The antiarrhythmic compound efsevin directly modulates voltage‐dependent anion channel 2 by binding to its inner wall and enhancing mitochondrial Ca2+ uptake

Background and Purpose The synthetic compound efsevin was recently identified to suppress arrhythmogenesis in models of cardiac arrhythmia, making it a promising candidate for antiarrhythmic therapy. Its activity was shown to be dependent on the voltage‐dependent anion channel 2 (VDAC2) in the outer mitochondrial membrane. Here, we investigated the molecular mechanism of the efsevin–VDAC2 interaction. Experimental Approach To evaluate the functional interaction of efsevin and VDAC2, we measured currents through recombinant VDAC2 in planar lipid bilayers. Using molecular ligand‐protein docking and mutational analysis, we identified the efsevin binding site on VDAC2. Finally, physiological consequences of the efsevin‐induced modulation of VDAC2 were analysed in HL‐1 cardiomyocytes. Key Results In lipid bilayers, efsevin reduced VDAC2 conductance and shifted the channel's open probability towards less anion‐selective closed states. Efsevin binds to a binding pocket formed by the inner channel wall and the pore‐lining N‐terminal α‐helix. Exchange of amino acids N207, K236 and N238 within this pocket for alanines abolished the channel's efsevin‐responsiveness. Upon heterologous expression in HL‐1 cardiomyocytes, both channels, wild‐type VDAC2 and the efsevin‐insensitive VDAC2AAA restored mitochondrial Ca2+ uptake, but only wild‐type VDAC2 was sensitive to efsevin. Conclusion and Implications In summary, our data indicate a direct interaction of efsevin with VDAC2 inside the channel pore that leads to modified gating and results in enhanced SR‐mitochondria Ca2+ transfer. This study sheds new light on the function of VDAC2 and provides a basis for structure‐aided chemical optimization of efsevin.


