Oncogenic H-RAS induces metformin resistance in head and neck cancer by promoting glycolytic metabolism

Abstract

Metformin administration has recently emerged as a candidate strategy for the prevention of head and neck squamous cell carcinoma (HNSCC). However, the intricate relationship between genetic alterations in HNSCC and metformin sensitivity is still poorly understood, which prevents the stratifications of patients harboring oral premalignant lesions that may benefit from the chemopreventive activity of metformin. In this study, we investigate the impact of prevalent mutations in HNSCC in response to metformin. Notably, we found that the expression of oncogenic HRAS mutants confers resistance to metformin in isogenic HNSCC cell systems and that HNSCC cells harboring endogenous HRAS mutations display limited sensitivity to metformin. Remarkably, we found that metformin fails to reduce activation of the mTOR pathway in HRAS oncogene expressing HNSCC cells in vitro and in vivo, correlating with reduced tumor suppressive activity. Mechanistically, we found that this process depends on the ability of HRAS to enhance glycolytic metabolism, thereby suppressing the requirement of oxidative phosphorylation to maintain the cellular energetic balance. Overall, our study revealed that HNSCC cells with oncogenic HRAS mutations exhibit diminished metformin sensitivity, thus shedding light on a potential mechanism of treatment resistance. This finding may also help explain the limited clinical responses to metformin in cancers with RAS mutations. Ultimately, our study underscores the importance of understanding the impact of the genetic landscape in tailoring precision cancer preventive approaches in the context of HNSCC and other cancers that are characterized by the presence of a defined premalignant state and, therefore, amenable for cancer interception strategies.

Introduction

Head and neck squamous cell carcinoma (HNSCC) arises most commonly in the oral cavity and oropharynx, and accounts for nearly 54,540 new cancer cases each year in the United States, resulting in 11,580 deaths in 2023 (1). The best-known risk factors for developing HNSCC include tobacco, alcohol consumption, and human papillomavirus (HPV) infection (2–4). Despite the implementation of innovative therapeutic modalities, including immunotherapy (5–7), there has been only marginal progress in the 5-year survival prognosis for individuals diagnosed with HNSCC at a late stage over the past few decades (1). Thus, early diagnosis and new preventive strategies are key to improving the prognosis of HNSCC patients, particularly in individuals who exhibit oral premalignant lesions (OPL), such as leukoplakia and erythroplakia who are at high risk of developing oral cavity HNSCC (8,9). While surgical excision is commonly the primary approach for treating oral leukoplakia, there is uncertainty regarding the effectiveness of various surgical interventions in preventing the development of oral HNSCC by removing potentially malignant oral lesions (10).

Early studies identified the sustained activation in the PI3K/mTOR signaling pathway as a widespread anomaly in HNSCC. While pivotal for HNSCC growth, this overreliance on PI3K/mTOR signaling presents a vulnerability that can be exploited for therapeutic targeting (11–15). Indeed, our prior studies supported that mTOR inhibitors display potent antitumor effects in multiple HNSCC models (16–18). A Phase II trial investigating mTOR blockade with rapamycin in newly diagnosed HNSCC patients revealed promising results and limited side effects, suggesting the potential efficacy of targeting the PI3K/mTOR pathway through pharmacological intervention in HNSCC (19). However, extending the use of rapamycin and rapalogs for cancer prevention in patients with OPL raises concerns about potential long-term immunosuppressive effects, especially in cases with increased malignancy risk (20). In this regard, our experimental studies suggested the benefit of repurposing metformin, a common type 2 diabetes treatment, as a preventive measure to hinder OPL progression to HNSCC by indirectly targeting mTOR (21). The introduction of metformin curbs HNSCC cell proliferation, partly by dampening mTOR activity within the mTORC1 complex (21,22). These suppressive effects are regulated by the AMP-activated protein kinase (AMPK) signaling pathway, a central mechanism in sensing cellular energy status (23). Furthermore, metformin may prevent the progression of HNSCC by targeting cancer-initiating cells and promoting cancer cell differentiation (24). Consistent with these experimental findings, epidemiological studies illustrated a reduced risk of HNSCC in individuals using metformin (25,26). Additionally, our recent Phase IIa metformin trial, which included participants with OPL, demonstrated mTOR signaling inhibition and a 60% histological response rate, with 17% complete and 43% partial responses, which was particularly notable among current and former smokers (27). Thus, metformin has garnered substantial attention for its potential to intercept HNSCC progression.

