Combined inhibition of HMGCoA reductase and mitochondrial complex I induces tumor regression of BRAF inhibitor-resistant melanomas

The BRAFi- and MEKi-sensitive A375 human melanoma cell line was from ATCC, from which the BRAFi/MEKi-acquired resistant A375R1 was generated [9]. The intrinsic BRAFi and MEKi-resistant human melanoma cell lines UCSD354L, MEL624, and WM1799 were acquired from MDACC Cell Line Core. BRAF and MEK inhibitor sensitivities of the cell lines were previously reported [9, 17, 18]. Cell lines were authenticated by STR fingerprinting, and all cells were grown in RPMI media containing 5% fetal bovine serum. IACS-010759 (IACS) was developed and synthesized at the MDACC Institute for Applied Cancer Science as described [13]. Compounds in the drug screen obtained from the Selleck L2000 customized bioactive compound library (Selleck Chem) consisted of 320 therapeutic agents or preclinical candidates with a wide variety of target specificities (Table S1). For in vitro treatments, all compounds were dissolved in DMSO. For in vivo treatments, suspensions of the compounds were prepared using the following vehicles: 0.5% methyl cellulose for IACS-010759 (IACS) and 35% PEG300 + 2% Tween-80 for Atorvastatin (STN).

The drug screen was performed at the High Throughput Research and Screening Center, Institute of Bioscience and Technology, Texas A&M University, following the methodology depicted schematically in Fig. 1A. Briefly, optimal seeding densities that maintain log-phase growth of melanoma cell lines over a 96-h assay window were determined using a cell growth curve analysis with Hoescht-33342 nuclear staining. This was followed by seeding of optimal cell numbers of each cell line and treatments with either the anchor compound (IACS), individual probe compounds, or their combinations with the anchor compound using 10-fold dose dilution ranges. Cell growth inhibition induced by the treatments over a period of 72 h was assessed using DAPI nuclear staining. The results were subjected to rigor and reproducibility analysis, and synergistic or additive effects of the combinations were determined using a bootstrapped bliss independence model. Detailed methodology of the drug screen is provided in the supplemental methods.

Cell proliferation inhibition was evaluated using Cell Titer Blue (Promega) or 0.5% crystal violet staining after 72 h inhibitor treatments of cells grown in 96-well plates. IC50 values of serially diluted inhibitor-induced cell growth inhibition were determined by nonlinear regression curve analysis in the GraphPad Prizm software. Cell death was determined by Propidium iodide–cell cycle analysis using flow cytometry and cytoplasmic histone-associated DNA fragment analysis using Cell Death Detection ELISA Plus kit (Roche Applied Science) as described previously [17].

Mitochondrial stress tests were performed to determine real-time oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) in melanoma cell lines as described earlier [9]. Final assay concentrations of inhibitors used in the mitochondrial stress test were 1.5 μM oligomycin, 0.5 μM FCCP, 0.5 μM rotenone, and 0.5 μM antimycin A. Metabolic Fuel Flex tests were performed to evaluate glucose, glutamine, and fatty acid dependencies of melanoma cell lines. Final assay concentrations of inhibitors used in this test were 3 μM BPTES, 4 μM etomoxir, and 2 μM of UK5099. The metabolic flux tests were performed in 96-well plates (2 × 10⁴ cells/well) using a Seahorse XFe96 analyzer, and data was normalized against cell numbers following the manufacturer’s protocol (Agilent Technologies).

Melanoma cells (3 × 10⁶ cells/150-mm dish) were seeded overnight, labeled with [U-¹³C]-glucose or [U-¹³C]-glutamine and treated with the indicated inhibitors for 12 h. [U-¹³C]-glucose and [U-¹³C]-glutamine tracing analyses were performed at the MDACC Metabolomics Core Facility as described before [7]. Detailed methodology is described in the supplemental methods.

Whole cell lysates from cell lines were prepared in RIPA lysis buffer, and protein lysates from tumors were prepared by homogenization of ~ 50 mg of tumor tissue in a bead homogenizer as previously described [14]. The protein lysates were denatured and used for western blotting using standard methods, or for Reverse-Phase Protein Array (RPPA) analysis at the MDACC Functional Proteomics Core Facility, and the data was analyzed as described previously [9]. Antibodies used for western blotting and RPPA are listed at the RPPA core website (www.​mdanderson.​org/​research/​research-resources/​core-facilities/​functional-proteomics-rppa-core.​html). RhoA (67B9) antibody was from Cell Signaling Technology.