| INTRODUCTION
Cardiovascular diseases represent the primary cause of death and hospitalization worldwide (Benjamin et al., 2018). Especially, Cardiac arrhythmias are difficult to treat due to major side effects of common antiarrhythmic drugs. It is thus a major focus of cardiovascular research to identify novel, safer therapies for cardiac arrhythmia.
Using a chemical suppressor screen on the zebrafish cardiac arrhythmia model tremblor (Ebert et al., 2005;Langenbacher et al., 2005), we have previously identified the synthetic compound efsevin, a dihydropyrrole carboxylic ester compound, which potently restored rhythmic cardiac contractions in otherwise fibrillating zebrafish embryonic hearts (Shimizu et al., 2015). We further demonstrated efficacy of efsevin in translational models for catecholaminergic polymorphic ventricular tachycardia (CPVT;Schweitzer et al., 2017). Here, efsevin reduced episodes of tachycardia in vivo in CPVT mice and suppressed arrhythmogenic events in induced pluripotent stem cell-derived cardiomyocytes from a CPVT patient (Schweitzer et al., 2017). These findings make efsevin a promising lead structure for human antiarrhythmic therapy.
In a pull-down assay with immobilized efsevin, the outer mitochondrial membrane (OMM) voltage-dependent anion channel 2 (VDAC2) was identified as the primary molecular target of efsevin (Shimizu et al., 2015). Voltage-dependent anion channels are large pore-forming proteins in the outer mitochondrial membrane. They represent the main pathway for ions and metabolites over the outer mitochondrial membrane. Three isoforms of voltage-dependent anion channels are expressed in vertebrates out of which VDAC2 was described to have a specific role in the heart. While a global knockout of VDAC2 in mice is embryonically lethal (Cheng, Sheiko, Fisher, Craigen, & Korsmeyer, 2003), a conditional heart-specific VDAC2 knockout mouse was reported to develop post-natal cardiac defects and to die shortly after birth (Raghavan, Sheiko, Graham, & Craigen, 2012). In cardiomyocytes, VDAC2 was described to interact with the ryanodine receptor (RyR; Min et al., 2012) and to modulate cytosolic Ca 2+ signals (Shimizu et al., 2015;Subedi et al., 2011). In arrhythmic tremblor zebrafish embryos, efsevin induced a restoration of rhythmic cardiac contractions. Transient knock-down of VDAC2 abolished the efsevin induced phenotype restoration while overexpression of VDAC2 recovered it. In cultured cells, efsevin enhanced uptake of Ca 2+ into mitochondria. A direct link between the enhanced mitochondrial Ca 2+ uptake and efsevin's anti-arrhythmic properties was established by two lines of experiments: (a) Pharmacological inactivation of mitochondrial Ca 2+ uptake by Ru360, an inhibitor of the mitochondrial Ca 2+ uniporter (MCU), abolished efsevin's protective effect in CPVT cardiomyocytes and (b) kaempferol, an activator of the mitochondrial Ca 2+ uniporter, reduced arrhythmogenic Ca 2+ signals comparable to efsevin (Schweitzer et al., 2017).
However, biophysical and structural determinants of the efsevin-VDAC2 interaction have remained unexplored. It is still unclear whether efsevin directly interacts with the VDAC2 peptide and thereby modulates the electrophysiological properties of the channel or if the observed effects require a yet unidentified protein partner. To address this issue, we expressed and purified recombinant zebrafish VDAC2 (zVDAC2) protein and inserted it into planar lipid bilayers. We found a pronounced effect of efsevin on channel gating and opening probability, indicating a direct effect of efsevin on the channel. To analyse the structural basis of the efsevin VDAC2 interaction, we performed computational proteinligand docking using the crystal structure of zVDAC2 (Schredelseker et al., 2014). VDAC2 is formed by a barrel-like structure consisting of 19 antiparallel β-sheets (β-sheets 1-19) and an N-terminal α-helix lining the inner channel wall (Schredelseker et al., 2014). We identified an efsevin binding site located in a groove between the inner channel wall and the pore-lining α-helix and formed by hydrogen bonds and hydrophobic interactions between efsevin and the zVDAC2 peptide. Replacement of three residues (N207, K236 and N238) from this binding site with alanines (zVDAC2 AAA ) resulted in a complete loss of efsevin sensitivity. To evaluate the physiological consequence of the observed efsevin-induced biophysical changes in VDAC2, we heterologously expressed wild-type zVDAC2 and the efsevin-insensitive zVDAC2 AAA mutant in cultured HL-1 cardiomyocytes. We demonstrate that the observed changes in zVDAC2 electrophysiology translate into enhanced mitochondrial Ca 2+ uptake and thereby explain the antiarrhythmic effect of efsevin. Our data provide novel insights into VDAC2 function and provide a basis for structureaided chemical optimization of efsevin as a lead structure for the development of novel antiarrhythmic drugs.

What is already known
• The synthetic compound efsevin suppresses arrhythmogenesis in cardiac arrhythmia models.

What this study adds
• Efsevin binds into a pocket formed by the channel wall and the pore-lining helix.
• Efsevin facilitates VDAC2 gating into a less anionselective state and enhances Ca 2+ flux.
What is the clinical significance • Identification of mode of action and binding pocket allows for future structure aided-drug optimization.

| Expression and purification of zVDAC2
Expression and purification of zVDAC2 were performed as previously described (Schredelseker et al., 2014) with minor modifications: Histidine-tagged zVDAC2 was purified on an Äkta pure chromatography system using a HisPrep FF 16/10 column. After refolding dialysis, size exclusion chromatography was performed on a HiLoad 16/600 Superdex200 pg column equilibrated with 150-mM NaCl, 1-mM DTT, 0.1% LDAO, and 20-mm Tris-HCl (pH = 8.0). Integrity of the purified zVDAC2 was confirmed by SDS-PAGE analysis on a 12% gel.