However, identifying the specific patient populations likely to benefit from metformin and the underlying mechanisms governing this response still needs to be discovered. We utilized an oncogene expression approach reflecting particular genetic alterations in HNSCC to investigate the molecular alterations that impact metformin sensitivity. These studies revealed that oncogenic HRAS mutations significantly diminish metformin sensitivity. These findings hold promise for improving the effectiveness of personalized cancer prevention approaches for HNSCC and may explain the limited clinical activity of metformin in many cancer types exhibiting a high frequency of activating mutations in RAS oncogenes.

Materials and methods

Cell culture.

Cell lines, which include the CAL27, CAL33, HN12, and Detroit562 cells as parental control and cells with endogenous HRAS mutations (SCC17B and ORL214), were obtained from the NIH/NIDCR Oral and Pharyngeal Cancer Branch cell collection (28,29). The cell lines used in this study were recovered from previous lab stocks preserved in liquid nitrogen. To eliminate potential mycoplasma contamination, the cells underwent treatment with the PlasmocinTreatment Removal Agent (InvivoGen). After recovery, cells were maintained using the Plasmocin Prophylactic Removal Agent to prevent any new mycoplasma contamination. Cultivation of these cell lines took place in DMEM (D6429, Sigma-Aldrich), or DMEM/ F-12 medium (Gibco) supplemented with 10% fetal bovine serum (FBS) (F2442, Sigma-Aldrich) and 1% antibiotic/antimycotic solution (A5955, Sigma-Aldrich) under standard conditions of 5% CO2 at 37°C. Collagen (Corning 354249) was applied during the culture of the ORL214 cell line.

DNA constructs.

The Cancer Pathways ORF Kit (30), generously provided by Kris Wood, includes cDNAs employed in a cancer cell screen aimed at identifying mutants that confer resistance to anticancer agents. The constructs were expressed in CAL27 cells individually using lentiviral infection, and underwent screening with metformin, with the goal of identifying those that demonstrated resistance.

Seahorse metabolic flux assays.

A seeding density of 10,000 cells per well was established in the Agilent Seahorse XF96 cell culture Microplate, allowing for a 12-hour incubation period before initiating the metabolic flux analysis. The Glycolysis stress testing and ATP rate assay were performed at the La Jolla Institute for Immunology (LJI) Immunometabolism Core facility.

Cell viability assay.

The cells were initially seeded in 96-well plates and subjected to treatment with Metformin once they adhered to the plates under low glucose (5 mM) DMEM conditions. Following a 72-hour treatment period, the culture medium was enriched with 1/100 of the culture volume of Aquabluer reagent (#6015, MultiTarget Pharmaceuticals LLC, Colorado Springs, CO, USA) for 3 hours. Absorbances were subsequently measured at 570 nm using the Spark microplate reader (Tecan, Switzerland).

Colony formation assay.

Cells were seeded at a density of 1,000 cells per well in 12-well plates and, upon attachment, were treated with Vehicle, metformin (1 mM). The 10-day treatment period involved changing the medium every 2 to 3 days. After completion of the treatment, colonies in the cell culture plates were gently washed twice with PBS and fixed for 5 minutes using a methanol-acetic acid solution (3:1). Subsequently, a 15-minute staining was performed with 0.5% crystal violet solution diluted in methanol. Excess stain was removed by repeated washing with PBS. The percentage of colony area was calculated using ImageJ (RRID:SCR_003070).

Western blotting.

Cells were washed in cold PBS after metformin treatment, lysed in RIPA buffer with protease and phosphatase inhibitors, and protein concentrations were determined. Equal protein amounts were separated by SDS-PAGE, transferred to PVDF membranes, and blocked in 5% milk-TBST. Membranes were probed with primary antibodies overnight, followed by secondary antibody incubation and visualization using chemiluminescence. Antibodies,such as against S6, phospho-S6 antibody were from Cell Signaling Technology. (RRID:AB_331355, RRID:AB_331679)

Immunohistochemistry staining.

Samples were fixed in zinc formalin, embedded in paraffin, and 5 μm sections were obtained for staining. Immunohistochemistry involved deparaffinization, antigen retrieval, and staining with Ki67 (M7240, Dako,1:200) and phospho-S6 antibody (CST#2211l,1:400). Image analysis was conducted using QuPath software (RRID: SCR_018257).

In vivo mouse experiments.

