Phenformin enhances the therapeutic benefit of BRAF^V600E inhibition in melanoma

Significance

Inhibitors of BRAF protein kinase, such as Vemurafenib and Dabrafenib, have shown remarkable antitumor activity in patients with BRAF mutant melanoma. However, most of the patients developed drug resistance during the course of treatment, leading to resumed tumor growth. This drug resistance challenge underscores the need to improve on current BRAF-targeted therapy. In this study, we have shown that phenformin, a biguanide used for treating type 2 diabetes, enhances the antitumor activities of BRAF inhibitors in both cultured melanoma cells and a genetically engineered BRAF^V600E-driven mouse model of melanoma. Our preclinical findings suggest that combining phenformin with a BRAF inhibitor may be a more effective treatment than a single-agent BRAF inhibitor for treating patients with melanoma whose tumor harbor BRAF mutations.

Abstract

Biguanides, such as the diabetes therapeutics metformin and phenformin, have demonstrated antitumor activity both in vitro and in vivo. The energy-sensing AMP-activated protein kinase (AMPK) is known to be a major cellular target of biguanides. Based on our discovery of cross-talk between the AMPK and v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) signaling pathways, we investigated the antitumor effects of combining phenformin with a BRAF inhibitor PLX4720 on the proliferation of BRAF-mutated melanoma cells in vitro and on BRAF-driven tumor growth in vivo. Cotreatment of BRAF-mutated melanoma cell lines with phenformin and PLX4720 resulted in synergistic inhibition of cell viability, compared with the effects of the single agent alone. Moreover, treatment with phenformin significantly delayed the development of resistance to PLX4720 in cultured melanoma cells. Biochemical analyses showed that phenformin and PLX4720 exerted cooperative effects on inhibiting mTOR signaling and inducing apoptosis. Noticeably, phenformin selectively targeted subpopulations of cells expressing JARID1B, a marker for slow cycling melanoma cells, whereas PLX4720 selectively targeted JARID1B-negative cells. Finally, in contrast to their use as single agents, the combination of phenformin and PLX4720 induced tumor regression in both nude mice bearing melanoma xenografts and in a genetically engineered BRAF^V600E/PTEN^null-driven mouse model of melanoma. These results strongly suggest that significant therapeutic advantage may be achieved by combining AMPK activators such as phenformin with BRAF inhbitors for the treatment of melanoma.

Melanoma is a major and deadly form of skin cancer that arises from the malignant transformation of melanocytes (1). One of the most commonly mutated genes in melanoma is v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) (∼50%), which encodes a member of the RAF protein kinase family and an intermediate in the RAS-RAF-MEK-ERK signaling cascade (2). More than 90% of BRAF mutations involve a single base substitution within the kinase domain, leading to a V600E amino acid change and a constitutively active kinase. BRAF selective kinase inhibitors such as Vemurafenib (also known as Zelboraf or PLX4032) and Dabrafenib have shown significant clinical antitumor activity in treating malignant melanoma bearing BRAF^V600E mutations (3). However, most patients developed resistance during the course of treatment with these agents (4, 5). Significantly, 20–30% of patients developed squamous cell carcinomas (SCC) as a major adverse effect (4–6). These limitations clearly necessitate the improvement of this BRAF-targeted therapy for melanoma (7).

Biguanides, such as metformin and phenformin, are common therapeutics for type 2 diabetics (8). Emerging evidence from retrospective population-based studies and preclinical studies using cultured cancer cells and mouse models have demonstrated that biguanides also possess antitumor activity (9). AMP-activated protein kinase (AMPK), an evolutionarily conserved energy sensor, is known to be activated in the liver in response to therapeutic doses of biguanides (9). The activity of AMPK is regulated by the binding of AMP or ADP to its regulatory subunit, followed by phosphorylation of the activation loop of the catalytic subunit by upstream activating kinases, including the tumor suppressor protein kinase liver kinase B1 (LKB1) (10). Metformin and phenformin activate AMPK by inhibiting complex I of the mitochondrial respiratory chain, thus increasing the cellular AMP (and ADP) to ATP ratio (9). We have recently shown that the oncogenic BRAF V600E mutant suppresses the activity of AMPK by promoting phosphorylation of LKB1 by ERK and p90Rsk on inhibitory sites, and that this inhibition is critical for melanoma cell proliferation and anchorage-independent growth (11). Based on these findings, we hypothesized that treatment of BRAF-mutant melanomas with a combination of a BRAF inhibitor and an AMPK activator could offer therapeutic advantages over BRAF inhibitor (BRAFi) single agent therapy. In this report, we show that combining phenformin with a BRAFi results in enhanced efficacy compared with either agent alone in both in vitro and in vivo models of BRAF mutant tumors. We show that this effect is, in part, explained by the ability of this combination to suppress the growth of the JARID1B-positive subset of tumor cells, which are resistant to single agent therapy.