| Planar lipid bilayer recordings
Painted planar lipid bilayers were formed on an Ionovation Bilayer Explorer lipid bilayer set-up across a 120-μm diameter opening with 1,2-diphytanoyl-sn-glycero-3-phosphatidylcholine (DPhPC) dissolved in n-decane at 12.5 mgÁml −1 . Both chambers were filled with 1-M KCl, 5-mM CaCl 2 , and 10-mM Tris-HCl (pH = 7.2) To facilitate insertion, purified zVDAC2 was inserted into lipidic bicelles (Ujwal, 2012). After formation of a stable bilayer, zVDAC2 containing bicelles were added to the cis chamber. After insertion of a channel, the membrane was clamped to 0 mV, and 10 s pulses to test potentials from −60 mV to +60 mV were applied. The signal was sampled at 10 kHz and filtered at 2 kHz. Efsevin (50-μM stock in recording solution) was added to the cis chamber to a final concentration of 8 μM. Data analysis was performed using Nest-o-Patch (Dr. V. Nesterov, https://sourceforge. netq/projects/nestopatch/). zVDAC2 ion selectivity was measured using folded planar membranes formed from opposition of two monolayers made of 5 mgÁml −1 solution of DPhPC in pentane, as described (Rostovtseva, Gurnev, Chen, & Bezrukov, 2012). Recordings were performed in the buffer described above. Channel insertion was achieved by adding zVDAC2 in a 2.5% Triton X-100 solution to the cis compartment while stirring.
After single channel was inserted in symmetrical solutions, the cis side was perfused with Tris-HCl (pH 7.2) buffer solution to achieve 0.2-M KCl. The exact KCl concentration of 0.2 M in the cis compartment after perfusion was verified at the end of each selectivity experiment using a conductivity meter CDM230 (Radiometer analytical). Ion selectivity of zVDAC2 conductance states was calculated from the reversal potential (V rev ). Permeability ratio between Cl − , and K + , I − /I + , was calculated according to the Goldman-Hodgkin-Katz equation (Hille, 2001): respectively (Lide, 2006).

| Molecular cloning
To introduce N207A, K236A and N238A into pQE60-zVDAC2 for recombinant expression and purification construct pQE60-zVDAC2 AAA was created by two mutagenesis PCR reactions using pQE60-zVDAC2 (Schredelseker et al., 2014) as a template. The two PCR fragments were fused by overlap extension PCR and the resulting product was ligated into pQE60-zVDAC2 with PstI and NheI.
For the creation of pCClc-CMV-zVDAC2 AAA -IRES-nlsEGFP, zVDAC2 was exchanged for zVDAC2 AAA in pCS2+-zVDAC2 (Shimizu et al., 2015) by PCR amplification of the zVDAC2 AAA open reading frame from PQE60-zVDAC2 AAA following ligation with BamHI and ClaI. The open reading frame of zVDAC2 AAA was PCR-amplified from pCS2+-zVDAC2 AAA fused to the IRES PCR-amplified from pCClc-CMV-zVDAC2-IRES-nlsEGFP by SOE PCR and the resulting fragment was fused into pCClc-CMV-zVDAC2-IRES-nlsGFP using the In-Fusion HD Cloning Kit (TaKaRa).

| HL-1 culture and creation of cell lines
HL-1 cells were cultured as described previously (Claycomb et al., 1998).

| Metabolic stability assay
Efsevin (10 mM in DMSO) was co-incubated with human liver microsomes at 37 C at an initial concentration of 1 μM. The reaction was initiated by addition of 1-mM NADPH; 0, 5, 10, 30, and 60 minutes after starting the reaction, small aliquots were transferred into icecold acetonitrile and centrifuged at 16000× g for 10 min. Supernatants were analysed by LC-MS/MS. Experiments were performed as contract research at 3D BioOptima Co., Ltd., Suzhuo, China.

| Data and statistical analysis
The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018). zVDAC2 channels were inserted into lipid bilayers and measured before and after addition of efsevin in recording solution, thus leading groups of equal size. In some instances,

| Materials
Efsevin was synthesized as described before (Henry et al., 2014

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), PHARMACOLOGY.

