Skip to main content
NCBI home page
UKPMC Funders Author Manuscripts logoLink to UKPMC Funders Author Manuscripts
. Author manuscript; available in PMC: 2017 Oct 20.
Published in final edited form as: Structure. 2012 Jul 26;20(9):1540–1549. doi: 10.1016/j.str.2012.06.015

β-barrel mobility underlies closure of the voltage-dependent anion channel

Ulrich Zachariae 1,2,#, Robert Schneider 3,4,#, Rodolfo Briones 1, Zrinka Gattin 3, Jean-Philippe Demers 3, Karin Giller 3, Elke Maier 5, Markus Zweckstetter 3, Christian Griesinger 3, Stefan Becker 3, Roland Benz 5,6, Bert L de Groot 1,§, Adam Lange 3,§
PMCID: PMC5650048  EMSID: EMS74568  PMID: 22841291

Abstract

The voltage-dependent anion channel (VDAC) is the major protein in the outer membrane of mitochondria, where it mediates efficient transport of ATP and ADP. Changes in its conformation and permeability, induced by voltage or apoptosis-related proteins, have been implicated in apoptotic pathways. The three-dimensional structure of VDAC has recently been determined as a 19-stranded β-barrel with an in-lying N-terminal helix. However, its mechanism of gating is a matter of ongoing debate. Using solid-state NMR spectroscopy, molecular dynamics simulations, and electrophysiology, we show that deletion of the rigid N-terminal helix sharply increases overall motion in VDAC’s β-barrel frame, resulting in elliptic, semi-collapsed shapes of the β–barrel. These states quantitatively reproduce conductance and selectivity of the closed VDAC conformation. Mutation of the N-terminal helix leads to a phenotype intermediate to the open wild-type and the closed channel. These data suggest that the N-terminal helix controls entry into elliptic β-barrel states which form the structural basis for the closed state of VDAC. Our results also indicate that β-barrel channels are intrinsically flexible and imply mechanosensitivity of VDAC channels.

Keywords: VDAC, membrane protein, gating, solid-state NMR, molecular dynamics

Introduction

The voltage-dependent anion channel (VDAC) is the most abundant protein in the outer mitochondrial membrane of eukaryotes, where it mediates permeation of ATP, ADP, and other essential metabolites (Benz, 1994; Colombini, 2004). It is also considered to be a key player in the mitochondrial pathway of apoptosis (Shoshan-Barmatz et al, 2010; Shoshan-Barmatz et al, 2008). VDAC exhibits a complex gating process which changes conductance and ion selectivity of the channel in the presence of transmembrane voltages higher than about 30 mV, independent of their polarity (Benz, 1994; Colombini, 1989; Colombini, 2004). After three decades of experimental study, recent work on crystallized and detergent-solubilized protein has revealed the three-dimensional structure of the VDAC1 isoform to be a 19-stranded β-barrel fold with a partly helical N-terminal region situated within the aqueous pore (Bayrhuber et al, 2008; Hiller et al, 2010; Hiller et al, 2008; Ujwal et al, 2008). So far, however, the nature of VDAC’s gating process has remained elusive from both experimental and computational studies (Bayrhuber et al, 2008; Choudhary et al, 2010; Hiller et al, 2010; Hiller et al, 2008; Lee et al, 2011; Rui et al, 2011; Ujwal et al, 2008). The N-terminus of VDAC has been shown to be crucial for channel gating (Song et al, 1998; Thomas et al, 1993) as well as for interaction with apoptosis-related proteins (Abu-Hamad et al, 2009; Shoshan-Barmatz et al, 2008). However, several studies have indicated that conformational rearrangements within the β-barrel may be involved in gating as well (Mannella, 1997; Peng et al, 1992; Zimmerberg & Parsegian, 1986). In particular, it was found that gating is strongly influenced by osmotic pressure applied on the bilayer. From this finding, a substantial internal volume change during gating was deduced, pointing to a rather large reconfiguration of channel geometry (Zimmerberg & Parsegian, 1986). Gating affects VDAC permeability for ATP and is regulated by small molecules as well as apoptotic and anti-apoptotic proteins (Benz, 1994; Colombini, 1989; Colombini, 2004; Shoshan-Barmatz et al, 2010; Shoshan-Barmatz et al, 2008). A molecular understanding of the gating process is thus not only of mechanistic interest, but also holds promise for further elucidation of VDAC’s role in mitochondrial apoptosis.

We have recently demonstrated that the N-terminus assumes a well-defined conformation and plays a stabilizing role for VDAC1 (Schneider et al, 2010a). β-strands in contact with the N-terminus are less dynamic than others (Villinger et al, 2010), and removal of the N-terminus affects the conformation of the β-barrel (Schneider et al, 2010a). On the other hand, VDAC1’s β-barrel frame can exhibit extensive breathing motions (Villinger et al, 2010). Mutations of a glutamate residue whose side chain points toward the lipid bilayer (Glu73) affect both the dynamic behavior of the β-barrel and the gating properties of the channel (Villinger et al, 2010; Zaid et al, 2005). Consequently, local mobility and stability within VDAC1 can be expected to play an important role for channel gating.

As the membrane environment is generally found to be important for the structure and function of membrane proteins, studies that explicitly take the lipid membrane into account are highly desirable (Ader et al, 2008; Hunte & Richers, 2008; Nietlispach & Gautier, 2011; Phillips et al, 2009). We have therefore investigated the dynamics of VDAC1 and its role in gating in lipid bilayers, using a combination of solid-state NMR spectroscopy, electrophysiology, and molecular dynamics (MD) simulations. We quantify molecular mobility in bilayer-embedded VDAC1 and examine the effect of removal of the N-terminus on the conformation of the β-barrel. Furthermore, we demonstrate the importance of specific contacts between the β-barrel and the N-terminal helix for the overall stabilizing role of the latter. We show that only a large-scale conformational transition of the barrel scaffold itself can explain the large drop in conductance upon closure and the concomitant change in ion selectivity. Changes in conductance and selectivity observed in simulations of VDAC1 at different degrees of closure are in quantitative agreement with experimental data. The combination of experimental and simulation results hence allows us to suggest both a mechanism of VDAC closure and a molecular model for the closed states of VDAC. This mechanism provides a simple explanation for the observed dependence of gating on membrane pressure. Our results challenge the common belief that β-barrel channels are generally characterized by high structural rigidity.

Results

The N-terminus does not exhibit large dynamics on the sub-ms timescale

Data reported in ref. (Schneider et al, 2010a) indicated that the N-terminus of human VDAC1 (hVDAC1) does not exhibit large-scale mobility. To investigate this finding quantitatively, we measured (13C,13C) dipolar order parameters (SCC) for residues in the N-terminus as well as in other parts of the molecule using double-quantum (2Q) spectroscopy as introduced in ref. (Schneider et al, 2010b). This technique is sensitive to dynamics on the pico- (ps-) to millisecond (ms) timescale. Due to the considerable size of hVDAC1 and the large number of residues in β-sheet conformation, spectral overlap precludes identification of residue-specific order parameters for a large part of the molecule. To estimate overall mobility in the β-barrel, we also analyzed overlapping signals of residues in β-strand secondary structure and determined average order parameters for these.

Results displayed in Figure 1 and Table S1 show that backbone SCC order parameters in the N-terminus are in the same range as or higher than average values found for residues in the β-barrel. Most measured values in the N-terminus range between 0.85 and 1, on the rigid end of the dynamics scale. Lower SCC values are found in residues Thr6, Leu10 in the helix kink, and Arg15. Thus, local mobility is present in the hVDAC1 N-terminus. However, our data show that, globally, the N-terminus is certainly not more floppy than the β-barrel on a sub-ms timescale. Peak broadening or doubling, which would indicate dynamics on slower timescales, is not observed, further confirming the well-defined structure of the hVDAC1 N-terminus (Schneider et al, 2010a; Villinger et al, 2010). However, it may be possible that other highly dynamic conformations not accessible to solid-state NMR experiments exist in equilibrium.

Fig. 1. (13C,13C) DQ-SQ order parameters (SCC) measured on hVDAC1 in lipid bilayers.

