MFF Regulation of Mitochondrial Cell Death Is a Therapeutic Target in Cancer

As signaling hubs in the eukaryotic cell (1), mitochondria control multiple forms of regulated cell death (2), including apoptosis (3) and necrosis (4). These processes involve a sudden increase in the permeability of the mitochondrial outer membrane (2), which, through as yet unclear mechanisms, dissipates inner membrane potential, uncouples the electron transport chain, and releases cell death–promoting proteins in the cytosol (5). The regulators of mitochondrial cell death have remained mostly elusive (6), and the composition of a putative permeability transition pore is still much debated (7). However, there is agreement that isoforms of voltage-dependent anion channel (VDAC; ref. 8) are key effectors of mitochondrial outer membrane permeability coupling to organelle cell death (9). These mechanisms are invariably subverted in cancer (10), where decreased mitochondrial permeability promotes tumor cell survival and confers treatment resistance (3).

A second process commonly exploited in cancer is mitochondrial dynamics (11), which controls the size, shape, and subcellular position of mitochondria. Although important to titrate organelle fitness under stress conditions (12), deregulated mitochondrial dynamics, in particular fission has been implicated in MAPK- (13) and Ras-dependent tumorigenesis (14), as well as tumor cell invasion and metastatic spreading, in vivo (15, 16). There is evidence that mitochondrial dynamics couples to apoptosis (17), but a direct, mechanistic link between mitochondrial fission and effectors of organelle cell death, especially in cancer, has not been demonstrated.

In this study, we uncovered a novel, therapeutically actionable interface between a regulator of mitochondrial fission, mitochondrial fission factor (MFF; ref. 18) and VDAC1 in the regulation of organelle cell death, selectively in cancer (19).

A clinically annotated series of 72 patients with histologically confirmed diagnosis of non–small cell lung cancer (NSCLC) was used in these studies (Supplementary Table S1). Archival tissues and clinical records were obtained from Fondazione IRCCS Ca′ Granda Hospital (Milan, Italy) as described previously (19) and written informed consent was waived for the retrospective nature of the study. NSCLC and normal bronchus tissue samples were arranged in tissue microarray blocks for IHC evaluations. For organotypic tissue cultures, one breast cancer and two NSCLC tissue samples (Supplementary Table S1) were subject to tissue slicing and ex vivo culturing as described previously (20). Cultures were treated with cell-permeable (D) MFF peptidomimetic or cell-permeable scrambled peptide (50–100 μmol/L) for 2 or 24 hours. At harvesting, tissues were formalin-fixed, paraffin-embedded and processed for IHC analyses.

Four-μm-thick sections from each tissue block were stained with an antibody to MFF (Protein Tech#17090-1-AP), Ki67 (clone 30‐9; Ventana Medical Systems, Roche Group) or cleaved caspase-3 (9661; Cell Signaling Technology) using diaminobenzidine. IHC was performed as described previously (19). Two pathologists (V. Vaira and S. Ferrero) blinded to clinical data evaluated and scored all slides. When discrepancies occurred, the case was further reviewed to reach an agreement score.

Human prostate adenocarcinoma (LNCaP, C4–2, C4–2B, PC3, and DU145), NSCLC (A549 and H460), breast adenocarcinoma (MDA-231 and MCF7), metastatic melanoma (A2058), normal prostate epithelial (RWPE-1), and human glioblastoma (LN229) cells were obtained from the ATCC, and maintained in culture according to the manufacturer's specifications. Benign prostatic hyperplasia (BPH-1) cells were a gift from Dr. Simon Hayward (Vanderbilt University, Nashville, TN) and primary human foreskin fibroblasts (HFF) were established at The Wistar Institute (Philadelphia, PA). WM983B and WM35 human metastatic melanoma cell lines were established at The Wistar Institute as described previously (21). Resistant cell lines were maintained in the presence of PLX4720 (3 μmol/L) plus PD0325901 (300 nmol/L) as described previously (22). In all conditions, cell passaging was limited to <40 passages from receipt and cell lines were authenticated by short tandem repeat profiling with AmpFlSTR Identifiler PCR Amplification Kit (Life Technologies) at the Wistar Institute's Genomics Shared Resource (19). Mycoplasma-free cultures were confirmed at the beginning of the studies, and every 2 months afterwards, by direct PCR of cultures using Bioo Scientific Mycoplasma Primer Sets (#375501) and Hot Start polymerase (Qiagen).