Subcutaneous xenograft tumors were generated for A375R1 and UCSD354 cell lines with 3 × 10⁶ cells/animal in the right flank of NOD scid gamma (NSG) mice. The mice were separated into two diet cohorts—regular diet (caloric profile of 62.1% carbohydrate, 24.7% protein, 13.2% fat) (PicoLab #5053), and high-fat ketogenic diet (caloric profile of 1.8% carbohydrate, 4.7% protein, 93% fat) (Bio-Serv #F3666). Inhibitor treatments were performed by oral gavage daily once for the indicated number of days. Tumor volumes and mice weight were recorded every 3 days. Animals were excluded if they showed overt toxicity or lost > 15% body weight over the treatment course. For molecular analysis of inhibitor effects, tumor-bearing mice were treated with the inhibitors as above, and tumors were harvested 3 h after drug administration on the second day of treatment. All animal experiments were approved by the Institutional Animal Care and Use Committee.

Rigor and reproducibility of the combinatorial drug screening assay was performed in accordance with the NCATS Assay Guidance Manual [19], and a bootstrapped bliss independence model was used to calculate drug synergy [20]. In vivo tumor growth studies consisted of nine mice per treatment group which provided significant power for tumor growth analysis. For functional proteomics analysis using RPPA, tumors were harvested from mice after 5 days of treatments, and protein lysates from three pieces of each tumor were analyzed separately. Hierarchical supervised clustering of significantly differing proteins in treated versus untreated samples was performed using Pearson correlation in Gene Cluster 3.0, and heatmaps were generated using Gene Treeview. For in vitro cell proliferation assays, significant differences between treatments and individual doses of treatments were analyzed by two-way ANOVA followed by post hoc Tukey multiple comparisons test in GraphPad Prizm. For tumor growth studies, cell death assays, metabolite and RPPA analyses, T tests were used for determining statistically significant differences (p < 0.05 for cell death and tumor growth studies, p < 0.005 for RPPA) between inhibitor treatments and mock/vehicle treatments. Where necessary, significant differences from mock/vehicle treatments or between treatments were designated with asterisks (*).

For identifying clinically relevant small molecules that significantly improve anti-melanoma activity of IACS, we performed an unbiased combinatorial drug screen that tested the combination of IACS (anchor) with each of 320 inhibitors (probes) that target 22 different signaling pathways (Fig. 1A and Table S1). We tested these combinations in two BRAF^V600E-mutant melanoma cell lines, UCSD354L and A375-R1, which possess intrinsic and acquired resistance respectively to BRAF/MEK inhibition [9, 17, 18]. The drug screen was performed as shown schematically in Fig. 1B, and cell growth inhibition induced by each probe and their combinations with IACS was determined after 72 h of incubation. Combination efficacy was determined using a bootstrapped bliss independence model and represented as “combo score” units, with ≥ 0.1 representing potentially synergistic growth inhibition. Values under 0.1 represent additive effects, and negative values potentially antagonistic effects. The combo scores revealed that combination with IACS induced mostly additive effects on cell growth inhibition by a majority of the 320 probes in both cell lines, with less than 20 probes showing antagonism (Fig. 1C). Less than ten probes showed potentially synergistic growth inhibition when combined with IACS in either cell lines (Fig. 1D and E).

The combinatorial drug screen results were validated with five probes that exhibited the highest combo scores with IACS in each of the two cell lines. Two probes with non-synergistic combo scores of < 0.1 were also tested to validate their lower efficacy observed in the screen. The cell lines were seeded in 96-well plates and treated with the indicated dose ranges of single agents or their combinations with IACS, and cell growth inhibition after 72 h was assessed using the Cell Titer Blue cell proliferation assay; IC50 values were determined. The results showed that among the various probes tested, the HMGCoA reductase inhibitor (HMGCRi), atorvastatin (STN), exhibited the most potent combination effect with IACS in both cell lines (Fig. 2A for A375R1 and Fig. 2F for UCSD354L). Compared with STN, other high-scoring probes from the screen exhibited lower combined efficacy with IACS (Fig. 2B–E for A375R1 and Fig. 2G–J for UCSD354L). Two non-synergistic probes from the screen, GSK690693 and BKM120, exhibited lower combined efficacy with IACS compared to IACS + STN in a separate validation study (Fig. S1A-B). IC50 values for all individual treatments and their combinations with IACS are shown in Table S2. We also evaluated the highest scoring IACS + STN in normal melanocytes and found that neither single agent nor their combination significantly inhibited the growth of these normal cells (Fig. S1C). Finally, to confirm that the Cell Titer Blue cell proliferation assay does not produce false-positive growth inhibitory effects, we also performed crystal violet staining of IACS + STN–treated cells and found a similar growth inhibitory profile as seen with Cell Titer Blue (Fig. S1D). Taken together, the drug screen and subsequent validation studies suggest that the HMGCRi, STN, could be a potent combination treatment for improving the efficacy of IACS treatment against BRAFi/MEKi-resistant, BRAF-mutant melanomas.

