(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Key residues in the VDAC2-BAK complex can be targeted to modulate apoptosis [1]

['Zheng Yuan', 'Walter', 'Eliza Hall Institute Of Medical Research', 'Parkville', 'Melbourne', 'Department Of Medical Biology', 'University Of Melbourne', 'Mark F. Van Delft', 'Mark Xiang Li', 'Peter Maccallum Cancer Centre']

Date: 2024-05 

BAK and BAX execute intrinsic apoptosis by permeabilising the mitochondrial outer membrane. Their activity is regulated through interactions with pro-survival BCL-2 family proteins and with non-BCL-2 proteins including the mitochondrial porin VDAC2. VDAC2 is important for bringing both BAK and BAX to mitochondria where they execute their apoptotic function. Despite this important function in apoptosis, while interactions with pro-survival family members are well characterised and have culminated in the development of drugs that target these interfaces to induce cancer cell apoptosis, the interaction between BAK and VDAC2 remains largely undefined. Deep scanning mutagenesis coupled with cysteine linkage identified key residues in the interaction between BAK and VDAC2. Obstructive labelling of specific residues in the BH3 domain or hydrophobic groove of BAK disrupted this interaction. Conversely, mutating specific residues in a cytosol-exposed region of VDAC2 stabilised the interaction with BAK and inhibited BAK apoptotic activity. Thus, this VDAC2–BAK interaction site can potentially be targeted to either inhibit BAK-mediated apoptosis in scenarios where excessive apoptosis contributes to disease or to promote BAK-mediated apoptosis for cancer therapy.

In this study, we use deep mutational scanning and chemical labelling experiments to identify key residues in the interface between VDAC2 and BAK. We establish the role these mutations have in stabilising/disrupting the BAK–VDAC2 and their consequences for apoptotic progression. We report that mutations of VDAC2 can stabilise the BAK–VDAC2 interaction and thereby impair BAK activation, rendering cells resistant to BH3 mimetic-induced apoptosis.

VDAC2 is one of 3 isoforms of the VDAC protein family that are present on the mitochondrial outer membrane (MOM) [ 5 , 6 ]. VDAC family proteins share the responsibility of transporting metabolites and ions across the MOM [ 7 , 8 ], but VDAC2 has drawn additional interest because of its role in apoptosis [ 4 ]. VDAC2 inhibits BAK-mediated apoptosis by sequestering BAK in its inactive form [ 4 ]. Hence, cells deficient in VDAC2 are hypersensitive to apoptotic stimuli [ 4 ]. This hypersensitivity is despite BAK being unable to localise to mitochondria effectively and that the mitochondria-associated pool of BAK alone cannot promote MOMP efficiently in VDAC2-deficient cells [ 9 , 10 ]. VDAC2 thus plays a role both in recruiting BAK to mitochondria and in restraining its activity once at mitochondria. At steady state, BAK is constitutively located on the MOM and assembles into a BAK–VDAC2 complex [ 11 ]. It has been suggested that the C-terminal transmembrane domain and the globular regions (α-helices 2 to 5) are involved in its interaction with VDAC2 [ 4 , 12 ]. However, the molecular details of this interaction remain unclear.

BAK and BAX are crucial effectors of intrinsic apoptosis, and their activation leads to mitochondrial outer membrane permeabilisation (MOMP) and apoptotic cell death [ 1 , 2 ]. Hence, understanding how BAX and BAK apoptotic activity is regulated is key for identifying opportunities to seize control of apoptosis. The focus of previous efforts has been primarily directed at their interactions with other BCL-2 family members [ 3 ]. However, non-BCL-2 family proteins are also emerging as important players in regulating BAX and BAK apoptotic function, in particular, the voltage-dependent anion channel 2 (VDAC2) [ 4 ].

(A) BAX/VDAC2-deficient MEFs expressing HA-VDAC2 or HA-VDAC2 A172W were treated with an approximate EC50 dose of BH3-mimetics (10 μm S63845 and 0.1 μm A1331852) and increasing doses of WEHI-9625 as indicated. Cell viability was determined by measuring PI negative cells using flow cytometry. Data are mean ± SD of 3 independent experiments. (B) Hypothetical model for how VDAC2 A172 variants may affect BAK stability on the MOM. WEHI-9625 and VDAC2 A172I/L stabilise the BAK–VDAC2 interaction such that BAK dissociation from VDAC2, and ensuing apoptotic activity, is restricted. In cells expressing VDAC2 A172W , WEHI-9625 is no longer able to stabilise the BAK–VDAC2. A possible molecular explanation for these observations is that VDAC2 A172 only partially occupies the region of the hydrophobic groove of BAK identified by our cross linking. WEHI-9625, and medium-sized hydrophobic side chains, may exploit the imperfect fit of A172, enhancing interaction by providing greater surface complementarity. The larger bulk of tryptophan at this position may destabilise the interaction relative to isoleucine or leucine through less ideal surface complementarity, at the same time restricting access to this region for WEHI-9625. The data including flow cytometry gating strategy underlying this figure is available in S1 Data . FACS data files are available in S2 Data . MEF, mouse embryonic fibroblast; MOM, mitochondrial outer membrane; PI, propidium iodide; VDAC2, voltage-dependent anion channel 2.