Fig. 1

Open in a new tab

All values refer to Cα-Cβ correlations except “Lys SC”, which corresponds to lysine Cδ-Cε sidechain correlations as a reference for mobile moieties. For comparison of mobility in the N-terminus and in the β-barrel, overall Cα-Cβ order parameters of alanine, leucine, threonine, and valine residues in β-strand conformation (whose resonances overlap) are also shown (labels “Aβ”, “Lβ”, “Tβ”, “Vβ”), as well as average Cα-Cβ order parameters for other β-strand residue types whose resonances overlap in two broad spectral regions (Cα shifts around 55 ppm, Cβ shifts around 35 ppm; labels “β1”, “β2”). The order parameter range from 0.85 to 1 is indicated by a gray shaded bar. A vertical dashed line separates SCC values of N-terminal residues (left) from values of residues in other parts of the molecule.

Additionally, on a timescale of about 300 ns, the dynamics of mouse VDAC1 (mVDAC1, see also SI Materials and Methods) was studied in MD simulations. mVDAC1, which differs from hVDAC1 in four amino acid residues, was chosen for MD simulations because of the availability of a high-resolution crystal structure (Ujwal et al, 2008). Figure 2A shows the backbone root mean square fluctuations (RMSF) of wild-type (WT-) mVDAC1 (black curve). RMSF minima correspond to the centers of β-strands 1-19. These define a baseline of maximal structural rigidity (cyan line in Fig. 2A), which is slightly raised across β1-6 and in β19. The N-terminal helix (red-shaded area) exhibits a low fluctuation level, comparable to the central residues of the β-strands. The RMSF of short inter-strand loops and some strand termini exceeds that of the body of the N-terminal helix by a factor of 2 to 3 (color coded in Fig. 2B). The MD results therefore show that the N-terminal helix is relatively rigid under equilibrium conditions on a timescale of up to hundreds of nanoseconds, in agreement with the NMR data.

Fig. 2. Flexibility and conformational changes of the mVDAC1 β-barrel.

Fig. 2

Open in a new tab

(A) RMSF distribution in WT- (black) and Δ(1-20)-mVDAC1 (red). Cyan: Baseline of minimal RMSF across the barrel. Inset: RMSD of WT- (black) and Δ(1-20)-mVDAC1 (red) with respect to the initial structure under a membrane surface pressure of -45 mN/m. (B) RMSF of mVDAC1 in MD simulations, color-coded on the structure from blue (low) to red (high). Parts of strands 9-18 are cut away for clarity (left panel). (C) Ellipticity of WT- (black), L10N- (cyan), and Δ(1-20)-mVDAC1 (red) without additional membrane pressure. (D) Partial collapse of the Δ(1-20)-mVDAC1 barrel under low uniaxial membrane stress (~ -10 mN/m) occurs preferentially along an axis approximately running through β-strands 1-9 rather than in the perpendicular direction. β-strands 1 to 19 are color-coded from blue to red.

Removal of the N-terminus induces structural instability

Recombinant WT-hVDAC1 exhibits a major open-state conductance of about 4 nS (Fig. S1) and a voltage-dependence of membrane current typical for VDAC channels (Benz, 1994; Colombini, 1989; Colombini, 2004). Deletion of most or all of the 20 N-terminal residues in Neurospora crassa VDAC (ncVDAC) has been shown to lead to more noisy recordings in electrophysiology, affect or abolish voltage gating, and reduce channel conductance (Popp et al, 1996; Runke et al, 2006). This also applies to hVDAC1 (De Pinto et al, 2008; Schneider et al, 2010a). Lipid bilayer measurements on a hVDAC1 N-terminal truncation variant (Δ(1-20)-hVDAC1) show noisy traces with no apparent voltage gating and a distribution of current steps with a broad maximum between G = 0.5 nS and G = 2.5 nS (Fig. S2). In solid-state NMR experiments on hVDAC1 in lipid bilayers, N-terminal truncation leads to disappearance of resonance signals from β-strand 9 which forms a hydrophobic contact with Leu10 in the N-terminal helix (Fig. S3) (Schneider et al, 2010a).

The findings of increased channel noise and disappearing resonance signals point to increased dynamics and reduced structural stability in N-terminally truncated hVDAC1. To test this hypothesis, we performed equilibrium MD simulations of Δ(1-20)-mVDAC1 in a dimyristoyl phosphatidylcholine (DMPC) lipid bilayer and compared its behavior to the wild-type. As Figure 2A shows, the overall fluctuation level is substantially raised in the deletion mutant (red curve). In particular, the difference in the RMSF baseline between β-strands 5-19 is remarkable, as it demonstrates that a major portion of the wild-type β-barrel had been stabilized by the presence of the N-terminal helix. Elliptic fits of the β-barrel scaffold demonstrate an increased tendency of Δ(1-20)-mVDAC1 to adopt a more elliptic geometry than the wild-type (Fig. 2C). The global breathing motion of the barrel (Villinger et al, 2010) is evident from fluctuations of the average ellipticity occurring on a timescale of 30-50 ns, both in Δ(1-20)- and WT-mVDAC1.

It has been demonstrated that tension of biological membranes can modulate the shape and function of membrane proteins (Perozo et al, 2002; Schmidt & MacKinnon, 2008). We therefore investigated the behavior of Δ(1-20)- and WT-mVDAC1 under low to medium membrane surface tension. A substantial difference between Δ(1-20)- and WT-mVDAC1 was observed. Figure 2A (inset) displays the development of the root mean square deviation (RMSD) from the initial structure at a membrane surface tension of -45 mN/m, applied isotropically in the x-y plane. These data show that Δ(1-20)-mVDAC1 is perturbed by membrane tension more strongly than the wild-type. Structural analysis revealed that this deviation results from a global transition of the β-barrel of the deletion mutant into a smaller, more elliptic shape (Fig. 2D). In WT-mVDAC1, this transition is largely inhibited by the N-terminal helix. Interestingly, in cases of uni-axial shear stress applied on the membrane, the transition is strongly direction dependent. The β-barrel of Δ(1-20)-mVDAC1 is much more susceptible to deformation along an axis passing through β-strands 1 and 9 than in the perpendicular direction connecting β-strands 4-14 (Fig. 2D). Substantial deformation along the former axis can already be induced by relatively small values of surface tension on the order of -10 mN/m. This magnitude of surface tension is in the range biologically relevant for, e.g., mechanosensitive channels (which are also voltage-dependent; (Martinac et al, 1987; Sukharev, 2002)) and close to values observed in thermal membrane undulations of planar lipid bilayers (Hirn et al, 1999). The wild-type N-terminal helix, running alongside of β-strands 10-17, is thus located at a position where it maximally stabilizes the barrel structure. The ellipticity of the β-barrel observed in low stress simulations converges to a value of about 0.5, which we will call a semi-collapsed state in the following. Note that this state corresponds to a broad ensemble of structures.

Electrophysiology calculations

We hypothesized that the elliptic deformation of the β-barrel seen in our simulations could provide an explanation for the changes in electrophysiology parameters observed experimentally in the N-terminal deletion mutant. MD simulations have been successfully used to characterize ion flux in bacterial porins (Pongprayoon et al, 2009; Tieleman & Berendsen, 1998). Using our newly developed computational electrophysiology scheme (Kutzner et al, 2011), we atomistically simulated ion flux through WT- and Δ(1-20)-mVDAC1 under transmembrane potential gradients close to experimental values (Fig. 3A, inset). From the individual fluxes of anions and cations, selectivity and conductance values of VDAC1 were calculated.

Fig. 3. Ion flux and conductance through WT and Δ(1-20)-mVDAC1.

Fig. 3

Open in a new tab

(A) Wild-type conductance, as measured (magenta) and calculated using computational electrophysiology (green). Blue: Δ(1-20)-mVDAC1 conductance in relation to elliptic distortion of the beta barrel scaffold. Ion selectivity of the respective states is shown as color code on the data points. The experimentally determined average conductance of the closed state of the wild-type is indicated by a red shaded bar. (B) Ion flux and selectivity of WT- (light colors) and Δ(1-20)-mVDAC1 (solid colors) for cations and anions at an ellipticity of 0.47 and a transmembrane potential elicited by a charge imbalance of 2 e- across the membrane.