| Efsevin reduces currents through zVDAC2
Recombinant zVDAC2 protein was purified from Escherichia coli as previously described (Schredelseker et al., 2014) and inserted into diphytanoylphosphatidylcholine (DPhPC) planar lipid bilayers. Currents were measured in response to 10 s test pulses to potentials from −60 to +60 mV in 1-M KCl. As shown in Figure 1a, typical average currents of 41 ± 3pA were recorded at 10 mV, where the channel almost exclusively resides in its open state (Colombini, 1989;Schredelseker et al., 2014). VDAC-typical flattening of the currentvoltage relationship (I-V curve) was observed starting at approximately ±30 mV (Figure 1b), where the channel starts gating until finally being preferentially in its closed state at potentials above +50 mV or below −50 mV (Figure 1a (c) Conductance-voltage relationship of zVDAC2 before (n = 9 individual channels, black circles) and after addition of efsevin (n = 6 individual channels, red triangles, Unpaired Student's t-test) relationship that is characteristic of VDACs (Figure 1c).  (Colombini, 1989;Guardiani et al., 2018;Menzel et al., 2009;Mertins et al., 2012). Under  (Waterhouse et al., 2018) since no experimentally determined structure for hVDAC2 is currently available. In eight out of 10 independent docking experiments, the conformation with the lowest binding energy was found in the same binding pocket like in zVDAC2 indicating that this binding pocket is conserved among species ( Figure S2).
3.4 | Elimination of residues N207, K236 and N238 abolishes the efsevin sensitivity of zVDAC2 To confirm the predicted efsevin binding site by mutational analysis, we created a zVDAC2 mutant in which residues N207, K236 and N238 were substituted by alanine residues, zVDAC2 N207A/K236A/N238A (zVDAC2 AAA ). Lipid bilayer recordings of this mutant showed that it forms a functional channel with a similar conductance-voltage relationship, single channel conductance and P O compared to wild-type zVDAC2 ( Figure S3). Most strikingly however, the channel was insensitive to efsevin, demonstrated by comparable values for the conductance-voltage relationship, single channel conductance and P O before and after addition of efsevin (Figure 4).

| Efsevin mediates SR-mitochondria Ca 2+ transfer by binding to the N207/K236/N238 binding site
To evaluate if the observed efsevin-induced electrophysiological changes can explain the enhanced mitochondrial Ca 2+ uptake observed previously for HeLa cells and HL-1 cardiomyocytes (Schweitzer et al., 2017;Shimizu et al., 2015), we developed a heterologous expression system for zVDAC2. To this aim, we created a stable HL-1 cardiomyocyte line in which the endogenous mouse VDAC2 (mVDAC2) was knocked down by stable expression of shRNA (Subedi et al., 2011; Figure S4) and overexpressed shRNA-insensitive zVDAC2 constructs by lentiviral transduction. Ca 2+ uptake into mitochondria upon caffeine-induced Ca 2+ release from the sarcoplasmic reticulum (SR) was then measured in permeabilized Rhod-2 stained cells ( Figure   5). In line with previous experiments, knock-down of the endogenous mVDAC2 eliminated transfer of Ca 2+ from the sarcoplasmic reticulum into mitochondria (Subedi et al., 2011). While wild-type HL-1 cardiomyocytes displayed maximum ΔF/F 0 values of 0.14 ± 0.02 that were enhanced to 0.38 ± 0.03 by 10-μM efsevin, shRNA-expressing cells displayed ΔF/F 0 max values of 0.04 ± 0.03, which were indistinguishable from those obtained from cells treated with the mitochondrial Ca 2+ uptake blocker ruthenium red (RuR; ΔF/F 0 max = 0.06 ± 0.02) or efsevin (ΔF/F 0 max = 0.06 ± 0.01, not significant). Though lack of VDAC2 was previously described to induce apoptosis and we can thus not rule out downstream effects induced by mVDAC2 knock down in these cells, the shmVDAC2 cell line was stable for the time of our experiments and thus suitable for heterologous overexpression experiments. Overexpression of zVDAC2 completely restored sarcoplasmic reticulum-mitochondria Ca 2+ transfer to ΔF/F 0 max = 0.14 ± 0.02, which was sensitive to modulation by efsevin (ΔF/F 0 max = 0.34 ± 0.02). Strikingly, zVDAC2 AAA likewise restored mitochondrial Ca 2+ uptake to levels indistinguishable from wild-type zVDAC2 (ΔF/F 0 max = 0.15 ± 0.02) but was insensitive to treatment with efsevin (ΔF/F 0 max = 0.16 ± 0.02). These data strongly indicate that the efsevin induced changes in zVDAC2 electrophysiology account for the enhanced mitochondrial Ca 2+ uptake in cardiomyocytes.

