Crucial Role of the Disulfide Bridge between Botulinum Neurotoxin Light and Heavy Chains in Protease Translocation across Membranes*

Clostridial botulinum neurotoxins (BoNTs) exert their neuroparalytic action by arresting synaptic exocytosis. Intoxication requires the disulfide-linked, di-chain protein to undergo conformational changes in response to pH and redox gradients across the endosomal membrane with consequent formation of a protein-conducting channel by the heavy chain (HC) that translocates the light chain (LC) protease into the cytosol. Here, we investigate the role of the disulfide bridge in the dynamics of protein translocation. We utilize a single channel/single molecule assay to characterize in real time the BoNT channel and chaperone activities in Neuro 2A cells under conditions that emulate those prevalent across endosomes. We show that the disulfide bridge must remain intact throughout LC translocation; premature reduction of the disulfide bridge after channel formation arrests translocation. The disulfide bridge must be on the trans compartment to achieve productive translocation of LC; disulfide disruption on the cis compartment or within the bilayer during translocation aborts it. We demonstrate that a peptide linkage between LC and HC in place of a disulfide bridge is insufficient for productive LC translocation. The disulfide linkage, therefore, dictates the outcome of translocation: productive passage of cargo or abortive channel occlusion by cargo. Based on these and previous findings we suggest a sequence of events for BoNT LC translocation to be HC insertion, coupled LC unfolding, and protein conduction through the HC channel in an N to C terminus orientation and ultimate release of the LC from the HC by reduction of the disulfide bridge concomitant with LC refolding in the cytosol.

emulate those prevalent at the endosome. We utilized a previously developed assay in Neuro 2A cells to monitor interactions between the HC and the LC during translocation with single molecule sensitivity (28). This assay led to the identification of intermediate channel conductances that reflect permissive stages during LC translocation for both BoNT/A and BoNT/E (28). Further, we showed that productive translocation requires proteolytic cleavage of LC cargo from the HC channel (28). In this work we use the assay to examine the consequences of disulfide linkage disruption by chemical reductants accessible to different sides of the membrane. The disulfide linkage emerges as a crucial determinant required for chaperone func-tion and LC translocation and release. These and previous findings indicate that BoNT translocation involves an acid pH-induced membrane insertion step coupled to LC unfolding and entry into the HC chaperone/channel, LC protein conduction through the HC channel in an N-to C-terminal orientation, and subsequent release of the LC cargo from chaperone by reduction of the disulfide bridge concomitant with LC refolding at the cytosol.

EXPERIMENTAL PROCEDURES
Unless otherwise specified, all chemicals were purchased from Sigma-Aldrich. Purified native BoNT/A holotoxin and holotoxin channels with addition of ␤ME after the onset of channel activity. Channel activity begins 5 min after G⍀ seal formation. Thirty seconds later, 1 mM ␤ME is added to the trans compartment. Multiple channel insertions occur with time in panels A and B. C, time course of channel ␥ change illustrated in panel B (average N/data point ϭ 340 events). Addition of ␤ME designated by the arrow. Thin red line represents results for holotoxin without ␤ME addition (average N/data point ϭ 686). ␥ values associated with raw data from panel B are indicated.
HC and BoNT/E holotoxin were from Metabiologics. The ING2 BoNT/A LC-specific monoclonal antibody was kindly provided by Dr. James Marks (University of California, San Francisco).
Cell Culture-Neuro 2A neuroblastoma cells were obtained from the American Type Culture Collection. Cells were passaged in Dulbecco's modified Eagle's medium (BioWhittaker) supplemented with penicillin 10 mM/streptomycin 10 g/ml/ glutamine 2 mM (Invitrogen), and 5% newborn bovine serum (Invitrogen). Cells were plated onto Matrigel (BD Biosciences)coated glass coverslips at a density of ϳ500 cells/coverslip and cultured at 37°C, 5% CO 2 for 1-3 days prior to patch clamp recordings.
