Lipophilic statins deplete GPX4 to promote ferroptosis and sensitize cancer cells to immune checkpoint blockade

Abstract

Statins have been reported to exert anticancer activity, varying with cancer type and specific statins. These findings suggest that more mechanistic insight into the anticancer effects of statins is needed. Here, we interrogated the ability of statins to induce cell death and ferroptosis in melanoma and colorectal cancer. First we showed that statins induce cell death in patient-derived melanoma cell lines and that lower expression of mevalonate pathway genes correlates with increased CD8+ T cell infiltration and improved overall survival in melanoma patients. We found that lipophilic statins induce cell death with features of ferroptosis. Transcriptional data also revealed system level changes to a variety of ferroptosis-related pathways. We found that mevalonate rescued statin-induced cell death. Mechanistically, mevalonate-derived isopentyl pyrophosphate is necessary for isopentylation of tRNA [Ser]Sec, which is required for efficient synthesis of the selenoprotein ferroptosis suppressor GPX4. Given the emerging role for ferroptosis in antitumor immunity, we tested lipophilic statins including simvastatin alone and in combination with α-PD1 in vivo and found that simvastatin and α-PD1 promoted tumor clearance and extended survival in 20% to 60% of mice alone but in nearly 100% of mice when administered together. Simvastatin also depleted GPX4 in vivo. These results highlight the therapeutic potential of statin use in combination with immunotherapies.

Introduction

Statins inhibit the rate-limiting step of the cholesterol synthesis pathway: the conversion of β-Hydroxy β-methylglutaryl-Coenzyme A (HMG-CoA) into mevalonate by HMG-CoA reductase (HMGCR). In addition to their role as widely prescribed, FDA-approved cholesterol-lowering drugs, statins have also been explored as anticancer agents. Mevalonate and its downstream metabolites can promote tumor growth, and statins have been shown to induce tumor cell death, reduce cancer risk, lower cancer grade and stage at diagnosis, and decrease the risk of recurrence [1–6]. Through genotype-selective, high-throughput combinatorial drug screening, we previously identified simvastatin as a therapeutic candidate for RAS-mutant melanomas [7]. We also showed that simvastatin synergizes with MAPK inhibition in melanomas by inhibiting isoprenylation and RAS-dependent AKT and Hippo signaling [8]. Other reports of the anticancer activity of statins, however, vary between cancer types and specific statins, so additional mechanistic insight into how statins target tumors and their microenvironments remains necessary.

Ferroptosis is an iron-dependent and lipid peroxide-driven regulated cell death modality that is morphologically, biochemically, and genetically distinct from apoptosis and other forms of programmed necrosis [9]. Since the discovery of ferroptosis, some studies have shown that statins induce it in cancer cells, while others report the opposite [10–14]. Multiple mechanisms have been put forth for the regulation of ferroptosis by statins, in cancer and non-cancer states, including increasing GSH, GPX4 and SLC7A11 levels (cystine/glutamate antiporter xCT) [15]. To better define statin anticancer activity, we interrogated the ability of the seven most commonly prescribed statins to induce cell death and ferroptosis in melanoma and colorectal cancer cell lines.

We demonstrate that in human melanoma, decreased expression of mevalonate pathway genes correlates with higher CD8+ T cell infiltration and improved response to immunotherapy. We show that lipophilic statins uniformly promote ferroptosis by decreasing GPX4, the primary enzymatic suppressor of ferroptosis, in a mevalonate-dependent manner. GPX4 is a selenocysteine-containing protein and its synthesis is dependent on mevalonate-derived isopentylation of tRNA [Ser]Sec by TRIT1 [16–17]. Consequently, the transcriptional profiles of cancer cells treated with statins and the ferroptosis inducer RSL3, which inhibits GPX4, show remarkable overlap. In murine models of melanoma and colorectal cancer, simvastatin stimulates antitumor immune responses on its own and improves responses to α-PD1 immunotherapy, accompanied by an increase in intratumoral lipid peroxidation. Altogether, these findings broaden our understanding of the molecular basis for statin anticancer activity and identify statins as an existing class of medications with the potential to augment immunotherapy responses.

