proceed for 16 h at 37°C before harvesting by centrifugation at 6000 ϫ g for 30 min at 4°C. The cell pellets were resuspended in 30 ml of lysis buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl) and lysed by addition of 5 ml of glass beads, followed by vigorous shaking for 10 min at 22°C. The broken cells were centrifuged at 9000 ϫ g for 30 min at 4°C, and the supernatant was collected. These SC TeTx WT and E234A lysates were separately mixed with 2 ml of Talon resin, previously equilibrated with the lysis buffer, and gently agitated at 22°C for 20 min. After the two mixtures had been loaded into mini-columns and washed with 100 ml of lysis buffer, each recombinant toxin was eluted with 8 ml of lysis buffer containing 50 mM imidazole, and 0.5-ml fractions were collected. The toxins were analyzed by SDS-PAGE and Western blotting, using antibodies specific for His tag, LC, or HC, as detailed elsewhere (9,14). The two SC toxins were nicked by incubation with EK (1 unit/30 g of protein) at 22°C for 1 h, and the reaction was stopped by the addition of 500 nM bovine pancreatic trypsin inhibitor. Recombinant HC was prepared as detailed in a previous study (9) and dialyzed into HEPESbuffered saline, pH 7.5.
Determination of the Proteolysis of Synthetic HV62 Polypeptide by Native TeTx and SC Preparations-An appropriate concentration of each sample was incubated at 37°C with 15 M HV62, a synthetic peptide corresponding to residues 33-94 of human synaptobrevin/ VAMP-2 (HV62) in 60 l of buffer (50 mM HEPES-NaOH (pH 7.4), 2 mM dithiothreitol (DTT), 0.2 mg.ml Ϫ1 bovine serum albumin (BSA), and 50 M ZnCl 2 ). Bovine pancreatic trypsin inhibitor (500 nM) was included in all incubations to prevent EK cleavage of HV62. The reactions were stopped after 30 min by addition of 60 l of 5 mM EDTA and 1% (v/v) trifluoroacetic acid (pH 2). Reverse-phase high pressure liquid chromatography (RP-HPLC) on a Micropax C 18 column equilibrated in 0.05% trifluoroacetic acid, and elution with a 0 -60% acetonitrile gradient separated the products; the peak height of one (residues 77-94) was used to calculate the percentage of HV62 hydrolyzed (for details see Ref. 26).
Assessment of the Neuromuscular Paralytic Activity and Lethality of Native TeTx and Recombinant Variants-Mouse phrenic-nerve hemidiaphragm preparations were dissected, bathed in Krebs-Ringer (KR) buffer (in mM; NaCl 118, KCl 4.7, MgSO 4 1.2, NaHCO 3 23.8, KH 2 PO 4 1.2, CaCl 2 2.5, and glucose 11.7, pH 7.4) containing 0.1% (w/v) BSA and gassed with 5% CO 2 /95% O 2 . After equilibration, TeTx, a recombinant sample (or buffer as a control) was added and toxin-induced neuromuscular paralysis was determined as the time taken for nerve-evoked muscle contraction to decrease to 10% of original value, as described previously (9,27). For competition studies, conditions designed to minimize toxin uptake into the nerve endings were used. The hemi-diaphragms were preincubated for 60 min at 4°C, in the absence and presence of a putative antagonist, in modified Krebs-Ringer buffer (MKR; equivalent to KR except for the addition of 3.8 mM MgCl 2 and the reduction of CaCl 2 to 0.5 mM) supplemented with 0.1% BSA and saturated with 95% O 2 /5% CO 2 . Native toxin (1 nM) was then added to the bath, and the incubation was continued for a further 30 min. The tissues were washed three times with MKR and then with KR before raising the temperature to 37°C, stimulating the nerve, and recording the evoked muscle twitch (9).
For quantitation of the lethality of each recombinant toxin relative to native TeTx, the samples (200 l) were diluted into 0.9% (w/v) sterile saline containing 0.25% (w/v) BSA and injected subcutaneously into the two flanks of each mouse (4 -5 weeks old, ϳ20 g of body weight, four mice/group), and the LD 50 values were determined, as described previously (28). In some experiments, native TeTx was mixed with mutated TeTx SC or nicked, or HC before co-injection into mice; these animals were continuously monitored for the appearance of tetanus symptoms over the following 96 h.