In our simulations, WT-mVDAC1 displayed a conductance of G = 4.2 ± 0.2 nS (s.e.m.) in 1M KCl, in close agreement with the experimental value for WT-hVDAC1 (about 4 nS, see Figs. 3A and S1, Refs. (Abu-Hamad et al, 2009; Benz, 1994; Runke et al, 2006)). The ion selectivity for WT-mVDAC1 was calculated as the ratio of permeating anions (p-) and cations (p+), yielding p-/p+ = 1.7 ± 0.4 (s.e.m.) (Fig. 3B), also in good agreement with experiment (p-/p+ = 2.2). Removal of the N-terminal helix, in the absence of further structural changes, results in a larger pore; accordingly, a raised conductance of 6.4 ± 0.3 nS (s.e.m.) was observed for Δ(1-20)-mVDAC1 at a similar p-/p+ ratio of 1.6 ± 0.4 (s.e.m.). We then went on to examine conductance levels of the partially collapsed, elliptic states of Δ(1-20)-mVDAC1 formed under low to medium membrane tension.

Figure 3A shows the conductance of Δ(1-20)-mVDAC1 relative to the ellipticity of the β-barrel. Near ellipticity values of 0.4-0.5, a sharp transition is observed that reduces the conductance to 1.5-2 nS (Fig. 3A). The fact that a drop in conductance is only seen upon such a large deformation of the barrel suggests that a significant structural change is required to elicit the magnitude of channel conductance decrease detected experimentally. In terms of conductance, semi-collapsed states with an ellipticity around 0.5 may thus explain voltage-induced entry into sub-conductance states of VDAC1. Importantly, these states can also explain the magnitude of volume loss upon closure, which has been measured to be in the order of 104 Å3 (Zimmerberg & Parsegian, 1986). In our simulations, semi-collapsed states exhibit a volume reduction of about 1×104 Å3 compared to WT-hVDAC1.

As mentioned above, the average conductance measured experimentally for Δ(1-20)-hVDAC1 closely resembles that reported for the closed state of VDAC1 variants (Benz, 1994; Colombini, 1989; Colombini, 2004). However, Δ(1-20)-hVDAC1 exhibits slightly increased anion selectivity in lipid bilayer experiments (p-/p+ about 2.6 in Δ(1-20)-hVDAC1, versus 2.2 in WT-hVDAC1), while the closed state of VDAC1 is typically reported to be cation-selective (Benz, 1994; Choudhary et al, 2010; Colombini, 2004). Notably, this trend is similar to the one observed in the N-terminal deletion mutant Δ(3-20) of ncVDAC (p-/p+ of about 2, versus 1.3 for WT-ncVDAC; refs. (Popp et al, 1996; Runke et al, 2006)). Consequently, we investigated whether the selectivity of the semi-collapsed conformations of Δ(1-20)-mVDAC1 could explain those observations. We found that the ionic selectivity of the elliptic states is crucially dependent on the precise geometry of the pore. The charge state of the N-terminal residue of Δ(1-20)-mVDAC1 (Gly21), which protrudes into the channel at the center of the pore, plays a minor, additional role. Importantly, Met1 in the wild-type is situated distant from the pore center, such that WT-hVDAC1 does not have a charge at this position.

Most of the elliptic states display clear anion selectivity, with selectivity values around p-/p+=1.5. However, the ionic selectivity switches to non-selective or slightly cation selective states near a conductance of about 1.7 nS. In particular, semi-collapsed forms with an ellipticity near 0.47 exhibit a conductance of 1.7 ± 0.1 nS (s.e.m.), and these states are found to be cation-selective with p-/p+=0.7 when the N-terminal residue is uncharged, resembling the wild-type situation (Fig. 3B). This ellipticity is close to the convergence value of the deletion mutant under low membrane stress. Depending on pore geometry and charge, semi-collapsed states can thus account for the selectivity of both the N-terminal deletion mutant and the closed state of the full-length protein.

The L10N mutant displays intermediate behavior between WT and Δ(1-20)-hVDAC1

Our data so far have shown that the N-terminus plays a stabilizing role for the entire VDAC1 β-barrel and that the conformation of the barrel is closely related to its conductance and selectivity. In previous studies, the N-terminus was found to be attached to the barrel wall via a network of hydrogen bonds as well as a hydrophobic contact between Leu10 in the N-terminus and a hydrophobic patch involving residues Val143 and Leu150 in β-strands 9 and 10 (Hiller et al, 2008; Schneider et al, 2010a; Ujwal et al, 2008). To investigate the relative importance of these interactions and the effect of mutations destabilizing N-terminal attachment to the barrel, we chose exchanging Leu10 for asparagine as a conservative mutation that selectively disrupts the hydrophobic contact (Fig. 4). The pattern of resonances, including those from the N-terminus, is very well preserved in solid-state NMR spectra of L10N-hVDAC1, indicating correct folding of the mutant protein (Fig. S3). However, striking differences in intensity are visible in signals from the N-terminus (Fig. 4A) as well as from β-strand 9 (Fig. 4B). A systematic analysis of signal intensities and linewidths shows that signals from the N-terminus and β-strand 9 are selectively and significantly attenuated (Figs. 4C and S4), suggesting increased dynamics in these residues. Signal linewidths, on the other hand, increase slightly, but in a similar manner for the N-terminus and other regions (Fig. 4D). These data confirm the crucial role of the hydrophobic contact between Leu10 and the region around Val143 for attachment of the N-terminus to the β-barrel wall and point to an elevated level of structural or dynamical disorder specifically in N-terminus and β-strand 9 of L10N-hVDAC1.

Fig. 4. Solid-state NMR and electrophysiological data from L10N-hVDAC1.

Fig. 4

Open in a new tab

(A, B) Comparison of cross-peak intensities in WT- and L10N-hVDAC1. Shown are regions of (13C,13C) PDSD correlation spectra (15 ms mixing time) displaying resonances from N-terminus (A) and β-strand 9 (B). Blue, WT; red, L10N-hVDAC1. (C) Ratios of normalized Cα-Cβ cross-peak volumes between (13C,13C) PDSD spectra (15 ms mixing time) recorded on L10N- and WT-hVDAC1. In L10N-hVDAC1, normalized cross-peak volumes are significantly more attenuated in the N-terminus than in control residues 105 and 253 (p = 0.0002, two-sided t-test, unequal variances assumed). (D) Ratios of average Cα-Cβ cross-peak linewidths between 15 ms (13C,13C) PDSD spectra of L10N and WT hVDAC1. Average peak linewidths increase slightly in L10N hVDAC1 spectra, but not differentially for residues in the N-terminus and control residues (p = 0.22, two-sided t-test, unequal variances assumed). (E) Single-channel recordings of L10N-hVDAC1 inserted into lipid bilayers. Applied voltage was 10 mV. (F) Histogram of conductance values (G) observed on L10N-hVDAC1 in lipid bilayers at a transmembrane voltage of 10 mV.

Next, we analyzed the electrophysiological characteristics of L10N-hVDAC1 in bilayer measurements. Figure 4E shows that the mutant forms stable, voltage-gated pores. Importantly, however, already at a transmembrane voltage of 10 mV, where WT-hVDAC1 is predominantly open with a major conductance of about 4 nS (Fig. S1), L10N-hVDAC1 exhibits a dominant conductance of 2 nS (Figs. 4E, F), a value resembling the closed state of WT-hVDAC1 and the observable conductance state of the N-terminal deletion mutant Δ(1-20)-hVDAC1. Also, L10N-hVDAC1 shows a voltage-dependence of conductance which is intermediate to WT- and Δ(1-20)-hVDAC1 (Fig. 5). These observations are in line with the hypothesis that disruption of the hydrophobic contact Leu10-Val143 facilitates dynamical and/or conformational changes that allow entry of hVDAC1 into a closed state. Note, however, that, similar to Δ(1-20)-hVDAC1, L10N-hVDAC1 exhibits increased selectivity for anions over cations (p-/p+ about 3.2 in L10N-hVDAC1, versus 2.2 in WT-hVDAC1).

Fig. 5. Lipid bilayer conductance values of the different hVDAC1 variants investigated.

Fig. 5

Open in a new tab

Shown are ratios of the conductance G at a given membrane potential (Vm) divided by the conductance G0 at 10 mV as a function of the membrane potential Vm. Circles, WT-hVDAC1; triangles, L10N-hVDAC1; squares, Δ(1-20)-hVDAC1. The membrane potential always refers to the cis-side of the membrane. Means ± SD of three membranes are shown for each hVDAC1 variant.

