Affixing N-terminal α-Helix to the Wall of the Voltage-dependent Anion Channel Does Not Prevent Its Voltage Gating*

Background: There is ongoing controversy concerning the location and mobility of the N-terminal α-helix in VDAC1 during voltage gating. Results: mVDAC1 with the N-terminal α-helix cross-linked to β-strand 11 forms typical voltage-gated channels. Conclusion: The N-terminal domain of VDAC1 does not move independently during voltage gating. Significance: This study dramatically alters the current view of voltage gating dynamic in VDAC1. The voltage-dependent anion channel (VDAC) governs the free exchange of ions and metabolites between the mitochondria and the rest of the cell. The three-dimensional structure of VDAC1 reveals a channel formed by 19 β-strands and an N-terminal α-helix located near the midpoint of the pore. The position of this α-helix causes a narrowing of the cavity, but ample space for metabolite passage remains. The participation of the N-terminus of VDAC1 in the voltage-gating process has been well established, but the molecular mechanism continues to be debated; however, the majority of models entail large conformational changes of this N-terminal segment. Here we report that the pore-lining N-terminal α-helix does not undergo independent structural rearrangements during channel gating. We engineered a double Cys mutant in murine VDAC1 that cross-links the α-helix to the wall of the β-barrel pore and reconstituted the modified protein into planar lipid bilayers. The modified murine VDAC1 exhibited typical voltage gating. These results suggest that the N-terminal α-helix is located inside the pore of VDAC in the open state and remains associated with β-strand 11 of the pore wall during voltage gating.

The voltage-dependent anion channel (VDAC) 4 serves as the primary conduit between the mitochondria and the rest of the cell, facilitating free exchange of ions and metabolites across the outer mitochondrial membrane. In addition to its metabolic and energetic functions, VDAC has a more complex role, serving as a receptor for molecules and proteins that modulate the organelle's permeability and thereby its function (1)(2)(3). This multifunctional channel has been implicated in the metabolic stresses of cancer, cardiovascular disease, and mitochondrialdependent cell death (4 -8); thus, understanding the structure and function of VDAC constitutes a critical objective for basic as well as medical research.
Single channel conductance experiments on VDAC1 at low membrane potential (Ͻ30 mV) reveal a high conductance (4.1 Ϯ 0.1 nanosiemens in 1 M KCl) indicative of a large pore, usually referred to as the "open" state of the channel (1). This conformer facilitates the passage of 10 6 ATP molecules (9) and displays a preference for monovalent anions over cations with the anion-to-cation permeability ratio of 2:1 in high salt (10) and 4:1 in physiological salt concentration (11). As voltage is increased (Ͼ30 mV) in either a positive or a negative direction, the channel switches into the lower conducting states (around 2 nanosiemens in 1 M KCl), termed as the "closed" state(s). These states are cation-selective with a negligible metabolite flux (9,12). The transitions between the open and the closed states involve large conformational changes (13,14) constituting a gating action that both hinders the passage of metabolites and alters ion selectivity (1). Colombini and co-workers (1, 14 -16) performed extensive electrophysiological and biochemical analyses of VDAC properties using site-directed mutagenesis and biotin modification studies. Based on these data, they developed a model for VDAC voltage gating where the N-terminal segment is an essential component of the mobile voltage sensor domain, which slides in and out of the channel lumen in response to the applied voltage (1,14,15). Several additional studies have reported alternative models in which the N-terminal segment is exposed to the cytoplasm (7,17,18) or resides on the membrane surface (19,20). Although the proposed mechanism of voltage gating differs among investigators, most models involve large conformational changes of the N-terminal segment.
In 2008, three groups reported, with increasing atomic detail, the long sought after structure of VDAC1 (21)(22)(23). All three structures showed a large pore composed of 19 ␤-strands with the high resolution structure of murine VDAC1 (mVDAC1) revealing the precise orientation of the N-terminal voltagesensing domain (Fig. 1). The ␣-helical portion of the N-terminal segment is positioned halfway through the pore, causing a narrowing of the cavity, where it is ideally situated to regulate metabolite flux. Thus, it was believed that the transition between conformations required a repositioning of the N-terminal segment. Here we report that, unexpectedly, the N-terminal segment of VDAC1, which lines the channel pore, does not undergo independent dynamic structural rearrangements during channel gating. We show that cross-linking of the ␣-helical portion of the N-terminal segment to the wall of the pore does not prevent VDAC1 voltage gating when inserted in planar lipid bilayers. These results suggest that the N-terminal ␣-helix is located inside the pore when VDAC1 is open and remains stably associated with residue 170 of the ␤-barrel wall during voltage gating.