As STN exhibited the most potent anti-melanoma combination efficacy with IACS among all probes with synergistic combo scores, we further explored the efficacy of this combination on melanoma cell death induction. Propidium iodide cell cycle analysis showed that a 72 h IACS treatment of A375R1 induced a small increase of G2/M and sub-G1 phase cells. STN treatment induced small increases of G2/M phase and a significant increase of sub-G1 phase, indicating cell death induction by the treatment, which further increased following IACS + STN treatment (Fig. 3A). In the UCSD354L, IACS induced G1 arrest and STN induced G2/M arrest, while the combination induced G1 arrest and sub-G1 accumulation (Fig. 3B). Combination of the standard of care BRAFi, dabrafenib (DAB), with IACS did not induce sub-G1 accumulation in either cell line, although DAB + STN induced a small, but insignificant increase compared with STN alone in A375R1, but not in UCSD354L (Figs. S1A and S1B). Cell death induction by IACS and STN combination treatments in both cell lines was confirmed using a cytoplasmic histone accumulation assay that sensitively detects cell death as increased cytoplasmic accumulation of histones in treated versus untreated cells (Fig. 3C and D).

For identifying the molecular determinants of sensitivity in the BRAFi-resistant melanomas in our study, we extracted protein lysates from the cells after 24 h treatments with 100 nM IACS, 1 μM STN, or their combinations and performed Reverse-Phase Protein Array (RPPA) analysis to assess functional status of over 200 cellular proteins in key oncogenic signal transduction pathways. We performed Pearson correlation and supervised clustering of the most significant (p < 0.005) treatment-induced alterations compared with vehicle treatments, and generated heatmaps of the treatment ratios for both cell lines (Fig. 3E and F). These results showed that IACS treatment activated phospho-AMPK_Thr172 and phospho-ACC_Ser79, inhibited phospho-S6_ Ser235_236_240_244 and phospho-MAPK_Thr202_Tyr204, as has been shown previously by us and others as targeted effects of inhibiting mitochondrial activity [14, 21, 22]. Interestingly in both cell lines, IACS treatment induced phospho-AKT_Thr308/Ser473, which is known to promote melanoma resistance to BRAFi/MEKi [23, 24]. Treatment with STN decreased phospho-AKT compared with vehicle treatment and also counteracted its IACS-induced increase in both cell lines. IACS + STN combination also downregulated pro-growth signaling proteins like phospho-Rb_Ser807_811 and upregulated growth inhibitory/cell death proteins (for example, cleaved caspase 7) (Fig. 3E and F). Some of the key IACS + STN–induced alterations revealed by RPPA were confirmed by western blotting analysis of phospho- and total proteins in protein lysates (Fig. 3G). RhoA protein upregulation was assessed as a marker of HMGCR inhibition by STN. Treatment-induced alterations in the levels of phospho-AKT_Thr308 and phospho-AMPK_Thr172 were quantified using NIH Image J software and represented as bar graphs (Fig. S1G-J).

The above protein analysis revealed that IACS + STN treatment induced significant alterations in phosphoproteins associated with mitochondrial metabolism like AMPK and ACC. STN inhibits the conversion of HMG-CoA to mevalonate, the first and rate-limiting step of cholesterol biosynthesis pathway [25, 26]. As the levels of HMG-CoA are dependent on acetyl CoA, a critical node in mitochondrial metabolism, we hypothesized that responsiveness to IACS + STN may be associated with mitochondrial metabolism. To test that, we performed the Seahorse Fuel Flex assay in parental BRAFi-sensitive A375 cells, BRAFi-acquired resistant A375R1 and intrinsic BRAFi-resistant UCSD354L cells. In this assay, we assessed the dependency of the cells on each of the three cellular fuels—glucose (GLC), glutamine (GLN), or fatty acids (FA)—and their flexibility to use either of the single fuels when the other two fuels are inhibited. The results showed that the parental BRAFi-sensitive A375 cells have the highest dependency on GLC (Fig. 4A), whereas the two BRAFi-resistant cells have the highest dependency on FA (Fig. 4B and C). A375 possessed low flexibility for compensatory utilization of any two alternate fuels when one fuel was inhibited, although the cells did possess as much capacity to oxidize FA as their basal dependency on this fuel (Fig. 4A). In comparison, the two resistant cell lines showed a comparatively higher flexibility to oxidize any single fuel when the other two were inhibited, with the highest flexibility for GLC utilization (Fig. 4B and C). The higher FA oxidation dependency of the two BRAFi-resistant melanoma cells indicates their altered metabolic requirement for FA compared with the sensitive cells and may influence their responses to metabolism targeting therapeutics. Two other BRAFi/MEKi-resistant cells that did not exhibit FA dependency were not sensitive to IACS + STN combination treatment (Fig. S2A-S2D).