In MD simulations of L10N-mVDAC1, no significant deviations from the behavior of WT-mVDAC1 were observed under equilibrium conditions (see, e.g., Fig. 2C, cyan curve). To emulate the effect of a transmembrane voltage on the N-terminus within the timescale accessible by MD, we investigated its behavior under forces exerted on charged residues by a transmembrane electric field raised by one to two orders of magnitude. The N-terminal helix has a surplus of two positive charges, rendering its position sensitive to an electric field. In force-probe MD simulations, we applied a constant force to the helix equivalent to that exerted by a transmembrane potential. In multiple 150 ns simulations in which a transmembrane voltage of about 2 V was modeled, the L10N-mutated helix detached from the mVDAC1 β-barrel with an average time constant of 66 ± 40 ns (s.e.m.). In contrast, no extraction of the N-terminal helix from the β-barrel scaffold was seen in WT-mVDAC1 at forces corresponding to this voltage. The wild-type helix could also be detached on a time-scale of a few hundreds of nanoseconds, but substantially higher forces corresponding to transmembrane voltages above 4 V were necessary. Thus, in experiment as well as simulation, the hydrophobic interaction between Leu10 and the hydrophobic patch around Val143 appears to be key to stabilizing the position of the wild-type N-terminal helix within the β-barrel, while the L10N mutant enters into a collapsed state more easily.

Discussion

One of the longest-standing questions related to the β-barrel channel VDAC is the nature of its gating process which can be induced by transmembrane voltages above ±30 mV and by apoptotic and anti-apoptotic proteins (Benz, 1994; Colombini, 1989; Colombini, 2004; Shoshan-Barmatz et al, 2010; Shoshan-Barmatz et al, 2008). In the present study, we aimed for mechanistic insight into the gating process, based on the atomic structure of VDAC.

In the first x-ray crystallographic study of a porin, it was noted that β-barrel channels should be deformable because of their hull-like architecture (Cowan et al, 1992). However, trimeric bacterial porins have a substantial hydrophobic core at their trimer interfaces, which explains their stability against denaturation and proteolysis (Cowan et al, 1992). The notion that β-barrel membrane proteins have a particularly high level of rigidity has thus entrenched over time, while it essentially stems from the stability of trimeric bacterial porins toward unfolding (Haltia & Freire, 1995; Wimley, 2003). Most trimeric porins show voltage-dependent gating in planar bilayer experiments at markedly raised transmembrane potentials (about ±100mV, Refs. (Dargent et al, 1986; Lakey & Pattus, 1989; Schindler & Rosenbusch, 1981)). In contrast to bacterial porins (Cowan et al, 1992; Zachariae et al, 2006), however, VDAC does not display a clearly localized area of intimate trimer contact (Bayrhuber et al, 2008; Ujwal et al, 2008). It was hypothesized earlier that this could enhance the dynamics of the entire barrel, which may in turn play a role in the propensity of VDAC to undergo gating (Mannella, 1997).

In this study, we first quantified and confirmed the overall rigid nature of the hVDAC1 N-terminus by measuring SCC order parameters of the backbone. Using MD simulations, we then developed a molecular description of the effects of N-terminal truncation on the VDAC1 β-barrel. In electrophysiology, deletion of the N-terminal helix is found to increase channel noise, reduce conductance, and most importantly, abolish voltage gating, while in solid-state NMR spectra, the deletion leads to loss of resonance signals also from residues within the barrel. Our results suggest that removal of the N-terminal helix results in a highly dynamic mutant VDAC1 channel which exhibits an increased propensity to enter semi-collapsed, elliptic states with conductance levels similar to those found in the closed state of the wild-type channel. These findings demonstrate not only the importance of the N-terminal helix in voltage gating, but also for stabilizing the open state of the pore by virtue of its rigidity. In addition, we found that this stabilizing effect hinges upon a major attachment point of the helix to the barrel wall at the hydrophobic Leu10-Val143 contact. Weakening of this contact in L10N-hVDAC1 leads to a phenotype characterized by a reduction of the gating voltage at which the channel enters predominantly into a closed state with a conductance of about 2 nS (Figs. 4E-F, 5).

Our observations suggest a scenario in which an ensemble of semi-collapsed, elliptic geometries of the VDAC1 β-barrel underlie the sub-conductance states observed experimentally under elevated transmembrane voltages. Our finding that large deformations of the VDAC1 β-barrel are required to account for the reduction in conductance observed experimentally is in excellent agreement with the observation of a volume reduction in the order of 104 Å3 during closure (Zimmerberg & Parsegian, 1986). Such a volume change would be inconsistent with a more local conformational change. Similarly, a recent VDAC1 study using continuum electrostatics calculations demonstrated that movement of the N-terminus alone, even if leading completely out of the β-barrel, is unlikely to account for VDAC1 gating (Choudhary et al, 2010). According to our data, such a movement of the N-terminus should also lead to destabilization of the barrel. Gating models involving only a movement of the N-terminus, without a concomitant effect on the barrel, thus appear unlikely. Conversely, our data are well in line with the observation of extensive dynamics and a pronounced breathing motion in the hVDAC1 β-barrel by both solution-state NMR and MD simulations (Villinger et al, 2010). A gating mechanism by semi-collapse of the pore is also consistent with the observed dependence of gating on osmotic pressure (Zimmerberg & Parsegian, 1986) and on the presence of non-lamellar lipids, which increase lateral pressure, in the surrounding membrane (Rostovtseva et al, 2006). Thus, a coherent mechanism underlying VDAC gating emerges that explains a wide range of experimental data.

Our simulations show that semi-collapsed barrel states can also account for the switch in ion selectivity observed upon gating, depending on pore geometry and charge. In our simulations of Δ(1-20)-mVDAC1, anion selectivity is preserved in most of the sub-conductance states, in agreement with experimental data. Δ(1-20)-mVDAC1 then switches to cation selectivity at barrel ellipticity values of about 0.47, where the barrel exhibits a conductance of about 2 nS, similar to the closed state of WT-hVDAC1.

Hence, our data are consistent with a voltage-dependent motion of the N-terminal helix, partially or fully detaching it from the β-barrel, as a possible mechanism to control entry of VDAC1 into the closed state (Fig. 6). More specifically, the attenuation of signals from the N-terminus in solid-state NMR spectra of the L10N-hVDAC1 mutant may be due to increased overall motion in this region or, alternatively, exchange between two states, namely the relatively rigid wild-type conformation and a more mobile or disordered N-terminus which is not visible in the solid-state NMR spectra. Notably, such a mobile population may well also be present in wild-type VDAC1, albeit to a smaller extent. The absence of chemical shift changes or linewidth increases in N-terminal resonances of L10N-hVDAC1 indicates that such exchange between different states of the N-terminus would have to be slow relative to the NMR chemical shift timescale (milliseconds), in agreement with electrophysiology data for the timescale of VDAC gating (Colombini, 1989). In our simulations, a force corresponding to a transmembrane voltage about two orders of magnitude higher than values used experimentally is sufficient to remove the N-terminus of hVDAC1 from the pore on a nanosecond timescale. Our results thus suggest that displacement of the charged N-terminal helix of hVDAC1 by transmembrane voltages in the range of tens of millivolts is possible on the timescale of gating. Previous accessibility studies have also indicated that it is in principle possible for the helix to leave the barrel (De Pinto et al, 1991; Guo et al, 1995).

Fig. 6. Suggested model of VDAC voltage-induced gating.

Fig. 6

Open in a new tab

At zero transmembrane potential, the VDAC pore (blue) most likely remains in the open state (upper panel, left). Increasing the membrane voltage beyond ± 30 mV (center) exerts a force on the N-terminal helix (red), which is attached to the barrel wall by the contact residue L10 (green). Detachment or removal of the N-terminal helix from the barrel wall at the L10-V143 contact leaves behind a labile, hull-like pore structure (right) which is more susceptible to undergo (semi-)collapse under membrane stress. At an ellipticity of 0.47, the semi-collapsed barrel geometry displays the conductance and ion selectivity found experimentally for the wild-type closed state (lower panel, right), while our calculations reproduce the wild-type open state values for non-collapsed structures containing the N-terminal helix (lower panel, left). Note that semi-collapse can also occur under membrane stress when the helix is not removed, yet with a smaller probability.