The following antibodies to MFF (Protein Tech #17090-1-AP), Hexokinase-I (HK-I, Cell Signaling Technology #2024), Hexokinase-II (HK-II, Cell Signaling Technology #2867), VDAC (Cell Signaling Technology #4866), PARP (Cell Signaling Technology #9532), Asp214-cleaved PARP (Cell Signaling Technology #9541), Asp175-cleaved Caspase-3 (Cell Signaling Technology #9661), HMGB1 (Cell Signaling Technology #6893), cyclophilin A (Cell Signaling Technology #2175), Flag (Sigma-Aldrich #F1804), β-tubulin (Sigma-Aldrich #T8328), or β-actin (Sigma-Aldrich #A5441) were used for Western blotting. Libraries of MFF peptides and their scrambled sequences used as control (purity >90%) were synthesized by Thermo Fisher Scientific or GenScript. Z-VAD-fmk and cyclosporin A were purchased from Selleckchem.

cDNA clones encoding human MFF1, MFF5, or control vector were purchased from GeneCopoeia (catalog no. EX-Z4766 and EX-Z0675). An MFF2 cDNA was obtained from Addgene. An MFF4 cDNA was constructed by mutagenesis of the MFF5 cDNA using QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies) and confirmed by DNA sequencing. Mutations in the MFF sequence were also generated using QuikChange II and confirmed by DNA sequencing. An MFF2 cDNA deleted in the VDAC-binding region ²²³SARGILSLIQSSTRRAYQQILDVL²⁴⁶ was generated by site-specific mutagenesis using QuickChange II and confirmed by DNA sequencing. Transfection of plasmid DNA (1 μg) was carried out using 2 μL X-Treme gene HP (Roche) for 24 hours.

PC3 cells were transfected with MFF1 cDNA, solubilized in lysis buffer (20 mmol/L Tris, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA) in the presence of 1% CHAPS, EDTA-free Protease Inhibitor Cocktail (Sigma-Aldrich), and Phosphatase Inhibitor Cocktail PhosSTOP (Roche). After sonication, lysates were centrifuged at 15,000 × g for 30 minutes, and cell extracts were incubated with anti-Flag–conjugated agarose beads (Sigma-Aldrich) for 4 hours at 4°C (19). After five washes in lysis buffer, MFF-associated proteins were separated by SDS gel electrophoresis. The entire gel region was excised and digested with trypsin, as described previously (23). Tryptic peptides were analyzed by LC/MS-MS on a Q Exactive HF Mass Spectrometer (Thermo Fisher Scientific) coupled with a Nano-ACQUITY UPLC System (Waters). Samples were injected onto a UPLC Symmetry Trap Column (180 μm i.d. × 2 cm packed with 5 μm C18 resin; Waters), and tryptic peptides were separated by RP-HPLC on a BEH C18 Nanocapillary Analytical Column (75 μm i.d. × 25 cm, 1.7-μm particle size; Waters) using a 95-minute gradient. Eluted peptides were analyzed by the mass spectrometer set to repetitively scan m/z from 400 to 2,000 in positive ion mode. The full mass spectrometry scan was collected at 60,000 resolution followed by data-dependent MS-MS scans at 15,000 resolution on the 20 most abundant ions exceeding a minimum threshold of 20,000. Peptide match was set as preferred, the exclude isotopes option and charge-state screening were enabled to reject singly and unassigned charged ions. Peptide sequences were identified using MaxQuant 1.5.2.8 (24). MS-MS spectra were searched against the UniProt human protein database (August 2015) using full tryptic specificity with up to two missed cleavages, static carboxamidomethylation of Cys, and variable oxidation of Met and protein N-terminal acetylation. Consensus identification lists were generated with FDR of 1% at protein and peptide levels.

Proteomics intensity data were floored to the minimum detected signal (intensity of 10⁵) and log₂ scaled. Proteins were then annotated as mitochondrial related using MitoCarta 2.0 database (25). Proteins detected in both MFF experiments with at least 10 peptides, as well as at least 10 MS-MS spectra counts and at least 10-fold higher intensity than controls were reported as significant interacting proteins (25). Ingenuity pathway analysis (IPA, Qiagen, www.qiagen.com/ingenuity) was used to identify all known protein–protein interactions between proteins in the list of candidates and their involvement in mitochondrial functions.

Cells were harvested and lysed with IP buffer (50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, and 1 mmol/L EDTA) containing 1% CHAPS, EDTA-free Protease Inhibitor Cocktail (Sigma-Aldrich), and Phosphatase Inhibitor Cocktail PhosSTOP (Roche). After sonication, lysates were centrifuged at 15,000 × g for 30 minutes, and cell extracts were incubated with anti-Flag–conjugated beads (Sigma-Aldrich). The precipitates were washed five times, separated by SDS-PAGE, and processed for Western blotting, as described previously (19). For analysis of direct protein–protein interaction, increasing concentrations (0–250 ng) of bead-coupled Flag-MFF1 proteins affinity-purified from PC3 cells were rinsed with reaction buffer (25 mmol/L Hepes, pH 7.5) and mixed with recombinant VDAC1 in 500 μL of 25 mmol/L Hepes (pH 7.5) for 2 hours at 4°C. Upon completion of the binding reaction, 50 μL of supernatant from each sample (unbound material) and 10 μL out of 100 μL eluate pool obtained from the beads were separated by SDS gel electrophoresis followed by Western blotting with antibodies to MFF or VDAC.