Materials and methods

Cell culture

The YUMMER1.7 (YR1.7) murine melanoma cell line (RRID: CVCL_A2AX) was established in our lab as previously described in 2017 [18]. YUSIK (RRID: CVCL_B487), YUCOT (RRID: CVCL_J066), and YUGASP (RRID: CVCL_J520) patient-derived melanoma cell lines were established as previously described and were a kind gift from from Dr. Ruth Halaban and Antonietta Bacchiocchi (Yale University, 2022) [19]. The MC38 murine colorectal cancer cell line (RRID: CVCL_B288) was obtained from the NIH NCI DCTD tumor repository in 2018. All cell lines tested negative for mycoplasma contamination and were used within 15 passages after thawing. Cell lines were not authenticated. YUSIK was established from a stage IV metastatic lesion on the right thigh of a 49-year-old female patient and YUCOT was established from a stage IV metastatic lesion on the right neck of a 33-year-old female patient both harboring BRAF mutations. YUGASP was established from a stage IV metastatic lesion on the right pretibial region of the leg of an 88-year-old female patient harboring a RAS mutation. Tumor samples were collected with written informed patient consent according to Health Insurance Portability and Accountability Act (HIPAA) regulations with Human Investigative Committee protocol. All studies involving human tissue were conducted in accordance with recognized ethical guidelines, including the Declaration of Helsinki, CIOMS, the Belmont Report, and the U.S. Common Rule, and were approved by the Yale University School of Medicine Institutional Review Board (IRB #0609001869). Murine cell lines were cultured in DMEM/F12 (ThermoFisher, #A4192001) while patient-derived cell lines were cultured in OptiMEM (ThermoFisher, #31985070). Both were supplemented with 10% FBS, 1% MEM non-essential amino acids (ThermoFisher, #11140050), and 1% penicillin/streptomycin (ThermoFisher, #15140163) and maintained in an incubator at 37°C and 5% CO₂.

Cell viability

Cells were grown until approximately 80% confluency and then harvested by trypsinization and seeded onto 96 well tissue culture plates at a density of 10,000 cells per well and allowed to adhere for at least 6 hours prior to treatment. After 24 hours of treatment, cells were incubated for 4 hours at 37°C with Aquabluer (Boca Scientific, #6015) and fluorescence intensity was measured at 540ex/590em via a SpectraMax M series microplate reader and normalized to the average of control wells. The following treatments were used throughout the study: atorvastatin – as indicated or experimentally determined IC₅₀ (YR1.7: 200 μM, MC38: 100 μM; Cayman Chemical, #10493), fluvastatin – as indicated or IC₅₀ (YR1.7: 250 μM, MC38: 50 μM; Cayman Chemical, #10010334), lovastatin – as indicated or IC₅₀ (YR1.7: 750 μM, MC38: 500 μM; Cayman Chemical, #10010338), pitavastatin – as indicated or IC₅₀ (YR1.7: 500 μM, MC38: 250 μM; Cayman Chemical, #15414), pravastatin – as indicated (Cayman Chemical, #10010342), rosuvastatin – as indicated (Cayman Chemical, #12029), simvastatin – as indicated or IC₅₀ (YR1.7: 75 μM, MC38: 50 μM; Cayman Chemical, #10010344), ferrostatin-1 – 50 μM (Cayman Chemical, #17729), liproxstatin-1 – 2 μM (Cayman Chemical, #17730), α-tocopherol vitamin E – 50 μM (Cayman Chemical, #25985), necrostatin-1 – 50 μM (Cayman Chemical, #11658), Z-VAD(OMe)-FMK – 50 μM (Cayman Chemical, #14463), chloroquine – 50 μM (Cayman Chemical, #14194), 4-Phenylbutyrate – 1.5 mM (Cayman Chemical #11323), MVA – 500 μM (Sigma Aldrich, #50838), CoQ₁₀ – 50 μM (Cayman Chemical, #11506), squalene - 50 μM (Cayman Chemical, #27058), cholesterol – 50 μM (Cayman Chemical, #9003100), 7DHC – 50 μM (Cayman Chemical, #14612), pioglitazone – 50 μM (Cayman Chemical, #71745), and RSL3 – 10 μM or otherwise indicated (Cayman Chemical, #19288). MVA, CoQ10, squalene, cholesterol, 7DHC, and pioglitazone were administered as pretreatments for 12 hours prior to treatment with statins.