Results

To examine the effects of metformin and phenformin on cell viability as single agents, we performed MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt] cell assays by using several BRAF mutated melanoma cell lines. Metformin had limited effect on reducing cell viability in these cell lines, when used at concentrations of up to 3 mM (Fig. 1A). In contrast, phenformin was much more potent, with an estimated IC₅₀ in the range of 0.5–1 mM (Fig. 1B). As expected, the effect of phenformin depended, at least partially, on AMPK, because shRNA knockdown of AMPKα1 in A375 melanoma cells (Fig. S1A) attenuated the reduction in cell viability in response to phenformin (Fig. 1C). A possible explanation for the observed difference in response seen here between these two biguanides is that metformin depends more on a family of organic cation transporters (OCT) for uptake into cells than phenformin (8, 12). Analysis of OCT2 protein expression levels in a panel of melanoma cell lines and MDA-MB-468 breast cancer cells, a metformin-sensitive breast cancer cell line, showed that melanoma cells express much lower levels of OCT2 than MDA-MB-468 cells (Fig. 1D). Knockdown of OCT2 by shRNA in MDA-MD-468 cells decreased the sensitivity of these cells to metformin (Fig. S1B). Conversely, overexpression of OCT2 in Sk-Mel-28 cells (Fig. S1C) increased the sensitivity of these cells to metformin (Fig. 1E), supporting the concept that OCT2 expression is a determinant of metformin sensitivity. These results together demonstrate that phenformin is much more potent than metformin in decreasing melanoma cell viability.

Fig. 1.

Phenformin and PLX4720 show a synergistic effect on reducing viability of BRAF-mutated melanoma cells. (A and B) Phenformin, but not metformin, reduces cell viability of Sk-Mel-28 and WM115 cells. Cells were treated with either drug at the indicated concentration for 3 d before MTS assays were performed. Data from three independent experiments were represented as mean ± SEM. (C) AMPK is important for the ability of phenformin to reduce cell viability. A375 cells stably expressing shRNA against AMPKα1 or scrambled control were treated with phenformin for 3 d before MTS assays were performed. (D) Melanoma cells express much lower levels of OCT2 proteins than MDA-MB468 breast cancer cells. Lysates were collected from various cell lines and blotted with indicated antibodies. (E) Expression of FLAG-OCT2 in Sk-Mel-28 cells increases its sensitivity to metformin. Sk-Mel-28 stably expressing pBabe-FLAG-OCT2 or control vector were treated with metformin for 3 d before MTS assays were performed. (F and G) Cotreatment of phenformin with PLX4720 increases the sensitivity of Colo829 (F) and WM115 (G) cells to PLX4720. Cells were treated with various concentration of PLX4720 as indicated in the absence or presence of 1 mM phenformin for 3 d before MTS assays were performed. (H) Melanoma cells bearing BRAF^V600E mutation or BRAF^WT were treated with various concentrations of PLX4720 and phenformin for 3 d before MTS assays were performed. Combination indexes were calculated for various cells at 3 μM PLX4720 and 0.5 mM phenformin by using the CompuSyn software, except 3 μM PLX4720 and 0.3 mM phenformin for WM115, and 0.3 μM PLX4720 and 1 mM phenformin for A375.

