Mcl-1 promotes lung cancer cell migration by directly interacting with VDAC to increase mitochondrial Ca2+ uptake and reactive oxygen species generation

Mcl-1 is an antiapoptotic member of the Bcl-2 family frequently upregulated in non-small cell lung carcinoma (NSCLC). We now report the physiological significance of an interaction between Mcl-1 and the mitochondrial outer membrane-localized voltage-dependent anion channel (VDAC) in NSCLC cell lines. Mcl-1 bound with high affinity to VDAC1 and 3 isoforms but only very weakly to VDAC2 and binding was disrupted by peptides based on the VDAC1 sequence. In A549 cells, reducing Mcl-1 expression levels or application of VDAC-based peptides limited Ca2+ uptake into the mitochondrial matrix, the consequence of which was to inhibit reactive oxygen species (ROS) generation. In A549, H1299 and H460 cells, both Mcl-1 knockdown and VDAC-based peptides attenuated cell migration without affecting cell proliferation. Migration was rescued in Mcl-1 knockdown cells by experimentally restoring ROS levels, consistent with a model in which ROS production drives increased migration. These data suggest that an interaction between Mcl-1 and VDAC promotes lung cancer cell migration by a mechanism that involves Ca2+-dependent ROS production.

The Bcl-2 proteins are a family of molecules comprised of both pro-and antiapoptotic members essential for the regulation of apoptotic cell death. In the classical paradigm, the antiapoptotic proteins Bcl-2, Bcl-x L and Mcl-1, inhibit cell death during receipt of apoptotic stimuli by binding and sequestering the proapoptotic members. 1 It is now appreciated, however, that in the absence of apoptotic stimuli, Bcl-2 proteins have numerous non-canonical interactions that influence diverse cellular functions, although the precise mechanisms are poorly understood. 2 Since antiapoptotic Bcl-2 family members are frequently upregulated in cancer, determining if and how these non-canonical interactions confer survival or other advantages to the cancer cell, will be an important step toward identifying new therapeutic targets. One such interaction is with the outer mitochondrial membrane-localized voltage-dependent anion channel (VDAC), a porin channel with three isoforms that serves as a major diffusion pathway for ions and metabolites, 3 and whose gating properties are affected by either Bcl-2 or Bcl-x L binding. [4][5][6] We recently identified an important role for Bcl-x L /VDAC interactions in the regulation of mitochondrial [Ca 2+ ]. 7 Moving Ca 2+ from the cytoplasm to the mitochondrial matrix requires transfer across the outer membrane by VDAC 3,8 and across the inner membrane by the Ca 2+ uniporter. 9 Our studies showed that Bcl-x L interacts with VDAC to facilitate Ca 2+ uptake into the mitochondrial matrix. It is not known if other Bcl-2 family members, particularly Bcl-2 and Mcl-1, which are also known VDAC binding partners impart the same physiological regulation on mitochondrial [Ca 2+ ]. Furthermore, the specific physiological consequences and significance of this regulation remain to be determined.
Increased production and reduced scavenging of reactive oxygen species (ROS) is frequently observed in cancer cells. 10 While excessive ROS levels are toxic, sub-lethal production serves an important signaling function, particularly in cancers, were ROS promote cell proliferation, migration and invasion. [11][12][13][14][15] A primary source of ROS are the mitochondria, and a number of mitochondrial signaling pathways are known to be remodeled and contribute to elevated ROS in cancer cells, including those involved in regulating the electron transport chain (ETC) function and metabolic activity. 11,[16][17][18] It is recognized that upregulation of antiapoptotic Bcl-2 proteins are also associated with a pro-oxidant intracellular environment. [19][20][21][22] Mechanistically, they are thought to act at the level of the mitochondria to affect the respiratory chain and increase production of ROS. Since matrix [Ca 2+ ] is an important regulator of mitochondrial metabolism, 23,24 and as such, contributes to the regulation of mitochondrial ROS production, 25 we reasoned that antiapoptotic Mcl-1/VDAC interactions could promote ROS generation by facilitating matrix Ca 2+ uptake.