WEHI-9625 is a small molecule that binds VDAC2 to inhibit mouse BAK-mediated apoptosis by stabilising the BAK–VDAC2 interaction [ 22 ]. However, the binding site for WEHI-9625 on VDAC2 is poorly defined. As VDAC2 A172L/I also stabilised the BAK–VDAC2 interaction and rendered cells resistant to death, thereby mimicking WEHI-9625 activity, we hypothesised that VDAC2 A172 may be involved in WEHI-9625 binding. To test this, we treated Bax -/- Vdac2 -/- MEFs expressing wild-type VDAC2 or VDAC2 A172W , both of which respond to BH3 mimetics ( Fig 6A ), with an approximate EC 50 concentration of combined BH3 mimetics together with an increasing concentration of WEHI-9625. While WEHI-9625 was able to inhibit the death of cells expressing wild-type VDAC2, it was unable to inhibit BAK apoptotic function in cells expressing VDAC2 A172W ( Fig 6A ). This suggests that VDAC2 A172 is involved in the binding site of WEHI-9625 within the BAK–VDAC2 complex, and that mutation at this site to tryptophan restricts compound binding ( Fig 6B ).

Most cells express both BAX and BAK. To assess if stabilising the BAK–VDAC2 interaction could inhibit the death of cells that express both BAK and BAX, we investigated cell death in Vdac2 -/- MEFs. Cells reconstituted with VDAC2 A172I/L variants were significantly more resistant to apoptosis induced by BH3 mimetics compared with Vdac2 -/- cells expressing wild-type VDAC2 ( Fig 5D ). In contrast, cells expressing VDAC2 A172W were more sensitive to BH3 mimetics ( Fig 5D ). This suggested that under certain circumstances, limiting BAK apoptotic function by stabilising its interaction with VDAC2 may be sufficient to inhibit cell death even in the presence of functional BAX.

VDAC2 has been implicated to have contrasting effects on BAK and BAX activity, as it can inhibit BAK-mediated apoptosis, but is required for BAX-mediated apoptosis [ 4 , 11 ]. This has been attributed to the differing subcellular localisation of BAX and BAK [ 11 , 28 ]. Unlike BAK, BAX is mainly cytosolic with redistribution from the cytosol to the MOM required for its apoptotic activity [ 29 , 30 ]. BAX utilises VDAC2 to target mitochondria because BAX cannot localise to mitochondria in cells devoid of VDAC2, and therefore its ability to kill cells is impaired [ 10 , 11 ]. As mutation of VDAC2 A172 influenced interaction with BAK, we wanted to understand if this residue was also involved in VDAC2’s interaction with BAX. We assessed the formation of the BAX–VDAC2 mitochondrial complex in Vdac2 -/- MEFs expressing VDAC2 A172 variants. Surprisingly, mutations to VDAC2 that stabilised its interaction with BAK (VDAC2 A172L/I ), actually destabilised its interaction with BAX as assessed by co-immunoprecipitation ( Fig 4F ) or BN-PAGE ( S3C Fig ). That the VDAC2-BAX complex was destabilised by VDAC2 A172L/I mutation in Bak -/- Vdac2 -/- MEFs ruled out an effect of competitive binding of BAK and BAX ( S3C Fig ). In Bak -/- Vdac2 -/- MEFs expressing VDAC2 A172L/I , BAX was less able to mediate cell death than in cells expressing wild-type VDAC2 or VDAC2 A172W ( Fig 5C ). Hence, while specific mutation of VDAC2 A172 had opposing effects on VDAC2 interaction with BAX and BAK, the consequence for BAX or BAK-mediated killing was similar, likely due to their respective cytosolic and mitochondrial localisation.