Based on these findings, we focused on phenformin as the biguanide AMPK activator of choice to examine the effects of its combination with BRAFi on melanoma cell viability. As shown in Fig. 1 F and G, cotreatment of Colo829 and WM115 melanoma cells with 1 mM phenformin and PLX4720 BRAFi led to a more pronounced decrease in cell viability compared with treatment with PLX4720 alone. In addition, a combination index of ∼0.3 and 0.5, respectively, as calculated by the CalcuSyn software, was found for these cells. The combination index is a measurement of the combined drug interaction and is defined as <1 for synergism and >1 for antagonism (13). Similar synergetic effects were found for several other melanoma cell lines harboring the BRAF V600E mutation (Fig. 1H).

Based on the synergistic effect of combining phenformin and PLX4720 BRAFi, we hypothesized that combining phenformin with BRAFi may also influence the emergence of acquired resistance to BRAFi in melanoma cells. We first assessed this possibility in colony formation assays in vitro. Mel1617 cells were seeded at a low density in the presence of PLX4720, either alone or together with metformin or phenformin, and were then analyzed for the growth of resistant colonies from single cells. As shown in Fig. 2A, treatment with PLX4720 plus phenformin (but not metformin) significantly reduced the number of colonies formed compared with treatment with PLX4720 alone. As a second approach to address this question, we adapted a well-established in vitro assay for the development of drug resistance (14), in which A375 cells were chronically treated with stepwise increased concentration of PLX4720 alone (A375-BR, BRAFi-resistant) or PLX4720 together with fixed 0.3 mM of phenformin (A375-BPR, BRAFi- and phenformin-resistant). Cells were selected at each step until they resumed the normal growth kinetics of the untreated parental line before moving to the next step. After 7 wk of treatment, cells from both treatment groups were evaluated for their sensitivities to PLX4720 as evaluated by the ability of PLX4720 (as a single agent) to suppress phosphorylation of ERK and to suppress cell growth. As shown in Fig. 2B, A375-BR were less sensitive to PLX4720 compared with parental cells in regard to cell growth suppression, and these cells were also less sensitive to PLX4720 in regard to inhibiting ERK phosphorylation (Fig. 2C). Interestingly, A375-BPR cells, which were selected for resistance to the combination therapy, retained sensitivity to PLX4720 as a single agent (Fig. 2B) and were intermediate between the parent cells and A375-BR cells in regard to PLX4720 inhibition of ERK activation (Fig. 2C). These results indicate that cotreatment with phenformin and PLX4720 suppresses the emergence of PLX4720 resistance in melanoma cells with BRAF V600E mutation.

Fig. 2.

Cotreatment of phenformin with PLX4720 delays the emergence of PLX4720 resistance in cultured melanoma cells with BRAF^V600E mutation. (A) Cotreatment of phenformin with PLX4720 delays the formation of Mel1617 PLX4720-resistant colonies. Mel1617 cells were seeded at a low density in the presence of PLX4720, either alone or together with metformin or phenformin, and were then analyzed for the growth of resistant colonies from single cells. Cells were stained with crystal violet after 15 d. (B) Cotreatment of phenformin with PLX4720 delays the emergence of PLX4720 resistance in A375 cells in the in vitro resistance development assays. A375 cells were treated with stepwise increased concentration of PLX4720 alone (A375-BR, BRAFi-resistant) or PLX4720 together with fixed 0.3 mM of phenformin (A375-BPR, BRAFi- and phenformin-resistant). Cells were selected at each step until they resumed the normal growth kinetics of the untreated parental line before moving to the next step. After 7 wk of treatment, cells from both A375-BR and A375-BPR groups were seeded in 96-well plates and treated with 0, 0.3, or 1 μM of PLX4720 for 3 d before MTS assays were performed. Data from three independent experiments were represented as mean ± SEM. (C) Cells from both A375-BR and A375-BPR groups were treated with 1 μM PLX4720 for 2 h. Cell lysates were used for Western blotting with indicated antibodies.