Understanding non-canonical roles of Mcl-1 is an important step toward identifying novel therapeutic targets, particularly in cancers where it is highly expressed, such as in non-small cell lung cancer (NSCLC). 26,27 Therefore, we hypothesized that Mcl-1 binding to VDAC promotes mitochondrial Ca 2+ uptake and ROS production in NSCLC cells and that this is essential in maintaining the cancer cell phenotype. To test this, we assessed the biochemical interaction between Mcl-1 and VDAC and examined the effects of manipulating Mcl-1 expression levels and Mcl-1/VDAC interactions on mitochondrial Ca 2+ uptake, ROS generation and NSCLC cell proliferation and migration.

Results
Mcl-1 binds robustly to VDAC1 and 3 and to VDAC1 with greater apparent affinity than Bcl-x L . To determine the relative binding affinity for the Mcl-1/VDAC interaction, a GST pull-down assay was performed. GST-fusion proteins of VDAC1, 2 and 3 were found to effectively pull down Mcl-1 from lysates of mouse embryonic fibroblasts (MEF) overexpressing human Mcl-1. Similar to the findings reported previously for Bcl-x L , 7 Mcl-1 bound more strongly to VDAC1 compared with VDAC3, and only weakly bound to VDAC2 (Figure 1a). This interpretation was confirmed by the reciprocal experiment in which immobilized His-tagged Mcl-1 robustly pulled down GST-tagged VDAC1 and 3 but not VDAC2 (Figure 1b). Several peptides based on the human VDAC1 sequence (VDAC-based peptides) were tested for their ability to inhibit binding between Mcl-1 and VDAC1. Inclusion of either VDAC-based peptide (N-ter or L14-15) in the reaction mixture effectively reduced both Mcl-1 and Bcl-x L pulldown by GST-VDAC1 ( Figure 1c). Intriguingly, compared with the amounts of Mcl-1 and Bcl-x L detected in the input lysate (lane 1, Figure 1c), VDAC1 appeared to pull down a greater proportion of the available Mcl-1 relative to Bcl-x L , suggesting that VDAC1 favors binding to Mcl-1 over Bcl-x L . This was verified using a GST-VDAC1 pull-down assay baited with equal amounts of purified recombinant Mcl-1 and Bcl-x L (200 ng). Under these conditions a greater proportion of the available Mcl-1 was pulled down by GST-VDAC1, as indicated by comparing the band density of the pulldown to input (Figure 1d; lane 5 and 1 for Mcl-1 and  lane 6 and 2 for Bcl-xL). These data confirm that Mcl-1 binds to all three VDAC isoforms, however, the interaction appears strongest with VDAC1 followed by VDAC3 and then only very weakly to VDAC2. Importantly, Mcl-1 binds VDAC1 with a much higher apparent affinity than Bcl-x L , suggesting that the Mcl-1/VDAC interaction could have important functional implications.
Mcl-1 and VDAC interaction increases [Ca 2+ ] mito uptake in A549 cells. The A549 cell line is a widely used NSCLC defined by high expression of Mcl-1. 26 We first confirmed that native Mcl-1 and VDAC1/3 interact in this cell line using a proximity ligation assay (PLA) to detect and quantify the endogenous protein complexes at the single cell level. 28 The Mcl-1/VDAC complexes were labeled using an antibody directed against Mcl-1 and an antibody known to react with both the VDAC1 and 3 isoforms. Since A549 cells do not express CFTR the assay was also performed using anti-Mcl-1 and anti-CFTR antibodies to control for background signal. Robust PLA signal was observed in the presence of Mcl-1 and VDAC1/3 antibodies compared with control (Figures 2a and b), indicating the presence of Mcl-1/VDAC complexes in vivo.