All animal studies were conducted following approval from the University of California San Diego (UCSD) Institutional Animal Care and Use Committee (IACUC), under protocol ASP #S15195, and adhered to relevant ethical regulations for animal testing and research. Mice (RRID: IMSR_JAX:007850) for the study were procured from the Jackson Laboratory. In xenograft models, 1 million cells were transplanted into both flanks. For short-term treatment for biochemical analysis, mice were randomized and administered metformin in drinking water (2.5 mg/ml) when the average tumor volume reached approximately 100 mm³. In the long-term tumor growth curve study, metformin (2.5 mg/ml) was included in the drinking water from the day of cell injection. In an orthotopic mouse model, 0.5 million cells were injected into the tongue of SCID NOD mice obtained from the UCSD in-house breeding colony. Metformin (2.5 mg/ml) was administered on the day of injection. The mice were euthanized at the indicated time points

Statistical analysis.

Data analysis was conducted using GraphPad Prism9 for MacOS (GraphPad Software, San Diego, CA, USA RRID:SCR_002798). Comparisons between experimental groups were performed using unpaired t-tests. Statistical significance is denoted by asterisks (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001), and non-significant results are represented as n.s (P > 0.05).

Data availability.

The data generated in this study are available upon request from the corresponding author.

Results

An oncogene open reading frame expression approach reveals that HRAS mutations reduce metformin sensitivity in HNSCC

To explore the impact of genetic alterations on metformin sensitivity in HNSCC cells, we individually introduced a selection of cDNAs from the Cancer Pathways ORFs Kit (30) into CAL27 cells, a typical HNSCC cell model system that does not harbor mutations affecting the PI3K-mTOR pathway (28). We focused on the expression of cDNAs encompassing mutations found in HNSCC and other cancer types, including, HRAS, STK11, PTEN, and AKT. Through this approach, we sought to identify genetic alterations conferring resistance to metformin, a crucial step in refining precision cancer prevention strategies. Cell viability for mass culture cells was assessed using the Alamar Blue assay, and a dose-response curve fit model was used to determine the IC₅₀ values along with a 95% confidence interval range. Within the PIK3CA/mTOR pathway, gene mutations, including HRAS(G12V), STK11(D194N), PTEN(R130Q), and AKT1(E17K) (Fig. 1A), emerged as promising candidates for predicting metformin resistance. Among these genes, we focused on HRAS, as 6% of HPV- HNSCC patients carry HRAS mutations, which exhibit mutational hotspots in G12, G13, and Q61 residues (Supplementary Fig. 1A).

Figure 1: Metformin sensitivity in response to gene alterations in the PIK3CA/mTOR signaling pathway.

A. To investigate how gene alterations affect the response to metformin, a collection of cDNAs, including HRAS(G12V), STK11(D194N), PTEN(R130Q), and AKT1(E17K) sourced from the Cancer Pathways ORFs Kit, were individually introduced into CAL27 cells. Subsequently, these lentiviral infected mass cell cultures were treated with metformin ranging from 0 to 30 mM for 72 hours. The resulting IC₅₀ values and corresponding 95% confidence interval ranges are compiled in the table. B. Selected CAL27 cell clones with isogenic HRAS G12V mutation (CAL27 HRAS) and cell lines with endogenous HRAS mutations, SCC17B(HRAS Q61L) and ORL214(HRAS G12C) were treated with metformin (0 to 30mM) for 72 hours. The table summarizes IC₅₀ values and 95% confidence interval ranges. C. Clonogenic assays were performed to evaluate the ability of single cells to form colonies. CAL27 Control and CAL27 HRAS cells were treated with metformin (1mM), and the colony area was quantified using ImageJ. (***p<0.001, ^nsp>0.05, unpaired t-test) D. Cell lysates were collected following a 24-hour treatment in vitro with indicated doses of metformin. Western blotting analysis was conducted to assess markers in the mTOR signaling pathway.

To confirm the potential association of HRAS hotspot mutations with metformin sensitivity in HNSCC cells, we employed a panel of cells, including isogenic CAL27 cells in which we stably expressed HRAS G12V mutations (CAL27 HRAS) by lentiviral gene delivery, and cells harboring endogenous HRAS mutations, specifically SCC17B (HRAS Q61L) and ORL214 (HRAS G12C) (28,31). Compared to the parental CAL27 cells (CAL27 Control), the isogenic HRAS mutant CAL27 cells displayed a notable insensitivity to metformin treatment in vitro (Fig. 1B). This resistance to metformin was not limited to the isogenic HRAS CAL27 cells; similar results were observed in cells with endogenous HRAS mutations, with SCC17B cells exhibiting an IC₅₀ of 12.71 mM and ORL214 of 12.42 mM. The table (Fig. 1B) summarizes these IC₅₀ values and their corresponding confidence interval ranges.