Identification and Site-directed Mutagenesis of a Protease-susceptible Bond in SC TeTx-In view of the appreciable toxicity of the un-nicked toxin in mice, its susceptibility to nicking was evaluated. SC TeTx (50 g) was treated at 37°C for 0.5 h with trypsin (final concentration, 1 g/ml), before the addition of SDS sample buffer and boiling for 5 min; 2 g of the same toxin (treated similarly except not exposed to protease) was used as a control. All the samples were subjected to SDS-PAGE, in the presence of reducing agent, electrotransferred onto an Immobilon-P membrane, and stained with Ponceau red to reveal the HC, which was cut out for N-terminal sequencing. Because this showed that the peptide bond between Arg 496 and Ser 497 was cleaved (see "Results"), this Arg was replaced with Gly using the QuikChange site-directed mutagenesis kit with the following two primers: a, 5Ј-AAG GAT CGA TGG GGA TCC TCT GGA TCA TTA ACA GAT TTA GGA GGA-3Ј; b, 5Ј-TCC TCC TAA ATC TGT TAA TGA TCC AGA GGA TCC CCA TCG ATC CTT-3Ј. The construct was analyzed, and the protein (R496G) was expressed and purified as detailed above for WT; its susceptibility to various trypsin concentrations was monitored by SDS-PAGE and protein staining of the products.
Neurons cultured for at least 10 days were washed four times with O 2 -gassed Krebs-Ringer HEPES (KRH) (in mM: 20 HEPES-NaOH (pH 7.4), NaCl 128, KCl 5, NaH 2 PO 4 1, CaCl 2 1.4, MgSO 4 1.2, D-glucose 10, and BSA 0.05 mg/ml (pH 7.4)) and 0.5 ml of the latter buffer containing 0.25 Ci/ml [ 14 C]glutamine (i.e. the glutamate precursor (30)) was added. All steps were performed at 37°C. After a 45-min labeling period, the neurons were washed four times as before and incubated for 5 min at 37°C in KRH containing either 1.4 mM Ca 2ϩ or 0.5 mM EGTA (i.e. to assess Ca 2ϩ -independent release); aliquots were retained for measurement of [ 14 C]glutamate content by ion exchange HPLC analysis and scintillation counting (30). A modified KRH buffer containing 50 mM KCl (with a reduced NaCl content of 83 mM to maintain osmolarity) and either 1.4 mM Ca 2ϩ or 0.5 mM EGTA was added for a 5-min stimulation period. The amounts of [ 14 C]glutamate in basal and stimulated samples were measured as above and expressed as a percentage of the total cell content. The quantities of KCl-induced [ 14 C]glutamate released into EGTA-containing buffer were subtracted from the values recorded from Ca 2ϩ -containing samples to calculate the Ca 2ϩ -dependent component of evoked release.

RESULTS
Generation and Characterization of pTrcHisA-SC TeTx WT and E234A Constructs-Two restriction enzyme sites (SalI and HindIII) were created inside the MCS and in front of the EK cleavage site in pTrcHisA vector (the original HindIII was replaced by SmaI). DNA encoding LC WT or E234A mutant (14) was inserted into pTrcHisA, and the stop codon at the end of each was removed before subcloning HC DNA into the precut plasmids to yield pTrcHisA-SC TeTx WT or E234A constructs ( Fig. 1), as detailed under "Materials and Methods." The two constructs gave the expected DNA fragments on agarose gel electrophoresis when the requisite restriction enzymes were used (cf., Fig. 1); for example, digestion of either with a combination of SalI and HindIII released LC DNA, whereas BamHI gave HC DNA (data not shown). This subcloning of the SC TeTx gene resulted in 66 extra nucleotides at the 5Ј-end; also, a 51-nucleotide linker encoding an EK substrate consensus motif was present (Fig. 1). The DNA sequences obtained for the two constructs confirmed that they were in the correct reading frame. As illustrated in Fig. 1, the predicted SC protein WT or E234A has 17 extra residues in the linker; nicking by EK (between Lys 489 and Asp 490 ) should create the dichain composed of a LC, with 22 and 10 extra amino acids at its N and C termini, and HC having a 7-residue N-terminal extension.