It is important to note, however, that our data do not imply that removal of the helix is required to obtain semi-collapsed conformations under membrane stress. Partial collapse may be achieved by conformational changes within the barrel alone, possibly also involving removal of β-strands from the barrel, as has long been hypothesized (Peng et al, 1992; Song et al, 1998). However, our data clearly show that destabilization of the N-terminus facilitates entry of VDAC1 into partially collapsed states which can explain the conductance and selectivity of the closed state. The exposed and charged nature of the N-terminus, its stabilizing role for the β-barrel, and the loss of voltage gating in its absence thus suggest that it functions as a switch in the gating process. Conformational changes in the N-terminus possibly involved in inducing voltage gating may however be more subtle than a full displacement, such as a partial movement, turning or unwinding of the helix (Fig. 6). Our data indicate that weakening the hydrophobic attachment of the N-terminus to the barrel via the Leu10-Val143 link may play a role in any such process.

The model we propose relies on observations made in a large number of experimental and computational studies. These – present and previous – data indicate that the N-terminal helix of VDAC serves as a voltage-dependent sensor and that the cylindrical β-barrel can undergo drastic conformational changes coupled to its membrane environment, in particular when the helix loses its rigid resting conformation (Mannella, 1997; Song et al, 1998; Thomas et al, 1993). The resulting elliptic pores show a wide range of subconductance states with varying selectivity including cation-selective conformations, in agreement with previous experimental data (Benz, 1994; Colombini, 1989; Colombini, 2004). Our model extends an earlier suggestion on the gating mechanism of VDAC, made before its atomic structure was available on the basis of electron microscopic images (Mannella, 1997; Mannella, 1998), and explains the long-standing observation that gating is pressure-dependent and involves a large volume change of the channel (Zimmerberg & Parsegian, 1986). The scenario we propose implies that the conductance state of VDAC is mechanosensitive, i.e. responds to changes in membrane osmotic pressure, as has indeed been observed (Zimmerberg & Parsegian, 1986).

Our results also shed new light on the structure and dynamics of β-barrel membrane proteins in general. We show that the geometry of a β-barrel can in fact be very labile and, as noted before, its dynamics can be sensitive to changes in the lipid environment (Villinger et al, 2010). This agrees well with the observation that VDAC gating can be influenced by osmotic pressure (Zimmerberg & Parsegian, 1986) as well as by lipids that modify membrane lateral pressure (Rostovtseva et al, 2006). Importantly, in many cases, the gating process in bacterial porins has also been reported to be sensitive to membrane pressure in addition to voltage (Lakey & Pattus, 1989; Le Dain et al, 1996). On a wider perspective, our results thus suggest that a closing mechanism by semi-collapse may be a general feature of β-barrel proteins.

Materials and Methods

Solid-state NMR

hVDAC1 was expressed, refolded, and purified according to the protocol described in SI Ref. (Engelhardt et al, 2007). For solid-state NMR measurements, the protein was reconstituted into DMPC liposomes at a protein/lipid ratio of 1/50 (mol/mol). Solid-state NMR experiments were conducted using 3.2 mm or 4 mm triple-resonance (1H,13C,15N) magic-angle spinning (MAS) probeheads at static magnetic fields of 18.8 T and 20.0 T (Bruker Biospin, Karlsruhe, Germany). Sample temperature was +5°C. Initial 1H-13C cross-polarization (CP) time was set to 600 µs. Typical proton field strength for 90° pulses and SPINAL64 (Fung et al, 2000) decoupling was 83 kHz. 13C-13C mixing was accomplished by proton-driven spin diffusion (PDSD) for 15 ms, using a MAS frequency of 10.6 kHz at a static field of 20.0 T. For measurement of dynamics, double quantum – single quantum (2Q,1Q) correlation spectra were recorded employing the SPC5 pulse sequence (Hohwy et al, 1999) as described previously (Schneider et al, 2010b), using a 13C field strength of 40 kHz at 8 kHz MAS and a static field of 18.8 T. Six different spectra with excitation and reconversion times each of 250, 400, 500, 600, 750, or 1000 µs were recorded. Spectra were processed in Topspin (Bruker Biospin, Karlsruhe, Germany) and analyzed using Sparky (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco).

Evaluation of (2Q,1Q) spectra for obtaining order parameters was done using custom-written scripts in MATLAB (The MathWorks, Natick, MA, USA) after importing processed spectra using the MatNMR (van Beek, 2007) add-on package. Cross-peak volumes were integrated to yield build-ups of cross-peak intensity. To extract order parameters from these, simulated build-ups for different order parameters (i.e. scaling factors of the dipolar 13C-13C coupling) were generated in an amino acid type-specific way using the GAMMA C++ program library (Smith et al, 1994) as described in ref. (Schneider et al, 2010b) and fit to the experimental build-ups using chi-square minimization. The order parameter used in the simulation yielding the smallest chi-square value in the fit was taken to be the order parameter of the corresponding experimental build-up. Data for which no good fits could be obtained (defined as a correlation coefficient between experimental data and simulation of less than 0.9 or a correlation coefficient between 0.9 and 0.95 together with a chi-square value above 20) were excluded from the analysis. Errors of order parameters were estimated based on Monte Carlo simulations, where estimates of the experimental error were obtained from absolute values of integrals in spectral noise regions, scaled by the size of the respective signal integration region. Extracted order parameters for build-ups that report on mobility of the same bond (e.g. CACB-CA and CACB-CB) were averaged if they differed, using standard error propagation or the difference of the two values, whichever was greater, as the error of the average.

For determining relative signal intensities in spectra of WT- and L10N-hVDAC1, PDSD spectra with 15 ms mixing time acquired and processed in the same fashion were used. Signal intensities were measured by integration of cross-peaks in Topspin (Bruker Biospin, Karlsruhe, Germany). For each signal in each spectrum, values from symmetry-related cross-peaks on both sides of the diagonal were added and normalized to the average value of four integrals within the same spectrum, using regions with large signal intensities containing overlapping resonances from different residue types (Ser and Leu Cα-Cβ as well as Lys Cγ-Cδ and Cδ-Cε resonances) to avoid systematic intensity variations in the signals used for normalization. Thus, only relative cross-peak intensities normalized within each spectrum were compared between spectra. This approach excluded effects of overall spectral intensity differences due to, e.g., differing amounts of labeled protein or different cross-polarization efficiencies. Cross-peaks exhibiting strong overlap with other signals were excluded from this analysis. This applied also to resonances Ala8 and Ala14, which appear well resolved in (13C,13C) PDSD spectra, but overlap with signals from outside the N-terminus, as can be seen in spectra from Δ(1-20)-hVDAC1 where spectral intensity can still be found at these positions. Signals that vanish in the spectrum of L10N-hVDAC1 (Tyr7, Leu144) were also excluded. Figures 4C and S4 show the ratios between these normalized peak intensity values between the two spectra.

For determining linewidths of cross-peaks, the same PDSD spectra of WT- and L10N-hVDAC1 were used as for measuring signal intensities. Linewidths were measured using the command ‘pe’ in Sparky (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco) and averaged over both spectral dimensions and both cross-peaks on each side of the diagonal. Only resolved cross-peaks were used. Values shown in Figure 4D are ratios between these averaged linewidth values in the two spectra.

Electrophysiology

The method used for the black lipid bilayer experiments has been described previously (Benz et al, 1978; Roos et al, 1982). Membranes were formed from a 1% (w/v) solution of diphytanoyl phosphatidylcholine (Avanti, Polar Lipids, Alabaster AL) in n-decane by painting onto a circular hole (surface area about 0.4 mm2) separating the two compartments of a Teflon cell. For standard single channel measurements, the Teflon chamber was filled with an unbuffered 1 M KCl solution. The temperature used was 20°C. The voltage across the membrane was applied through silver/silver chloride electrodes (with salt bridges) inserted into the aqueous compartments on both sides of the membrane. The membrane current was measured with a current amplifier (Keithley 427). The amplified signal was monitored with a storage oscilloscope and recorded on a strip chart recorder. hVDAC1 and its mutants were added in a concentration of about 10 ng/ml to one or both sides of the membranes because there was no obvious asymmetry of the pores. The channel-forming activity of the proteins was approximately the same.

The probability P(G) for the occurrence of a given conductivity unit in single-channel experiments (Figs. 4F, S1B, and S2B) was calculated by dividing the number of fluctuations with a given conductance increment by the total number of conductance fluctuations. The applied membrane potential was 10 mV. Histograms were derived from at least 100 conductance steps. For voltage-dependence measurements, the refolded recombinant hVDAC1 was added at a concentration of about 100 ng/ml to one or both sides of a black diphytanoyl-phosphatidylcholine/n-decane membrane. After about 20 min the reconstitution of porin channels into the membrane reached equilibrium. Then, different potentials were applied to the cis-side of the membrane starting with ±10 mV. These experiments were repeated with ±20 to ±80 mV in steps of ±10 mV.