Flow cytometry

Cells were grown until approximately 80% confluency and then harvested by trypsinization and seeded onto 6 well tissue culture plates and allowed to adhere for at least 6 hours prior to treatment. After 18 hours of treatment, cells were collected and resuspended in sterile PBS before undergoing the following staining with 10 μM Liperfluo (Dojindo, #L-248-10) for 30 minutes at 37°C. After staining, cells were washed 3 times with PBS, analyzed on the FITC channel of an LSRII flow cytometer, and normalized to mean fluorescence intensity of control samples. Tumors were harvested 12 days after initial tumor challenge and dissociated in dissociation buffer rotating at 37°C for 30 minutes, passed through a 70 μM filter, treated with ACK lysis buffer (ThermoFisher, #A1049201), quenched with RPMI and 10% FBS, and washed in sterile PBS. Dissociation buffer consisted of RPMI (ThermoFisher, #11875101), collagenase (Sigma-Aldrich, #C5138), and 1x DNase (Qiagen, #79254). Cells were then stained with CD45 (Biolegend, #103155) and Liperfluo (described above) for 30 minutes at 37°C, washed 3 times with PBS, analyzed on an LSRII flow cytometer using the CD45⁻ population as a proxy for tumor cells, and normalized to mean fluorescence intensity of control samples.

Metabolite quantification

Glutathione (Cayman Chemical, #600360) and iron content (BioAssay Systems, #DIFE-250) in approximately 1.2 × 10⁶ cells per sample were measured using cell-based detection kits according to manufacturer protocols. MDA content in cells and tumors was measured using a Thiobarbituric Acid Reactive Substances (TBARS) assay kit (Cayman Chemical, #10009055). Approximately 1.2 × 10⁶ cells or 25 mg tumor tissue (harvested 12 days after initial tumor challenge) was homogenized in 250 μL RIPA buffer containing protease inhibitors. Tissue homogenates were centrifuged at 1,600g for 10 min and 100 μL supernatant was used for MDA analysis. For in vitro glutathione, iron, and MDA quantification, statins were administered at a dose of 50 μM for 18 hours.

Mouse experiments

All animal experiment protocols were followed according to the Yale Office of Animal Research Support Committee guidelines. 7-week-old C57BL/6J male mice (IMSR_JAX:000664) were purchased from the Jackson Laboratory (Bar Harbor, ME) and allowed to acclimate for at least a week prior to use. For tumor challenge, YR1.7 and MC38 cells were grown until approximately 80% confluency, harvested by trypsinization, washed in sterile PBS, and counted. 500,000 cells in 100 μL of sterile PBS were injected into the shaved rear flank of each mouse using a 27G needle. Tumor measurements via digital calipers were initiated 7 days after tumor challenge and repeated every 3 to 4 days. Length, width, and height measurements were recorded, and tumor volume was calculated as 0.5233 × length × width × height. Mice were euthanized when their tumor reached 1000 mm³ or ulcerated. Cages were randomized to treatment group after tumor challenge. 10 mg/kg α-PD1 (RMP1–14, Bio X Cell, #BE0146) or rat IgG2a isotype control (Bio X Cell, #BE0089) was administered via intraperitoneal (IP) injection every 3 to 4 days beginning 7 days after initial tumor challenge. 10 mg/kg simvastatin was administered daily via IP injection for 14 days beginning 7 days after initial tumor challenge.