Expression and Purification of SC TeTx WT and E234A-After transformation of E. coli with either of the above-noted constructs, expression of the proteins was induced by IPTG; in each case, a SC protein was obtained in the bacterial lysate but only after induction, as demonstrated by Western blotting with an anti-His tag antibody ( Fig. 2A). Moreover, the same single band was also recognized by polyclonal antibodies specific for TeTx LC or HC ( Fig. 2A). This evidence together with the expected size (ϳ150 kDa) on SDS-PAGE under reducing conditions confirmed the identity of SC TeTx; an absence of smaller components excluded any nicking, degradation, or premature translation. SC TeTx was purified by nickel-affinity chromatography, as judged from the protein pattern on a Coo- massie Blue-stained SDS-PAGE gel run under non-reducing conditions (Fig. 2B). A doublet was seen for SC TeTx E234A (Fig. 2B) and WT (not shown), migrating slightly behind native TeTx as expected due to the extra residues; only a single band (molecular mass ϭ 150 kDa) was observed under reducing conditions, indicative of the doublet seen in the absence of DTT being attributable to the disulfide in the HC being present in only a fraction of both the native and recombinant molecules (see below). In contrast to the documented SC toxin, native TeTx converted to HC and LC after reduction (Fig. 2B).
Having shown that the recombinant TeTx existed exclusively as an SC protein both in the E. coli lysate and in the purified state, its complete conversion to the dichain was achieved in vitro using EK to cleave the engineered site in the inter-chain linker (Fig. 1). After SC TeTx was incubated with EK, it still migrated in non-reducing SDS-PAGE as the 150-kDa doublet but was totally converted to HC and LC after reduction (Fig. 2,  B and C). This established that the inter-chain disulfide had formed in all of the recombinant toxin. After purification, ϳ4.5 mg of either SC TeTx WT or E234A was generally obtained from 500 ml of culture.
SC TeTx WT Gives a Near-maximal Rate of Cleavage of a Synaptobrevin Peptide: Nicking Enhances Its Protease Activity to Match That of Native TeTx-An established RP-HPLC method, based on the proteolysis of HV62, a 62-mer polypeptide corresponding to residues 33-94 of human synaptobrevin-2, was used to measure the protease activities of the recombinant toxin preparations and investigate the influence of nicking. The initial rates of cleavage (Table I) were determined by quanti-fying one of the separated products (residues 77-94). Reduced SC TeTx gave a value of 8 nmol min Ϫ1 mg Ϫ1 , which represents 39% of that obtained for native dichain TeTx; this demonstrates, for the first time, that nicking is not an absolute prerequisite for the toxin's catalytic activity. Recombinant nicked WT toxin displayed comparable proteolytic activity to its native counterpart (Table I), establishing that its LC moiety is folded correctly when expressed as an SC in E. coli. Thus, the additional linker and N-and C-terminal residues present in the recombinant TeTx are not detrimental to its proteolytic activity.
Nicked Recombinant TeTx Displays the Same Neuromuscular Paralytic Activity and Mouse Lethality as the Native Toxin: SC TeTx WT Displayed Somewhat Lower Potencies-To assess the abilities of un-nicked and nicked TeTx to bind motor nerve terminals, undergo local internalization/translocation, and block acetylcholine release, they were tested in vitro on the mouse nerve-diaphragm. The nicked WT and TeTx caused neuromuscular paralysis in the same time (Table I), demonstrating that the bacterially expressed TeTx is equipotent to its native counterpart in this multistep intoxication. Likewise, subcutaneous injection into mice of equal quantities of either recombinant nicked or native TeTx caused an indistinguishable pattern of spastic paralysis, due to the blockade of transmitter release at inhibitory synapses in the spinal cord, and showed the same specific neurotoxicity, ϳ10 8 LD 50 /mg (Table I). Notably, the un-nicked TeTx showed lower lethality and neuroparalytic potency than the recombinant un-nicked or native toxin, the longer time observed for neuromuscular paralysis (Table I) being equivalent to 2-to 2.5-fold less activity (i.e. un-nicked toxin must be present at 2-2.5 ϫ the concentration of nicked toxin to induce paralysis in an equivalent time; data not shown). Although activity of SC cannot be ruled out, partial nicking of SC TeTx was observed after its application to neurons in culture (see below). Therefore, these findings are indicative of nicking being required for toxicity and that this occurs to a large extent in the diaphragm and the whole animal. Thus, despite heterologous expression in E. coli and modifications to its primary sequence, the recombinant TeTx retains maximal biological activities.