Previous studies showed that statins interfere with calcium homeostasis to inhibit mitochondrial complexes I and III [27], and can weakly inhibit OCR in cancer cells [28]. To determine if STN could inhibit OCR in the BRAFi-resistant melanoma cells in our study, we performed Seahorse bioenergetics stress tests in A375R1 cells. The results showed that 24 h treatment with 1 μM STN minimally inhibited basal cellular oxygen consumption rate (OCR), interpreted as an indirect readout for cellular OxPhos (Fig. 4D). STN also induced a small but significant inhibition of basal extracellular acidification rate (ECAR), an indirect readout for glycolytic activity (Fig. 4E). On the other hand, treatment with 100 nM IACS completely inhibited basal and maximal OCR and slightly increased ECAR levels as shown previously [14]. As A375R1 cells show elevated FA dependence, we also tested etomoxir (ETMR), a carnitine palmitoyltransferase inhibitor that inhibits FA β-oxidation. ETMR induced significant inhibition of basal and maximal OCR compared with STN. However, unlike STN, it did not inhibit A375R1 cell proliferation (Fig. S2E). Taken together, these results suggest that the weak OCR inhibition by STN is unrelated to its cell growth inhibition. As IACS treatment induced a complete inhibition of OCR, with no increase from FCCP treatment, we replicated the above IACS and the mitochondrial inhibitor treatments in a separate 96-well plate and assessed cell viability at the end of each subsequent treatment using trypan blue dye exclusion. The results showed < 5% decrease of viability in IACS-treated cells following FCCP treatment, suggesting that the complete OCR inhibition by IACS was not a result of altered cell viability (Fig. S2F).

We then performed ¹³C-labeled GLC and GLN tracing analyses to mechanistically assess the above molecular and metabolic effects of IACS and STN, and to potentially identify a metabolic basis for melanoma cell growth inhibition by their combination. We treated A375R1 cells grown in [U-¹³C]-GLC or [U-¹³C]-GLN media with 100 nM IACS, 1 μM STN, or their combination for 12 h and detected relative incorporation of ¹³C-labeled metabolites in glycolysis, tricarboxylic acid cycle (TCA cycle), FA synthesis, and mevalonate/HMGCoA pathways using LC-MS (Fig. 5A). The results showed that IACS (I) treatment stimulated relative GLC incorporation into the glycolysis metabolites—glucose 6 phosphate, phosphoenolpyruvate, and pyruvate (G6P, PEP, and PYR)—compared to vehicle (V), while STN (S) induced a small increase and IACS + STN (I + S) inhibited incorporation into G6P, suggesting that the combination treatment inhibits the initial step of glycolysis (Fig. 5B). Similar effects were observed in the TCA cycle, where IACS stimulated relative GLC and GLN incorporation into citrate and α-ketoglutarate (CIT and αKGA), and combination with STN inhibited these effects (Fig. 5C and D). Interestingly, STN increased relative GLC incorporation into acetyl CoA (Ac-CoA), while IACS increased relative GLN incorporation into Ac-CoA (Fig. 5E and F), which reveal the unique metabolic effects of the two inhibitors. Also, interestingly, IACS treatment increased relative GLC incorporation but inhibited relative GLN incorporation into HMGCoA, the substrate for HMGCR enzyme in the mevalonate pathway (Fig. 5G and H). STN treatment induced relative accumulation of GLC and GLN carbons in HMGCoA, an expected effect of substrate accumulation in the wake of HMGCR inhibition, and a corresponding decrease in mevalonate (MEV) (Fig. 5G and H). Combination treatments inhibited incorporation of GLN into HMG-CoA and also inhibited GLC and GLN incorporation into MEV (Fig. 5G and H). Finally, treatment-induced alterations in the relative incorporation of GLC and GLN into malonyl CoA (MAL-CoA) and its downstream fatty acid product, lipoic acid (LIPO), suggest that IACS treatment inhibits FA biosynthesis, STN increases it, and the combination inhibits it (Fig. 5I and J). As an earlier study showed that upregulation of acetoacetate (AcAc), a ketogenic metabolite of Ac-CoA promotes the growth of BRAF^V600E tumors [29], we evaluated treatment effects on relative GLC and GLN carbon incorporation into AcAc. The results showed that IACS and IACS + STN potently inhibited relative GLC incorporation into AcAc, while IACS induced a small decrease of relative GLN incorporation (Fig. S2H and I). The individual and combination treatments induced an increase of total pooled Ac-CoA levels (Fig. S2G). Taken together, these results suggest that although individually, IACS and STN have differential effects on cellular utilization of GLC and GLN; their combination ultimately inhibits relative incorporation of the two fuels into FA synthesis and mevalonate pathways.