The experiments were analyzed in the following way: the membrane conductance (G) as a function of voltage, Vm, was measured when the opening and closing of channels reached an equilibrium, i.e. after the exponential decay of the membrane current following the voltage step Vm. G was divided by the initial value of the conductance (G0, which was a linear function of the voltage) obtained immediately after the onset of the voltage. The data of Fig. 5 correspond to the symmetric voltage-dependence of WT-hVDAC1, the mutant L10N-hVDAC1, and the N-terminal deletion mutant Δ(1-20)-hVDAC1 (mean ± SD of three membranes).

Molecular Dynamics Simulations

All simulations were based on the crystal structure of murine VDAC1 (mVDAC1, pdb code 3EMN, ref. (Ujwal et al, 2008)), whose sequence is 99% identical with that of human VDAC1 (hVDAC1), differing in only four sequence positions (Thr55Asn, Met129Val, Ala160Ser, and Ile227Val). For equilibrium simulations, WT-mVDAC1, Δ(1-20)-mVDAC1, and L10N-mVDAC1 were each inserted in simulation boxes with a fully hydrated and equilibrated membrane consisting of 176 dimyristoylphosphatidylcholine (DMPC) molecules and about 13,000 water molecules using the tool g_membed (Wolf et al, 2010). The amber99sb force field was used for the protein and ions (Hornak et al, 2006), and parameters for DMPC were derived from Berger et al. (Berger et al, 1997). The solvent was modeled using the SPC/E water model (Berendsen et al, 1987). Water bond distances and angles were constrained using SETTLE (Miyamoto & Kollman, 1992) All simulations were carried out with the Gromacs simulation software, version 4 (Hess et al, 2008). Electrostatic interactions were calculated explicitly at a distance smaller than 1.0 nm, long-range electrostatic interactions were treated by particle-mesh Ewald summation at every step (Darden et al, 1993). Lennard-Jones interactions were calculated using a cutoff of 1.0 nm. The LINCS algorithm was employed to constrain all protein and lipid bonds (Hess et al, 1997). The simulation temperature was kept constant by weakly (t = 0.1 ps) coupling the system to a temperature bath of 320 K using the velocity rescale method (Bussi et al, 2007). The pressure was kept constant by semi-isotropic Berendsen coupling of the system to a pressure bath of 1 bar, separately for the xy- and for the z-direction (Berendsen et al, 1984). By employing a virtual sites model (Berendsen et al, 1999), an integration time-step of 4 fs was used. The combination of force-field and water model was selected based on the results of Hess and van der Vegt (Hess & van der Vegt, 2006). The protonation states of titratable residues of VDAC, especially that of the Glu73 side chain which points towards the lipid bilayer, were determined by Monte Carlo sampling (Beroza & Case, 1996) based on results from Poisson-Boltzmann calculations using MEAD (Bashford & Gerwert, 1992). According to these calculations, a neutral Glu73 side chain is most probable at pH 7. Therefore, we used the protonated form of Glu73 in our simulations (E730).

The ellipticity of the barrels was calculated by fitting an ellipse to the barrel Cα atoms in the x-y plane after alignment along the z axis according to the ratio (b-a)/a, where b and a are the minor and major axes of the ellipse, respectively. Simulation of ion flux through WT-mVDAC1 and Δ(1-20)-mVDAC1 and calculation of conductance and ion selectivity values was based on the computational electrophysiology method implemented in Gromacs 4.5 (Kutzner et al, 2011). In brief, transmembrane voltages were applied using a small charge imbalance across two lipid bilayers containing WT or mutant VDAC and separating two aqueous compartments, formed by duplication of the systems described above. The deterministic ion-interchange scheme as described in Kutzner et al. (Kutzner et al, 2011) was used. Conductance and selectivity were calculated from the channel cation and anion flux within 20-ns time slices. The transmembrane potential was determined by solving the Poisson equation during the same time windows, imposing equal voltage at the z boundaries of the simulation box (Tieleman & Berendsen, 1996). Values were averaged over the two channels.

In simulations of isotropic membrane surface tension, the pressure in the xy-plane was increased to 150 bar, while that in z direction was kept at 1 bar. This corresponds to a membrane surface tension of -45 mN/m. To model anisotropic membrane stress, a uniaxial pressure of 40 bar was applied in either x or y direction. Force probe simulations mimicking the effect of an electric field of ~6.7×108 V/m on the N-terminal helix, carrying a surplus of two positive charges, were performed by exerting a constant force of 120 pN on the Cα atom of sequence position 10, chosen as it is located halfway between the N-terminus of WT-VDAC1 and the Δ(1-20)-deletion.

Supplementary Material

Supplementary information is available at The EMBO Journal Online.

Supplementary Information

Acknowledgements

We thank Sergei Sukharev for helpful discussions and Carsten Kutzner for creating Fig. 6. Financial support by the Max Planck Society, the DFG (Emmy Noether Fellowship to A.L., SFB 487, SFB 803), the Rudolf Virchow Center for Experimental Medicine (R. Be.), and the NSERC of Canada (PhD fellowship to J.-P.D.) is gratefully acknowledged. Z.G. was supported by a Marie Curie fellowship within the 7th EU Framework Programme.

Footnotes

Author contributions

U.Z. performed and analyzed MD simulations. R.S. performed and analyzed ssNMR experiments. R.B. analyzed MD simulations. Z.G. and J.-P.D. performed ssNMR experiments. K.G. and S.B. produced VDAC samples. E.M. and R.B. performed electrophysiology studies. M.Z. and C.G. analyzed data. B.G. supervised the MD work and A.L. the ssNMR work.

Conflict of Interest

The authors declare that no conflict of interests exists.