Limited Trypsinolysis of SC TeTx WT Reveals Multiple Nicking Sites between LC and HC: Mutating R496 to G Ablates One Scissile Bond but Does Not Reduce Its Biological Activities-
With the aim of engineering a toxin that would be more resistant to nicking in vivo, and yet whose neurotoxicity could be controlled by EK treatment in vitro, the peptide bonds of the SC susceptible to trypsin were first examined. A fixed time of incubation with various trypsin concentrations, followed by SDS-PAGE and protein staining, revealed that several proteolytic sites are present in the LC-HC junction (see Fig. 3A, also reported in Ref. 31). Higher trypsin concentrations yielded single forms of both LC and HC (Fig. 3A), and Edman analysis of the resultant HC gave the partial N-terminal amino acid sequence SLTXL, which only corresponds to residues 497-501 of recombinant TeTx WT (Fig. 3B, denoted by asterisks). Therefore, the Arg 496 -Ser 497 bond is cleaved by trypsin (Fig. 3B). When Arg 496 was mutated to Gly, expressed, and purified (Fig.  3C), again a single protein (SC TeTx R496G) of 150-kDa molecular mass was obtained. Notably, its trypsin fragmentation pattern was altered giving one fewer HC and LC precursor than WT at the lower enzyme concentrations, but higher amounts yielded the fully nicked chains; thus, one of three cleavage sites observed with the WT was removed by this mutation. The un-nicked R496G mutant showed no decrease in toxicity (relative to un-nicked WT) or the extent of its activation by trypsin, as determined by the neuroparalysis and After purification of SC TeTx on Talon resin and incubation with and without EK (see "Materials and Methods" and for details), SDS-PAGE was carried out, followed by protein staining of the gel with Coomassie Blue (B) or Western blotting with an antibody recognizing TeTx HC (C). A sample of native TeTx was also run as a control. mouse lethality assays (Table I); also, it was as active as WT in blocking neuroexocytosis from cerebellar neurons (Fig. 4). Surprisingly, SC R496G gave a significantly higher rate of proteolysis of HV62 than WT (Table I) and a larger enhancement upon nicking than that seen with the non-mutated toxin; the basis of this difference remains unclear.
SC-or EK-nicked Protease-deficient TeTx E234A Mutant Is Atoxic but Protects Mice against a TeTx Challenge-As expected from studies with LC (14), purified SC TeTx, in which the catalytic glutamate at position 234 was replaced by an alanine, failed to show any detectable proteolysis of HV62 substrate, before or after nicking with EK (Table I). Accord-ingly, SC-and EK-nicked E234A proved to be devoid of toxicity in mice (Table I) and unable to inhibit transmitter release at the neuromuscular junction (Table I) or from cerebellar neurons (Fig. 4). In view of the intended use of E234A as a central nervous system-targeted drug vehicle, we measured the abilities of SC or nicked E234A to antagonize lethal challenges of native TeTx (either 3 or 10 LD 50 ) in mice. Importantly, SC-E234A completely prevented tetanus poisoning caused by 3 LD 50 units of native toxin when the largest dose was used, and 300-to 900-fold molar excess significantly delayed the onset of symptoms (Table II). The observed concentration dependence of the ability of E234A to delay/prevent onset of tetanus would be expected for an antagonist competing with TeTx for binding FIG. 3. Characterization of a trypsin-sensitive bond in SC TeTx: mutation of Arg 496 to Gly alters the nicking pattern. SC TeTx (6 g at 0.2 mg/ml) was incubated for 30 min at 37°C (in 10 mM HEPES-NaOH, pH 7.4, and 145 mM NaCl) with various trypsin concentrations; control samples, lacking the protease, were treated identically. After boiling for 5 min and addition of DTT (50 mM final concentration), the reaction mixtures were subjected to SDS-PAGE and Coomassie Blue staining (A). B, following exposure of SC TeTx WT to 1 g ml Ϫ1 trypsin for 30 min at 37°C, SDS-PAGE under reducing conditions was performed followed by transfer onto Immobilon-P membrane and protein staining with Ponceau red. The 100-kDa HC band was excised, and its N terminus was sequenced. After five rounds of Edman analysis, four amino acids were identified (marked with asterisks) giving a sequence SLTXL (where X is an unidentified residue; D occurs at this position in the native toxin), which corresponds to residues 497-501 of TeTx, indicating that the Arg 496 -Ser 497 bond is sensitive to trypsin. After mutagenesis to preclude cleavage at this position, tryptic digestion of the R496G mutant was performed (C), exactly as for TeTx WT in A.