A full-length human VDAC1 cDNA (hVDAC1) was cloned into pET28b vector containing an N-terminal hexahistidine-SUMO fusion tag, with a flexible linker cleavable by tobacco etch virus (TEV) protease. The hVDAC1 was expressed in E.coli ScarabXpress T7 lac–competent cells (Scarab Genomics) for 16 hours at 16°C using 1 mmol/L IPTG (Denville Scientific Inc.). The cells were harvested by centrifugation and lysed on ice via sonication in buffer containing 25 mmol/L Tris-HCl (pH 7.5), 1 mol/L urea, 1 mol/L KCl, 5% glycerol, 1 mmol/L benzamidine, and 1 mmol/L PMSF (Ni Buffer A). After centrifugation at 18,000 rpm for 20 minutes at 4°C, the cell pellet was washed extensively in Ni Buffer A with 1% Triton X-100 and then solubilized in buffer containing 20 mmol/L Tris-HCl (pH 7.9), 500 mmol/L NaCl, 4 mol/L guanidine-HCl, and 10% glycerol for 45 minutes with gentle stirring. The supernatant was collected following centrifugation at 20,000 rpm for 10 minutes at 4°C. The protein was purified over nickel-nitrilotriacetic acid (Ni-NTA, Qiagen) column, buffer-exchanged to 25 mmol/L Tris-HCl (pH 7.5), 500 mmol/L KCl, 5% glycerol, 1 mmol/L benzamidine, and 1 mmol/L PMSF (Ni buffer C) with 2% n-Octylglucoside (Research Products International), eluted with 300 mmol/L imidazole, and treated overnight with TEV at 4°C to cleave the His-SUMO tag. The protein was then buffer exchanged to buffer C with 100 mmol/L salt and loaded onto tandem HS(poros)-HQ(poros) column to remove the TEV and the His-SUMO fusion tag. The cleaved, full-length hVDAC1 was collected from the HS-HQ flow through, concentrated using amicon ultra filter (10 kDa cutoff), and used for further experiments.

Isothermal titration calorimetry (ITC) experiments were performed using MIcroCal iTC200 (Malvern). Purified full-length hVDAC1 was buffer exchanged into 25 mmol/L HEPES-KOH (pH 7.5), 0.1 mol/L KCl, 5% glycerol, 2% n-Octylglucoside, and 1 mol/L TCEP (ITC buffer). Wild-type (WT) MFF peptide #8–11 corresponding to the minimal VDAC binding site, ²²³SARGILSLIQSSTRRAYQQILDVL²⁴⁶, and its scrambled control, SSQLRYLARSQRITIQLIAGS (see below) were also prepared in ITC buffer. The ITC binding experiments were carried out at 20°C. Peptides at a concentration of 100 μmol/L were added by 2.47 μL injections to 10 μmol/L hVDAC1. The data collected was processed in MicroCal Origin Software (Malvern).

The hVDAC1-MFF model was generated using the CABS-dock server, which uses an efficient protocol for the flexible docking of proteins and peptides (26, 27). The coordinates of hVDAC1 (PDB ID: 2JK4; ref. 28) and the WT MFF peptide sequence (SARGILSLIQSSTRRAYQQIL) were provided for the modeling. The MFF peptide docking into hVDAC1 structure was carried out in three steps as described previously (26, 27). In this study, we used the best binding mode of the peptide from the 10 top scored.

A library of partially overlapping synthetic peptides duplicating the entire MFF1 sequence is presented in Supplementary Table S1. A library of deletion mutant peptides based on MFF peptide #8 sequence ²¹⁷DGANLSSARGILSLIQSSTRRAYQQILDVL²⁴⁶ was also synthesized (Supplementary Table S2). The minimal MFF interacting sequence with VDAC, designated peptide #8–11 with the sequence ²²³SARGILSLIQSSTRRAYQQIL²⁴³, and its corresponding scrambled version, SSQLRYLARSQRITIQLIAGS were also synthesized. To target the MFF–VDAC complex in tumor cells, the MFF peptide #8–11 was made cell permeable with the addition of an NH₂-terminus biotin-Ahx linker and HIV Tat cell-penetrating sequence RQIKIWFQNRRMK. A cell-permeable control scrambled peptide #8–11 and an MFF peptide #8–11 mutant containing the double mutation Arg225Asp/Arg236Asp (DD) were also synthesized. To generate a clinical candidate of the MFF-VDAC pathway, in vivo, a retro-inverso d-enantiomer peptidomimetic of MFF #8–11 sequence was synthesized containing all d-amino acid in the reverse orientation, as described previously (29). A scrambled d-enantiomer peptide was also synthesized as control. All peptides were synthesized with >95% purity. For analysis of intramitochondrial accumulation using the Colorimetric Biotin Assay Kit (Sigma #MAK171), PC3 cells were incubated with biotin-conjugated, cell-permeable MFF (D) 8–11 peptidomimetic (10 μmol/L), or cell-permeable scrambled peptide (10 μmol/L) for 30 minutes at 22°C. Isolated mitochondrial extracts were then treated with HABA (2-(4-Hydroxyphenylazo) benzoic acid)/avidin assay mixture for 5 minutes at 22°C and absorbance was quantified at 500 nm. In this assay, accumulation of the biotinylated peptidomimetic in mitochondrial samples displaces HABA from HABA/Avidin complex, thus leading to reduction in absorbance.