(A) Stabilisation of the BAK–VDAC2 interaction inhibits apoptosis. Bax -/- Vdac2 -/- MEFs ectopically expressing VDAC2 A172 variants were treated with 10 μm S63845 (MCL-1 inhibitor) and escalating doses of A1331852 (BCL-xL inhibitor) for 24 h. Cells were stained with PI and analysed by flow cytometry. Data are mean ± SD of 3 independent experiments. (B) Stabilisation of the BAK–VDAC2 interaction inhibits BAK activation. Bax -/- Vdac2 -/- MEFs ectopically expressing VDAC2 A172 variants were treated with BH3 mimetics (10 μm S63845 and 1 μm A1331852) for 2 h. BAK activation was assessed by intracellular flow cytometry with the BAK conformation-specific antibody G317-2. Representative histograms are shown in the left panel, fold increase in MFI plotted in the right panel. Data in the right panel are mean ± SD of 3 independent experiments. (C) In the absence of BAK, destabilisation of the BAX–VDAC2 interaction inhibits apoptosis. Bak -/- Vdac2 -/- MEFs and Vdac2 -/- MEFs ectopically expressing VDAC2 A172 variants were treated with 10 μm S63845 (MCL-1 inhibitor) and escalating doses of A1331852 (BCL-xL inhibitor) for 24 h. Cells were stained with PI and analysed by flow cytometry. Data are mean ± SD of 3 independent experiments. The data including flow cytometry gating strategy underlying this figure is available in S1 Data . FACS data files are available in S2 Data . MEF, mouse embryonic fibroblast; MFI, mean fluorescence intensity; PI, propidium iodide; VDAC2, voltage-dependent anion channel 2.

To understand the consequences of stabilising the BAK–VDAC2 interaction and the influence of this on cell fate, cell death assays were following treatment with BH3-mimetic compounds to specifically induce intrinsic apoptosis in Bax -/- MEFs to isolate BAK-apoptotic function [ 24 – 26 ]. Bax -/- Vdac2 -/- MEFs expressing wild-type VDAC2 or VDAC2 A172W could still undergo apoptosis in response to combined treatment with BH3 mimetics S63845 (MCL-1 inhibitor) and A1131852 (BCL-xL inhibitor) ( Fig 5A ). In contrast, MEFs expressing VDAC2 A172L or VDAC2 A172I were highly resistant to combined BH3 mimetics. Consistently, in BH3 mimetic-treated cells expressing VDAC2 A172L/I , BAK remained in an inactive conformation based on binding of the conformation-specific antibody G317-2 [ 27 ], but was readily activated in cells expressing wild-type VDAC2 or VDAC2 A172W (Figs 5B and S5 ). This suggested that dissociation of the BAK–VDAC2 complex is a key step in BAK-mediated apoptosis, and that preventing BAK dissociation is sufficient to inhibit BAK activation and subsequent BAK-induced apoptosis.

To test if this stabilisation impacted BAK’s ability to mediate MOMP and cytochrome c release, supernatant and membrane fractions were fractionated following treatment of mitochondria-enriched heavy membrane with tBID ( Fig 4H ) in cell lines expressing wild-type VDAC2 or VDAC2 A172I/L/W . Cytochrome c release was delayed from mitochondria isolated from cells expressing VDAC2 A172I or VDAC2 A172L , suggesting that a single point mutation on VDAC2 is sufficient to stabilise the BAK–VDAC2 complex and restrain BAK’s capacity to mediate MOMP ( Fig 4H ). As BAK is constitutively mitochondrial and BAX is predominantly cytosolic argues that BAK is likely the dominant effector of MOMP in these mitochondrial assays. Moreover, VDAC2 A172I/L mutation caused BAX to further dissociate from mitochondria, at least in cells lacking BAK ( S3C Fig ), consistent with our previous report that BAX requires either VDAC2 or BAK to efficiently target mitochondria [ 10 ]. However, we could not exclude that residual mitochondrial BAX contributed to cytochrome c release in these assays. Hence, to assess the impact of mutations to VDAC2 A172 in the complete absence of BAX, we assessed cytochrome c release in Bax -/- Vdac2 -/- MEFs, reconstituted with VDAC2 variants, treated with BH3 mimetics (Figs 4I , S3A , and S4 ). When BAK is the sole apoptosis effector, constraining mutations of VDAC2 A172I/L mutation inhibited cytochrome c release ( Fig 4I ).

We next tested the effect of other VDAC2 A172 variants, focussing on mutants that either increased (VDAC2 A172C/I/L ) or decreased (VDAC2 A172W ) BAK levels in our DMS assay. We expressed these VDAC2 variants in Vdac2 -/- MEFs and Bax -/- Vdac2 -/- MEFs and assessed BAK subcellular localisation by both SDS-PAGE and BN-PAGE (Figs 4D–4G , S3A , and S3B ). In all cell lines tested, BAK was able to localise to the mitochondria (Figs 4D, 4G , and S3B ), suggesting that these mutants could still interact with BAK to mediate its mitochondrial targeting. BN-PAGE showed that the BAK–VDAC2 complex was stabilised by VDAC2 A172C/I/L (Figs 4E and S3B ). We additionally performed immunoprecipitation with the HA-tag on VDAC2 and found that more BAK co-precipitated with VDAC2 A172C/I/L variants further suggesting the stabilisation of the BAK–VDAC2 complex by these VDAC2 variants ( Fig 4F ).