FIG. 4. Nicked R496G and native TeTx are equipotent in blocking transmitter release from rat cerebellar neurons: nicking enhances the inhibitory activity, which is absent from the E234A mutant. Cerebellar neurons, prepared as outlined under "Materials and Methods" and maintained for 10 days in vitro, were washed using KRH and exposed to the specified concentrations of native TeTx (q), EK-nicked TeTx R496G (E), SC TeTx R496G (), or EK-nicked TeTx E234A (ƒ). After 30 min at 37°C, the wells were washed twice as above and incubated for a further 30 min before replacement of the buffer with KRH containing a glutamate precursor, 14 C-labeled glutamine, which was used for the measurement of glutamate release, as detailed under "Materials and Methods." Data presented are means Ϯ S.D. (n ϭ 3 or 6).
FIG. 5. Innocuous, protease-inactive TeTx E234A antagonizes neuromuscular paralysis caused by native toxin as effectively as HC. For assessing competition, mouse hemi-diaphragms were incubated for 1 h at 4°C in MKR containing 0.1% BSA only, or the latter plus 100 nM nicked E234A (closed bar) or 100 nM HC (open bar), before the addition of 1 nM native TeTx. Following 30-min exposure to the latter, the tissues were washed three times with MKR and then twice with KR. The temperature was raised to 37°C, the nerve was stimulated (0.2 Hz, 1.5-2.5 V), and the evoked muscle twitch was recorded, as outlined under "Materials and Methods." The times taken for each nerve-muscle to be reduced to 10% of the original tension were recorded, and the values for the antagonists were plotted relative to that for TeTx alone.
to its neuronal acceptors. EK nicking of E234A prior to injection gave improved antagonism; even with a challenge of 10 LD 50 , the nicked E234 extended the delay before onset of tetanus from 2 to 3.5 days (Table II). In contrast, large (9000-fold molar excess) doses of TeTx HC proved almost ineffective in delaying tetanus poisoning (Table II). Thus, proof of principle is provided for the suitability of E234A for the intended purpose of central nervous system-targeted vehicle in vivo; also, the data highlight the limited ability of HC to antagonize the intoxication process of the intact toxin in the whole animal, which entails retrograde trafficking (see the introduction).
SC or Nicked E234A TeTx Antagonizes the "Local" Inhibitory Action of the Native Toxin on Central and Peripheral Neurons: HC Is Neurotoxic-In view of TeTx E234A proving to be much more effective than the HC in preventing tetanus in vivo, their levels of antagonism of the inhibition of transmitter release in vitro were compared. Surprisingly, both delayed the onset of TeTx-induced neuromuscular paralysis to similar extents (Fig.  5). Likewise, preincubation of cerebellar neurons with a modest excess of E234A antagonized the inhibition of transmitter release caused by subsequent addition of 0.2 nM native toxin at 37°C; nicking caused a minimal increase in the antagonism (Fig. 6A). Unlike SC-or EK-nicked E234A, HC proved to be toxic after only short exposure to the cultured neurons; it induced a dose-dependent increase in the basal level of [ 14 C]glutamate efflux, and a reduction in Ca 2ϩ -dependent K ϩevoked release (Fig. 6B). Although HC reduced evoked release, it also afforded some protection from the additional inhibition by native TeTx (Fig. 6C). Due to its intrinsic toxicity, the ability of HC to antagonize native TeTx was assessed by expressing the difference between the K ϩ -evoked Ca 2ϩ -dependent release remaining following 0.2 nM TeTx treatment in the presence or absence of HC, relative to the evoked release retained following exposure to HC alone (Fig. 6D). This calculation revealed similar antagonism of native TeTx-induced inhibition of neurotransmitter release by E234A and HC.