| DISCUSSION
Here, we describe the efsevin-zVDAC2 interaction at a biophysical and structural level. Efsevin was previously shown to enhance mitochondrial Ca 2+ uptake and to be a powerful modulator of cardiac rhythmicity and a potent suppressor of cardiac arrhythmia. Efsevin suppresses arrhythmogenesis in both, models for Ca 2+ overload and models for inherited arrhythmias like catecholaminergic polymorphic  (Subedi et al., 2011), was identified as the primary target of efsevin (Shimizu et al., 2015). These findings make VDAC2-mediated mitochondrial Ca 2+ uptake a promising therapeutic target for the development of a novel class of antiarrhythmic drugs with efsevin as a lead candidate. However, efsevin was identified in a chemical screen using 168 newly synthesized diversity-oriented compounds (Shimizu et al., 2015) without further compound optimization. In our HL-1-based sarcoplasmic reticulum-mitochondria Ca 2+ transfer assay, efsevin showed a half-maximal activity at 2.2 μM ( Figure S5A). Preliminary results on efsevin's pharmacokinetics show that it is hydrolysed rapidly in liver microsomes ( Figure S5B). Furthermore, binding of efsevin to other targets, including VDAC isoforms 1 and 3, was never tested. Thus, its unknown selectivity, the low stability of efsevin and the relatively high EC 50 value indicate the need for chemical optimization of the compound before further preclinical and clinical studies are performed. In this study, we present a molecular model of the efsevin binding site on zVDAC2 and thereby provide a basis for structure-based drug optimization in future experiments.
When discussing VDAC2 as a potential drug target for cardiac arrhythmia, it should be noted that the transfer of Ca 2+ from the sarcoplasmic reticulum into mitochondria is most likely a specialized role of VDAC2 in cardiomyocytes, presumably accomplished by a functional or even physical coupling to the ryanodine receptor (Min et al., 2012;Shimizu et al., 2015;Subedi et al., 2011) and might be less relevant or maybe accomplished by other VDAC isoforms in other cell types. In fact, VDAC1 was previously reported to promote Ca 2+ transfer from the endoplasmic reticulum into mitochondria in non-excitable cells through coupling to the IP 3 receptor (De Stefani et al., 2012;Rapizzi et al., 2002;Szabadkai et al., 2006). It is thus conceivable that despite the ubiquitous expression of VDACs in the body, efsevinmediated effects are most pronounced or even limited to cardiomyocytes due to the specialized role of VDAC2 in this tissue.
This effect might also explain the lack of major side effects that were observed previously in translational models (Schweitzer et al., 2017).
However, future studies are needed to evaluate the role of the efsevin-mediated effects on other VDAC2-mediated effects like F I G U R E 5 Sarcoplasmic reticulum (SR)-mitochondria Ca 2+ transfer upon heterologous expression of zVDAC2 and zVDAC2 AAA in HL-1 cardiomyocytes. (a) Representative recordings of mitochondrial Ca 2+ upon application of 10-mM caffeine to induce SR calcium release from permeabilized HL-1 cardiomyocytes. Traces from control conditions (black), recordings in the presence of 10-μM ruthenium red to block mitochondrial Ca 2+ uptake (RuR, grey) and in the presence of 10-μM efsevin (red) are shown for native HL-1 cells (native), cells transduced with shRNA targeting the endogenous mouse mVDAC2 (shmVDAC2), and cells overexpressing zVDAC2 and zVDAC2 AAA , respectively. (b) Statistical analysis of SR-mitochondria Ca 2+ transfer experiments. While native HL-1 cardiomyocytes showed an efsevin-sensitive uptake of Ca 2+ into mitochondria (n = 21 for control, n = 7 for RuR, and n = 18 for efsevin), this uptake was abolished upon knock-down of the endogenous mVDAC2 (shmVDAC2, n = 24 for control, n = 8 for RuR, n = 15 for efsevin). Subsequent heterologous expression of zVDAC2 (shmVDAC2, n = 21 for control, n = 7 for RuR, n = 15 for efsevin) and zVDAC2 AAA (shmVDAC2 AAA , n = 18 for control, n = 6 for RuR, n = 18 for efsevin) revealed restoration of SR-mitochondria Ca 2+ transfer. However, only zVDAC2 but not zVDAC2 AAA was sensitive to efsevin (Kruskal-Wallis test with Dunn's post hoc test) apoptosis or regulation of bioenergetics (for a review, see Naghdi & Hajnóczky, 2016). Furthermore, although VDAC2 was identified as the primary target of efsevin and efsevin-induced antiarrhythmic effects were shown to depend on VDAC2 (Shimizu et al., 2015), interaction of efsevin with other VDAC isoforms were never investigated. Although these data do not rule out the possibility that efsevin promotes its effects on the two distinct channels differently, it argues against it.