Patch Clamp Recordings-Patch pipettes were pulled from borosilicate glass (Hilgenberg, Lambrecht, Germany), fire-polished, and used at 3.5-7.0 M⍀ resistance when immersed in recording solution. Excised patches in the inside-out configuration were used (29). After gigaohm (G⍀) seal formation, the patch was excised from the cell, removed, and reimmersed through the air-water interface to achieve an inside-out configuration. Current recordings were obtained under voltage clamp conditions by the application of consecutive voltage steps 800 ms in duration from ϩ50 mV to Ϫ150 mV at a sampling frequency 20 kHz. Records were acquired and analyzed using the patch clamp amplifier system (List EPC-9; HEKA Electronik) fitted with an ITC-16 interface (Instrutech, Port Washington, NY) and the Pulse/PulseFit acquisition and analysis software (HEKA). Data were further analyzed using Clampfit v.9.2 software (Molecular Devices, Sunnyvale, CA), Microsoft Excel, and IGOR Pro (Wavemetrics, Portland, OR). All experiments were conducted at 22 Ϯ 2°C.
Solutions-To emulate endosomal conditions the trans compartment (bath) solution contained (in mM) NaCl 200, NaMOPS [3-(N-morpholino) propanesulfonic acid] 5, (pH 7.0 with HCl), tris-(2-carboxyethyl) phosphine (TCEP) 0.25, ZnCl 2 1, and the cis compartment (pipette) solution contained (in mM) NaCl 200, NaMES [2-(N-morpholino) ethanesulfonic acid] 5, (pH 5.3 with HCl). The osmolarity of both solutions was determined to be 390 mosM. ZnCl 2 was used to block endogenous channel activity specific to Neuro 2A cells (30,31). BoNT reconstitution and channel insertion was achieved by supplementing 5 g/ml BoNT holotoxin or HC to the pipette solution, which was set to an endosomal pH of 5.3. BoNT/E experiments were performed with 40 M trypsin in the trans compartment. G⍀ seal formation was optimized as follows: The pipette tip was first dipped in the pipette solution in the absence of BoNT and then back-filled with solution containing BoNT.
Data Analysis-Analysis was performed on single bursts of each experimental record; a single burst is defined as a set of openings and closings lasting Ն50 ms bounded by quiescent periods of Ն50 ms before and after. BoNT channel activity occurs in bursts, and only single bursts were analyzed due to the random duration of quiescent periods between these bursts (32). The single channel conductance (␥) was calculated from Gaussian fits to current amplitude histograms. Fab Generation-The Pierce ImmunoPure Fab Preparation kit (44885) was used to generate Fabs from BoNT/A LC antibody ING2. In short, 3 mg/ml ING2 was incubated in the equiv- alent volume of immobilized papain at 37°C for 5 h. The digested ING2 was transferred to a Protein A column to separate the Fabs from the Fc and undigested ING2: the Fabs flow through the column whereas the Fc and undigested IgG are retained and elute later. BoNT/A LC Fabs were quality-checked using SDS-PAGE prior to the single molecule assays. Purified Fabs were concentrated to 1 mg/ml in patching bath solutions prior to incubation with BoNT/A holotoxin.