Since we previously demonstrated that regulation of mitochondrial Ca 2+ ([Ca 2+ ] mito ) uptake was an important functional consequence of the Bcl-x L /VDAC interaction, 7 we asked if the Mcl-1/VDAC interaction similarly affected mitochondrial Ca 2+ handling. This was first assessed by examining the effect of Mcl-1 knockdown on [Ca 2+ ] mito uptake. The effectiveness of siRNA treatment was confirmed by western blot (Figure 2c). Control and Mcl-1 knockdown cells were loaded with the Ca 2+ indicator Rhod-2 AM and permeabilized with digitonin to release cytoplasmic dye and enable application of solutions directly to the mitochondria. Cells were then bathed in Ca 2+ -free intracellular like media (ICM) containing   ] mito is associated with increased mitochondrial ROS production. 25 To This was done to ensure that differences in RH123 fluorescence were specifically due to differences in ROS and not due to artifacts introduced by changes in membrane potential upon Mcl-1 knockdown. Interestingly, a recent study using a knockout MEF model reported a slight decrease in membrane potential with loss of Mcl-1 measured using a similar approach. 30 In A549 cells, however, Mcl-1 knockdown using siRNA was without effect on membrane potential (Figure 3c), suggesting that either complete ablation is required or the effect of Mcl-1 on membrane potential is cell-type dependent. To further confirm that Mcl-1 specifically affects mitochondrial ROS, MitoSOX was employed as an alternative mitochondrial redox probe. In MitoSOX-loaded cells, the difference in ROS levels between control and Mcl-1 knockdown was found to be qualitatively similar to that measured with RH123 ( Figure 3c). We next assessed the [Ca 2+ ] mito -dependence of ROS production. First, intact A549 control and Mcl-1 knockdown cells were incubated with BAPTA-AM, to chelate intracellular Ca 2+ and block cytoplasmic [Ca 2+ ] signals Increased ROS production by Mcl-1/VDAC interactions promotes NSCLC cell migration. We hypothesized that increased ROS production, mediated specifically through the regulation of [Ca 2+ ] mito uptake by Mcl-1/VDAC interactions, was necessary and sufficient for Mcl-1 to promote cell migration. To test this, the movement of individual cells was monitored over time by visualizing the tracks left by cells migrating on a colloidal gold-coated surface. Control and stable Mcl-1 knockdown NSCLC cell lines were treated for 24 h under the various conditions described below and the track area created by moving cells was measured as an index of motility (Figure 6a). First, the effect of disrupting the Mcl-1/VDAC interaction using the VDAC-based peptides (N-ter and L14-15) was assessed. As summarized in Figure 6b, application of N-ter and L14-15 significantly reduced ROS production in control A549, H1299 and H460 cells but had no effect after Mcl-1  Figure 6d). Importantly, elevating ROS levels in the Mcl-1 knockdown cells restored migration to levels similar to those observed in control cells (Figure 6e).

Discussion
A direct interaction between VDAC and antiapoptotic Bcl-2 and Bcl-x L is well-established. [34][35][36] Both Bcl-2 and Bcl-x L bind to the N terminus of VDAC1, 5,36,37 and as demonstrated using Bcl-x L , requires contact with a highly conserved region spanning alpha helices 5 and 6 of the Bcl-2 protein. 38 The involvement of the BH4 domain of Bcl-2 members has also been implicated, 39,40 however, this region may be more important in mediating the functional effects rather than contributing to robust binding. 38,39 We recently demonstrated that in addition to VDAC1, Bcl-x L also binds well to VDAC3 but only very weakly to VDAC2. 7 To our knowledge there has been only one previous report of an interaction between Mcl-1 and VDAC. 41 In that study, however, the authors reported that only an N-terminally-truncated fragment of Mcl-1 bound to VDAC1. 41 This is in contrast to the current study in which immobilized VDAC1 and 3 is readily pulled down by full-length Mcl-1 from cell lysates (Figure 1a). In addition, His-tagged Mcl-1 interacts robustly with recombinant VDAC1 and 3, suggesting direct protein-protein interaction (Figure 1b). This is further supported by the observation that peptides based on the VDAC sequence (N-ter and L14-15) are able to disrupt the interaction (Figure 1c). The VDAC-based peptides used in the current study were originally designed based on the binding determinants of the VDAC1/Bcl-2 interaction, 35 and later identified as being effective at disrupting VDAC/Bcl-x L . 