To explore the molecular mechanisms underlying the synergistic effects of phenformin and PLX4720, we performed Western blot analyses to evaluate the activities of relevant signaling pathways in melanoma cells treated with PLX4720 and phenformin in combination compared with treatment with each drug individually. These analyses revealed that PLX4720 alone inhibited ERK activity dramatically, and cotreatment of phenformin with PLX4720 led to further reduction of pERK levels. In addition, combining phenformin and PLX4720 cooperatively activates AMPK (Fig. 3A). Because both AMPK and BRAF kinases are capable of regulating the mTOR signaling pathway (10), we further examined the effects of the phenformin and BRAFi combination on mTOR-dependent phosphorylation of S6. As shown in Fig. 3A, using either drug alone resulted in only modest inhibition of pS6 phosphorylation levels. However, when applied in combination, they essentially abolished activation of pS6. These results support the presence of a strong synergistic effect on mTOR inhibition for the phenformin/BRAFi combination.

Fig. 3.

Mechanisms underlying the synergistic inhibitory effects of phenformin and PLX4720 in melanoma cells. (A) Phenformin and PLX4720 act cooperatively to inhibit mTOR signaling. Colo829 and WM115 cells were treated with DMSO, 0.25 mM phenformin, 1 μM PLX4720 alone, or in combination for 24 h. Cell lysates were used for Western blotting with indicated antibodies. (B) Phenformin and PLX4720 cooperatively induce apoptosis in Colo829 melanoma cells. Colo829 cells were treated with DMSO, 0.25 mM phenformin, 1 μM PLX4720 alone, or in combination for 72 h, before used in the annexin V staining assay. Data from three independent experiments were represented as mean ± SEM; **P < 0.01. The Student t test was performed to compare between the control or parental vs. treated group. (C and D) WM115 (C) or BP01 (D) cells were treated with DMSO, 0.25 mM phenformin, 1 μM PLX4720 alone, or in combination for 3 or 5 d, respectively. Cells were stained with anti-JARID1B antibodies, and the numbers of JARID1B⁺ and JARID1B⁻ cells were analyzed by FACS and plotted; *P < 0.05. (E) Representative FACS plots from three independent experiments were shown for WM115 cells. (F) A375 and A375-BR (BRAFi-resistant) cells were stained with anti-JARID1B antibodies before being subjected for FACS analysis. A375-BR cells were treated with or without 2 mM phenformin for 3 d. (G) WM115 cells expressing shAMPKα1 or pLKO vector control were cultured with or without 0.25 mM phenformin for 3 d before being subjected for FACS analysis.

We next used annexin V staining assays to further explore the mechanism underlying the synergistic effect of the combination. Although PLX4720 modestly induced apoptosis in Colo829 cells, phenformin alone did not. However, the combination of phenformin with PLX4720 greatly enhances the apoptotic activity of PLX4720 (Fig. 3B). Previously, metformin has been shown to selectively target cancer stem cell populations of several breast cancer cell lines (15, 16). Although it is not yet well established whether melanoma cancer stem cells exist, a slow-cycling subpopulation of melanoma cells, marked with expression of JARID1B, a H3K4 demethylase, has been described (17). Expression of JARID1B was shown to be critical for long-term maintenance of melanoma tumor growth (17). To examine the effect of phenformin/PLX4720 on JARID1B-positive and JARID1B-negative cells separately, we performed FACS analyses on WM115 cells exposed to phenformin, PLX4720, or both in combination. Treatment with PLX4720 alone for 3 d led to a decrease in the number of JARID1B-negative cells and an increase of JARID1B-positive cells (Fig. 3 C and E). In contrast, after 3 d of treatment with phenformin alone, the number of JARID1B-positive cells, but not that of JARID1B-negative cells decreased significantly compared with the control treatment group (Fig. 3 C and E). Importantly, shRNA knockdown of AMPKα1 increased the percentage of JARID1B-positive cells and abolished the inhibitory effect of phenformin on JARID1B expression (Fig. 3G), suggesting that AMPK is critical for the regulation of JARID1B expression by phenformin. Similar effects on JARID1B-positive and JARID1B-negative populations were also observed in BP01 cells (Fig. 3D), a mouse melanoma cell line derived from a genetically engineered BRAF^V600E/PTEN^null-driven mouse model (18). Interestingly, we detected increased JARID1B-positive cell populations in the BRAFi-resistant A375 cells (A375-BR), compared with the parental BRAFi-sensitive cells in FACS analyses (Fig. 3F), suggesting that BRAFi-resistant cells were enriched with JARID1B-positive cells upon long-term treatment with PLX4720. Interestingly, treatment of A735-BR cells with phenformin reduced the percentage of JARID1B-positive cells (Fig. 3F). A similar increase in JARID1B-positive cells was observed in BRAFi-resistant Mel617 cells compared with the parental Mel1617 cells (Fig. S2). These results are consistent with the short-term effect of BRAFi on increasing the JARID1B-positive cell population (Fig. 3 C and D). Importantly, combination treatment with PLX4720 and phenformin reduced the numbers of both JARID1B-positive and JARID1B-negative cells (Fig. 3 C–E). Overall, these findings indicate that PLX4720 preferentially targets the JARID1B-negative cells, whereas phenformin plus PLX4720 suppresses growth and survival of both JARID1B-positive and JARID1B-negative cells, most likely by suppressing the ability of the JARID1B-negative cells to switch to the more resistant JARID1B-positive subtype.