PC3 cells were treated with cell-permeable scrambled peptide or cell-permeable MFF #8–11 peptide, washed three times in PBS, pH 7.4, and analyzed on a FACSCalibur flow cytometer, with the TMRE signal as FL1. Intact cells were gated in the FSC/SSC plot to exclude small debris. Cells treated with 20 μmol/L FCCP for 10 minutes prior to TMRE staining were used as a negative control. The resulting FL1 data were plotted on a histogram. For quantification of membrane potential of purified mitochondria, mitochondria isolated from PC3 cells (50 μg) were suspended in 400 μL of SB buffer (0.2 mol/L sucrose, 10 mmol/L Tris-MOPS, pH 7.4, 5 mmol/L succinate, 1 mmol/L sodium phosphate, 10 μmol/L EGTA-Tris, and 2 μmol/L rotenone). Samples were treated with control scrambled peptide or the various MFF peptides for 20 minutes at 22°C, and incubated with 100 nmol/L TMRE for 30 minutes at 22°C. After two washes with SB buffer, mitochondria were recovered by centrifugation at 6,000 × g for 10 minutes, suspended in 100 μL of SB buffer, and mitochondrial membrane potential was quantified on a florescence microplate reader (Ex/Em = 535/595 nm, Beckman Coulter). Samples treated with 2 mmol/L CaCl₂ were used as control for the lowest membrane potential (fully depolarized state).

Primary, patient-derived glioblastoma (GBM) neurospheres were isolated and maintained in culture as described previously (30). Tissue samples were obtained from chemotherapy- or radiotherapy-naïve patients with confirmed histologic diagnosis of GBM undergoing treatment at the Neurosurgery Division of Fondazione IRCCS Ca′ Granda Ospedale Maggiore Policlinco (Milan, Italy). Written informed consent was obtained from all patients. Neurospheres were cyto-spinned on charged slides (Thermo Fisher Scientific) and treated with cell-permeable scrambled peptide (50 μmol/L) or cell-permeable retro-inverso d-enantiomer MFF peptide (10–50 μmol/L) in PBS, pH 7.4, for up to 24 hours. Control-normalized changes in cell viability under the various conditions were determined by the ratio of calcein (live cells) and To-Pro (dead cells) staining and fluorescence microscopy (Eclipse Ti-E, Nikon Instruments).

Studies involving vertebrate animals (rodents) were carried out in accordance with the Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011). Protocols were approved by the Institutional Animal Care and Use Committee of The Wistar Institute (protocol #112625 and 112610). Groups of 4- to 6-week-old male athymic nude mice [Crl:NU(NCr)-Foxn1^nu, Charles River Laboratory] were injected subcutaneously with PC3 cells (5 × 10⁶ cells in 50% Matrigel) and superficial tumors were allowed to grow to approximately 50–75 mm³ volume. Tumor-bearing animals were randomized into two groups receiving cell-permeable scrambled peptide (50 mg/kg) or cell-permeable retro-inverso d-enantiomer MFF peptide (10 or 50 mg/kg) as daily intraperitoneal injections in PBS, pH 7.2. Tumor growth in the animal groups was measured with a caliper during a 2-week time interval and quantified. For in vivo studies using patient-derived xenograft (PDX) melanoma models, NOD SCID γ (NSG) mice were implanted with WM3960 tumor tissue. The WM3960 model was derived from a patient with metastatic (stage IV) melanoma who progressed while treated with the standard-of-care combination of the BRAF inhibitor dabrafenib and the MEK inhibitor trametinib. When implanted tumors reached a palpable volume (100–200 mm³), tumor-bearing animals were randomized into three groups receiving cell-permeable scrambled peptide (50 mg/kg), dabrafenib/trametinib chow (150 and 1.5 mg/kg, respectively), or cell-permeable retro-inverso d-enantiomer MFF peptide (50 mg/kg) plus dabrafenib/trametinib chow as daily intraperitoneal injections in PBS, pH 7.2.

Data were analyzed using the two-sided unpaired t, χ², or Kruskal–Wallis (with P value correction for multiple testing) tests using a GraphPad software package (Prism 8.1) for Windows. Data are expressed as mean ± SD of replicates from a representative experiment out of at least two or three independent determinations or as mean ± SD of three individual experiments. A P < 0.05 was considered as statistically significant.

We began this study by examining the expression of MFF, a mitochondrial outer membrane receptor for the fission effector, Dynamin-Related Protein-1 (Drp1; ref. 18), in a clinically annotated cohort of patients with NSCLC (Supplementary Table S3). We found that MFF was differentially overexpressed in patients with NSCLC, including cases of adenocarcinoma and squamous cell carcinoma (SCC), compared with normal bronchus (Fig. 1A and B). In this cohort, MFF expression did not correlate with tumor grade (Fig. 1C), nodal status (Supplementary Fig. S1A), or tumor size (Supplementary Fig. S1B). In independent confirmatory studies, MFF was also differentially overexpressed in patient samples of prostatic adenocarcinoma, compared with normal prostate (19). Similarly, MFF was highly expressed in multiple tumor cell lines, including prostate cancer (Supplementary Fig. S1C).