Recombinant protein expression and purification were carried out as described previously (23). Briefly, a two-step purification procedure consisting of Talon affinity and size exclusion chromatography was performed to isolate a homogenous population of murine His-tagged VDAC1 WT and VDAC-Cys. Purified protein was washed using an Amicon 30-kDa cut-off concentrator to replace the size exclusion chromatography buffer with 20 mM Tris, pH 8.0, 50 mM NaCl, 0.1% LDAO. Aggregated protein was removed by ultracentrifugation at 430,000 ϫ g for 30 min, and the concentration was determined using absorption at 280 nm.
Cross-linking-Cysteine mutants described above (VDAC-Cys) were used to cross-link the mVDAC1 N-terminal region to the ␤-barrel using a mix of copper/phenanthroline as a crosslinking reagent. The metal chelator ␣-phenanthroline, in combination with trace amounts of cupric ion, catalyzed the oxidation of the sulfhydryl groups of cysteines to form disulfide bonds (24). This reaction could be reversed by reducing agents, such as DTT. The cross-linking reaction was performed by using a copper and ␣-phenanthroline mix (CuSO 4 -phenanthroline; Cu-Ph). Briefly, 0.005 g of CuSO 4 was dissolved in 75 l of double-distilled water, and 0.012 g of phenanthroline was dissolved in 25 l of 100% ethanol and 25 l of double-distilled water. Both solutions were then mixed stepwise (i.e. 25-l aliquots of CuSO 4 were added to the phenanthroline solution and mixed by vortex) in order to avoid the formation of precipitates. Serial dilutions were then prepared from the stock solution of 150 mM Cu-Ph.
Cross-linking reactions, "in tube," were performed by mixing 20 l of 0.1 mg/ml VDAC-Cys with 0.02 mM Cu-Ph followed by incubation for 10 min at room temperature. The rest of the experiment was performed on ice. Cross-linking was reversed by the addition of 100 M DTT to the VDAC-Cys or to the VDAC-Cys with 0.02 mM Cu-Ph mix. To monitor the crosslinking efficiency, cross-linked samples were analyzed on SDS-PAGE after tetramethylrhodamine 5-maleimide (TMRM) labeling (described below).
TMRM Labeling-TMRM (Sigma) was dissolved in 100% DMSO. Purified mVDAC1 was incubated with a 10-fold molar excess of TMRM for 15 min at room temperature. The reaction was stopped by the addition of SDS-PAGE loading buffer. 10 g of protein samples were run on a 12% SDS-PAGE. Detection of TMRM was carried out using UV light (Gene Flash Syngene Bio Imaging). Gels were subsequently stained with R-250 Coomassie Brilliant Blue to visualize loading quantities.
VDAC Reconstitution and Conductance Measurements-Planar lipid membranes were formed on a 70 -90 m diameter orifice in a 15-m-thick Teflon partition that separated two chambers as described previously (25). Lipid monolayers used for membrane formation were made from a 5 mg/ml solution of diphytanoylphosphatidylcholine (Avanti Polar Lipids, Inc.) in pentane. Channel insertion was achieved by adding 0.3 Ϯ 0.1 l of mVDAC1 WT (stock: 0.01 mg/ml in LDAO buffer) or 1.5 Ϯ 0.8 l of VDAC-Cys (stock: 0.1 mg/ml in LDAO buffer) into the 1.2-ml aqueous phase of the cis compartment while stirring. Aqueous solutions of 1 M KCl were buffered with 5 mM HEPES at pH 7.4. Potential was defined as positive when it was greater at the side of the VDAC addition (cis-side). Current recordings were performed as described previously (25) using an Axopatch 200B amplifier (Axon Instruments, Inc., Foster City, CA) in the voltage-clamp mode. Single-channel data were filtered by a low pass 8-pole Butterworth filter (model LPF-8, Warner Instrument Corp.) at 15 kHz and directly saved with a sampling frequency of 50 kHz.