Finally, we investigated the effects of the phenformin/PLX4720 combination on BRAF-driven melanoma tumor growth in vivo by using both xenograft and GEM models. Colo829 and A375 melanoma cells were grown as xenograft tumors in nude mice to assess their sensitivities to various treatment options. Once tumor volumes reached between 80 and 100 mm³, animals were randomly assigned to four groups that were administered vehicle, PLX4720, phenformin, or PLX4720/phenformin combination by oral gavage twice per day. As shown in Fig. 4A and Fig. S3, although tumors in the vehicle-treated animals progressed steadily over the 2-wk course analyzed, those tumors treated with phenformin or PLX4720, respectively, showed slight and significant inhibition of tumor growth, but no evidence of tumor regression. In contrast, animals treated with the combination of PLX4720 and phenformin showed significant reduction in tumor size. We next examined the efficacy of the phenformin/PLX4720 combination on a genetically modified, BRAF^V600E/PTEN^null-driven melanoma mouse model (Tyr::CreER; Braf^CA/+; Pten^lox/lox) (18). This mouse model allows the simultaneous conditional conversion of a Braf^WT allele to the active Braf^V600E allele and deletion of the Pten allele in a melanocyte-specific fashion by administration of 4-hydroxitamoxifen (4-HT). Upon 4-HT–induced Cre expression, these mice develop melanoma with a high penetrance and a short latency (∼3 wk) (18). For the GEM model, analyses similar to those conducted for the xenograft model were used to determine the effect of these drugs on tumor growth. We again observed significant tumor regression with treatment of the phenformin/PLX4720 combination in these mice (Fig. 4B). PLX4720 alone significantly reduced the rate of tumor progression, whereas phenformin alone only modestly attenuated tumor growth in these mice (Fig. 4B). Immunohistochemical analyses of these tumors indicated that the phenformin/PLX4720 combination dramatically induced apoptotic cell death and attenuated tumor cell proliferation in vivo (Fig. 4C). In addition, this combination strongly inhibited the pS6 levels in the tumors. Importantly, we also observed that the residual tumor from mice treated with PLX4720 exhibited enhanced expression of JARID1B, whereas the tumor from mice treated with phenformin or the phenformin/PLX4720 combination exhibited reduced expression of JARID1B (Fig. 4C). These results confirmed findings observed in melanoma cells in vitro (Fig. 3). Noticeably, we did not observe significant differences among different treatment groups in body weight, levels of glucose, IGF1, or insulin in blood plasma (Fig. S4), suggesting that the effect of the combination is unlikely to be mediated indirectly through potential changes in whole-body metabolism. In summary, these data from both xenograft and BRAF^V600E/PTEN^null-driven GEM models demonstrate that the combination of the BRAF inhibitor PLX4720 and phenformin induces significant tumor shrinkage and has stronger antitumor activities than PLX4720 alone in vivo.

Fig. 4.