DISCUSSION
Induction of spastic paralysis by TeTx is mediated by a variety of functionally important, neuronal components and poorly understood protein trafficking processes (4,5); all these can be investigated by employing TeTx or derivatives as unique probes. A pre-requisite for such studies is an efficient means of producing adequate quantities of recombinant fully active TeTx, as well as variants lacking one or more of its above-noted activities. Thus, a new strategy was devised herein for fast and efficient expression of soluble TeTx as a tagged single chain in E. coli, which not only obviates the time-consuming and costly alternative of anaerobically culturing C. tetani but also facilitates a genetic engineering approach to preparing novel TeTx forms.
Recombinant TeTx was produced in high yield (typically ϳ9 mg/liter) and easily isolated as a soluble concentrated protein (ϳ 4 mg/ml) devoid of contaminants, degradation, or premature-translation products, as determined by SDS-PAGE followed by protein staining and Western blotting. The purified toxin was exclusively in the SC form, and its controlled nicking with EK in vitro gave quantitative conversion to dichain with its disulfide link intact. Most importantly, the nicked toxin matched or exceeded the biological properties of native TeTx in terms of specific activities for its proteolytic and neuroparalytic actions, as well as lethality due to spastic paralysis in mice. Thus, this is the first recombinant preparation of TeTx with abilities to undergo all of the above-mentioned steps required to elicit the symptoms of tetanus, with the same potency as the natural toxin. These much needed advances avoid all the reported difficulties of minimal solubility, aggregation, and low yields when expressing or handling individual HC and LC in the absence of fusion tag (9,14,32) and the additional losses during their reconstitution as well as the reduced biological activity of the resultant dichain (9,14,19). The much lower specific neurotoxicity of the latter suggests that suboptimal, independent folding of the separate domains occurs and that a Un-nicked and nicked samples were shown to be completely in the SC and dichain form, respectively, both before and after incubation with substrate, by SDS-PAGE and protein staining.
b Initial rates of proteolysis were measured using the RP-HPLC-based method, detailed previously (26). Incubations with 15 M of a synthetic peptide corresponding to residues 33-94 of human synaptobrevin 2 (HV62) were performed at 37°C in 50 mM HEPES.NaOH, pH 7.5, containing 2 mM DTT, 0.2 mg.ml Ϫ1 BSA, and 50 M ZnCl 2 , using the appropriate concentration of each reduced toxin required to proteolyse 10 -15% of the substrate during a 30-min period. Data are means ϮS.D. (n ϭ 4).
c LD 50 unit is the amount of toxin that killed 50% of the injected mice within 4 days. d Toxin preparations were nicked with EK (1 unit/30 g) at 22°C for 1 h. e This v 0 value represents the detection limit of the RP-HPLC assay; no proteolysis of HV62 was observed using prolonged incubations. the optimal surfaces for interaction between HC and LC of TeTx are only created upon folding of the intact SC, as occurs in C. tetani. Thus, this first high level production of soluble, intact TeTx by recombinant means represents a major advance.