BH3-only proteins such as BIM or caspase-8-cleaved BID (truncated BID, tBID) dissociate BAK from its interaction with VDAC2. In this scenario, BAK maintains its localisation at the MOM where it undergoes conformational change and oligomerisation leading to MOMP [ 10 , 22 ]. Whether BH3-only proteins compete with VDAC2 for a shared binding site on BAK, or whether induced changes in BAK conformation allosterically promote BAK–VDAC2 dissociation is unclear. Given its apparent enhanced stability on BN-PAGE, we hypothesised that the stabilised BAK–VDAC2 A172C complex would be more resistant to dissociation induced by BH3-only proteins. Membrane fractions isolated from Vdac2 -/- MEFs expressing VDAC2 variants were incubated with recombinant tBID and the BAK–VDAC2 complex assessed on BN-PAGE. Following tBID treatment, BAK completely dissociated from wild-type VDAC2 to form homo-oligomers ( Fig 4C ). The same was observed for cells expressing VDAC2 D170E ( Fig 4C ). However, tBID was less able to promote dissociation of BAK from VDAC2 A172C and consequently BAK activation and oligomerisation were limited ( Fig 4C ).

To test whether the stability of this BAK–VDAC complex can be affected by single point mutations in VDAC2, we first expressed VDAC2 D170E , which has been previously reported to disrupt the BAK–VDAC2 interaction [ 16 ]. Consistent with published findings, re-expression of VDAC2 D170E in Vdac2 -/- MEFs did not restore the formation of the BAK–VDAC2 complex ( Fig 4B ). We found that the expression of VDAC2 A172C increased the stability of the BAK–VDAC2 complex compared with wild-type VDAC2 as evidenced by the enriched complex and the virtual absence of monomeric BAK on BN-PAGE ( Fig 4B ), consistent with our DMS data that indicated that cells expressing this VDAC2 mutant also displayed increased BAK levels overall due to stabilisation at the MOM ( Fig 1B and 1C ).

BAK is in complex with VDAC2 on mitochondria in healthy cells [ 4 ]. During apoptosis, BAK must dissociate from VDAC2 in order to form dimers and high-order oligomers that permeabilise the MOM [ 10 , 20 – 23 ]. Therefore, restraining BAK dissociation from VDAC2 might be a way to limit apoptosis. Indeed, it has recently been shown that a small molecule, WEHI-9625, can inhibit BAK-mediated apoptosis by stabilising the BAK–VDAC2 complex [ 22 ]. On blue native polyacrylamide gel electrophoresis (BN-PAGE), BAK forms a multi-protein complex with VDAC2, which also includes VDAC1 and VDAC3 [ 11 ]. This complex is absent in Vdac2 -/- MEF cells and is a destabilised in Vdac1 -/- or Vdac3 -/- MEFs [ 11 ]. However, direct interaction has only been observed between BAK and VDAC2 [ 4 , 12 ]. VDAC1 and 3 do not appear to bind BAK directly and might instead serve as binding partners of VDAC2 within this complex.