To uncover the potential function(s) of MFF in cancer, we next carried out a proteomics screen for MFF-associated proteins in prostate adenocarcinoma PC3 cells. We identified 42 proteins that associate with MFF (Supplementary Fig. S2A), including regulators of mitochondrial cell death (2), such as VDAC1, -2, and -3 (8), metaxin-1 and -2 (MTX1 and MTX2), and hexokinase-I and -II (HK-I and HK-II; Fig. 1D; Supplementary Fig. S2A). Other MFF-associated proteins comprised mitochondrial receptors for protein import (TOMM70 and TOMM40), protein sorting (SAM50), and cristae remodeling (DNAJC11, APOOL-MIC27 and CHCHD3; Fig. 1D; Supplementary Fig. S2A). Therefore, we focused on VDAC1 as the most abundant regulator of mitochondrial outer membrane permeability in cancer (8).

Processing of the human MFF locus is complex and predicted to generate at least five protein isoforms by alternative splicing (Supplementary Fig. S2B). Of these, MFF1 and MFF2 were the most prominently expressed in PC3 and DU145 prostate cancer cells (Supplementary Fig. S2C). Consistent with the results of the proteomics screen, Flag-MFF1 or Flag-MFF2 bound VDAC1 in coimmunoprecipitation experiments, in vivo (Fig. 1E; ref. 19). HK-II, which associates with VDAC at the mitochondrial outer membrane and is important for cell survival (6), was also present in MFF–VDAC1 complexes (Fig. 1E). Similar results were obtained in analysis of endogenous proteins, as endogenous MFF associated with endogenous HK-I, HK-II, and VDAC1 in coimmunoprecipitation experiments from PC3 cells (Fig. 1F). We next asked whether this interaction was direct, and we mixed increasing concentrations of recombinant VDAC1 with affinity-purified Flag-MFF1. In these experiments, Flag-MFF1 directly bound recombinant VDAC1 in a concentration-dependent manner, whereas uncoupled beads had no effect (Fig. 1G). Next, we looked at the role of other MFF isoforms (Supplementary Fig. S2B) in VDAC1 binding. In coimmunoprecipitation experiments of differentially tagged recombinant proteins, we found that MFF isoforms assembled in homodimers, such as MFF1-MFF1 (Fig. 1H), as well as heterodimers, including MFF1-MFF2 (Fig. 1I; Supplementary Fig. S2D) and MFF1-MFF4 (Supplementary Fig. S2E and S2F) in the VDAC1 complex.

On the basis of these results, we next asked whether MFF affected the function of VDAC1 in mitochondrial outer membrane permeability (9). To test this hypothesis, we examined partially overlapping synthetic peptides duplicating the entire MFF1 sequence (Supplementary Table S1) in a high-throughput assay that measures depolarization of isolated tumor (PC3) mitochondria (Fig. 2A). We found that peptide #8 corresponding to the MFF sequence Asp217-Leu246 depolarized tumor mitochondria, whereas none of the other MFF peptides had an effect (Fig. 2B). This response was selective because MFF peptide #8 depolarized isolated mitochondria from PC3 cells in a concentration-dependent manner, whereas BPH-1 cells were not affected (Supplementary Fig. S3A). Among MFF isoforms (Supplementary Fig. S2B), only MFF1 and MFF2 contained the peptide #8 region (Supplementary Fig. S3B), consistent with the ability of these isoforms to directly bind VDAC1, in vivo (Fig. 1E and F) and in vitro (Fig. 1G).

To quantify the interaction between MFF peptide #8 and VDAC1, we next performed ITC studies. In these experiments, MFF peptide #8 bound human recombinant VDAC1 with a Kd of approximately 12 μmol/L, whereas a control peptide with the scrambled sequence had no effect (Fig. 2C). Next, we asked whether this MFF sequence bound VDAC1, in vivo. Accordingly, bead-immobilized MFF peptide #8 affinity purified HK-I, HK-II, and VDAC1 from PC3 cell extracts, in vivo (Fig. 2D). Molecular modeling studies using the CABS dock server (26, 27) placed the MFF peptide #8 sequence in the interior hole of the VDAC1 ring (Fig. 2E), with an average root mean square deviation of 4.1, cluster density of 36, and number of contact elements of 148. This model identified three MFF residues that coordinate the binding interface with VDAC1: MFF Arg225 is predicted to contact VDAC1 Glu40 and Glu59, MFF Arg236 contacts the side chains of VDAC1 Thr6, Gln154, and Asp128, and MFF Gln241 is within coordinating distance of VDAC1 Asn185 and Glu189 (Fig. 2F).

On the basis of these data, we next asked whether MFF peptide #8 acted as a competitive inhibitor of the MFF–VDAC1 complex. Consistent with this, addition of MFF peptide #8 to DU145 (Fig. 2G, top) or PC3 (Fig. 2G, bottom) cell extracts inhibited the association of MFF with VDAC1 or HK-I, in coimmunoprecipitation experiments, in vivo (Fig. 2G). To independently test our molecular modeling predictions, we next generated variants of the MFF1 protein containing single amino acid substitutions, Arg225Asp, Arg236Asp, or Gln241Ala predicted to disrupt the contact sites with VDAC1 (Fig. 2F). In coimmunoprecipitation experiments, each of the three MFF1 mutants showed significantly reduced binding to VDAC1 and HK-II, compared with WT MFF1 (Fig. 2H and I).