Our data indicate that efsevin binds to the inner wall of the β-barrel in close proximity to the N-terminal α-helix and thereby affects channel conductance and gating. Although still not fully resolved, various models were proposed to explain gating of VDACs.
Almost all models include a role of the N-terminal α-helix ranging from relatively small movements of the helix inside the barrel (Mertins et al., 2012;Shuvo, Ferens, & Court, 2016) to a large movement that place the helix outside of the barrel, which then collapses to induce channel closure (Choudhary et al., 2010;Zachariae et al., 2012). Furthermore, a fixation of the helix inside the pore through disulfate bridges was reported as a possible mechanism to regulate channel gating by redox sensing (Okazaki et al., 2015;Reina et al., 2016). Our data identified a binding pocket for efsevin located in a groove between the N-terminal α-helix and the inner channel wall. In all our predicted binding conformations, efsevin is bound to the channel wall through hydrogen bonds provided by N19, N207, R218, K236 or N238 plus additional hydrophobic interactions and interacts with a binding pocket located on the hinge of the α-helix. Interestingly, one of the residues in the channel wall, K236, was previously reported to form a hydrogen bond with F18 in the α-helix (Ujwal et al., 2008), a residue that interacts with efsevin upon efsevin binding. It is thus conceivable that efsevin interferes with the interaction between the barrel and the helix in zVDAC2 and consequently affects the movement of the α-helix required to modulate the channel. However, it should be noted that the binding of efsevin might induce a conformational change in the channel, which is not considered in our docking model. Further experimental and computational investigations, like molecular dynamics simulations and determination of the ligand-bound VDAC2 structure are thus needed to clarify the exact mode of gating and ion conduction through the channel in the presence and absence of efsevin.
Using an electrophysiological approach, we found efsevinpromoted gating of zVDAC2. Already at low potentials, the channel preferentially gates into the two closed states and only rarely returns to the open state. We demonstrate that the efsevin-induced low conducting states display less anion selectivity compared to the open state which translates into a higher Ca 2+ flux and consequently into higher Ca 2+ uptake into mitochondria in cardiomyocytes. This is in line with previous experimental (Tan & Colombini, 2007)  to −27 mV (Lemeshko, 2006) and a study based on pH measurements suggests a potential of approximately −40 mV (Porcelli et al., 2005).
This most likely explains the enhanced Ca 2+ uptake into mitochondria seen in the presence of efsevin. The data presented in this study are thus in agreement with the concept of increased Ca 2+ flux upon channel closure.
In this study, we present a binding pocket in zVDAC2 and an associated biophysical mechanism, namely, a modification of voltage gating and a shift towards the cation-selective closed states. VDACs were previously suggested as promising drug targets mainly because of their prominent role in apoptosis (for a review, see . However, it remained unclear whether the role of VDAC in apoptosis was directly associated with the channel's gating behaviour or rather with a modified interaction with partner proteins such as hexokinase or members of the Bcl-2 family. In this study, we present evidence that the efsevin-induced change in gating facilitates transfer of Ca 2+ from the SR into mitochondria. Interestingly, we never observed an involvement of the drug in apoptosis, neither in this study nor in previous studies (Schweitzer et al., 2017;Shimizu et al., 2015) despite a previously identified isoform-specific role for VDAC2 in apoptosis through recruitment of Bax (Lauterwasser et al., 2016;Ma et al., 2014). Efsevin could thus serve as a tool to dissect these functions in future studies.
Taken together, we provide functional and structural data that explains the interaction between zVDAC2 and its modulator efsevin on a molecular basis. Our data provide new insights into VDAC2 function as well as a basis for computer-aided design of optimized compounds that could serve as research compounds and therapeutics for cardiac arrhythmia.