We assessed in vivo effects of the individual and combination treatments in subcutaneous A375R1 and UCSD354L tumors in mice fed with normal carbohydrate-rich chow (62% calories from carbohydrate) (Fig. 6A). Additionally, as our above results suggested that these tumor cell lines have FA fuel dependency and that IACS + STN treatments potently inhibit lipid metabolism, we functionally assessed the effect of the treatments on tumor growth in mice fed with a high-fat ketogenic diet (93% calories from fat) (Fig. 6A). RPPA proteomics analysis of protein lysates from untreated mice with A375R1 (Fig. 6B) and UCSD354L (Fig. S3A) tumors showed that in comparison with regular diet, the high-fat diet significantly (p < 0.005) downregulated fatty acid synthase (FASN), growth factor signaling (phospho-IGFR, IGFBP2, phospho-Src, VEGFR2, etc.), and cell division/survival proteins (phospho-Rb, phospho-Wee1, phospho-p90RSK, phospho-NFκB, p-Rictor), but surprisingly activated P-AKT without the consequent downstream phosphorylation of GSK3. The high-fat diet also activated markers of autophagy (LC3A/B, WIPI1), cell cycle inhibition (p21, TIGAR), and cellular stress (Phospho-AMPK, Phospho-RPA32, P38-MAPK).

In mice fed with regular diet, IACS treatment induced complete stasis of A375R1 tumors and robust inhibition of UCSD354L tumor growth over 30 days of treatment (Fig. 6C and D). Over the same time course, STN treatment induced minimal inhibition of A375R1 and UCSD354L tumor growth. IACS + STN induced complete regression of A375R1 and stasis of UCSD354L tumors (Fig. 6C and D), suggesting potent tumor growth inhibition by the combination treatment. In mice fed with high-fat keto diet, A375R1 and UCSD354L tumor growth in vehicle-treated mice was significantly less (p < 0.001) than the growth observed in mice fed with regular diet (Fig. 6E and F). While this is consistent with ketogenic diet-induced ketosis in certain conditions [30, 31], one study showed that high-fat diet selectively promotes tumor growth of BRAF^V600E-dependent human melanoma cells [29]. As our models are resistant to BRAFi, it is possible that they are not dependent on mutant-BRAF protein, which may potentially alter their response to high-fat keto diet. So, we tested the effect of high-fat keto diet on subcutaneous growth of parental A375 tumors that are dependent on BRAF^V600E and sensitive to BRAFi. Interestingly, the A375 tumors also showed significantly lower growth in mice fed with high-fat keto diet versus regular diet (Fig. S3B). In the high-fat keto diet–fed mice, STN treatment induced minimal inhibition of A375R1 and UCSD354L tumor growth compared with vehicle, as was also observed in mice fed with regular diet. IACS treatment however induced potent tumor regressions within 20 days, and IACS + STN completely eradicated the tumors of both models (Fig. 6E and F). Longer treatment times were not pursued as IACS and IACS + STN treatments induced > 15% weight loss within 25 days in mice fed with high-fat diet versus regular diet (Fig. S3C and S3D).

RPPA analysis of A375R1 tumor lysates from regular and high-fat keto diet–fed mice showed similar treatment-induced alterations compared with vehicle treatments (Fig. 6G and H). For example, IACS treatment activated phospho-AMPK and inhibited phospho-MAPK, while STN treatment inhibited IACS-induced phospho-AKT. The downstream alterations induced by these molecular effects were also similar between the regular and high-fat keto diet tumor samples. The higher potency of tumor growth inhibition in the high-fat keto diet cohorts could be a combination of ketogenesis and FA nutritional cutoff by the treatments. Compared with the in vitro cell line results, the tumor RPPA results additionally showed that IACS and IACS + STN inhibited the pro-tumorigenic hypoxia protein, HIF1α. Also, STN treatment upregulated PAR, and IACS counteracted it (Fig. 6G and H). These results suggest that IACS and STN cancel one another’s pro-tumorigenic protein signaling and enhance antitumor signaling.