5,7 The ability of VDAC-based peptides to function as general inhibitors of antiapoptotic Bcl-2 protein-VDAC interactions suggests that the structural determinants governing VDAC binding to Bcl-2, Bcl-x L and Mcl-1 are highly conserved. That being said, we did discover that individual Bcl-2 proteins differ with respect to their relative binding affinity to VDAC. Although binding of Bcl-2 itself was not examined in the current study, we show that Mcl-1 has a much higher apparent affinity for VDAC compared with Bcl-x L (Figures 1c and d). It is likely then that the relative binding affinity in addition to the expression level of each Bcl-2 member governs the structural architecture, and presumably the functional outcome, of the VDAC-Bcl-2 protein complex. We previously determined that [Ca 2+ ] mito uptake was tightly regulated by Bcl-x L through direct interaction with both VDAC1 and VDAC3. 7 We now show that the VDAC/Mcl-1 interaction confers similar regulation (Figure 2), however, we did not investigate the relative importance of specific VDAC isoforms, choosing instead to focus on defining the functional consequences of enhanced [Ca 2+ ] mito uptake. Nevertheless, we do show that Mcl-1 and Bcl-x L bind similarly, in that they both interact more strongly with VDAC1 and VDAC3 compared with VDAC2. The observation that Mcl-1 binds more avidly to VDAC suggests that in cancer cells characterized by Mcl-1 overexpression, the Mcl-1/VDAC interaction could be a major determinant in remodeling mitochondrial physiology. We now provide evidence for this. In A549 cells, known to have high endogenous levels of Mcl-1, 27 we show that Mcl-1 is in complex with VDAC1/3 (Figures 2a and b), and that either knockdown of Mcl-1 or Mcl-1/VDAC disruption with inhibitor peptides decreases [Ca 2+ ] mito uptake by the same amount (Figures 2d and e). However, A549 cells do express Bcl-x L , and the Bcl-x L /VDAC interactions, which are also sensitive to inhibitor peptides, should contribute to the regulation of [Ca 2+ ] mito uptake, as we described. 7 In that case we might have expected inhibitor peptides to be more effective in lowering [Ca 2+ ] mito uptake than Mcl-1 knockdown alone. However, Mcl-1 knockdown and inhibitor peptides are equally effective (Figure 2e), supporting the interpretation that Mcl-1/VDAC, rather than Bcl-x L /VDAC, is the dominant interaction in this cell type.
Mitochondrial ROS generation is largely coupled to respiration and occurs when electrons leak from the ETC at complex I and III and react with O 2 to produce superoxide. 10,25 Increased ROS generation has also been observed upon Bcl-2, Bcl-x L or Mcl-1 overexpression and linked to mechanisms that result in increased respiration and complex III modulation. 19,20,42,43 The current study describes a novel mechanism that involves Mcl-1/VDAC interactions. Our conclusion that ROS production is dependent on the Mcl-1/VDAC interaction is supported by the observation that VDAC-based inhibitor peptides decrease ROS in cells expressing Mcl-1 but are without effect when Mcl-1 is knocked down (Figure 3d). These data rule out the possibility that the effects of Mcl-1 on ROS are mediated by Mcl-1 localized to the mitochondrial matrix or at the inner membrane. This is important in light of recent findings that implicate a role for matrix-localized Mcl-1 in regulating bioenergetics and membrane potential. 30 Evidence for the Ca 2+ dependence of Mcl-1/VDAC effects on ROS is provided by the observation that Ca 2+ chelation by BAPTA-AM has the same effect on ROS regardless of Mcl-1 expression or the application of inhibitor peptides (Figure 3d). Moreover, by carefully controlling [Ca 2+ ] mito uptake we demonstrate that ROS production is critically dependent on the Ca 2+ load but not Mcl-1 expression (Figure 3e). These data effectively eliminate alternative interpretations that include the potential for Mcl-1/VDAC to affect ROS by either controlling their release from mitochondria through VDAC, 44 regulating mitochondrial substrate uptake 6 or by direct remodeling of the respiratory chain. 