Next, we wished to further narrow the MFF-VDAC1 binding interface, and we synthesized variants of peptide #8 with amino- or carboxy-terminus deletions (Supplementary Table S2). Using our high-throughput assay of mitochondrial depolarization (Fig. 2A), we found that variants #1, #2, and #6 of peptide #8 depolarized isolated tumor mitochondria (Supplementary Fig. S3C). None of the other MFF peptide #8-derived sequences was effective (Supplementary Fig. S3C). Therefore, we synthesized a new MFF peptide #8–11 containing the sequence Ser223-Leu243, corresponding to a minimal binding site for VDAC1 (Supplementary Fig. S3B).

To understand the function of an MFF–VDAC1 complex in tumor cells, we next made the MFF peptide #8–11 cell-permeable via the addition of an NH₂-terminus RQIKIWFQNRRMK HIV-Tat cell-penetrating sequence (31). The cell-permeable MFF peptide, but not a cell-permeable scrambled sequence, induced sudden loss of mitochondrial membrane potential, by single-cell time-lapse videomicroscopy (Fig. 3A and B), as well as analysis of whole tumor cell populations, by flow cytometry (Supplementary Fig. S4A and S4B). Consistent with increased mitochondrial outer membrane permeability (6), treatment with the cell-permeable MFF peptide induced cytochrome c release from PC3 cells (Fig. 3C), whereas BPH-1 cells were minimally affected (Fig. 3D). An unrelated MFF #1 peptide did not induce cytochrome c release from PC3 cells (Fig. 3C). In line with acute loss of mitochondrial integrity, treatment with the cell-permeable MFF peptide #8–11 killed (IC₅₀ ∼10 μmol/L) multiple prostate cancer cell types in a concentration- and time-dependent manner (Fig. 3E). Normal prostatic epithelial cells or fibroblasts were not affected, and a cell-permeable scrambled peptide did not kill normal or prostate cancer cells (Fig. 3E). In addition, treatment with the cell-permeable MFF peptide #8–11 comparably killed isogenic pairs of drug-sensitive and drug-resistant melanoma cells (Fig. 3F; Supplementary Fig. S4C). To validate the specificity of these findings, we next introduced the Arg225Asp/Arg236Asp double mutation (DD) that prevents VDAC1 binding in the cell-permeable MFF peptide. This mutant peptide entirely lost the ability to induce tumor (PC3) cell death, whereas WT MFF peptide #8–11 efficiently killed tumor cells in a time- and concentration-dependent manner (Fig. 3G).

Next, we used these reagents to characterize more in detail the type of mitochondrial cell death induced by MFF-VDAC1 targeting. In these experiments, a pan-caspase inhibitor, Z-VAD-fmk did not significantly reduce tumor cell death induced by the cell-permeable MFF peptide #8–11 (Supplementary Fig. S4D). Instead, an inhibitor of regulated necrosis (4), cyclosporine A, partially reversed tumor cell killing induced by MFF peptide #8–11 (Fig. 3H; Supplementary Fig. S4E). Biochemically, tumor cell death in these settings was accompanied by loss of plasma membrane integrity, another marker of regulated necrosis (6), as judged from the release of HMGB1 and cyclophilin A from tumor cells (Supplementary Fig. S4F). Conversely, a cell-permeable scrambled peptide or MFF-DD–mutant peptide #8–11 did not cause HMGB1 release (Supplementary Fig. S4F and S4G). The mitochondrial matrix protein, Cyclophilin D (CypD) is an essential effector of regulated necrosis (32), and its role in this pathway was next investigated. Treatment with the cell-permeable MFF peptide #8–11 killed immortalized WT mouse embryonic fibroblasts (MEF) in a concentration-dependent manner (Fig. 3I). In contrast, CypD^−/− MEFs were resistant to cell death induced by the MFF peptide (Fig. 3I). Finally, we tested the effect of MFF targeting on immortalized VDAC1^−/− MEF. Consistent with the data above, treatment with MFF peptide #8–11 efficiently killed WT MEF in a time- and concentration-dependent response (Fig. 3J, left). Conversely, VDAC1^−/− immortalized MEF were insensitive to cell death in these settings (Fig. 3J, right).

From these proof-of-concept studies, we next generated a clinical candidate to target the MFF–VDAC1 complex for cancer therapy, in vivo. To accomplish this, we synthesized a cell-permeable retro-inverso peptidomimetic of the Ser223-Leu243 MFF sequence, containing all d-amino acids in the reverse orientation (Fig. 4A) to improve stability during systemic administration (33). This retro-inverso d-enantiomeric MFF peptidomimetic, designated MFF (D) 8–11 rapidly (<20 minutes) accumulated intracellularly (Fig. 4B) and became associated with isolated mitochondria (Supplementary Fig. S5A). Functionally, the MFF (D) 8–11 peptidomimetic was more active than the l-enantiomeric (L) MFF peptide #8–11 at depolarizing tumor mitochondria (Fig. 4C) and killed tumor cells in a time- and concentration-dependent manner also more efficiently than the MFF (L) peptide #8–11 (Fig. 4D). By time-lapse videomicroscopy, the MFF (D) 8–11 peptidomimetic killed tumor cells within 15 minutes of exposure with morphologic features of regulated necrosis (Supplementary Movie S1). A cell-permeable scrambled peptidomimetic had no effect (Supplementary Movie S2).