To further evaluate the in vitro proliferative capacity of cells bearing HRAS mutations, clonogenic assays were conducted to assess the ability of single cells to form colonies, thereby gauging their potential for unlimited cell division. The results unveiled that the colony-forming ability of CAL27 HRAS cells remained significantly greater than when compared to the parental control cells (Fig. 1C). At the same time, metformin treatment also had only a limited impact on colony formation in SCC17B and ORL214 cells (Supplementary Fig. 1B). Additionally, We overexpressed HRAS in HNSCC cell lines (CAL33, Detroit 562, HN12), and observed reduced metformin sensitivity compared to wild-type cells. HRAS-mutant cells had higher IC₅₀ values, indicating resistance to metformin, as confirmed by dose-response curves showing higher viability at increasing drug concentrations. (Supplementary Fig.2)

To investigate the impact of HRAS expression on metformin-induced signaling events, we first examined mTOR signaling using the isogenic HRAS mutant HNSCC cell line, in which HRAS is expressed in a comparable genetic background. In these experiments, cells were subjected to in vitro metformin treatment for 24 hours, and lysates were collected to analyze signaling pathway markers using western blotting (Fig. 1D). From our analysis, we were able to confirm the expression of the HRAS G12V mutation through the use of a mutation-specific antibody. Importantly, our studies demonstrated that the expression of OCT3, a metformin transporter potentially affecting metformin sensitivity (22), remained unaltered. Examining the mTOR signaling pathway, we observed consistent results in parental cells, with the accumulation of pAMPK, the active phosphorylated form of AMPK, and pACC, a downstream target of AMPK (32), leading to the inhibition of pS6, the most downstream target of mTOR signaling, in response to metformin treatment (22). However, in cells expressing HRAS, the mTOR signaling pathway remained active, with no activation of AMPK (Fig. 1D). The results underscore that HRAS mutant HNSCC cells exhibit an absence of metformin responsiveness, as indicated by the continued activity of the mTOR signaling pathway and the lack of AMPK activation, aligned with a reduced sensitivity to the growth suppressive effects of metformin.

Effect of oncogenic HRAS on glycolytic metabolism in HNSCC cells

To further investigate the mechanism by which HRAS drives metformin resistance, we focused on the metabolic pathway controlling AMPK signaling. AMPK is a critical cellular energy sensor that monitors the AMP-to-ATP ratio, activating when cellular energy levels drop to restore energy balance. The absence of activation of AMPK implies that metformin does not stimulate the cellular energy sensor in HRAS mutant cells. In search of the underlying mechanism, we hypothesized that enhanced cellular metabolism, such as glycolysis, might compensate for OXPHOS inhibition in these cells. To explore this possibility, we utilized the Seahorse XF Glycolysis Stress assay to measure the extracellular acidification rate (ECAR) in real-time in CAL27 control and CAL27 HRAS cells. We observed an immediate glycolytic response upon introducing glucose to both cell lines (Fig. 2A). Interestingly, HRAS mutant cells exhibited a significantly higher glycolytic flux than parental cells. This initial finding indicates a potential link between the oncogenic HRAS mutation and enhanced glycolytic metabolism. Subsequent addition of oligomycin further accentuated the glycolytic response in both cell lines. Notably, HRAS mutant cells maintained a substantially higher glycolytic capacity. Lastly, adding the glycolysis inhibitor 2-DG effectively halted glycolysis in both cell lines, confirming that the changes in ECAR were indeed due to glycolytic activity. Sequentially, with enhanced glycolytic activity, an enhanced basal Oxygen Consumption Rate (OCR) was also observed in HRAS mutant cells (Supplementary Fig. 3A). The calculated glycolysis and glycolytic capacity (Fig. 2B) provide quantitative insights into these findings. Putting together, the results suggest that the presence of oncogenic HRAS mutations significantly enhances glycolytic metabolism in HNSCC cells.

Figure 2: Oncogenic HRAS on glycolytic metabolism.