References

  1. Abu-Hamad S, Arbel N, Calo D, Arzoine L, Israelson A, Keinan N, Ben-Romano R, Friedman O, Shoshan-Barmatz V. The VDAC1 N-terminus is essential both for apoptosis and the protective effect of anti-apoptotic proteins. J Cell Sci. 2009;122:1906–1916. doi: 10.1242/jcs.040188. [DOI] [PubMed] [Google Scholar]
  2. Ader C, Schneider R, Hornig S, Velisetty P, Wilson EM, Lange A, Giller K, Ohmert I, Martin-Eauclaire MF, Trauner D, Becker S, et al. A structural link between inactivation and block of a K+ channel. Nat Struct Mol Biol. 2008;15:605–612. doi: 10.1038/nsmb.1430. [DOI] [PubMed] [Google Scholar]
  3. Bashford D, Gerwert K. Electrostatic calculations of the pKa values of ionizable groups in bacteriorhodopsin. J Mol Biol. 1992;224:473–486. doi: 10.1016/0022-2836(92)91009-e. [DOI] [PubMed] [Google Scholar]
  4. Bayrhuber M, Meins T, Habeck M, Becker S, Giller K, Villinger S, Vonrhein C, Griesinger C, Zweckstetter M, Zeth K. Structure of the human voltage-dependent anion channel. Proc Natl Acad Sci USA. 2008;105:15370–15375. doi: 10.1073/pnas.0808115105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Benz R. Permeation of hydrophilic solutes through mitochondrial outer membranes: review on mitochondrial porins. Biochim Biophys Acta. 1994;1197:167–196. doi: 10.1016/0304-4157(94)90004-3. [DOI] [PubMed] [Google Scholar]
  6. Benz R, Janko K, Boos W, Läuger P. Formation of large, ion-permeable membrane channels by the matrix protein (porin) of Escherichia coli. Biochim Biophys Acta. 1978;511:305–319. doi: 10.1016/0005-2736(78)90269-9. [DOI] [PubMed] [Google Scholar]
  7. Berendsen HJC, Feenstra KA, Hess B. Improving efficiency of large time-scale molecular dynamics simulations of hydrogen-rich systems. J Comput Chem. 1999;20:786–798. doi: 10.1002/(SICI)1096-987X(199906)20:8<786::AID-JCC5>3.0.CO;2-B. [DOI] [PubMed] [Google Scholar]
  8. Berendsen HJC, Grigera JR, Straatsma TP. The missing term in effective pair potentials. J Phys Chem. 1987;91:6269–6271. [Google Scholar]
  9. Berendsen HJC, Postma JPM, Vangunsteren WF, Dinola A, Haak JR. Molecular dynamics with coupling to an external bath. J Chem Phys. 1984;81:3684–3690. [Google Scholar]
  10. Berger O, Edholm O, Jahnig F. Molecular dynamics simulations of a fluid bilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant temperature. Biophys J. 1997;72:2002–2013. doi: 10.1016/S0006-3495(97)78845-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Beroza P, Case DA. Including side chain flexibility in continuum electrostatic calculations of protein titration. J Phys Chem. 1996;100:20156–20163. [Google Scholar]
  12. Bussi G, Donadio D, Parrinello M. Canonical sampling through velocity rescaling. J Chem Phys. 2007;126:014101. doi: 10.1063/1.2408420. [DOI] [PubMed] [Google Scholar]
  13. Choudhary OP, Ujwal R, Kowallis W, Coalson R, Abramson J, Grabe M. The electrostatics of VDAC: implications for selectivity and gating. J Mol Biol. 2010;396:580–592. doi: 10.1016/j.jmb.2009.12.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Colombini M. Voltage gating in the mitochondrial channel, VDAC. J Membr Biol. 1989;111:103–111. doi: 10.1007/BF01871775. [DOI] [PubMed] [Google Scholar]
  15. Colombini M. VDAC: The channel at the interface between mitochondria and the cytosol. Mol Cell Biochem. 2004;256:107–115. doi: 10.1023/b:mcbi.0000009862.17396.8d. [DOI] [PubMed] [Google Scholar]
  16. Cowan SW, Schirmer T, Rummel G, Steiert M, Ghosh R, Pauptit RA, Jansonius JN, Rosenbusch JP. Crystal structures explain functional properties of two E. coli porins. Nature. 1992;358:727–733. doi: 10.1038/358727a0. [DOI] [PubMed] [Google Scholar]
  17. Darden T, York D, Pedersen L. Particle Mesh Ewald - an N.Log(N) method for Ewald sums in large systems. J Chem Phys. 1993;98:10089–10092. [Google Scholar]
  18. Dargent B, Hofmann W, Pattus F, Rosenbusch JP. The selectivity filter of voltage-dependent channels formed by phosphoporin (PhoE protein) from E. coli. EMBO J. 1986;5:773–778. doi: 10.1002/j.1460-2075.1986.tb04280.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. De Pinto V, Prezioso G, Thinnes F, Link TA, Palmieri F. Peptide-specific antibodies and proteases as probes of the transmembrane topology of the bovine heart mitochondrial porin. Biochemistry. 1991;30:10191–10200. doi: 10.1021/bi00106a017. [DOI] [PubMed] [Google Scholar]
  20. De Pinto V, Reina S, Guarino F, Messina A. Structure of the voltage dependent anion channel: state of the art. J Bioenerg Biomembr. 2008;40:139–147. doi: 10.1007/s10863-008-9140-3. [DOI] [PubMed] [Google Scholar]
  21. Engelhardt H, Meins T, Poynor M, Adams V, Nussberger S, Welte W, Zeth K. High-level expression, refolding and probing the natural fold of the human voltage-dependent anion channel isoforms I and II. J Membr Biol. 2007;216:93–105. doi: 10.1007/s00232-007-9038-8. [DOI] [PubMed] [Google Scholar]
  22. Fung BM, Khitrin AK, Ermolaev K. An improved broadband decoupling sequence for liquid crystals and solids. J Magn Reson. 2000;142:97–101. doi: 10.1006/jmre.1999.1896. [DOI] [PubMed] [Google Scholar]
  23. Guo XW, Smith PR, Cognon B, D'Arcangelis D, Dolginova E, Mannella CA. Molecular design of the voltage-dependent, anion-selective channel in the mitochondrial outer membrane. J Struct Biol. 1995;114:41–59. doi: 10.1006/jsbi.1995.1004. [DOI] [PubMed] [Google Scholar]
  24. Haltia T, Freire E. Forces and factors that contribute to the structural stability of membrane proteins. Biochim Biophys Acta. 1995;1241:295–322. doi: 10.1016/0304-4157(94)00161-6. [DOI] [PubMed] [Google Scholar]
  25. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM. LINCS: A linear constraint solver for molecular simulations. J Comput Chem. 1997;18:1463–1472. [Google Scholar]
  26. Hess B, Kutzner C, van der Spoel D, Lindahl E. GROMACS 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation. Journal of Chemical Theory and Computation. 2008;4:435–447. doi: 10.1021/ct700301q. [DOI] [PubMed] [Google Scholar]
  27. Hess B, van der Vegt NFA. Hydration thermodynamic properties of amino acid analogues: A systematic comparison of biomolecular force fields and water models. J Phys Chem B. 2006;110:17616–17626. doi: 10.1021/jp0641029. [DOI] [PubMed] [Google Scholar]
  28. Hiller S, Abramson J, Mannella C, Wagner G, Zeth K. The 3D structures of VDAC represent a native conformation. Trends Biochem Sci. 2010;35:514–521. doi: 10.1016/j.tibs.2010.03.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Hiller S, Garces RG, Malia TJ, Orekhov VY, Colombini M, Wagner G. Solution structure of the integral human membrane protein VDAC-1 in detergent micelles. Science. 2008;321:1206–1210. doi: 10.1126/science.1161302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Hirn R, Benz R, Bayerl TM. Collective membrane motions in the mesoscopic range and their modulation by the binding of a monomolecular protein layer of streptavidin studied by dynamic light scattering. Phys Rev E Stat Phys Plasmas Fluids Relat Interdiscip Topics. 1999;59:5987–5994. doi: 10.1103/physreve.59.5987. [DOI] [PubMed] [Google Scholar]
  31. Hohwy M, Rienstra CM, Jaroniec CP, Griffin RG. Fivefold symmetric homonuclear dipolar recoupling in rotating solids: Application to double quantum spectroscopy. J Chem Phys. 1999;110:7983–7992. [Google Scholar]
  32. Hornak V, Abel R, Okur A, Strockbine B, Roitberg A, Simmerling C. Comparison of multiple Amber force fields and development of improved protein backbone parameters. Proteins. 2006;65:712–725. doi: 10.1002/prot.21123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Hunte C, Richers S. Lipids and membrane protein structures. Curr Opin Struct Biol. 2008;18:406–411. doi: 10.1016/j.sbi.2008.03.008. [DOI] [PubMed] [Google Scholar]
  34. Kutzner C, Grubmüller H, de Groot BL, Zachariae U. Computational electrophysiology: the molecular dynamics of ion channel permeation and selectivity in atomistic detail. Biophys J. 2011;101:809–817. doi: 10.1016/j.bpj.2011.06.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Lakey JH, Pattus F. The voltage-dependent activity of Escherichia coli porins in different planar bilayer reconstitutions. Eur J Biochem. 1989;186:303–308. doi: 10.1111/j.1432-1033.1989.tb15209.x. [DOI] [PubMed] [Google Scholar]
  36. Le Dain AC, Häse CC, Tommassen J, Martinac B. Porins of Escherichia coli: unidirectional gating by pressure. EMBO J. 1996;15:3524–3528. [PMC free article] [PubMed] [Google Scholar]