The voltage-dependent properties of a VDAC-containing membrane were assessed following the protocol devised by Colombini and co-workers (10,26) in which gating is inferred from the channel response to slowly changing periodic voltage waves applied. A symmetrical 5 mHz triangular voltage wave with Ϯ60 mV amplitude from a Function Waveform Generator 33120A (Hewlett Packard) was used. Data were acquired with a Digidata 1322A board (Axon Instruments Inc.) at a sampling frequency of 1Hz and analyzed using the pClamp 10.2 software (Axon Instruments, Inc.). Analysis of VDAC voltage gating was performed following Ref. 10 and as described in Ref. 25. Two gating parameters (n, which is the effective gating charge and is the measure of the voltage-dependence steepness, and V 0 , the voltage at which half of the channels are open and half are closed) were calculated from the open probability plots following Ref. 10 as described previously (25). Plot fitting was performed through the logarithmic version of the Boltzman distribution equation (25).
After VDAC-Cys was incubated with different concentrations of Cu-Ph in tube, the cross-linked VDAC was then reconstituted into the planar lipid bilayers as described above. Crosslinking reactions were reversed by the addition of 100 M DTT to VDAC-Cys in tube before channel reconstitution into the membrane or by the addition of 100 M DTT "in bath," after membrane reconstitution of VDAC-Cys. Cross-linking reactions in bath were performed on VDAC-Cys already reconsti-tuted into planar membrane. When single or multiple channels were incorporated into the membrane and their conductance and voltage gating properties were established, Cu-Ph was added directly to the membrane-surrounding aqueous solutions to both sides of the membrane under constant stirring, and channel properties were monitored 10 min after the Cu-Ph addition.

Cross-linking of ␣-Helical Portion of N-terminal Segment to
␤-Strand 11-Previously, we resolved the structure of mVDAC1 at 2.3 Å resolution using the technique of lipidic bicelle crystallization (27,28). The structures of mVDAC1 reveal a unique 19-strand ␤-barrel ( Fig. 1A) arranged in an antiparallel pattern with the exception of strands ␤1 and ␤19, which associate in a parallel manner to seal the barrel. The exterior of the ␤-barrel primarily consists of hydrophobic residues that are exposed to the lipid environment, whereas the interior is extensively hydrophilic, making it conducive for the passage of ions and large metabolites through the outer mitochondrial membrane. From both sides of the membrane, the channel's opening resembles an hourglass with a maximal inner dimension of 27 ϫ 24 Å. Approximately halfway through the pore, the N-terminal ␣-helix segment (aligned nearly parallel to the plane of the membrane) causes a partial narrowing of the cavity to 27 ϫ 14 Å (Fig. 1C), yet ample space remains for the passage of metabolites in this arrangement. Thus, this structure probably represents the open conformation, a claim substantiated by our recent computations (29). Numerous studies have established the importance of the N-terminal ␣-helix in regulating the flux of ions and metabolites through the channel, and the observation that it is localized within the pore (21-23) makes it ideally suited to regulate their passage. Furthermore, the N-terminal region is widely believed to be flexible with a potential for large structural rearrangements, in response to stimuli that alter the channel's properties (7, 16 -18, 20, 23, 30 -32).
Closer examination of the mVDAC1 structure shows that the N-terminal segment (amino acids 1-26) has extensive hydrophobic interactions and hydrogen bonds with the interior wall of the pore (adjacent to ␤-strands 8 -19), facilitating its orientation (Fig. 1B). In the current structure, Leu-10 (located on the N-terminal ␣-helix) and Ala-170 (located on the ␤-strand 11) are ideally positioned to form a disulfide bond (Fig. 1B), which would "lock" the ␣-helix to the pore wall. To accomplish this, an initial Cys-less mVDAC1 construct (CL-mVDAC1) was generated by mutating the two endogenous cysteine residues to alanines (C127A and C232A). In this Cys-less background (CL), cysteines at position 10 (L10C-CL) and at position 170 (A170C-CL) were introduced, and their accessibility was monitored using the fluorescent chromophore TMRM, which forms a covalent bond with accessible Cys residues. This reaction is easily monitored in gels by following the fluorescence of TMRM (33). Fig. 2B (left) shows that TMRM fluorescence is eliminated in the CL background (lane 2). When Cys was engineered at position 170 or 10 (lanes 3 and 4, respectively), the TMRM fluorescence was reestablished. The cysteine accessibility was also monitored in the double cysteine mutant L10C/ A170C (VDAC-Cys) (Fig. 2B, right), which still retained its proper function (Figs. 3B, 4B, and 5).