A. Seahorse XF glycolysis stress test in CAL27 Control and CAL27 HRAS cells, measured as the extra-cellular acidification rate (ECAR) in response to glucose (Glu), oligomycin (Omy), and 2-deoxy-D-glucose (2DG) injections as indicated. B. The bar plot demonstrated the quantification of glycolysis, glycolytic capacity, and glycolytic reserves. (***p < 0.001, ^nsp > 0.05, unpaired t-test) C. Seahorse XF real-time ATP rate assays were conducted to quantify the total ATP production. (***p < 0.001, unpaired t-test) D. This schematic illustrates the role of HRAS in enhancing glycolysis metabolism. (Created with BioRender.com)

We next investigated the role of enhanced glycolysis in ATP production in HRAS mutant cells. To achieve this, we used the Seahorse XF Real-Time ATP rate assay, which allowed for concurrent measurement of glycolysis and oxidative phosphorylation (OXPHOS), ultimately calculating the total rate of cellular ATP production and the proportional contribution from each pathway. The total ATP production rate exhibited a noteworthy increase in HRAS mutant CAL27 cells compared to parental control cells(Fig. 2C) and this increase was primarily attributed by the enhanced glycolytic pathway, although oxidative phosphorylation (OXPHOS) also contributed to a portion of ATP production (Supplementary Fig. 3B and 3C). Our results highlight the elevated glycolytic activity in HRAS mutant CAL27 cells, even under normoxic conditions (Fig. 2D).

Figure 3: Restoration of metformin sensitivity by glycolysis blockade.

A. The schematic illustrates the experimental approach of suppressing glycolytic activity in cells by substituting glucose with galactose (10 mM).(Created with BioRender.com) B. Measurement of glycolytic activity suppression by seahorse XF Glycolysis Stress Test, indicated by the extra-cellular acidification rate (ECAR). (***p < 0.001, unpaired t-test) C. Representative stained images of clonogenic assays under 1mM metformin treatment in CAL27 Control and CAL27 HRAS cells in normal or glycolytic suppression (Galactose) condition. D. The bar chart illustrates the quantification of clonogenic assays. (***p < 0.001, ^nsp > 0.05, unpaired t-test)

Blockade of glycolysis restores metformin sensitivity in HRAS mutant cells.

We next aimed at exploring whether the heightened glycolytic metabolism plays a critical role in the development of metformin insensitivity in HRAS mutant CAL27 cells. To address this question, we suppressed glycolytic activity in the cells by substituting glucose with galactose in the cell culture medium (Fig. 3A), a common approach known for diminishing glycolysis while preserving oxidative phosphorylation (OXPHOS) functionality (33). Our observations yielded an immediate decrease in glycolytic activity when subjected to the galactose condition. This reduction was evident in CAL27 control and CAL27 HRAS cells (Fig. 3B). Furthermore, we noted a substantial reduction in ATP production driven by glycolysis in the galactose condition (Supplementary Fig. 3D). At the same time, OXPHOS remained unaffected (Supplementary Fig. 3E). These data support the efficacy of galactose in suppressing glycolysis, thus preserving OXPHOS.

After glycolytic suppression, we assessed the sensitivity of cancer cells to metformin treatment using clonogenic assays. Notably, CAL27 control cells remained responsive to metformin treatment, while HRAS mutant CAL27 cells exhibited a significant restoration of sensitivity to metformin (Fig. 3C and 3D). These results suggest that the resensitization of HRAS mutant cells to metformin may be directly linked to the suppression of glycolysis. In summary, these results support a direct association between elevated glycolytic metabolism and the development of metformin insensitivity in HRAS mutant HNSCC cells.

As we restored metformin sensitivity in HRAS mutant HNSCC cells in galactose medium, we next aimed to validate the therapeutic potential of combining metformin with the glycolysis inhibitor dichloroacetate (DCA). Both drugs target metabolic pathways, with metformin inhibiting mitochondrial complex I and DCA promoting oxidative phosphorylation by inhibiting pyruvate dehydrogenase kinase (PDK) (34). In vitro experiments were conducted on SCC17B cells with endogenous Q61L HRAS mutation and CAL27 HRAS cells to assess cell viability and synergy using the SynergyFinder tool(35). The combination of metformin and DCA demonstrated synergy in reducing cell viability. In SCC17B cells (Fig. 4A), the viability heatmap showed a dose-dependent reduction, particularly at higher metformin concentrations (≥1 mM). CAL27 HRAS cells (Fig. 4B) exhibited a broader synergy across both low and high drug concentrations. The synergy score distribution (Fig.4C) shows that both SCC17B (Q61L) and CAL27 HRAS cells demonstrate overall synergistic effects with the combination of metformin and DCA. These findings support the notion that RAS-induced glycolysis limits the effectiveness of metformin as a cancer preventing agent.