42 Although not investigated in the current study, precedents exist that link [Ca 2+ ] mito and ROS generation. Physiological Ca 2+ uptake into the mitochondrial matrix promotes activation of TCA cycle and generates ROS as a result of increased flux through the ETC. 25 In addition, Ca 2+ can enhance electron leak by indirectly inhibiting the ETC at complex III or IV. 45,46 ROS promotes cancer cell migration and invasion by impinging on a number of signaling pathways, including actin cytoskeleton, cell adhesion and extracellular matrix degradation. 47 Using a scratch wound-healing assay ( Figure 4) we show that wound closure is inhibited by Mcl-1 silencing in all three NSCLC cell lines (A549, H1299 and H460). Since Mcl-1 has no effect on proliferation in these cells ( Figure 5), decreased wound closure is most likely mediated by changes in migration (Figure 6a). Importantly, the effect of Mcl-1 knockdown on migration is recapitulated by VDACbased peptide exposure; moreover, peptides have no additional effect when applied to Mcl-1-knockdown cells (Figure 6c). These data support a model in which Mcl-1 expression in NSCLC cells promotes migration through Mcl-1/ VDAC interactions. The conclusion that this is driven by increased ROS is supported by the observation that decreased migration in Mcl-1 knockdown cells is restored by maneuvers designed to elevate ROS (Figure 6e). The delicate balance between the stimulatory and toxic effects of ROS was also noted when experimentally increasing ROS actually decreased migration in control cells (Figure 6e). This is entirely consistent with previous observations in these cell types, demonstrating that excessive ROS leads to decreased proliferation and cell death. 48 In conclusion, we describe a novel mechanism governing ROS generation and migration in NSCLC cells. These findings are significant because there is a growing body of evidence that links increased mitochondrial ROS generation to increased migration, invasiveness and metastasis in a variety of cancers, 11,14,18,49 including NSCLC. 12 Since Mcl-1 is elevated in about 60% of NSCLCs, at levels greater than that observed in any other cancer types, 50 our data identify the Mcl-1/VDAC interaction as a possible therapeutic target that could limit the metastatic potential in lung cancer.  Proximity ligation assay. A proximity ligation assay kit (Olink Bioscience, Uppsala, Sweden) was used to study the in vivo interaction between Mcl-1 and VDAC1/3. A549 cells were washed in PBS at room temperature and fixed by incubation in Buffer A (1 mM MgCl 2 , 1 mM EGTA, 100 mM PIPES and 3.7% paraformaldehyde, pH 6.5) and Buffer B (100 mM Na 2 B 4 O 7 , 1 mM MgCl 2 , 3.7% paraformaldehyde, pH 11.0) for 5 and 10 min, respectively. Cells were permeabilized with 0.1% Triton X-100 for 30 min and then incubated for 15 min with 50 mM NH 4 Cl before being blocked for 1 h (10% goat serum, 1% BSA in PBS). The remainder of the protocol was carried out following the manufacturer's instructions using rabbit anti-Mcl-1 (Abcam) and mouse anti-VDAC1/3 (Abcam) antibodies to detect Mcl-1/VDAC interactions and mouse monoclonal anti-CFTR (Ab #596; Dr J. Riordan, University of North Carolina, Chapel Hill, NC, USA) as control. The cellular PLA signal was visualized using the PlanApo 60 × , 1.42 NA oil immersion objective of an Olympus IX71 inverted microscope (Olympus America Inc., Center Valley, PA, USA) coupled to a VT-Infinity 3 confocal system (VisiTech International, Sunderland, UK) and quantified using ImageJ software. 51 Mcl-1 overexpression and knockdown. Human Mcl-1 was transiently overexpressed as described previously. 52 Transient transfection of control or Mcl-1 siRNAs (Santa Cruz, Dallas, TX, USA) was carried out 24 h prior to experimentation using Lipofectamine RNAiMAX (Life Technologies) following the manufacturer's instructions. Stable knockdown was achieved using lentiviral transduction particles carrying control or Mcl-1 shRNAs and puromycin (2 μg/ml) selection. Knockdown was confirmed by western blot.