RESULTS AND DISCUSSION
Single Molecule Translocation Assay-Translocation of BoNT/A LC by the BoNT/A HC channel can be monitored in real time and at the single molecule level in excised membrane patches from Neuro 2A cells (28). Translocation requires pH 5.3 on the cis compartment, defined as the compartment containing BoNT/A, and pH 7.0 on the trans compartment, which is supplemented with the membrane-nonpermeable reductant TCEP, conditions that emulate those prevalent across endosomes. Translocation is then observed as a time-dependent increase in Na ϩ conductance through the HC channel, as illustrated by the control experiment shown in Fig. 1A. The time course of change of the single channel conductance ␥ after insertion of BoNT/A holotoxin into the membrane displays multiple discrete transient intermediate conductances before achieving a ␥ of 67.1 Ϯ 2.0 pS (Fig. 1C, red trace). The top panel of Fig. 1A shows that at the onset of translocation small, discrete events with a ␥ ϳ 12 pS are clearly discerned, as indicated by the segment of the record designated with a black bar that is dis-  OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40 played at higher time resolution under the trace. Progressively, ␥ increases and, as shown in the middle panel of Fig. 1A, reaches a value of ϳ40 pS; note the insertion of two channels into the membrane, designated O 1 and O 2 , which even during the short segment displayed undergo a continuous increase in conductance, ultimately reaching a stable value of 67 pS (bottom panel), a conductance at which they remain for the duration of the experiment (Fig. 1C, red trace). A ␥ of 67.1 Ϯ 2.0 pS is also the characteristic conductance of isolated HC recorded under identical conditions; therefore it represents the conductance of the unoccluded HC in holotoxin experiments after translocation is complete. We interpret these different conductance events as reporters of discrete intermediate stages during the translocation of the LC across the membrane. During protease translocation, the protein-conducting channel progressively conducts more Na ϩ around the polypeptide chain before entering an exclusively ion-conductive state. This typical pattern of channel activity for holotoxin proceeds under conditions that mimic those across endosomes and lead to LC translocation and retrieval of protease activity after completion of transloca-tion (27). Thus, we have used this assay to examine the role of the interchain disulfide linkage on the translocation process.

Disulfide Bridge Drives Botulinum Neurotoxin Translocation
Premature Reduction of the Disulfide Bridge after Channel Formation Arrests LC Translocation-Previous work has demonstrated the important role of the disulfide bridge in the translocation process (27). We used the differential accessibility of the disulfide linkage between the HC and the LC to TCEP to identify requirements for translocation and showed that LC translocation requires both a pH gradient and a redox gradient, acidic and oxidizing on the cis compartment and neutral and reducing on the trans compartment. Significantly, addition of TCEP only to the cis compartment after acidification fails to evoke channel activity. This is indicative of disulfide shielding arising from the onset of the LC translocation through the HC channel. Here, we pursue this strategy to determine how the disulfide bridge affects the progress of LC translocation. ␤-mercaptoethanol (␤ME) is a powerful tool for this task. First, ␤ME does not modify HC channel activity, and preincubation of BoNT/A  Fig. 4B (black) and average time course of channel ␥ change for holotoxin preincubated with Fab (average N/data point for Fab preincubated with BoNT/A without addition of ␤ME ϭ 1,003 events and with addition of ␤ME ϭ 180 events; n ϭ 5 for each condition, magenta). Addition of ␤ME designated by the green arrow. Thin red line represents results for holotoxin without ␤ME addition; ␥ values associated with raw data from panel B are indicated.
with reductants results in HC channel activity (data not shown). Additions to the cis compartment cannot be directly made after seal formation; therefore membrane-permeable reagents that equilibrate across both compartments are required. In contrast to TCEP (33), ␤ME can traverse the lipid bilayer and reduce the disulfide bridge from either side of the membrane. If the disulfide bridge were translocated first across the membrane and were confined to the TCEPcontaining trans compartment during the early steps of translocation, then addition of ␤ME should have no effect on channel activity and growing conductance would end invariably in an unoccluded channel with ␥ ϳ 67 pS. The conductance growth of holotoxin channels is interrupted by addition of ␤ME immediately after its onset, as shown in Fig. 1B. The top panel shows a single BoNT/A channel undergoing a progressive increase in conductance, comparable with the entry events characteristically displayed by holotoxin (Fig. 1A). Addition of ␤ME, however, precludes entry into the higher ␥ intermediates, as evidenced in the middle and bottom panels of Fig. 1B in which two channels that inserted into the membrane prior to ␤ME addition remain in one of the low conductance states for the remainder of the experiment (Fig. 1C, black trace). The current transitions are faster and shorterlived than those typical of unmodified, unoccluded holotoxin, giving the appearance of flickering between low and high conductance states, a characteristic feature of channel block (27,34). Analysis of the results of five experiments of this type (Fig. 2) shows that LC translocation is arrested by reduction of the disulfide linkage after the initiation of translocation. Under these conditions, the holotoxin channel preferentially resides in one of the following conductance states ( Fig. 2A): 10, 17, or 31 pS. The lowest two states detected correspond to the entry event identified for holotoxin if translocation is unperturbed (28); however, the 31-pS intermediate may correspond to a non-productive, dead-end state for the LC and HC. The increased P o of the low conductance states (␥ ϳ 10 and 30 pS, Fig. 2B) and the preponderant occupancy of these occluded intermediate states (Fig. 2C) further demonstrate that intermediate steps in the growing conductance of the holotoxin channel are stabilized, thereby precluding completion of LC translocation: the HC channel is occluded by the LC and translocation is arrested.

