Identification of the mitochondrial binding partner of MetF
DARTS is based on the changes in resistance to proteolysis due to conformational changes induced when a compound of interest and proteins bind13. In our previous study, we used DARTS-LC-MS/MS to identify targets of autophagy-related compounds27–30. Using this technique, we identified the potential mitochondrial target proteins of unmodified MetF.
To focus on the effects of MetF on mitochondria, mitochondrial proteins were isolated in HepG2 cells. MetF was then reacted with the protein pool at a concentration of 300 uM for 1 h at 4 ℃, and proteolysis of the proteome was performed by Pronase (Fig. 2a). LC-MS/MS analysis was performed on the control group, the pronase-treated group, and the pronase-and MetF-treated groups. Of the total 212 proteins, 64 proteins were degraded by more than 5% by pronase. Among these, there were 18 candidate target proteins that resisted pronase degradation due to their binding to MetF, resulting in an increase in sequence coverage of more than 5% (Fig. 2a,b). Reactome pathway analysis using the STRING database (https://string-db.org/) revealed that the candidate target proteins were involved in specific cellular processes, including selective autophagy, metabolism, PINK1-PRKN-mediated mitophagy, gluconeogenesis and the TCA cycle (Fig. 2c). To elucidate the molecular mechanisms underlying the effect of MetF on autophagy and cancer metabolism, VDAC1 was selected as a direct target. The sequence coverage of VDAC1 increased by 11.4% when combined with MetF, which is a higher FC value compared to other target candidates.
Next, we compared the expression levels of VDAC1 in HCC (Huh7, HepG2) and normal liver cells (Chang, LX-2). As shown in Fig. 2d, VDAC1 was significantly more highly expressed in HCC cells compared to Chang and LX-2 cells. In addition, the expression of VDAC1 between HCC and normal tissues was also examined to determine if there was a difference existed. RNA sequencing data including 115 HCC tissue samples and 52 adjacent non-tumor tissues (ANTTs) were obtained from the GEO database (http://www.ncbi.nlm.nih.gov/geo/). This analysis revealed a 2-fold upregulation of VDAC1 expression in HCC primary tumor tissues compared to adjacent non-tumor tissues, highlighting its strong association with HCC (Fig. 2e).
Identification of hotspots on VDAC1 for MetF binding using computational alanine scanning (CAS) mutagenesis
Next, we investigated the possibility of MetF binding to VDAC1 by computational analysis. Based on the results obtained by ensemble docking and cluster analysis using SIFt, two major clusters with the best average docking scores were selected for further analysis. From each cluster, the best docking pose based on docking scores were selected as representative structures for further investigation, namely Poses 1 (Fig. 3a) and 2 (Fig. 3b). These structures were equilibrated at 303.15 K and 1 atm using conventional MD simulation, and the last 10 ns of the 150 ns MD simulations were used to calculate the binding free energy of Poses 1 and 2. Per-residue energy decompositions were performed for the two selected docking poses to elucidate residues with significant contributions to the binding free energy of the VDAC1-MetF complex (Fig. 3c-f, Supplementary Fig. 5). These residues were then subjected to computational alanine scanning mutagenesis (CAS) to confirm their involvement in supporting the binding of MetF to VDAC1. For Pose 1, the following residues were considered for CAS: Pro4, Pro5, Asp9, Val143, Leu150, Val171, and His181 (Fig. 3c,e). For Pose 2, the following residues were considered: Asp9, Lys12, Ser13, Asp16, Gln179, Tyr195, Glu203, and Thr204 (Fig. 3d,f). The ΔΔG calculated for the residues in Poses 1 and 2 are shown in Fig. 3c and Fig. 3d respectively.
Based on the CAS calculation performed, four hotspots were proposed to be critical for MetF binding to VDAC1. The CAS performed for Pose 1 led to the identification of two possible binding hotspots, as shown by the significant destabilization of MetF binding upon mutation to alanine. These mutations are D9A and H181A, which resulted in ΔΔG of -20.77 and − 2.70 kcal/mol, respectively. On the other hand, the calculation performed for Pose 2 suggested three possible binding hotspots. One of the hotspots identified for Pose 2 was Asp9, which was also predicted by the CAS performed for Pose 1. However, the D9A mutation in Pose 2 caused less destabilization of MetF binding, as compared to Pose 1, with reported ΔΔG of -1.79 kcal/mol. The other predicted hotspots for Pose 2 are Asp16 and Glu203, which when mutated to alanine yielded ΔΔG of -2.14 and − 11.99 kcal/mol respectively. Therefore, we have proposed two hotspots with the lowest ΔΔG in each position for further investigation by experimental mutation studies. These two residues are Asp9 (Pose 1) and Glu203 (Pose 2).