A cross-link was generated between positions 10 and 170 by incubating the sample with 0.02 mM Cu-Ph, an oxidizing agent that induces disulfide formation. After the addition of the Cu-Ph, the accessibility was again tested, and no fluorescence was observed, indicative of a disulfide bond involving the two cysteine residues (Fig. 2B, right). In addition, after cross-linking, a large gel shift was observed, resulting in the L10C/A170C cross-linked sample running lower on the gel (Fig. 2). This gel shift was reversible upon the addition of the reducing agent DTT ( Fig. 2A), and TMRM accessibility was restored. Similar features have been observed in other outer membrane proteins that exhibited cross-linking (34,35).

Characterization of L10C/A170C Double Cysteine Mutant in Planar Lipid
Bilayers-A representative current record of the VDAC1 WT reconstituted into planar lipid bilayer is shown in Fig. 3A (VDAC WT). Increasing applied voltage, in either the positive or negative direction, resulted in the characteristic VDAC voltage gating where the channel switches from a high conducting state referred to as the "open" conformation, which facilitates the passage of ATP and other nucleotides, to the variety of the lower conducting or "closed" states that are essentially impermeable to ATP. Similarly, the L10C/A170C mutant protein in the Cys-less background (VDAC-Cys) showed the characteristic VDAC voltage gating as it transitioned between open and multiple closed states (Fig. 3B). Cross-linking was initiated by the addition of 0.02, 0.2, and 0.8 mM Cu-Ph to the membrane-surrounding buffer. All cross-linked proteins, even at the highest concentration of Cu-Ph, also formed normal fully functioning channels in planar membranes (Fig. 3, C-E). These results strongly indicate that the N-terminal segment does not undergo independent conformational changes during voltage gating.
To establish that the gating behavior recorded was representative for the majority of mVDAC1 channels undergoing this transition, we repeated these experiments using multichannel recordings on membranes with an average of 20 channels reconstituted at once (Fig. 4). Again, the WT and the L10C/ A170C mutant proteins with and without Cu-Ph (0.6 and 1.3 mM) in the membrane-bathing solutions consistently behaved analogously. Analysis of the probability of VDAC to be open or closed at varying potentials (Fig. 5A) showed similar voltage gating for VDAC-Cys regardless of the presence of cross-linker in the bathing solution, once again indicating that the crosslinking does not inhibit or reduce voltage gating as one would anticipate if the N-terminal segment undergoes independent large conformational changes. Furthermore, the VDAC-Cys showed higher voltage sensitivity at positive potentials, as observed by a shift in the open probability curves to the lower potentials ( Fig. 5A and Table 1).
It is possible that once inserted into the bilayer, the VDAC-Cys assumes a different conformation (36), which is not permissive for disulfide formation. To ensure that a proper crosslink was formed between the ␤-strand and the ␣-helix, the sample was preincubated with 0.02, 0.8, and 10 mM Cu-Ph prior to reconstitution into the planar membranes (Fig. 5B). This reaction was monitored by TMRM labeling (Fig. 2B). Once again, these cross-linked samples maintained typical VDAC voltage gating in planar membranes and shifted the open probability to the lower potentials with respect to the WT protein ( Fig. 5B and Table 1). Alternatively, the VDAC-Cys was treated with 0.1 mM DTT, in order to reduce the disulfide bond, prior to protein reconstitution into the membrane. Interestingly, the open probability curve and voltage-gating parameters obtained under these conditions were nearly identical to those of the VDAC WT (Fig. 5B and Table 1). The same effect was observed when 0.1 mM DTT was added directly to the membrane-surrounding buffer solution after VDAC-Cys (pretreated with 0.02 mM Cu-Ph in tube) was reconstituted into the membrane (supplemental Fig. 1 and Table 1). In conclusion, cross-linked VDAC forms functional channels in planar lipid membranes with typical conductance and voltage gating. These new data indicate that the N-terminal domain does not move independently during voltage gating, resulting in large structural rearrangements, as suggested by several previous studies (7,17,18,20,23,32). The N terminus remains affixed to ␤-strand 11 of the pore wall, or if in motion, it moves in coordinated motion with the rest of the voltage sensor domain.