Bulk RNA sequencing

Approximately 8 × 10⁶ cells per sample were treated with 10 μM RSL3 or 100 μM simvastatin for 18 hours. Cells were then harvested by trypsinization, washed, and RNA was isolated using RNeasy kit, supplemented with QIAshredder and DNase treatment (Qiagen). RNA concentration and fragment sizes were evaluated for QC using BioAnalyzer or Agilent TapeStation, then library prep performed and sequenced using Ilumina 97 HiSeq4000 by the Yale Stem Cell Center Genomics Core. Reads were aligned using STAR244 (v2.7.9) to mm10. Briefly, counts were normalized to library size using cpm() and differential expression analysis calculated by Exact test statistic using exactTest() in a pairwise-manner. Trimmed mean of M-values (TMM) normalization was performed during differential expression analysis with edgeR (v4.36.0). Gene set enrichment analysis was performed using desktop software (v4.2.3) on pre-ranked lists, calculated by sign (logFC) * −log10(pval); and the HALLMARK gene sets from MSigDB (v2022.1) [20–21].

scRNA-seq analysis

We analyzed published single-cell RNA sequencing data from human melanoma tumor samples obtained from Pozniak et al. [22]. The dataset included dissociated tumors, along with their immune microenvironment, from patients treated with immune checkpoint blockade, sampled before treatment (BT) and on-treatment (OT), and stratified by response status (responders vs. non-responders). Data were processed using Seurat v5.3.0 in R v4.4.1 (RStudio v2023.06.0+421) [23]. To assess expression of mevalonate pathway genes, we calculated a module score using key pathway genes: MVK, MVD, FDPS, and FDFT1. Gene expression values were scaled by subtracting the mean and dividing by the standard deviation for each gene across all cells, and the mevalonate pathway composite expression score for each cell was defined as the mean of these scaled values. Melanoma tumor cells were identified using a panel of established markers (MITF, PMEL, TYR, DCT, MLANA, S100B), and a melanoma score was similarly calculated for each cell. Cells in the top 20th percentile of melanoma scores were classified as tumor cells, yielding 11,834 cells for downstream analysis. Differences in mevalonate pathway composite expression score between BT and OT samples were assessed within responder and non-responder groups using Wilcoxon rank-sum tests. Violin plots with overlaid box plots (showing medians and interquartile ranges) were generated using ggplot2 v3.5.2. All analyses were performed using Seurat (v5.3.0), dplyr (v1.1.4), tidyr (v1.3.1), and patchwork (v1.3.0) in R v4.4.1 [23].

Statistical analysis

No statistical methods were used to predetermine sample size. For cell-based experiments, a minimum of three biological replicates were used. For animal experiments, a minimum of 5 mice per condition were used. Statistical analyses were performed using GraphPad Prism versions 8, 9, or 10 (GraphPad Software, Inc.). Data were assessed for normality using the Shapiro-Wilk test. For comparisons between two groups, unpaired two-tailed t-tests were used for normally distributed data, and Mann-Whitney U-tests were applied when normality assumptions were not met. For comparisons involving more than two groups, one-way ANOVA was performed followed by Dunnett’s multiple comparisons test. For tumor survival curves, log-rank (Mantel-Cox) tests were used, followed by the Benjamini-Hochberg multiple comparisons correction.

TCGA analysis

Gene expression profiling interactive analysis (GEPIA) was used to generate Kaplan-Meier overall survival curves for skin cutaneous melanoma (SKCM) tumors with high and low expression of MVK, MVD, FDPS, and FDFT1 by quartile based on sequencing data from the TCGA. P-values from associated log-rank tests are displayed on the figure panels. GEPIA was also used to assess for correlation between expression of GPX4, CD8, and various mevalonate pathway genes. The TIMER2.0 web portal and EPIC were used to evaluate correlations between mevalonate pathway genes and CD8+ T cell infiltration [22, 24].