Validation of binding hotspots between MetF and VDAC1 in vitro
To validate whether MetF binds to VDAC1, we monitored changes in the stability of VDAC1 to pronase depending on MetF concentration (Fig. 4a). The dose-dependent curve allowed us to evaluate the binding affinity of MetF and VDAC1. The half maximal effective concentration (EC50) of ligand and protein binding was estimated to be 300 µM.
According to the results of in silico modeling, we performed alanine mutagenesis for two residues, Asp9 and Glu203. The negative charge properties of Asp (D) and Glu (E) were lost by substitution with Ala (A). The resistance of VDAC1 to proteolysis by MetF was not restored in cells expressing VDAC1-D9A or VDAC1-E203A (Fig. 4b), suggesting that D9 and E203 are critical hotspot amino acid residues for the ionic interaction of MetF to bind to VDAC1.
Consistent with previous research indicating that MetF reduces mitoROS31, it effectively inhibits mitoROS in non-transfected cells or in cells transfected with VDAC1 WT. However, in cells transfected with mutant VDAC1, the binding of MetF to VDAC1 was blocked, preventing the ROS reduction effect (Fig. 4c). Taken together, our data demonstrate that VDAC1 is a target protein that directly binds MetF and particularly highlight the importance of Asp9 and Glu203 as binding sites.
MetF reduces ER-mitochondria interaction by targeting VDAC1 and increasing autophagy.
VDAC1 plays a role in regulating the flow of ions and molecules across the mitochondrial outer membrane, thereby affecting mitochondrial function. MetF treatment dose-dependently reduced mitoROS production (Fig. 5a,c) and also decreased ATP levels (Fig. 5e) in HepG2 cells.. When VDAC1 was genetically knocked down, similar phenotypic changes were observed (Fig. 5b,d,f), suggesting a potential regulatory role of MetF in modulating mitochondrial function through its interaction with VDAC1. Since an increase in the intracellular AMP:ATP ratio activates the AMPK pathway32, we examined the effect of MetF on AMPK-mTOR signaling. MetF treatment induces energy depletion, which leads to the activation of AMPK and subsequent the inhibition of mTOR. (Fig. 5g). As shown in Fig. 5h, when VDAC1 was knocked down, an increase in AMPK and a decrease in mTOR were also observed, supporting the notion that MetF directly binds to VDAC1 and regulates autophagy through the AMPK-mTOR signaling pathway.
Mitochondria-associated ER membranes (MAM), an interface connecting mitochondria and the ER, has received attention in regulating mitochondrial homeostasis including Ca2+ and lipid transport, mitochondrial dynamics, autophagy, apoptosis. Several protein interactions occur in the MAM and in particular, the IP3R-GRP75-VDAC1 complex is a key factor in strengthening the MAM and enables Ca2+ exchange between ER and mitochondria. It is well known that the interaction of VAPB (an ER protein) with PTPIP51 (a mitochondrial protein) is a component of the MAM and is involved in autophagy33. Our study focused on investigating whether MetF treatment alters the interaction between mitochondria and ER by targeting VDAC1. Proximity ligation assay (PLA) analysis between VAPB (ER) and PTPIP51 (Mito) in HepG2 cells showed that MetF decreased organelle contacts in a dose-dependent manner (Fig. 6a,c). MetF also reduced the interaction between IP3R-GRP75 as well as GRP75-VDAC1 (Supplementary Fig. 6). Consistent results were obtained after VDAC1 knockdown, with reduced interaction between the two organelles (Fig. 6b,d), suggesting that the effect of MetF on MAM is likely due to its direct interaction with VDAC1. Furthermore, considering the critical role of MAM in the calcium crosstalk between ER and mitochondria, we examined the calcium levels in mitochondria and cytosol upon treatment with MetF. In HepG2 cells loaded with Rhod-2, a mitochondrial calcium indicator, Met treatment decreased Rhod-2 intensity, which is a result of reduced ER -Mitochondria interaction (Fig. 6e). Under the same conditions, cytosolic calcium increased due to a decrease in calcium influx into the mitochondria. (Fig. 6g). Calcium transport from the ER to the mitochondria was also impaired in VDAC1-depleted cells, suggesting that MetF targets VDAC1 to regulate the organelle interactions. (Fig. 6f, h).
Elevated Ca2+ levels in the cytosol activate calcineurin, leading to transcription factor EB (TFEB) dephosphorylation and its nuclear translocation34. To further investigate the effect of cytosolic calcium upregulation by MetF binding to VDAC1, we examined whether MetF treatment regulates TFEB nuclear translocation. ICC results indicate that MetF treatment leads to the nuclear colocalization of TFEB (Fig. 6i). Given that TFEB is a master regulator of lysosomal biogenesis and autophagy induction, our result suggests that the increase in TFEB nuclear translocation may potentially contribute to the autophagy-inducing effect of MetF treatment.
Taken together, the biological validations demonstrated that VDAC1 knockdown mimicked the key activities induced by MetF, supporting the conclusion that VDAC1 is a biologically relevant target of MetF in regulating autophagy in HCC.