(A) Bax -/- Bak -/- Vdac2 -/- MEF cells expressing HA-VDAC2 ΔCys and BAK ΔCys variants were permeabilised with 0.025% digitonin and treated with PEG-maleimide (5 kDa, 0.5 mM) at room temperature for 30 min. Immunoprecipitations with HA-tag were performed with 1% digitonin solubilised membrane fractions post-labelling. Immunoblots are representative of 3 independent experiments. (B) VDAC2 mutants affect the stability of the BAK–VDAC2 complex. Heavy membrane fractions of MEF cells were solubilised by 1% digitonin and analysed by BN-PAGE. Immunoblots are representative of 3 independent experiments. (C) BAK activation (dissociation from the VDAC2 complex) induced by tBID (10 nM) was delayed in cells expressing VDAC2 A172C . Heavy membrane fractions from Vdac2 -/- MEFs expressing wild-type VDAC2 or VDAC2 mutants were solubilised with 1% digitonin post-tBID treatment and analysed on BN-PAGE. Immunoblots are representative of 3 independent experiments. (D–G) VDAC2 A172 variants stabilise the interaction with BAK, but destabilise the interaction with BAX. Subcellular fractionation of VDAC2-deficient MEFs ectopically expressing HA-VDAC2 A172 mutants assessed by subcellular fractionation (D and G), BN-PAGE (E), and immunoprecipitation (F) of solubilised mitochondria. Results are representative of 3 independent experiments. (H) VDAC2 A172 variants that stabilise the interaction with BAK delay the release of cytochrome c as assessed by western blot analysis. Mitochondria-enriched heavy membrane fractions from VDAC2-deficient MEFs expressing HA-VDAC2 A172 mutants were treated with 10 nM tBID over a time course up to 30 min. Membrane and cytosol fractions were separated and analysed by SDS-PAGE. Immunoblots are representative of 2 independent experiments. (I) VDAC2 A172 variants that stabilise the interaction with BAK reduce the release of Cytochrome c as assessed by flow cytometry. Bax -/- Vdac2 -/- MEFs expressing VDAC2 A172 mutants were treated with BH3 mimetics (10 μm S63845 and 0.1/1 μm A1331852) for 2 h to induce apoptosis. Intracellular flow cytometry was performed using APC-conjugated cytochrome c antibody. Histograms are representative of 3 independent experiments, and % cells with released cytochrome c were plotted, data are mean ± SD. The data including flow cytometry gating strategy underlying this figure is available in S1 Data . FACS data files are available in S2 Data . BN-PAGE, blue native polyacrylamide gel electrophoresis; MEF, mouse embryonic fibroblast; VDAC2, voltage-dependent anion channel 2.

Cells lacking VDAC2 are primed to undergo BAK-mediated apoptosis, although counter-intuitively, mitochondrial targeting of BAK is impaired [ 11 ]. Previous mutagenesis studies have suggested that the BAK BH3 domain mutant L76E disrupts the BAK–VDAC2 interaction while the BAK BH1 domain triple mutant W123A/G124E/R125A reduces it [ 4 ]. While these mutagenesis studies, and our DMS and linkage analysis informed residues involved in the BAK–VDAC2 interaction, they did not inform the consequence of acute disruption of the interaction of BAK with mitochondrial VDAC2. To determine this, we employed an obstructive chemical labelling approach that we have previously used to decipher how BAK is activated by BH3-only proteins [ 17 ], to investigate the effect of acute disruption of the BAK–VDAC2 interaction. We designed BAK variants that introduce a single cysteine residues at positions on the exposed surface of pre-activated BAK ( Fig 2A and 2B ) [ 18 ] that have been implicated in BAK oligomerisation: the N-terminus (G4), the BH3 domain (R88), the hydrophobic groove (N124), and the α6 helix (R156) [ 19 , 20 ]. To ensure that any observed effect was due to labelling of cysteine introduced into BAK, as opposed to labelling of cysteines in VDAC2, we engineered an HA-tagged VDAC2 mutant that lacked all cysteines (HA-VDAC2 ΔCys ). Mitochondria-enriched membrane fractions from Bak -/- Bax -/- Vdac2 -/- mouse embryonic fibroblasts (MEFs) expressing HA-VDAC2 ΔCys and BAK variants were isolated and incubated with 5 kDa PEG-maleimide (PEG-mal) to label exposed cysteines on BAK. Digitonin-solubilised membrane fractions were then subjected to immunoprecipitation of HA-tagged VDAC2 to assess the interaction of BAK. Following treatment of membrane fractions with PEG-maleimide, a Cys-null variant of BAK (BAK ΔCys ) was unaffected in its capacity to coimmunoprecipitate with HA-VDAC2 ΔCys ( Fig 4A ), consistent with previous findings that VDAC2 ΔCys retains its interaction with BAK [ 16 ] and confirming the specificity of this obstructive labelling approach. Labelling of BAK at its N-terminus, distal to the hydrophobic groove (G4C), or at a site on the α6 helix located on the opposite surface of BAK to the hydrophobic groove (R156C) had no effect on the interaction with VDAC2 ( Fig 4A ). In contrast, labelling at R88C and N124C within the hydrophobic groove of BAK, reduced co-immunoprecipitation with VDAC2 ( Fig 4A ). Interestingly, mutation of R88 to cysteine disrupted BAK–VDAC2 interaction even in the absence of PEG-mal labelling ( Fig 4A ). These results suggested that the canonical hydrophobic groove of BAK is involved in its interaction with VDAC2.

Models of the BAK:mVDAC2 interaction were generated by multiple rounds of docking initially of the BAK α9 peptide to mVDAC2 β7–β11, followed by docking a minimised BAK α1–8 model to the mVDAC2 solvent exposed A121 loop. Several models were obtained with discrete interaction sites, and each subjected to molecular dynamics simulation. Of the several models generated, this model was selected as the most representative of the cross-linking data. MOM, mitochondrial outer membrane; VDAC2, voltage-dependent anion channel 2.