Western blots

YR1.7 and MC38 cells were treated with 50 μM statins for 18 hours, harvested by trypsinization, washed in ice-cold sterile PBS, and lysed in 1× RIPA lysis buffer (ThermoFisher, #89900) with 1× protease and phosphatase inhibitor cocktail (ThermoFisher, #78440). Lysates were incubated on ice for 30 minutes and cleared by centrifugation at 15,000 g for 20 minutes. Protein concentration was quantified using a BCA protein assay kit (ThermoFisher, #23225). 15 μg total protein was mixed with sample buffer (Thermo Fisher, B0007) and denatured at 95°C for 10 minutes. Sample was separated by Criterion TGX Stain-Free Protein Gel (Bio-Rad, #5678085), and transferred to a Nitrocellulose membrane (Bio-Rad, #P1620112). Membranes were blocked with 5% w/v nonfat dry milk and incubated with primary antibodies overnight at 4°C and HRP-conjugated secondary antibodies (Cell Signaling Technology (CST) Danvers, MA) for 1 hour at room temperature. Signal was detected using Clarity and Clarity Max Western ECL Blotting Substrates (Bio-Rad) and captured using ChemiDoc Imaging System (Bio-Rad). Antibodies used were rabbit anti-ACSL4 (1:1000) (ThermoFisher, #PA5–27137, RRID: AB_2544613), anti-GAPDH (1:1000) (Cell Signaling, #5174, RRID: AB_10622025), and anti-GPX4 (1:1000) (ThermoFisher, #MA5–32827, RRID: AB_2810103). Densitometry analysis was performed using ImageJ (v1.54, NIH); band intensities were first normalized to GAPDH for each sample and then to the DMSO control to calculate relative GPX4 expression values.

IHC

YR1.7 tumors treated with simvastatin and control YR1.7 tumors were harvested for IHC analyses. Tumors were fixed in 10% neutral-buffered formalin for 24 hours and then embedded in paraffin. For GPX4 IHC, mouse testis and stomach were used as positive controls. GPX4 (rabbit monoclonal (EPNCIR144), 1:100) (Abcam # ab125066, RRID: AB_10973901) staining was manually performed, with a 40 minute heat-based antigen retrieval, overnight 4°C incubation and horseradish peroxidase based secondary antibodies. H&E and IHC slides were scanned at 40X magnification with the Hamamatsu NanoZoomer and visualized with the NDP.View2Plus software for brightfield microscopy cell counting. For GPX4 quantification, the H score was used, which combines staining intensity and number of cells stained, for manual scoring by board-certified pathologists (SFR and MWB).

Data Availability

Mouse bulk RNA-sequencing data generated in this study are publicly available at the NCBI Sequence Read Archive (PRJNA1346137). Human expression data has been previously published [22], and data is available at https://rdr.kuleuven.be/dataset.xhtml?persistentId=doi:10.48804/GSAXBN. All other raw data generated in this study are available from the corresponding authors upon request.

Results

Expression of mevalonate pathway genes is prognostic and predictive of immunotherapy responses in melanoma

Through genotype-selective, high-throughput combinatorial drug screening, we had previously identified simvastatin as a therapeutic option for RAS-mutant melanomas [7]. We then aimed to assess the importance of HMGCR to cancer cells in a broad manner using the DepMap portal [25]. DepMap is a comprehensive resource that contains genome-wide CRISPR-Cas9 and shRNA screens in over 1,000 cell lines and calculates gene effect score for individual genes in each cell line. A gene effect score of −1 is the median score for all panessential genes, while a score of 0 indicates that a gene is not essential. The median gene effect score for HMGCR across the 1,078 cell lines contained in DepMap was −1.14, and the gene effect score for all but 1 cell line was below 0 (Figure 1A, Table S1). These findings suggest that dependency on HMGCR is a relatively universal feature of cancer cells. Next, we tested whether five commonly prescribed statins, all of which inhibit HMGCR, induce cell death in multiple patient-derived melanoma lines. Atorvastatin, fluvastatin, lovastatin, pitavastatin, and simvastatin each induced dose-dependent cell death in all three lines tested (Figure 1B). Building on the observed sensitivity of melanoma lines to statin treatment, we next investigated whether endogenous mevalonate pathway gene expression could predict treatment outcomes in human melanoma. Using bulk RNA-sequencing data, we found that lower expression of mevalonate decarboxylase (MVD), mevalonate kinase (MVK), farnesyl diphosphate synthase (FDPS), and farnesyl diphosphate farnesyltransferase 1 (FDFT1) was independently associated with improved survival (Figure 1C–D). Each of these genes, along with HMGCR, negatively correlated with CD8 expression, and EPIC analysis confirmed a negative association between CD8+ T cell infiltration for MVD and MVK (Figure S1A–B) [26–27]. To assess whether mevalonate gene expression changes in response to immunotherapy, we reanalyzed single-cell RNA-sequencing (scRNA-seq) data from melanoma tumors collected before and after immune checkpoint blockade (Pozniak et al.) [22]. Uniform Manifold Approximation and Projection (UMAP) analysis revealed distinct clustering of pre- and on-treatment tumor cells (Figure 1E). Among patients who responded to immunotherapy, the mean mevalonate pathway composite expression score (calculated using MVK, MVD, FDPS and FDFT1 expression) decreased from 0.409 (±0.636 SD) before treatment to 0.127 (±0.546 SD) on-treatment (p = 8.27 × 10−¹⁸, Mann–Whitney U test; Figure 1F). In contrast, non-responders exhibited the opposite pattern, with scores increasing from 0.213 (±0.648 SD) to 0.421 (±0.776 SD) on-treatment (p = 6.80 × 10−⁴⁹, Mann–Whitney U test; Figure 1F). Notably, this divergence was evident at baseline: mevalonate scores were significantly higher in eventual responders than in non-responders (0.409 vs. 0.213), suggesting that tumors with higher baseline expression of mevalonate pathway genes may be more responsive to immune checkpoint blockade.