Combination treatment of phenformin and PLX4720 lead to tumor regression in mouse models. (A) Nude mice bearing Colo829 xenograft tumors were treated with vehicle, PLX4720 (20 mg/kg), phenformin (100 mg/kg), or the combination of PLX4720 and phenformin twice per day when tumor volume reached between 80 and 100 mm³. The two-way ANOVA test was performed to compare between the PLX4720 group vs. the combination group. (B) Tyr::CreER; BRAF^CA/+; PTEN^lox/lox mice bearing single tumor induced by 4-HT were treated with vehicle, PLX4720 (50 mg/kg), phenformin (100 mg/kg), or the combination of PLX4720 and phenformin twice per day when tumor volume reached between 80 and 100 mm³; ****P < 0.0001. (C) Combination treatment with phenformin and PLX4720 cooperatively induces apoptosis and activation of AMPK, and inhibits cell proliferation, activation of ERK and S6, and reduced JARID1B expression in melanoma tumors. Representative images of mouse tumor samples subjected for various immunohistochemical analyses are shown. (Scale bar: 50 μm.)

Discussion

In this study, we have demonstrated that the combination of phenformin and the BRAF inhibitor PLX4720 offers a therapeutic advantage against BRAF mutant melanoma over either agent alone in both cell culture and animal models. Metformin, an analog of phenformin, is widely used as a first-line therapy for type 2 diabetes. Recent epidemiological studies have found that the subset of patients with type 2 diabetes who were treated with metformin had lower cancer risk and lower cancer-related mortality rates compared with patients treated with other therapeutics (9). Moreover, both metformin and phenformin have antitumor activities in various xenograft, carcinogen-induced, and genetically modified mouse models, raising strong interest in repurposing these drugs for cancer therapy (9). Consistent with previous reports in other models (19, 20, 24), we found that phenformin is much more potent than metformin in suppressing tumor growth, apparently because metformin requires an organic cation transporter to enter tumor cells while phenformin does not. Although metformin was used in combination with BRAF inhibitors in melanoma cells in an in vitro study, the IC₅₀ values for metformin were found to be in the ranges of 10–30 mM (21). This range of doses appears to be much higher than what can be achieved in humans with oral administration of metformin (22). In our in vivo studies, we found that phenformin by itself possesses weak activity against BRAF V600E melanoma tumors in both xenograft and GEM melanoma models. We have also examined the effect of metformin in these mouse models (Fig. S5) and did not observe any significant effect of metformin compared with the vehicle treatment. Metformin monotherapy was also recently shown to be ineffective in A375 melanoma xenografts in nude mice (15). However, in another recent study, metformin alone was found to promote the growth of A375 xenografts in nude mice, possibly through ERK-dependent up-regulation of VEGF-A (19). The reason for the discrepancy among these studies regarding the effect of metformin on BRAF^V600E- driven melanoma tumor growth remains to be resolved.

Understanding mechanisms underlying the antitumor activities of metformin and phenformin remains an active research topic. AMPK-dependent inhibition of mTOR signaling and protein synthesis has been proposed as a major mechanism supporting their antiproliferative effects (10). However, it has been suggested that metformin and phenformin may also inhibit mTOR signaling in an AMPK-independent manner (23). More recently, phenformin was shown to induce apoptosis in LKB1-deficient nonsmall cell lung cancer cells independent of AMPK activation (24). In addition to these cancer cell-autonomous mechanisms, it is possible that the systemic metabolic effects of metformin and phenformin, such as lowering circulating insulin/IGF1 levels, may also play a role (9). These systemic effects of metformin and phenformin have been recently observed in some mouse model studies, such as one using a tobacco carcinogen-induced lung cancer mouse model (25), but not in others, including a study using the Pten^+/− spontaneous lymphomamouse model (20) and the LKB1^−/− NSCLC model (24). It is therefore possible that different mouse cancer models or the approaches of drug administration may affect systemic metabolic responses. Our data from the GEM melanoma model are in agreement with the second group of studies. Moreover, using the shRNA knockdown approach, we concluded that at least part of sensitivity to phenformin in melanoma cells is mediated by AMPK activation in the tumor cells. Future investigations should focus on further exploring both the tumor cell-intrinsic and extrinsic effects of phenformin in melanoma.