Although xenograft models are suboptimal due to the lack of immune competence and superficial, as opposed to visceral tumor engraftment, daily systemic administration of the MFF (D) 8–11 peptidomimetic (10–50 mg/kg i.p.) inhibited PC3 tumor growth in immunocompromised mice in a concentration-dependent manner (Fig. 4E; Supplementary Fig. S5B). Systemic administration of a cell-permeable scrambled peptidomimetic had no effect (Fig. 4E; Supplementary Fig. S5B). Next, we examined the effect of MFF-VDAC1 targeting in PDX models of melanoma resistant to the combination therapy of dabrafenib (mutant BRAF inhibitor) plus trametinib (MEK inhibitor). When exposed to vehicle or the combination dabrafenib/trametinib, these PDX tumors grew exponentially in immunocompromised mice (Fig. 4F). Conversely, treatment with the MFF (D) 8–11 peptidomimetic significantly inhibited the growth of drug-resistant PDX melanoma (Fig. 4F), without overt signs of toxicity or animal weight loss (Supplementary Fig. S5C).

In a third preclinical model, treatment with the MFF (D) 8–11 peptidomimetic inhibited cell proliferation (loss of Ki-67⁺ cells) and induced markers of cell death (increased cleaved caspase-3–positive cells) in patient-derived NSCLC (Fig. 4G and H) and breast adenocarcinoma (Fig. 4I; Supplementary Fig. S5D) 3D organoids in culture. A cell-permeable scrambled peptidomimetic had no effect on 3D tumor organoids (Fig. 4G–I; Supplementary Fig. S5D). Consistent with these results, matched tumor cell lines, NSCLC A549 and H460 or breast adenocarcinoma MDA-231 and MCF-7 cells expressed MFF1, MFF2, and MFF4 (Supplementary Fig. S5E), and exhibited mitochondrial depolarization and nearly complete killing after treatment with the MFF (D) 8–11 peptidomimetic, but not a scrambled sequence (Supplementary Fig. S5F). Finally, we looked at patient-derived GBM neurospheres as a preclinical model enriched in cancer stem cells. In validation experiments, the cell-permeable MFF (D) 8–11 peptidomimetic efficiently killed GBM LN229 cells in a time- and concentration-dependent manner (Supplementary Fig. S6A). Normal HFFs were not affected and a control scrambled sequence did not cause LN229 cell killing (Supplementary Fig. S6A). Consistent with these data, treatment with the MFF (D) 8–11 peptidomimetic, but not a scrambled peptide, was highly cytotoxic for patient-derived GBM neurospheres (Fig. 4J; Supplementary Fig. S6B), resulting in time- and concentration-dependent loss of viability, as early as 1 hour after drug exposure (Fig. 4K).

In this study, we have shown that a regulator of mitochondrial fission, MFF is overexpressed in NSCLC and assembles in isoform-specific (MFF1 and MFF2) homo- and heterodimeric complexes with VDAC1 at the mitochondrial outer membrane. Peptidyl mimicry of the MFF–VDAC1 interaction is feasible and a cell-permeable MFF Ser223-Leu243 d-enantiomeric peptidomimetic disrupted MFF–VDAC1 complex(es), in vivo, triggered mitochondrial outer membrane permeability, and killed genetically disparate tumor types, including drug-resistant melanoma, predominantly by regulated necrosis. Treatment with the MFF peptidomimetic was well-tolerated and delivered potent anticancer activity in multiple preclinical, patient-derived tumor models.

The molecular details of how MFF regulates VDAC1-directed mitochondrial outer membrane permeability (2), including potential posttranslational modifications, remain to be fully elucidated (19). Although only MFF1 and MFF2 contain the sequence that directly binds VDAC1, structure–function studies demonstrated that other MFF isoforms are recruited in a multimeric complex with VDAC1, in vivo. Nuclear magnetic resolution and reconstitution studies with isolated mitochondria have also shown that VDAC1 forms oligomers that assemble in “open” or “closed” configurations (34) to regulate channel conductance (35). In this context, the predicted insertion of MFF multimers in the interior hole of the VDAC1 ring appears ideally suited to shut off oligomeric channel conductance and oppose cell death initiation in tumors (2). A similar scenario of “channel closure” has been proposed for other VDAC-binding proteins with prosurvival functions, including HK-I (36) and HK-II (37), which are also recruited to MFF–VDAC1 complex(es).