DISCUSSION
There is considerable evidence demonstrating that VDAC undergoes large conformational changes during voltage-dependent gating (1,13,14), and many studies have attributed these changes to movements of the N-terminal segment (7,17,18,20,23,32,37). By a combination of site-directed mutagenesis, induced disulfide bond formation, and electrophysiological characterization in planar lipid bilayers, we have shown that VDAC with the N-terminal ␣-helix portion engineered to covalently bind the pore wall behaves indistinguishably from endogenous protein, suggesting that the N-terminal segment is either not mobile or does not move independently of the rest of the voltage-sensor domain during gating.
In addition to changes in membrane potential, different factors shift the dynamic equilibrium of VDAC toward the closed conformation, including protein-protein interactions (2, 38 -41), low pH (10,42), or the binding of NADH (10). To date, there are a number of proposed mechanisms for achieving gating through the N-terminal segment (23,30,43). In our initial report of the mVDAC1 structure, the stimuli mentioned above were believed to disrupt the interactions between the N-terminal helix and the ␤-barrel wall, resulting in a reorientation within the channel lumen to adopt the more occlusive conformation of the pore (23). Further evidence in support of alterations in the position of the ␣-helix came from the NMR structure of human VDAC1 (21), which showed an NADH interaction site on ␤-strands 17 and 18. The position of the NADH site flanks a flexible region (Gly-21, Tyr-22, Gly-23, Phe-24, and Gly-25) of the N-terminal segment, and the mVDAC1 structure clearly demonstrates that there is insufficient space for NADH binding in this region. Thus, the N-terminal segment would have to be repositioned to accommodate NADH. Colombini and colleagues (16,36) have previously generated a topology model of VDAC, derived from biochemical and electrophysiological assays, where 13 ␤-strands and a single ␣-helix form the pore of the channel. In this model, the voltage-sensitive domain of VDAC is mobile and forms a large portion of the pore wall, including the N-terminal segment. They further asserted that upon application of voltage, large conformational changes would be required for changing the pore's dimension, conductance, and reversal of ion selectivity. However, according to this two-dimensional topology, the N-terminal ␣-helix is part of the pore's wall (including position Leu-10), and Ala-170 is located in a loop in the presumably intermembrane space between ␤-strands 8 and 9 (according to the numbering in Refs. 16 and 36). If this were the case, residues Leu-10 and Ala-170 would be too far to form a disulfide bond.
It should be noted that only a small population of protein is inserted into the bilayer, and, although unlikely, there remains the possibility that the protein measured in bilayers has not been cross-linked. We have attempted to address this issue by demonstrating nearly complete cross-linking (Fig. 2) before channel reconstitution into the bilayer and performing multiple measurements (a minimum of three replicates) on membranes reconstituted with multiple channels (ϳ20 channels) (Figs. 4 and 5). Furthermore, only the L10C/A170C mutant, before or after the cross-linker addition, shows a lower open probability at positive potentials than VDAC WT (Fig. 5 and Table 1), demonstrating that there is a quantitative difference in gating parameters between wild type and cross-linked protein. Indeed, V 0 , the characteristic voltage at which half-channels are open, is ϳ20 mV lower for the cross-linked VDAC-Cys (16 -19 mV) than for VDAC WT (37 Ϯ 2 mV) ( Table 1). The V 0 value of non-cross-linked VDAC-Cys lies between those of WT and cross-linked protein (25 mV), suggesting that part of VDAC-Cys was spontaneously cross-linked without Cu-Ph treatment. Importantly, this shift in open probability at positive potentials was completely reversed by the addition of DTT to the VDAC-Cys prior to reconstitution of channels (Fig. 5B) as well as after their reconstitution when DTT was added to the membrane-surrounding buffer solution (supplemental Fig. 1 and Table 1). These data, together with the Coomassie Blue gel that shows two bands of VDAC-Cys without Cu-Ph or DTT treatment (Fig. 2), strongly suggest that a significant portion of VDAC-Cys have generated disulfide bonds even before the addition of Cu-Ph. The spontaneous formation of a disulfide bond is reasonable, considering the close proximity of Leu-10 and Ala-170 (Fig. 1B). Analogous results have been reported for the bacterial porin, OmpF (35), where a series of Cys mutations were generated between residues in the constriction loop and the wall of the ␤-barrel in which no alterations in voltage gating were observed upon cross-linking.