To further explore residues on BAK that may interface with VDAC2, we then tested residues on the periphery of the groove in which R88, Y89, and N124 are located, including in the α2, 3, 4, and 8 ( Fig 2A and 2B ). Interestingly, the BAK variants D84C, I85C, E92C, S117C, and S121C were all able to crosslink with VDAC2 A172C , but I123C, which orientates away from N124, and A179C which is further from the centre of the groove were both unable to do so ( Fig 2D ). These cross-linking data ( Fig 2E ) identified a surface on BAK formed by its α3, 4, and 5 that interfaces the cytosol-exposed β10–11 loop on VDAC2 ( Fig 3 ).

(A) Domain architecture of human BAK. Relative positions of selected single cysteine mutations (on a BAK ΔCys–C14S/C166S background) used in this study are labelled. (B) Structure of human BAK (PDB code: 2IMT) [ 18 ] shown in cartoon and surface representations with α helices 2–5 (hydrophobic groove) coloured in light blue. Residues selected for cysteine mutagenesis and cross-linking are shown as sticks. (C and D) . Heavy membrane fractions containing mitochondria were isolated from Bax -/- Bak -/- Vdac2 -/- MEFs ectopically expressing HA-VDAC2 S58C or HA-VDAC2 A172C (both on a VDAC2 ΔCys background) in combination with different BAK single cysteine mutants. Cysteine cross-linking was induced by incubation with 1 mM CuPhe at 4°C for 30 min and analysed by non-reducing SDS-PAGE. Cross-linked BAK and VDAC2 are denoted by arrowheads. Immunoblots are representative of at least 3 independent experiments. (E) Densitometric analysis of CuPhe cross-linking. Cross-linked BAK–VDAC2 (%) was calculated by determining the intensity of cross-linked BAK over the sum intensity of total BAK. Data are mean ± SD of 3 independent experiments. The data underlying this figure is available in S1 Data . MEF, mouse embryonic fibroblast; VDAC2, voltage-dependent anion channel 2.

DMS implicated A171 on the VDAC2 β10–11 loop (equivalent to A172 in mouse VDAC2) as an interfacing residue with BAK. We next employed cysteine cross-linking to investigate residues on BAK that interact with mmVDAC2 A172. A172 was mutated to cysteine (VDAC2 A172C ) on a mouse VDAC2 ΔCys background and stably expressed in Bak -/- Bax -/- Vdac2 -/- MEFs. As a control, we expressed VDAC2 with a single cysteine introduced on the opposite side of the β-barrel (VDAC2 S58C ) ( S2 Fig ). Following stable expression of selected BAK single cysteine variants ( Fig 2A and 2B ), we treated membrane fractions with the oxidant copper phenanthroline (CuPhe) to induce disulphide bond formation and analysed the linkage adducts by non-reducing SDS-PAGE. Efficient disulphide-linkage was detected between VDAC2 A172C and BAK R88C /BAK Y89C , and to a lesser extent BAK N124C ( Fig 2C ). No detectable linkage was observed between VDAC2 A172C and BAK ΔCys , BAK G4C , BAK R156C , or BAK D160C ( Fig 2C ). None of the BAK variants exhibited specific linkage to VDAC2 S58C ( Fig 2C ). These data suggested that BAK hydrophobic groove residues are proximal to the VDAC2 β10–11 loop.

Next, we expressed BAK with an N-terminal green fluorescent protein (GFP) in VDAC2-deficient cells. In the absence of VDAC2, the GFP signal was low, but upon co-expressing wild-type VDAC2 the GFP signal was markedly elevated as GFP-BAK was recruited to and stabilised at the MOM ( S1D Fig ). Thus, we introduced the VDAC2 β10–11 DMS library to these cells and used GFP fluorescence as a measure for BAK recruitment and/or stabilisation at the MOM ( S1E Fig ). This DMS approach highlighted residues that either positively or negatively influenced GFP-BAK levels. The most pronounced effects were observed for substitutions of Ala171 ( Fig 1B and 1C ). In some cases, such as mutation to the midsized hydrophobic residues valine, isoleucine, and leucine, GFP-BAK levels increased, suggesting that these substitutions stabilised the BAK–VDAC2 interaction. Conversely, mutation to some larger residues, such as tryptophan and arginine, resulted in reduced GFP-BAK levels, indicative of destabilisation. These finding implicate Ala171 as being a key residue involved in VDAC2 interaction with BAK and that this interaction could be modulated to affect BAK apoptotic function.