Figure 1. Expression of mevalonate pathway genes distinguishes treatment response in melanoma tumor cells.

A The gene effect score plotted against gene expression for HMGCR for every DepMap cell line. B Percent viability curves of YUSIK, YUCOT, and YUGASP patient-derived melanoma cell lines after increasing doses of statins. C An overview of mevalonate metabolism. D Kaplan-Meier survival curves of human melanoma patients with high and low expression of MVK, MVD, FDPS, and FDFT1. E UMAP projection of single-cell RNA-sequencing data from melanoma tumor cells before treatment (BT) and on-treatment (OT) with immune checkpoint blockade (data from Pozniak et al., Cell, 2024; analysis by the authors). F Violin plots of mevalonate pathway composite expression scores in melanoma tumor cells from responders and non-responders BT and OT with immune checkpoint blockade (data from Pozniak et al., Cell, 2024; analysis by the authors).

Lipophilic statins induce cell death in a dose- and time-dependent manner consistent with ferroptosis

Because reduced mevalonate pathway expression scores correlated with favorable immunotherapy responses, and given that several statins induced cell death in patient-derived melanoma lines, we next sought to determine which pharmacologic features distinguish those statins with anticancer effects. To address this, we tested whether each of the seven most prescribed statins induce cell death in YUMMER1.7 (YR1.7) melanoma and MC38 colorectal cancer cell lines. We found that atorvastatin, fluvastatin, lovastatin, pitavastatin, and simvastatin induced cell death in a dose- and time-dependent manner, but pravastatin and rosuvastatin did not (Figure 2A, S2A). This discrepancy could not be explained by their reported IC₅₀ for HMGCR, their ability to lower cholesterol in patients, or by classifying them as type 1 statins, which bind HMGCR by a decalin ring structure, and type 2 statins, which bind by their fluorophenyl group (Figure S2B–C). Instead, we observed that the atorvastatin, fluvastatin, lovastatin, pitavastatin, and simvastatin are lipophilic statins, whereas pravastatin and rosuvastatin are hydrophilic (Figure 2B). We reasoned that the hydrophilic nature of pravastatin and rosuvastatin alters pharmacokinetic properties, including subcellular concentrations of drug in some cell types. After establishing that lipophilic statins induce cell death in YR1.7 and MC38, we next assessed statin-induced cell death for hallmarks of ferroptosis. We found that treating YR1.7 and MC38 with statins increased levels of lipid peroxides and the lipid peroxide intermediate malondialdehyde (MDA) (Figures 2C–D). We also found that treatment of YR1.7 and MC38 with statins decreases cellular reduced glutathione and ferrous iron in accordance with other ferroptosis inducers (Figures 2E–F). Since statin-induced cell death exhibited multiple hallmarks of ferroptosis, we next tested whether ferroptosis inhibitors could rescue YR1.7 or MC38 cells from this process. We found that the ferroptosis inhibitors ferrostatin-1 and liproxstatin-1 rescued cell death in YR1.7 and MC38 cell lines from a subset of statins, including most lipophilic statins tested. (Figures 2G–H). The apoptosis inhibitor Z-VAD-FMK, the necroptosis inhibitors necrostatin-1, the autophagy inhibitor chloroquine, and the endoplasmic reticulum stress inhibitor 4-phenylbutyrate also did not prevent cell death in either cell line (Figures S2D–E).