There are now a number of lines of evidence that the N terminus does not undergo independent dynamic rearrangements. The first indications of a stable N terminus arose from NMR studies on human VDAC1 by Wagner's group (37). Spectra were collected in the "open" conformations as well as under conditions known to "close" the channel (in the presence of NADH). Although significant spectral shifts were observed, the assignment for the external portion of the N terminus remained the same in both open and closed conformations. A more recent solid state NMR study provided a more complete assignment of human VDAC1, including the N-terminal segment (44). The authors were able to demonstrate that the N-terminal segment stabilizes the integrity of the ␤-barrel by monitoring major spectral shifts between full-length protein and the N-terminal deletion protein. Supporting these data, Popp et al. (45) observed that an N-terminal deletion did not alter VDAC voltage gating but produced noisy channels with a tendency to close, suggesting that the N-terminal region might also be involved in the stabilization of the open state. Finally, the crosslinking data presented here strongly suggest that the ␣-helix remains inside the channel's pore attached to ␤-strand 11 when VDAC is open. During voltage-induced gating, the N-terminal region either stays inside the pore or moves in a concerted fashion with the rest of the voltage sensor rather than exerting large independent structural rearrangements to achieve voltage-dependent gating. It should be noted that the data presented here still permit the translocation of the N terminus in and out of the lumen upon gating, as suggested previously (16,20,32,46), as long as this movement is coordinated with the voltage sensor.
If voltage gating does not involve rearrangements of the N-terminal segment, then how is voltage gating achieved? There is still no clear structural understanding of VDAC dynamics, but a new mechanism is emerging; the barrel wall becomes elongated, and the pore narrows around the intact N-terminal segment to exclude nucleotides. Using a combination of NMR, hydrogen/deuterium exchange, and MD simulations, Villinger et al. (30) showed pronounced dynamics in ␤-strands 1-6 and 16 -19. The analysis of the B-factors from our original crystal structure supports these data where these same regions have increased B-factors (Ͼ20%) compared with the ␤-strand flanking the N-terminal helix (␤-strands 8 -15), indicating that ␤-strands 8 -15 and the ␣-helix have a more rigid structure than the remainder of the pore. Also, initial computational attempts to model the voltage dependence of gating through movements in the N-terminal helix alone failed to produce large enough gating charge values (29), hinting at the possibility that the rearrangement is more extensive.
Until recently, ␤-barrel proteins were considered to be rigid structures due to their vast hydrogen bonding network, but large conformational changes have recently been observed for the outer membrane protein FimD (47). In the apo conformation, this 24-stranded ␤-barrel protein has an oval shape with a 52 ϫ 28 Å diameter. Upon activation, FimD forms a complex with FimC-FimH, and the pore becomes more circular, with a 44 ϫ 36 Å diameter. It is worth mentioning that the apo form of FimD also has segments of high B-factors corresponding to the regions with the largest movements. It is thus tempting to envision a similar mechanism for VDAC where the pore would collapse around the N terminus to achieve closure. Interestingly, it was recently suggested that the conformational transition of VDAC during voltage gating could be sensitive to pressure changes in the hydrocarbon area of the membrane (25), which supports the hypothesis of the soft or flexible part of the VDAC ␤-barrel. Taking together the current understanding of VDAC structure/function, there is a reasonable probability that the barrel undergoes partial constriction (and possibly simultaneous elongation along the axis perpendicular to the membrane narrowing the pore). Importantly, this does not contradict previous models where part of the ␤-barrel together with the ␣-helix move in concert upon gating and slide out of the channel lumen toward the membrane interface (1,36). The correct model for the transition mechanism between the open and closed conformations of the channel that reconciles all the reliable experimental data will be achieved after further evaluation of the data from molecular dynamics, structural, and biophysical techniques.