Availability of SC TeTx that could be readily converted to the dichain allowed an investigation of the effects of nicking on each stage of the intoxication process. In terms of the intracellular phases of action, dichain TeTx proved ϳ2-fold more effective than SC in the proteolytic cleavage of synaptobrevin; this finding dismisses the widely held notion that nicking is essential for the enzymic activity (reviewed in Ref. 33). Such a minimal increase in the protease activity upon nicking contrasts markedly with the absolute requirement for reduction of the inter-chain disulfide (Table I ( 11)). The crystal structure of BoNT/A reveals that its active site is occluded by a loop between the LC and the translocation domain of HC (34); elucidation of the three-dimensional structure of TeTx should, likewise, give a basis for the need to reduce the interchain disulfide, as well as for the slight enhancement brought about by nicking. The relative affinity of SC-and EK-nicked TeTx for acceptor binding was tested by using the recombinant protein to antagonize the productive interaction of native TeTx with acceptors. This protocol required the development of a recombinant TeTx isoform that retains the binding properties of the WT while being devoid of toxicity, an achievement accomplished by site-directed mutagenesis, expression, and purification of SC TeTx E234A. The resultant mutant proved to be unable to cleave synaptobrevin, inhibit transmitter release, or cause neuromuscular paralysis and lacked neurotoxicity in mice, consistent with the known absence of these activities from dichain reconstituted from mutated LC E234A and purified native HC (9,14). Clearly, TeTx E234A retains high affinity for binding to the ecto-acceptors on both peripheral and central neurons because of its potent antagonism of TeTxinduced neuromuscular paralysis and inhibition of transmitter release from cerebellar granule cells. Importantly, both SC and dichain TeTx E234A attenuated the induction by native toxin of spastic paralysis in mice. Although nicking only slightly improved this antagonism, the importance of this remains unclear, because nicking has been shown to occur in vivo (35). SC TeTx is clearly susceptible to nicking in vitro, and such cleavage was observed with cultured neurons (data not shown); thus, it seems likely that the reduced paralytic activity recorded for SC TeTx WT result from partial nicking by cellular and/or tissue proteases. In support of the latter, native TeTx is readily nicked following exposure to brain tissue, by enzymes released following tissue damage, and during or after its retrograde axonal transport to the spinal cord (35). Thus, to elucidate the importance of nicking in the internalization and intracellular poisoning steps, it is desirable to create a variant resistant to proteases in vivo. Identification of a scissile bond in SC TeTx and its mutation (Arg 496 3 Gly) removed one cleavage site but, at least, two other trypsin nicking sites remained, and TeTx R496G was as toxic as WT in vitro and in vivo. Moreover, Krieglstein and co-workers (31) revealed numerous peptide bonds in the LC-HC linker of native TeTx susceptible to other proteases. Nevertheless, by the gradual replacement of amino acids in protease-labile bonds with other less sensitive pairings, this protein engineering approach may eventually generate a SC with significantly increased resistance to activation in vivo. Such an achievement in the future would attenuate the safety hazard of preparing TeTx.
Exclusive removal of the protease activity leaves intact the domains responsible for all the internalization and retrograde trafficking, thus, affording potential application of TeTx E234A as a neuronally targeted transporter for the delivery of research or therapeutic agents into the central nervous system. Importantly, the observed striking superiority of E234A over HC in protecting mice from TeTx demonstrates that the latter fragment cannot efficiently target the receptor responsible for the retrograde trafficking pathway in vivo. In contrast, the binding component, which leads to local release of TeTx into peripheral or central nerve terminals, is antagonized by HC as effectively as SC or dichain, suggesting that distinct acceptors may mediate the retrograde TeTx trafficking. A feasible explanation for these findings is that TeTx binds the locally trafficking acceptor via its HC, whereas the acceptor for the retrograde pathway additionally requires LC domains for productive binding, a situation found for the related botulinum neurotoxin (36 -38). Alternatively, the poor antagonism by HC of TeTx in vivo may result from its membrane insertion behavior (a characteristic that is removed by the presence of LC (39)), which may limit its targeting efficiency to central neurons, or that LC association may exert an allosteric effect on the three-dimensional structure of the HC that improves binding to the trafficking acceptor. The H c fragment is more soluble than HC, but it is not as good an antagonist of acceptor binding by TeTx (9,14). Accordingly, the targeting efficiency for delivery of H c and H c -conjugates into the central nervous system has been found to be rather poor (17,18,40), and much of the internalized hybrid protein is rapidly metabolized (41). An additional advantage of dichain TeTx E234A is that it retains the capability to translocate its LC across the membrane into the neuronal cytosol and possesses the N-terminal moiety of HC that is essential for cytosolic delivery of the LC (at least) via a putative low pH-induced membrane penetration event (42); also, the disulfide linking LC to HC (in both TeTx and BoNT/A) is crucial for internalization (36,37,43). Furthermore, the SC TeTx expression strategy allows ligation of the DNA encoding a therapeutic protein to the 5Ј-end of the SC TeTx E234A transporter gene, prior to high yield bacterial expression; the protein adduct attached to the LC would not only be transported to central neurons, but could "piggy-back" into the cytosol of its target cell. Finally, the transporter can be optimized by downsizing the LC to afford the largest cargo while retaining the stability, longevity, neuronal targeting, and delivery properties of TeTx in vivo.