First, we constructed a retroviral library of FLAG-tagged human VDAC2 single amino acid substitution variants spanning the residues Q165 to F180 and evaluated how these substitutions impacted VDAC2 expression level ( S1A and S1B Fig ). Impaired expression was observed for truncated variants, several variants with proline substitutions, and variants where membrane-facing hydrophobic amino acids were substituted for polar or charged residues (e.g., positions M166, F168, and F180) ( S1C Fig ). Most other substitutions appeared to be well tolerated as measured by VDAC2 expression level.

(A) AlphaFold2 predicted structure of human VDAC2 (Uniprot entry: P45880) [ 13 , 14 ] with residues within the β strands 10–11 region evaluated with DMS coloured in orange. Human and mouse VDAC2 show high sequence identity in this region but have 1 amino acid numbering difference. T167 and D169 (D168 and T170 in mouse VDAC2) were previously identified as residues for its interaction with BAK [ 16 ]. ( B) Heatmap representation of DMS screen to identify residues within hsVDAC2 β10–11 that influence BAK recruitment and stabilisation at the MOM. Bax -/- Bak -/- Vdac2 -/- Mcl1 -/- MEF cells were engineered with retroviruses to express GFP-tagged mouse BAK and a library of uniquely barcoded FLAG-tagged human VDAC2 variants with amino acid substitutions between positions Q165 and F180. Cells were sorted into GFP low and GFP high populations and the log2-fold difference in distribution for each barcode was determined by Illumina sequencing. Mann–Whitney p-values were calculated comparing the log2-fold differences of barcodes associated with each coding substitution to those associated with wild-type VDAC2 coding sequence. Data are represented as signed log transformed p-values to reflect the direction of change: negative/blue for variants skewed towards the GFP low fraction (i.e., impaired stabilisation of GFP-BAK); positive/red for variants skewed towards the GFP high fraction (i.e., enhanced stabilisation of GFP-BAK). (C) Additional detail on the site-specific mutational tolerance of VDAC2 at position A171. Each dot represents a uniquely barcoded independent clone within the DMS library and values are log2[GFP high barcode reads]–log2[GFP low barcode reads]. Aliphatic substitutions (Ile, Leu, and Val) have positive effects on BAK stability (skewed to GFP high fraction), while substitutions to smaller (Gly) or bulkier residues (Trp, Arg) appear to destabilise BAK in cells (skewed to GFP low fraction). Mann–Whitney p-values reflect the comparison of each substitution variant to barcodes associated with wild-type VDAC2 coding sequence (p: < 0.05*; < 0.01**, < 0.001***, < 0.0001****). The data underlying this figure is available in S1 Data . DMS, deep mutational scanning; GFP, green fluorescent protein; MEF, mouse embryonic fibroblast; VDAC2, voltage-dependent anion channel 2.

BAK is normally localised to the MOM where, prior to activation, it interacts with VDAC2 [ 4 , 11 ]. Genetic deletion of Vdac2 impairs the ability of BAK to target mitochondria leading to a reduction in BAK cellular abundance [ 10 , 11 ]. VDAC2 is an integral membrane protein with a β-barrel architecture [ 13 – 15 ], previous work has identified β-strands 7–11 are involved in interactions with BAK [ 16 ]. To further narrow down the region on VDAC2 responsible for BAK targeting, deep mutational scanning (DMS) was performed on the cytosol-exposed region of VDAC2 β10–11 ( Fig 1A ).

Discussion

VDAC2 facilitates the mitochondrial localisation of BAK where it acts to mediate apoptosis [10,16]. However, when BAK is on the MOM, VDAC2 sequesters BAK [4]. To drive apoptosis, BAK must dissociate from VDAC2 to oligomerise and form the apoptotic pore [4]. Here, we have sought to map the interface between VDAC2 and BAK and explore the consequence of modulating this interaction. We show that the VDAC2-complex can be acutely disrupted by obstructive labelling of the BAK hydrophobic groove, that mutation to A172 within the VDAC2 β10–11 loop affects BAK localisation and the stability of the VDAC2–BAK complex, and that residues in the groove of BAK can be cross-linked to A172. This provides strong evidence that VDAC2 A172 forms part of an interaction interface within the VDAC2–BAK complex. Previously, we showed that a modified BH3 peptide that could bind the BAK groove with high affinity and displace BAK from the VDAC2 complex [21]. However, we did not know if this was due to a direct competition for the BAK groove or a result of allosteric effects due to the groove opening upon BH3 binding. Our results here show that the BAK groove does indeed bind directly to VDAC2 and that BH3 proteins likely compete with VDAC2 for the groove, explaining our previous observations.