Figure 4: Synergistic effects of Metformin and Dichloroacetate (DCA) in HRAS-mutant cells.

A. The viability heatmaps show the effects of increasing concentrations of Metformin and DCA on SCC17B (Q61L) or B. CAL27 HRAS, and the synergy score heatmaps demonstrate the synergistic effect of the combination treatment, where positive values (pink) represent synergy and negative values (blue) represent antagonism. The combination index (CI) was determined using the HSA method (CI > 10 synergism, 0 < CI < 10 additivity, CI < 0 antagonism). C. Violin plot displaying the distribution of HSA synergy scores across SCC17B (Q61L) and CAL27 HRAS cells, with scores above zero indicating overall synergistic effects.

In addition, we conducted tests on other HRAS overexpressing HNSCC cell lines, including Detroit 562 HRAS, CAL33 HRAS, and HN12 HRAS (Supplementary Fig. 4). The viability heatmaps for these cell lines revealed a dose-dependent reduction in cell viability with increasing concentrations of metformin and DCA. Our synergy analysis showed varying degrees of synergy across all cell lines. The violin plot comparing the synergy score distribution across these cell lines indicates consistent positive synergy in all tested cells, further supporting the therapeutic potential of combining metformin with glycolysis inhibitors in HRAS-mutant cancers.

To test the hypothesis that hypoxia induces metformin resistance by complementary approach, we conducted experiments using cobalt chloride (CoCl₂) in CAL27 cells. CoCl₂ is commonly used as a hypoxia-mimetic agent because it stabilizes hypoxia-inducible factor 1-alpha (HIF1α) under normoxic conditions by inhibiting prolyl hydroxylase, which prevents HIF1α degradation (36). This allows for the activation of hypoxia-related signaling pathways even in the presence of normal oxygen levels, mimicking the hypoxic conditions often observed in tumor microenvironments. The results demonstrated that CoCl₂ treatment upregulated HIF1α expression, confirming the establishment of hypoxic-like conditions (Fig. 5A). Under these conditions, metformin treatment showed reduced efficacy in decreasing cell viability compared to normoxic conditions (Fig. 5B). Furthermore, analysis of TCGA data (Fig. 5C) reveals that HRAS-mutant cancers exhibit a significantly higher hypoxia signature compared to wild-type cancers. This suggests that HRAS mutations may exacerbate the hypoxic response in tumors, potentially contributing to further metabolic shifts toward aerobic glycolysis (the Warburg effect) and reducing the efficacy of metabolism-targeted therapies, such as metformin.

Figure 5: Hypoxic-like conditions induce metformin resistance in HNSCC cells.

A. Western blot showing the dose-dependent induction of HIF1α in CAL27 cells treated with increasing concentrations of cobalt chloride. (CoCl₂) B. Bar graph comparing the viability of CAL27 cells treated with 1 mM Metformin under normoxic conditions (Control) and hypoxia-mimetic conditions induced by 100 μM CoCl₂. A significant reduction in metformin sensitivity is observed under hypoxic-like conditions (***p < 0.001, unpaired t-test). C. comparing the Reactome Hypoxia Signature ssGSEA scores between wild-type (WT) and HRAS-mutant head and neck squamous cell carcinoma (HNSCC) samples from the TCGA dataset (n=520). HRAS-mutant tumors show a significantly higher hypoxia signature. (p = 0.030, unpaired t-test)

HRAS mutations decrease the tumor suppressive effect of metformin in vivo.

Based on our in vitro findings, we next extended our studies into xenograft mouse models. In these experiments, we transplanted parental CAL27 cells and SCC17B cells into the flanks of nude mice, subsequently subjecting them to treatment with either metformin in their drinking water or placebo (Fig. 6A). For these studies, we administered 2.5 mg/ml of metformin in the drinking water, which we have previously shown to achieve metformin blood levels comparable to those achieved in diabetic patients (22) and in our recent Phase IIa metformin trial (27). The results demonstrated that, in the case of CAL27 tumors, metformin treatment resulted in a significant delay in tumor progression, as reflected by the substantial tumor size and weight reduction within the metformin treatment group (Fig. 6B and 6C). Conversely, the growth pattern of HRAS mutant tumors (SCC17B HNSCC xenografts) under metformin treatment mirrored that of the untreated group (Fig. 6B and 6C), which was illustrated by the tumor growth curve. Thus, although concentrations of metformin used in in vitro models are higher than those achieved in vivo as they do not fully recapitulate the metabolic state of HNSCC, these findings using HNSCC xenograft models support the resistance of HRAS mutant tumors to metformin treatment in the in vivo situation.