  37. Lee KI, Rui H, Pastor RW, Im W. Brownian dynamics simulations of ion transport through the VDAC. Biophys J. 2011;100:611–619. doi: 10.1016/j.bpj.2010.12.3708. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Mannella CA. Minireview: on the structure and gating mechanism of the mitochondrial channel, VDAC. J Bioenerg Biomembr. 1997;29:525–531. doi: 10.1023/a:1022489832594. [DOI] [PubMed] [Google Scholar]
  39. Mannella CA. Conformational changes in the mitochondrial channel protein, VDAC, and their functional implications. J Struct Biol. 1998;121:207–218. doi: 10.1006/jsbi.1997.3954. [DOI] [PubMed] [Google Scholar]
  40. Martinac B, Buechner M, Delcour AH, Adler J, Kung C. Pressure-sensitive ion channel in Escherichia coli. Proc Natl Acad Sci U S A. 1987;84:2297–2301. doi: 10.1073/pnas.84.8.2297. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Miyamoto S, Kollman PA. Settle - an analytical version of the Shake and Rattle algorithm for rigid water models. J Comput Chem. 1992;13:952–962. [Google Scholar]
  42. Nietlispach D, Gautier A. Solution NMR studies of polytopic alpha-helical membrane proteins. Curr Opin Struct Biol. 2011;21:497–508. doi: 10.1016/j.sbi.2011.06.009. [DOI] [PubMed] [Google Scholar]
  43. Peng S, Blachly-Dyson E, Forte M, Colombini M. Large-scale rearrangement of protein domains is associated with voltage gating of the VDAC channel. Biophys J. 1992;62:123–135. doi: 10.1016/S0006-3495(92)81799-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Perozo E, Cortes DM, Sompornpisut P, Kloda A, Martinac B. Open channel structure of MscL and the gating mechanism of mechanosensitive channels. Nature. 2002;418:942–948. doi: 10.1038/nature00992. [DOI] [PubMed] [Google Scholar]
  45. Phillips R, Ursell T, Wiggins P, Sens P. Emerging roles for lipids in shaping membrane-protein function. Nature. 2009;459:379–385. doi: 10.1038/nature08147. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Pongprayoon P, Beckstein O, Wee CL, Sansom MS. Simulations of anion transport through OprP reveal the molecular basis for high affinity and selectivity for phosphate. Proc Natl Acad Sci U S A. 2009;106:21614–21618. doi: 10.1073/pnas.0907315106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Popp B, Court DA, Benz R, Neupert W, Lill R. The role of the N and C termini of recombinant Neurospora mitochondrial porin in channel formation and voltage-dependent gating. J Biol Chem. 1996;271:13593–13599. doi: 10.1074/jbc.271.23.13593. [DOI] [PubMed] [Google Scholar]
  48. Roos N, Benz R, Brdiczka D. Identification and characterization of the pore-forming protein in the outer membrane of rat liver mitochondria. Biochim Biophys Acta. 1982;686:204–214. doi: 10.1016/0005-2736(82)90114-6. [DOI] [PubMed] [Google Scholar]
  49. Rostovtseva TK, Kazemi N, Weinrich M, Bezrukov SM. Voltage gating of VDAC is regulated by nonlamellar lipids of mitochondrial membranes. J Biol Chem. 2006;281:37496–37506. doi: 10.1074/jbc.M602548200. [DOI] [PubMed] [Google Scholar]
  50. Rui H, Lee KI, Pastor RW, Im W. Molecular dynamics studies of ion permeation in VDAC. Biophys J. 2011;100:602–610. doi: 10.1016/j.bpj.2010.12.3711. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Runke G, Maier E, Summers WAT, Bay DC, Benz R, Court DA. Deletion variants of Neurospora mitochondrial porin: Electrophysiological and spectroscopic analysis. Biophys J. 2006;90:3155–3164. doi: 10.1529/biophysj.105.072520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Schindler H, Rosenbusch JP. Matrix protein in planar membranes: clusters of channels in a native environment and their functional reassembly. Proc Natl Acad Sci U S A. 1981;78:2302–2306. doi: 10.1073/pnas.78.4.2302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Schmidt D, MacKinnon R. Voltage-dependent K+ channel gating and voltage sensor toxin sensitivity depend on the mechanical state of the lipid membrane. Proc Natl Acad Sci U S A. 2008;105:19276–19281. doi: 10.1073/pnas.0810187105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Schneider R, Etzkorn M, Giller K, Daebel V, Eisfeld J, Zweckstetter M, Griesinger C, Becker S, Lange A. The native conformation of the human VDAC1 N terminus. Angew Chem Int Ed Engl. 2010a;49:1882–1885. doi: 10.1002/anie.200906241. [DOI] [PubMed] [Google Scholar]
  55. Schneider R, Seidel K, Etzkorn M, Lange A, Becker S, Baldus M. Probing molecular motion by double-quantum (13C,13C) solid-state NMR spectroscopy: application to ubiquitin. J Am Chem Soc. 2010b;132:223–233. doi: 10.1021/ja906283h. [DOI] [PubMed] [Google Scholar]
  56. Shoshan-Barmatz V, De Pinto V, Zweckstetter M, Raviv Z, Keinan N, Arbel N. VDAC, a multi-functional mitochondrial protein regulating cell life and death. Mol Aspects Med. 2010;31:227–285. doi: 10.1016/j.mam.2010.03.002. [DOI] [PubMed] [Google Scholar]
  57. Shoshan-Barmatz V, Keinan N, Zaid H. Uncovering the role of VDAC in the regulation of cell life and death. J Bioenerg Biomembr. 2008;40:183–191. doi: 10.1007/s10863-008-9147-9. [DOI] [PubMed] [Google Scholar]
  58. Smith SA, Levante TO, Meier BH, Ernst RR. Computer simulations in magnetic resonance. An object-oriented programming approach. Journal of Magnetic Resonance, Series A. 1994;106:75–105. [Google Scholar]
  59. Song J, Midson C, Blachly-Dyson E, Forte M, Colombini M. The sensor regions of VDAC are translocated from within the membrane to the surface during the gating processes. Biophys J. 1998;74:2926–2944. doi: 10.1016/S0006-3495(98)78000-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Sukharev S. Purification of the small mechanosensitive channel of Escherichia coli (MscS): the subunit structure, conduction, and gating characteristics in liposomes. Biophys J. 2002;83:290–298. doi: 10.1016/S0006-3495(02)75169-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Thomas L, Blachly-Dyson E, Colombini M, Forte M. Mapping of residues forming the voltage sensor of the voltage-dependent anion-selective channel. Proc Natl Acad Sci U S A. 1993;90:5446–5449. doi: 10.1073/pnas.90.12.5446. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Tieleman DP, Berendsen HJ. A molecular dynamics study of the pores formed by Escherichia coli OmpF porin in a fully hydrated palmitoyloleoylphosphatidylcholine bilayer. Biophys J. 1998;74:2786–2801. doi: 10.1016/S0006-3495(98)77986-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Tieleman DP, Berendsen HJC. Molecular dynamics simulations of a fully hydrated dipalmitoyl phosphatidylcholine bilayer with different macroscopic boundary conditions and parameters. J Chem Phys. 1996;105:4871–4880. [Google Scholar]
  64. Ujwal R, Cascio D, Colletier JP, Faham S, Zhang J, Toro L, Ping PP, Abramson J. The crystal structure of mouse VDAC1 at 2.3 angstrom resolution reveals mechanistic insights into metabolite gating. Proc Natl Acad Sci USA. 2008;105:17742–17747. doi: 10.1073/pnas.0809634105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. van Beek JD. matNMR: A flexible toolbox for processing, analyzing and visualizing magnetic resonance data in Matlab((R)) J Magn Reson. 2007;187:19–26. doi: 10.1016/j.jmr.2007.03.017. [DOI] [PubMed] [Google Scholar]
  66. Villinger S, Briones R, Giller K, Zachariae U, Lange A, de Groot BL, Griesinger C, Becker S, Zweckstetter M. Functional dynamics in the voltage-dependent anion channel. Proc Natl Acad Sci U S A. 2010;107:22546–22551. doi: 10.1073/pnas.1012310108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Wimley WC. The versatile beta-barrel membrane protein. Curr Opin Struct Biol. 2003;13:404–411. doi: 10.1016/s0959-440x(03)00099-x. [DOI] [PubMed] [Google Scholar]
  68. Wolf MG, Hoefling M, Aponte-Santamaria C, Grubmuller H, Groenhof G. g_membed: Efficient insertion of membrane protein into an equilibrated lipid bilayer with minimal perturbation. J Comput Chem. 2010;31:2169–2174. doi: 10.1002/jcc.21507. [DOI] [PubMed] [Google Scholar]
  69. Zachariae U, Kluhspies T, De S, Engelhardt H, Zeth K. High resolution crystal structures and molecular dynamics studies reveal substrate binding in the porin Omp32. J Biol Chem. 2006;281:7413–7420. doi: 10.1074/jbc.M510939200. [DOI] [PubMed] [Google Scholar]
  70. Zaid H, Abu-Hamad S, Israelson A, Nathan I, Shoshan-Barmatz V. The voltage-dependent anion channel-1 modulates apoptotic cell death. Cell Death Differ. 2005;12:751–760. doi: 10.1038/sj.cdd.4401599. [DOI] [PubMed] [Google Scholar]
  71. Zimmerberg J, Parsegian VA. Polymer inaccessible volume changes during opening and closing of a voltage-dependent ionic channel. Nature. 1986;323:36–39. doi: 10.1038/323036a0. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information
NIHMS74568-supplement-Supplementary_Information.doc (1MB, doc)

RESOURCES