Premature Reduction of the Disulfide Bridge of Single-chain BoNT/E before Channel Formation Arrests LC Translocation-
The finding that interruption of BoNT/A LC translocation results from premature reduction of the disulfide bridge leads to the hypothesis that the LC must be anchored to the HC during translocation. Is any linkage between the LC and HC sufficient to promote LC translocation or is the intact disulfide bridge specifically required? Whereas BoNT/A is cleaved to the mature di-chain within the Clostridium bacteria, BoNT/E is not cleaved before secretion. We previously demonstrated that, for the single-chain BoNT/E, completion of LC translocation proceeds only after proteolytic cleavage by trypsin and disulfide reduction in the trans compartment (28). Single-chain BoNT/E holotoxin is, therefore, an appropriate system to explore the linkage requirements for LC translocation. Accordingly, BoNT/E was incubated with 10 mM TCEP for 30 min at room temperature before the translocation assay. Prereduced BoNT/E displays channel activity as illustrated in Fig. 3A. Channel openings exhibit a ␥ similar to that of the early intermediates detected for unmodified BoNT/E (28); however, these low ␥ events do not undergo a transition to the higher ␥ intermediates or to unoccluded states (Fig. 3A). Despite the fact that trypsin is present in the trans compartment, the channel remains in low ␥ states for the lifetime of the experiment.
Analysis of four separate experiments under these conditions demonstrates that the channel activity of prereduced singlechain BoNT/E is different from that of the intact, disulfide cross-linked BoNT/E, never reaching the unoccluded channel state ␥ ϳ 65 pS. BoNT/E with an intact disulfide bridge transi- BoNT/A preincubated with Fab without ␤ME addition Gaussian fit illustrated in pink, data not shown (N ϭ 22,113 events). B, P o as a function of ␥ for Fab BoNT/A complex (pink) (n ϭ 5; the average N/data point ϭ 7,350 events), Fab BoNT/A complex with ␤ME addition (black) (n ϭ 5; the average N/data point ϭ 8,750 events), and unoccluded HC (blue) (n ϭ 1; the average N/data point ϭ 1,024 events). C, average occupancy time of conductance states for BoNT/A-Fab complex with (black) and without (pink) addition of ␤ME after onset of channel activity (with ␤ME, n ϭ 4, average N/data point ϭ 2,840 events; without ␤ME, n ϭ 5, average N/data point ϭ 5,170 events).
tions from occluded to unoccluded channel activity within 300 s (Fig. 3B, cyan); reduction of the disulfide prior to channel insertion results in a permanently occluded channel with intermediate conductance states at ␥ ϳ 12 and 26 pS (Fig. 3B, black,  and 3C). The conductance intermediate at 26 pS is similar to the prematurely reduced BoNT/A intermediate at 31 pS and may correspond to a dead-end intermediate state. The persistent occupancy of these low ␥ states and decreased P o with respect to unmodified BoNT/E further demonstrate the requirement for an intact disulfide bridge before channel insertion and throughout LC translocation (Fig. 3, D and E). The peptidic linkage between the LC and HC is insufficient to promote productive LC translocation. The disulfide bond may serve to maintain a translocation-competent conformation of the LC and its close proximity to the HC channel; reduction of the disulfide cross-link may dissociate such dynamic interaction.