Figure 6: In vivo validation of HRAS mutation-mediated resistance to metformin.

A. Schematic representation of the experimental setup involving transplantation of CAL27 cells and SCC17B cells into the flanks of the nude mice, followed by treatment with either metformin(2.5mg/ml) in drinking water or placebo. B. Tumor progression results from the xenograft mouse model experiment. (***p < 0.001, ^nsp > 0.05, unpaired t-test, compared tumor volume at day 30) C. Tumor mass measured at day 30, illustrating the substantial reduction in tumor size and weight within the metformin treatment group. (*p < 0.05, ^nsp > 0.05, unpaired t-test) D. Representative images of the Immunohistochemistry (IHC) staining to assess the response of cancer cell signaling transduction to metformin treatment. E. The bar chart illustrates the quantification analysis of IHC staining for biomarkers, including Ki67 and pS6 (*p < 0.05, **p < 0.01, ^nsp > 0.05, unpaired t-test)

Additionally, we utilized an orthotopic mouse model that mirrors the human cancer origin, by implanting parental CAL27 cells and CAL27 cells with isogenic HRAS mutation into the tongue of SCID NOD mice. These mice were subsequently exposed to either metformin in their drinking water or placebo (Supplementary Fig. 5A). CAL27 tumors positively responded to metformin treatment, significantly retarding tumor progression and reducing tumor size. Conversely, the growth of isogenic HRAS mutant tumors under metformin treatment closely paralleled that of the untreated group (Supplementary Fig. 5B). The tumor growth curve underscored the resistance of HRAS mutation cells to metformin.

Furthermore, we examined whether the response of cancer cell signaling to metformin in HRAS mutant cells correlates with resistance in vivo. Tissue samples were embedded in paraffin and processed for immunohistochemistry (IHC) analysis. The sections were stained for KI67, as a proliferative marker, and pS6, which reflects mTOR pathway activation (Fig. 6D and 6E). Stained tissues were analyzed by an automated positive cell detection algorithm within the Qupath software (37). The quantification results demonstrated a statistically significant decrease in pS6-positive stained cells in the metformin-treated group of parental CAL27 cells, indicating a marked inhibition of the mTOR signaling pathway. Additionally, KI67 cell proliferation marker staining provided supporting evidence that metformin effectively hinders tumor proliferation of parental CAL27 cells in the in vivo setting. In contrast, the IHC analysis revealed no significant changes in the expression levels of the mTOR marker pS6 and the cell proliferation marker KI67 in the HRAS mutant cells, specifically in SCC17B cells and CAL27 cells expressing HRAS mutations (Supplementary Fig. 5C). In summary, the in vivo experiments conducted with xenograft mouse model systems supported that HRAS mutations confer metformin resistance.

Discussion

Metformin is emerging as a potential cancer preventive agent, yet selecting suitable patient candidates for ongoing and future clinical trials remains challenging. In our study, we investigated key oncogenic events frequently associated with dysregulated mTOR pathway signaling, including AKT1, STK11, PTEN, and HRAS, to begin addressing their potential contribution to metformin sensitivity. While HRAS, PTEN, and AKT mutated proteins act upstream of mTOR (23), STK11 mutations, a frequent event in lung cancer, impair AMPK-mediated inhibition, resulting in decreased AMPK activity and relieving the inhibition on mTOR (23,38). Our analysis suggests that each of these mutations confer insensitivity to metformin, aligned with the central role of mTOR activity in the response to metformin treatment. We focused specifically on the link between oncogenic HRAS mutations and metformin sensitivity, which may enhance the opportunity to impact the outcome of current and planned cancer prevention trials. HRAS, as part of the RAS oncogene family alongside KRAS and NRAS, plays a pivotal role in cancer development, including HNSCC (29,39,40). Our previous clinical study also unveiled HRAS mutations in the OPL samples (27), suggesting that this gene may undergo mutations at the early stages of HNSCC development. Although limited sample size prevented the evaluation of the impact of HRAS mutations in our metformin trial, one of the two HRAS mutant lesions was among the very few cases that progressed in tobacco-associated OPL lesions (27). Thus, these results and our new experimental findings strongly suggest that HRAS mutations may adversely impact metformin sensitivity, which can be evaluated in future precision prevention trials.