Mitochondrial [Ca 2+
]. Cells were cultured on glass coverslips and loaded with 3 μM Rhod-2 AM by incubation at 37 o C for 30 min. Cells were then permeabilized by 3-4 min exposure to digitonin (25 μg/ml) applied in Ca 2+ -free ICM. The permeabilized preparation was then allowed to equilibrate in regular Ca 2+ -free ICM for 15 min prior to experimental recording. Coverslips were mounted in a recording chamber on the stage of an inverted IX71 microscope (Olympus America Inc.) and excited at 548 nm. Emitted fluorescence was filtered at 605 nm and collected using a CCD-based imaging system running SimplePCI software (Hamamatsu Corporation, Sewickley, PA, USA). The chamber was continuously perfused with ICM at room temperature, and a rapid solution changer was used to switch to the Ca 2+ containing solution bathing the cells under study.
Mitochondrial ROS and membrane potential. Cells were cultured on black optical-bottom 96-well plates. To monitor ROS, cells were incubated in culture medium with 5 μM DHR 123 for 30 min at 37 o C, or 10 μM MitoSOX for 30 min followed by 1 h wash out at 37 o C. For membrane potential measurements, TMRE (15 nM) was added to the medium for 30 min at 37 o C before measurement. In permeabilized cells, 5 μM DHR 123 was added to Ca 2+ -containing ICM and incubated at 37 o C for 30 min. Images were acquired using wide-field fluorescent microscopy. The emission and excitation wavelengths were: 500/535 nm (DHR 123); 548/588 nm (MitoSOX and TMRE).
Scratch wound-healing assay. Cells were cultured on 24-well plates and wounds were made by scratching the monolayer with a pipette tip. Images of the wounds were taken using a 4 × objective at 0 and 24 h after scratch and the open wound area was measured as a percentage of the total area by TScratch software developed by the Swiss Federal Institute of Technology, Zürich, Switzerland. 53 Proliferation assay. Proliferation was assessed using CyQUANT NF Cell Proliferation Assay Kit (Life Technologies) according to the manufacturer's instructions. Briefly, 5000 cells were plated in each well of a 96-well plate. At 24-h intervals, the medium was removed and 50 μl assay mix was added to each well and incubated at 37 o C for 1 h, and florescence intensity measured using a POLARstar Omega plate reader (BMG Labtech Inc., Cary, NC, USA).
Single cell track migration assay. Colloidal gold-coated 24-well plates were prepared as described in the study by Nogalski et al. 54 One thousand cells were seeded in each coated plate with cultured medium and treatments and cultured for 24 h. Then images were taken using a 1.25 × objective and areas of each single cell track were measured using WIS-PhagoTracker software developed by the Weizmann Institute of Science, Rehovot, Israel. 55, 56 Only tracks containing one cell were used in the analysis.
Analysis and statistics. Fluorescence microscopy data were collected using a 20 × objective enabling capture of~50 cells per image field. For all experiments, multiple fields were acquired from each coverslip or well of 96-well plate, and the data were pooled from three to four independent coverslips or wells acquired on at least two different days from independent cultures. When comparison of different cell lines was required, the cells were cultured at the same density and passaged in parallel, and data were acquired on the same day. All fluorescence intensities were background subtracted. Rhod-2 signal was normalized to the initial fluorescence value F 0 and expressed as F/F 0 . Data were summarized as mean ± S.E. and differences between means assessed using the Student's t-test for unpaired comparisons. A one-way ANOVA with Fisher's least significant difference post-hoc analysis was used for multiple comparisons. For all tests the differences between means were accepted as statistically significant at the 95% confidence level (Po0.05).