Our analyses have revealed several potential mechanisms that might be involved in the synergistic effects of phenformin and BRAFi. We have demonstrated that phenformin and BRAFi exert cooperative activities in inhibiting mTOR signaling and inducing apoptosis in both BRAF V600E-mutant melanoma cells in vitro and tumors in vivo. Moreover, BRAFi appears to target the JARID1B-negative population of melanoma cells, which represents the majority of cells in the tumors, and spares the JAR1D1B-positive slowing cycling population. Conversely, phenformin suppresses the JARID1B-positive population, either by inhibiting their growth or by suppressing the switch from the JARID1B-negative to JARID1B-positive phenotype. Although additional studies are needed to understand how phenformin suppresses this population of cells, it is clear that combining phenformin with PLX4720 reduces both cell populations and results in tumor regression. Previously, metformin was shown to specifically target the cancer stem cell population of breast cancer cells by inhibiting the inflammatory response associated with cell transformation (15, 16). A combination of metformin with doxorubicin resulted in tumor mass reduction and delayed tumor relapse in xenograft tumor models (15, 16). Intriguingly, two recent reports have shown that treatment with a BRAFi in BRAF mutant melanoma enhanced oxidative phosphorylation capacity through up-regulation of PGC1α, a master transcription regulator for mitochondria biogenesis and function (26, 27). Because treatment of BRAFi appeared to render certain melanoma cells more addicted to oxidative phosphorylation, it is possible that BRAFi-treated melanoma cells would be more sensitive to phenformin, an inhibitor of mitochondrial oxidative phosphorylation (28), which may contribute to the synergistic effects of the phenformin/BRAFi combination. Future studies on the relationship among AMPK, JARID1B, and mitochondrial energy metabolism may reveal additional insight into the mechanisms underlying the synergistic effects of phenformin and BRAFi against melanoma.

BRAF-selective kinase inhibitors have shown great clinical benefits in malignant melanoma with BRAF V600E mutations in the initial phase of treatment. However, most of the responsive melanoma tumors treated with BRAF inhibitors developed resistance during the course of treatment, leading to resumed tumor growth. Recent studies have shown that reactivated ERK signaling due to amplification or mutation of proteins in the RAS-RAF-MEK-ERK pathway (i.e., NRAS, COT1, MEK1/2, and BRAF splicing variants) play a major role in acquired resistance to BRAFi (29). In addition, hyperactivation of the RTK (receptor tyrosine kinase)-PI3K-Akt survival pathway has also been identified as an alternative mechanism (29). Combination therapy strategies using BRAFi together with MEK inhibitors or the PI3K-Akt pathway inhibitors have been proposed to overcome the drug resistance and side effects associated with BRAFi single agent therapy. We demonstrated here that cotreatment of phenformin with BRAFi delayed the development of acquired resistance to BRAFi in melanoma cells in vitro. Moreover, our data suggest that JARID1B-positive, slow cycling cells could be involved in the development of acquired resistance to BRAFi. The effect of phenformin on limiting the switch to these slow-cycling cells may make combination with phenformin more advantageous than other combinatorial therapy strategies. Consistent with this notion, we found that treatment of WM115 cells with the MEK inhibitor GSK1120212, similar to PLX4720 BRAFi, only reduced the number of JARID1B-negative cells, but not JARID1B-positive cells (Fig. S6).

The main serious adverse effect of both metformin and phenformin in diabetic patients is lactate acidosis, which is predominantly found in those with impaired renal function. However, phenformin has a higher incidence rate of lactate acidosis than metformin (∼64 versus ∼3 cases per 100,000 patient years) (30). Because of this adverse effect, phenformin was discontinued for use treatment of type 2 diabetes in the United States in the late 1970s, whereas it is still being used in some other countries. Multiple clinical trials are underway to explore the utility of metformin for cancer therapy. In addition to metformin, the adoption of phenformin, a more potent biguanide, for cancer treatment is also worthy of exploration, because in comparison with many commonly used cancer chemotherapy and adjuvant therapies, it possesses relatively lower toxicity and might be more acceptable for treatment of cancer than diabetes. In summary, our preclinical findings presented here demonstrate that the combination of phenformin and BRAFi for treating melanoma with BRAF V600E mutations, including in the adjuvant therapy setting, warrant future clinical evaluation.