These results indicate that the disulfide bridge must be on the trans (cytosolic) compartment to achieve productive translocation of the LC; disulfide disruption on the cis compartment or within the bilayer during translocation aborts it. Disulfide disruption at different intermediate ␥ states resulted in arrested LC translocation; therefore, we infer that completion of LC translocation occurs as the disulfide bridge, C terminus of the LC, enters the cytosolic compartment. This analysis supports a model of N-to C-terminal orientation of cargo during translocation with the C terminus as the last portion to be translocated and exit the channel. We propose that an intact disulfide bridge is a necessary condition for translocation but not for channel insertion, as demonstrated by the facts that the isolated HC channel is unperturbed by chemical reductants (27) and that prematurely reduced single-chain BoNT/E exhibits low conductance channel activity. The tight coupling of translocation completion with disulfide reduction strongly argues in favor of the view that LC refolding precludes retrotranslocation. From this viewpoint, refolding in cytosol may be interpreted as a trap that prevents retrotranslocation and dictates the unidirectional nature of the translocation process. The disulfide linkage is, therefore, a crucial aspect of the BoNT toxicity and is required for chaperone function, acting as a principal determinant for cargo translocation and release.
An LC-specific Antibody Arrests LC Translocation-To selectively restrict the location of the disulfide linkage to the entry site into the HC and to probe its accessibility to ␤ME, we exploited an LC-specific monoclonal antibody previously documented to block LC translocation (28). For this type of experiment Fab fragments are preincubated with BoNT/A for 1 h at pH 7 before the translocation assay. Under these conditions, low conductance channels are detected within a few minutes after patch excision (Fig. 4A, top panel). The initial conductances are comparable with those characteristic of the early events in unmodified LC translocation; however, the channels remain in the low conductance states throughout the experiment (Fig. 4A, bottom panel). We interpret these intermediates as early steps in translocation in which the HC has formed a channel that is partially occluded by the LC. Fab binding to the LC allows channel formation and early translocation, presumably stabilizing intermediate protein-protein interactions. However, it locks the channel and the LC in a translocating conformation that is irreversibly incomplete.
Reduction of the Disulfide Bridge Releases the Fab-induced HC Channel Block by LC-Reduction of the disulfide bridge between the LC and the HC may facilitate release of the Fab-LC complex, thereby unoccluding the HC channel. The disulfide on the Fab will be concurrently reduced; accordingly, only if the complex remains intact under reducing conditions will release of HC channel block ensue. This model was tested by preincubating the Fab with the BoNT/A, followed by supplementing ␤ME to the trans compartment after the onset of the channel activity. Within minutes of ␤ME addition the low conductance channel does enter the unoccluded channel state (Fig. 4B, bottom panel, and 4C, black) in sharp contrast to the Fab-induced block of channel activity (Fig. 4C, pink). The latency period for release of the HC channel from block by the LC estimated from these measurements is ϳ730 s (Fig. 4C, black) as compared with ϳ150 s for unabated LC translocation in unmodified holotoxin (Fig. 4C, red). Holotoxin channels under these conditions exhibit discrete transient intermediate conductances at ␥ Х 13, 19, 50, and 57 pS before entering the unoccluded state at ␥ Х 65 pS, as evidenced from analysis of four separate experiments (Fig. 5, A and B). These ␥ and P o (Fig. 5C, pink) values approximate the low conductance intermediate states of holotoxin during productive LC translocation (Fig. 5C, black). However, the occupancy time in each conductance intermediate is longer as compared with unperturbed holotoxin (Fig. 5C) (28). The low ␥ intermediate states are stabilized by the Fab (Fig. 5A); however, the prolonged residency in the larger ␥ intermedi- Fab with ␤ME addition (black), unmodified holotoxin (red), and unoccluded HC (blue). V1 ⁄2 for holotoxin bound to Fab ϩ ␤ME is Ϫ59.0 Ϯ 4.5 mV, for unmodified holotoxin is Ϫ58.3 Ϯ 1.7 mV, and for HC is Ϫ63.4 Ϯ 2.4 mV (3 Յ n Յ 11/data point; average N/data point ϭ 2,260 events). B, structure of BoNT/A holotoxin (2): LC, purple, translocation domain, orange, and receptorbinding domain, red, prior to insertion in the membrane (gray bar with magenta boundaries) ➀; then schematic representation of BoNT/A inserted during translocation of the LC through the HC channel (orange) with intact disulfide bridge (green) while located in the cis-compartment ➁ and the unoccluded HC channel in the membrane after LC dissocia-tion➂. Occluded HC channels can be trapped in low conductance states under two experimental conditions: ➃ LC translocation arrested by Fab and ➄ LC translocation arrested by premature reduction of the LC-HC disulfide bridge. Channels formed by holotoxin bound to an LC/A-specific Fab can be unoccluded by the subsequent ␤ME-induced release of the LC-Fab complex into the acid pH compartment ➅. ates (Fig. 5B) is a unique aspect of this experimental condition. These findings indicate that although the LC is released from the HC channel, the kinetics are different from that of unabated translocation. The release of channel block is consistent with the strong binding affinity of the Fab for the LC: even though the disulfide bridge holding the two Fab subunits has been reduced the binding affinity is sufficient to facilitate dissociation of the LC from the HC channel. In the presence of Fab, therefore, translocation is aborted and disulfide reduction releases the LC to the cis compartment, thereby unblocking the HC channel.
Importantly, the unoccluded HC channels that result from this experimental strategy (Fig. 6A, black) are indistinguishable from channels produced by isolated HC/A (Fig. 6A, blue) and by holotoxin/A after productive translocation of LC (Fig. 6A,  red). These three experimental conditions lead to a channel entity that exhibits comparable voltage dependence: the V1 ⁄ 2 is ϳϪ60 mV (Fig. 6A). In the presence of Fab translocation is aborted (Fig. 6B, pathway ④) and disulfide reduction releases the LC on the cis compartment, thereby unblocking the HC channel (Fig. 6B, pathway ⑥).
Concluding Remarks-The new findings here described and our interpretation are summarized in the scheme shown in Fig.  6B. The initial condition ①, illustrated with a surface representation of the crystal structure of BoNT/A (2), depicts the holotoxin as an aqueous soluble protein at neutral pH. Exposure to a pH gradient and a redox gradient comparable with those across endosomal membranes leads to a conformational change of the holotoxin with the consequent insertion of the HC into the membrane and the onset of LC translocation, manifested as occluded channels with a characteristic small ␥ (Fig.  6B, pathway ②). Productive translocation leads to the release of the LC in the trans compartment, with the ultimate refolding of the LC and retrieval of its protease activity, and the consequent unblocking of the HC (Fig. 6B, pathway ③). Two experimental conditions arrest LC translocation and yield occluded HC channels. Preincubation of holotoxin with an LC-specific Fab allows HC channel formation yet it aborts LC translocation (Fig. 6B, pathway ④). Premature reduction of the disulfide linkage of BoNT/A holotoxin after channel formation terminates LC translocation (Fig. 6B, pathway ⑤); in accordance, premature reduction of single-chain BoNT/E holotoxin before channel formation allows channel insertion yet aborts LC translocation. Pathway ⑥ highlights the result that reduction of the disulfide bridge on the cis compartment releases the Fab-induced HC channel block by LC with the ensued dissociation of the LC-Fab complex into the cis compartment, thereby unblocking the HC channel. The unoccluded HC channel entity that results from the productive translocation of the LC (pathway ① 3 ② 3 ③) (32) or from the disulfide reductionmediated relief of the Fab-induced HC channel block (pathway ① 3 ④ 3 ⑥) is indistinguishable from that produced by the isolated HC (27) in terms of single channel conductance, selectivity, and voltage-dependence features.