Our findings may also shed light on the challenges encountered in exploring metformin as a potential cancer prevention and treatment option, particularly concerning its efficacy in the context of RAS mutations. For instance, in pancreatic cancer, which is characterized by a remarkable prevalence of KRAS alterations (88% of the cases (41)), the effectiveness of metformin was evaluated in two separate trials. A double-blind, randomized, placebo-controlled phase II trial failed to demonstrate the significant benefit of metformin use in advanced pancreatic cancer receiving gemcitabine and erlotinib (42). Similarly, metformin treatment did not impact the disease-free or overall survival of metastatic pancreatic cancer patients (43). These findings underscore the limited effectiveness of metformin in pancreatic cancer therapy. RAS mutation frequencies are also notably enriched in other malignancies, including lung adenocarcinoma (32% KRAS mutations) and colorectal cancers, including colon adenocarcinoma (50% KRAS mutations) and rectal adenocarcinoma (50% KRAS mutations) (41). Clinical trials evaluating the efficacy of metformin in these cancer types have also yielded disappointing results (44–46). Thus, collective findings from studies across multiple cancer types can be reinterpreted in light of our current observations, supporting the limited effectiveness of metformin in cancers exhibiting widespread genomic alterations in RAS genes.

Although our study has not addressed the potential immune modulatory effects of metformin in the tumor microenvironment, our experimental approaches have helped identify an intrinsic metabolic mechanism driving metformin resistance in HRAS-mutant HNSCC cells, Mechanistically, we obtained evidence that elevated glycolytic metabolism in HRAS mutant cells may drive their insensitivity to metformin treatment. We found a significant up-regulation of glycolysis metabolism in HRAS mutant cells, consistent with prior studies in KRAS mutant cell models (47). Furthermore, by inhibiting glycolytic metabolism, we restored metformin sensitivity, while mimicking hypoxia and consequently HIF1-α expression, which enhances glycolysis (48) promoted resistance. These interventions support the role of enhanced glycolysis in developing resistance to metformin treatment. Thus, we can hypothesize that enhanced metabolic rate and glycolysis that characterize most solid tumors (49) may explain, at least in part, why metformin treatment may not have met the initial expectations for cancer treatment. Instead, our observations reinforce the potential use of metformin in cancer chemoprevention, particularly by intervening at earlier stages, such as before cancer initiation or at the earliest stages of cancer development (50).

This is especially crucial given the common occurrence of hypoxia in cancer cells as tumors progress, where hypoxia arises when tumors surpass their vascular capacity, resulting in inadequate oxygen supply to the tumor mass (51). Consequently, this hypoxic environment often triggers a metabolic shift known as aerobic glycolysis or the Warburg effect (49,51), which, in light of our study, could potentially result in metformin resistance.

Indeed, promising results have emerged from our study in subjects with OPL (27) and using metformin to prevent colon cancer recurrence (52), both involving potential pre-cancer lesions, rather than using metformin in established solid tumors. Specifically, the latter trial enrolled patients who had previously undergone endoscopic resection of colorectal adenomas or polyps. Low-dose metformin was associated with reduced prevalence and number of metachronous adenomas or polyps, which may have arisen from small residual premalignant lesions. This emphasizes the need to explore the clinical benefits of metformin for cancer prevention in cancer types for which an established premalignant state can be identified, rather than once tumor progress to a state in which they undergo a metabolic, glycolytic switch that may render metformin ineffective.

In summary, our study supports a link between oncogenic mutations in HRAS with elevated glycolytic metabolism and metformin resistance. Ultimately, predicting metformin sensitivity based on specific genetic alterations can significantly enhance chemoprevention trial effectiveness, enabling the inclusion of individuals most likely to benefit from metformin treatment. Our findings also underscore the importance of genetic alterations in cancer prevention strategies, thus offering insights to select suitable participants based on their genetic profiles, thereby potentially boosting future metformin trial outcomes.

Prevention Relevance Statement:

Our findings highlight the challenges of using metformin for cancer prevention in RAS-mutant cancers, where elevated glycolysis may reduce drug efficacy. This underscores the need to explore metformin’s potential in early, premalignant stages, before metabolic shifts render it less effective