The hydrophobic groove of BCL-2 family proteins is the primary site of interaction with other BCL-2 family member proteins. However, in those cases the interaction interface involves an amphipathic helix, the BH3 domain of the partner, binding into the groove on the BCL-2 fold [31–33]. This interaction is mediated by at least 4 hydrophobic residues on consecutive turns of the helix interacting with hydrophobic pockets within the groove, and a conserved salt bridge between an aspartate on the helix and an arginine in the groove [33]. The interaction topology adopted by the VDAC2 β10–11 loop with the BAK hydrophobic groove interface is unlikely to be similar, primarily because the VDAC2 β10–11 loop is only 6 amino acids so not long enough to form a helix of 4 consecutive turns (each turn in an α helix requires 4 residues). It is more likely that this region of VDAC2 instead inserts into the hydrophobic groove as a loop. Recently, a structure of p53 bound to the pro-survival family member BCL-2 has been reported as having such an interaction [34]. However, to our knowledge, such an interface has not been described for a pro-apoptotic BCL-2 executioner family member.

Interestingly, we observe that the type of substitution made to VDAC2A172 can have variable impact on the stability of the VDAC2–BAK complex. Mutations to medium-sized hydrophobic sidechains (Ile, Leu, Cys) stabilise the complex, while mutation to a large hydrophobic residue (Trp) is less able to restrain BAK. It is notable that generally the hydrophobic residues on BH3 domains that interact with BCL-2 family hydrophobic grooves are medium-sized, particularly the h2 position which is invariably a leucine. It is possible that the medium-sized mutations to VDAC2 A172 mimic such an interaction to increase the affinity for the groove, thus limiting competition by activating BH3 domains. Alternatively, mutation to the larger tryptophan may not be as readily accommodated, thus reducing its affinity. Mutation from Leu to Ala at the BIM BH3 and BID BH3 h2 positions significantly impairs binding to BAX [35] and BAK [31], respectively. However, in our view this does not necessarily rule out the h2 interacting pocket as being the site of interaction for VDAC2 A172. Such a substitution is likely to be less tolerated in the context of a BH3 helix where backbone conformers are restricted, to maintain hydrogen-bonding in the helix, compared with an inserted flexible loop that can adopt various orientations.

The observation that mutation of VDAC A172 to tryptophan renders cells less responsive to the BAK inhibitor WEHI-9625, suggests that this interface is also the likely site of interaction for this compound. It is not known precisely how WEHI-9625 functions to limit mouse BAK apoptotic activity other than it appears to stabilise the BAK–VDAC2 complex. Our results suggests that this molecule takes advantage of a pocket within the BAK–VDAC2 complex. We hypothesise that when this pocket is occupied by a medium-sized sidechain such as leucine, isoleucine, or cysteine, it phenocopies the stabilising effect of WEHI-9625. However, when A172 is mutated to a large sidechain such as tryptophan, it can compete with WEHI-9625 for BAK pocket binding, and because the tryptophan does not stabilise the complex an increase in killing is observed. Structures of the BAK-VDAC2 complex with compound bound are ultimately required to resolve this mechanism.

That mutation to VDAC2A172 impacts both BAX and BAK activity suggests that this site of interaction is conserved for these 2 proteins. When VDAC2A172 is mutated to tryptophan, it decreases the stability of the interaction with BAK, and more BAX becomes associated with VDAC2. Conversely, mutation of VDAC2A172 to Leu/Ile had opposing effects on VDAC2 interaction with these 2 proteins—destabilising interactions with BAX, but stabilising interactions with BAK. For BAX this results in reduced recruitment to the mitochondria, especially when BAK is absent, for BAK it results in less freed protein on the mitochondria. In both scenarios, there is a reduction of free BAX and/or BAK on the mitochondria available to induce MOMP, and therefore, the ultimate consequence on cell death is the same. Excitingly, this suggests that this site on VDAC2 can be potentially targeted to inhibit apoptosis in cells expressing both BAK and BAX. While this site on VDAC2 mediates interactions with both BAK and BAX, our data suggest that the effectiveness of targeting this site on VDAC2 to inhibit cell death will likely be more profound in cells, or with apoptotic stimuli, that predominantly rely on BAK.

While it is well established that VDAC2 interacts with BAK and regulates BAK apoptotic activity [4,9,10,12,16], the precise nature of this interaction has remained unclear. Here, we have mapped a new interaction interface between VDAC2 and BAK, providing insight into the nature of this complex at the MOM. Excessive apoptosis is implicated in contributing to a range of disease states including autoimmune disease and acute or chronic neurodegenerative conditions, stabilising the VDAC2–BAK interaction may be a potential therapeutic strategy for inhibiting apoptosis in such settings.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002617

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/