Although the role of VDAC in channel conductance is established (35), how this pathway connects to downstream steps of mitochondrial cell death remains controversial (6). In fact, earlier mouse-knockout studies suggested that none of the three VDAC isoforms may be required for cell death, in vivo (38). On the other hand, this pathway may be differentially exploited to preserve mitochondrial integrity in transformed cells. Accordingly, VDAC1 is a recognized regulator of tumor cell death (8), and actionable therapeutic target (39), VDAC2 participates in Bax-mediated apoptosis during tumor progression (40), and cysteine oxidation of VDAC3 may control pore conductance and ROS sensing (41), including in cancer. A similar context-specific rewiring may apply to MFF. Although MFF-knockout mice showed better tissue protection and reduced mitochondrial apoptosis after ischemia-reperfusion injury (42), MFF was recently characterized as a critical survival factor in prostate cancer, preserving the cancer stem cell pool (43), and in-line with the findings presented here. In addition, a prosurvival function of MFF via VDAC1 binding as described here may be important to oppose the propensity to apoptosis associated with mitochondrial fragmentation (17), and fully enable the role of deregulated fission in primary (13, 14) and metastatic (15, 16) tumor growth.

Targeting protein–protein interactions at the mitochondrial outer membrane is a validated therapeutic strategy, and has produced the first apoptosis modifier clinically approved for cancer therapy (44). Here, comprehensive studies of peptidyl mimicry identified a cell-permeable MFF Ser223-Leu243 d-enantiomeric peptidomimetic as a potential clinical candidate of this pathway. Treatment with this agent induced hallmarks of acute mitochondrial dysfunction, with sudden depolarization of the organelle inner membrane, release of cytochrome c in the cytosol (6), and massive cell death in disparate tumor types, including drug-resistant melanoma. This cell death response, which was dependent on VDAC1, had hallmarks of both apoptosis, with cytochrome release and caspase activation, in vivo, as well as regulated necrosis (4), as characterized by loss of plasma membrane integrity, insensitivity to caspase inhibition, and requirement of CypD (32). The coexistence of both apoptosis and necrosis is common after acute mitochondrial dysfunction (45) and reflects mechanistic overlap of organelle cell death pathways (46).

In our preclinical evaluation, treatment with the cell-permeable MFF d-enantiomeric peptidomimetic was well-tolerated, and delivered potent anticancer activity in multiple, patient-derived tumor models representative of acquired drug resistance (dabrafenib/trametinib-resistant melanoma), 3D tumor architecture (breast adenocarcinoma and NSCLC organoids), and cancer stem cells (GBM neurospheres). Although the long-term safety of this regimen remains to be established, the encouraging therapeutic index observed here may reflect the differential distribution of MFF in cancer, compared with normal tissues, but also a general dependence on this pathway to preserve mitochondrial integrity in transformed cells. This is consistent with the importance of mitochondrial fitness to sustain bioenergetics (47), buffer ROS production (48), and elevate the threshold of “mitochondrial priming” in tumors (49). Compared with approved therapies that target Bcl2 family proteins at the mitochondrial outer membrane (3), disruption of the MFF–VDAC1 complex may offer broader efficacy in multiple, genetically heterogeneous malignancies and bypass drug resistance mechanisms maintained by prosurvival Bcl2 proteins.

No potential conflicts of interest were disclosed.

Conception and design: J.H. Seo, Y.C. Chae, D.C. Altieri

Development of methodology: Y.C. Chae, E. Agarwal

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.C. Chae, Y.G. Lee, H.-Y. Tang, D.W. Speicher, P.K. Shastrula, S. Ferrero, M. Caroli, D. Tosi, M. Giroda, V. Vaira, V.W. Rebecca, M. Herlyn, M. Xiao, D. Fingerman, A. Martorella, E. Skordalakes

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.V. Kossenkov, H.-Y. Tang, D.W. Speicher, A.M. Storaci, G. Gaudioso, V. Vaira

Writing, review, and/or revision of the manuscript: J.H. Seo, Y.C. Chae, A.V. Kossenkov, D.I. Gabrilovich, L.R. Languino, D.W. Speicher, D.C. Altieri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Agarwal

Study supervision: Y.C. Chae

Other (conceived the project - performed experiments of MFF–VDAC interaction, mitochondrial permeability transition, peptide mapping, bioenergetics, site-directed mutagenesis, and preclinical characterization of anticancer activity): J.H. Seo

We thank James Hayden and Frederick Keeney for assistance with time-lapse videomicroscopy in the Imaging Core Facility as well as the Proteomics and Metabolomics Core Facility for LC-MS/MS analysis. This work was supported by NIH grants P01 CA140043 (to D.C. Altieri, L.R. Languino, and D.W. Speicher), R35 CA220446 (to D.C. Altieri), R50 CA221838 (to H.-Y. Tang), and R50 CA211199 (to A.V. Kossenkov), by Fondazione Cariplo (2014–1148 to V. Vaira), by the Italian Minister of Health- Ricerca Corrente program 2017 (to S. Ferrero), and by the National Research Foundation of Korea funded by the Ministry of Education (2018R1D1A1B07048104 and 2018R1A6A1A03025810) and the Ministry of Science and ICT (2014M3A9D8034459). Y.C. Chae is the recipient of the Research Fund (1.170074.01) of Ulsan National Institute of Science and Technology and the National Research Foundation of Korea (2014M3A9D8034459) funded by the Ministry of Science and ICT. A.M. Storaci is supported by a fellowship from the Doctorate School in Molecular and Translational Medicine at the University of Milan. Support for core facilities utilized in this study was provided by Cancer Center Support Grant CA010815 to The Wistar Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.