Materials and Methods

Flow Cytometric Analysis.

Apoptotic cells were detected by using BD FITC Annexin V Apoptosis Detection Kit I according to the manufacturer’s protocol and analyzed by BD FACSCalibur. For JARID1B expression FACS analysis, cells were fixed in 1% formalin in PBS for 20 min, followed by incubation with 90% methanol for 30 min at −20 °C. After wash with 1% FBS in PBS, anti-JARID1B antibody (for human cells) or anti–PLU-1 (for mouse cells) was added to the cells and incubated for 30 min at room temperature. After wash, the cells were stained with Alexa Fluor 647 conjugated anti-rabbit IgG. The samples were analyzed by BD LSRII Flow Cytometer. All flow cytometry data were analyzed with FlowJo software.

Animal Studies.

All animal experiments were performed by following Columbia University's Institutional Animal Care and Use Committee guidelines. For xenograft models, 6-wk-old female athymic mice (NCr^nu/nu) were purchased from Taconic Farms or Charles River. Animals were allowed a 1-wk adaptation period after arrival. A375 (2.5 × 10⁶) or Colo829 (2 × 10⁶) cells in 0.2 mL of basal culture medium were injected s.c. into the right lateral flank. For the genetically engineered mouse model, 6–8 wk-old Tyr::CreER; Braf^CA/+; Pten^lox/lox mice were topically administrated 4-hydroxytamoxifen (70% Z-isomer; Sigma H6278) in ethanol to induce tumor formation. Treatment began when the tumor volume reached between 80 and 100 mm³. Tumor volumes were calculated from caliper measurements by using the following ellipsoid formula: (D × d²)/2, in which D represents the large diameter of the tumor, and d represents the small diameter. Animals were randomly assigned to four groups that were administered vehicle [10% (vol/vol) DMSO in 1% (wt/vol) carboxymethylcellulosae], PLX4720, Phenformin, or PLX4720/Phenformin combination (same dose as used in the single-agent groups) by oral gavage twice per day for the duration of the experiment. Mice were weighed daily, and drug doses were adjusted accordingly. Levels of insulin and insulin-like growth factor-1 (IGF-I) in plasma were measured by using the insulin ELISA Kit (Millipore) and the IGF-I Quantikine ELISA Kit (R&D Systems), according to manufacturers’ protocol. Plasma glucose levels were measured by using OneTouch Ultra (LifeScan).

Immunohistochemistry Analysis.

Harvested mouse tissues were fixed in 10% neutral buffered formalin or in 70% ethanol and embedded in paraffin. Formalin fixed tissues were used for Ki67 and pAMPK staining, and ethanol fixed tissues were used for pERK, pS6, and TUNEL staining. The slides were deparaffinized by using HistoChoice clearing reagent (Amresco) and then rehydrated with water. Antigen retrieval for formalin fixed tissue sections was performed by heating slides in a pressure cooker for 10 min in citrate antigen retrieval solution. After wash with PBS, endogenous peroxidase activity was quenched with 3% hydrogen peroxide in PBS for 10 min at room temperature. For Ki67, slides were blocked with 5% normal goat serum in 0.3 M glycine and 0.25% Triton X-100 in PBS overnight at 4 °C and then incubated with anti-Ki67 antibody for 90 min, followed by incubation with biotinylated anti-rabbit IgG for 60 min. For pAMPK, pERK, and pS6 staining, slides were blocked with 5% normal goat serum in PBS for 30 min, followed by incubation with biotinylated anti-rabbit IgG for 30 min (Vector Laboratories). For TUNEL staining, slides were treated with Proteinase K in PBS for 10 min at room temperature, followed by endogenous peroxidase quenching. The slides were then incubated with TdT reaction buffer for 10 min, followed by an incubation with terminal deoxynucleotidyl transferase reaction mix for 1 h at 37 °C and rinse with stop buffer for 10 min. All slides were then incubated with avidin-biotin peroxidase complex for 30 min, and the signals were visualized by using DAB Substrate Kit (Vector Laboratories). The tissue sections were counterstained with Gill’s hematoxylin QS and mounted with VectaMount after dehydration.