Figure 2. Lipophilic statins induce cell death in a dose- and time-dependent manner consistent with ferroptosis.

A Percent viability curves of YR1.7 and MC38 cells after increasing doses of statins. B Diagram of statin chemical structures classified according to their polarity. C Measurements of lipid peroxides, D MDA, E GSH, and F ferrous iron in YR1.7 and MC38 cells after statin treatment compared to control. G Percent viability of YR1.7 and H MC38 cells after treatment with statins alone or in combination with ferrostatin-1 or liproxstatin-1. * p < 0.05, ** p < 0.01., *** p < 0.001, **** p < 0.0001

Statins alter gene expression in several ferroptosis-related pathways

Since specific inhibitors of various forms of regulated cell death failed to rescue statin-induced cell death, we performed RNA sequencing on YR1.7 and MC38 cells treated with simvastatin for additional insight into how statins might elicit cell death. Given that statin-induced cell death featured multiple hallmarks of ferroptosis, we also performed RNA sequencing on the same cell lines treated with RSL3 to determine the transcriptional changes associated with ferroptosis induction in these cells. When we analyzed the differentially expressed genes (DEGs) between control cells and each treatment condition, we identified substantial overlap between the DEGs in each condition and 64 DEGs that were present in in both conditions across both cell lines. (Figure 3A). These included PTGS2, an established marker of ferroptosis, SLC7A11, which encodes one subunit of the key ferroptosis suppressor System X_c⁻, PLA2G4A, a cytosolic phospholipase A2 responsible for cleaving arachidonic acid from the membrane, and several other ferroptosis-related genes. Next, we performed gene set enrichment analysis (GSEA) for each of these conditions. Out of the top 20 gene sets enriched in each condition, 9 gene sets, including “Heme metabolism” and “ROS pathway”, were common to all 4 conditions and an additional 3 gene sets, including “Cholesterol synthesis” were enriched in 3 out of 4 conditions (Figure 3B, S3A–B). This prompted us to evaluate ferroptosis-related gene changes following simvastatin treatment in more detail using the FerrDb database. In doing so, we detected extensive changes pertaining to glutathione synthesis (GCLC, GCLM, GLUL, GLS2, SLC3A2, SLC7A11), iron regulation (FTH1, FTL1, TFRC), lipid metabolism (ACSL1, ACSL3, ACSL4, AGPAT3, CARS, SQSTM1), redox homeostasis (DHODH, KEAP1, POR, AIFM2, GCH1, NFE2l2, NQO1, and TXNRD1), and other ferroptosis-related pathways in both YR1.7 and MC38 following treatment with simvastatin (Figure 3C). Coordinated changes of expression of key genes involved in other forms of regulated cell death such as pyroptosis, necroptosis, and apoptosis were notably absent (Figure S3C–E).

Figure 3. Statins alter gene expression in several ferroptosis-related pathways.

A Venn diagram displaying overlap between upregulated genes in RSL3- and simvastatin-treated YR1.7 and MC38 samples compared to control with a sample of common ferroptosis-related genes of interest listed in a textbox. B Top 20 enriched gene sets by normalized enrichment score (NES) in YR1.7 cells treated with simvastatin with pathways of interested bolded and italicized. Table listing NES for common enriched gene sets in RSL3- and simvastatin-treated YR1.7 and MC38 samples compared to control. C Heatmaps showing differentially expressed ferroptosis genes between simvastatin-treated YR1.7 and MC38 samples and control.

Statins deplete GPX4 expression and mevalonate rescues statin-induced cell death

The remarkable overlap between the transcriptional changes induced by RSL3 and simvastatin in both YR1.7 and MC38 led us to question whether statins might also act on GPX4 to promote ferroptosis. Others have reported that the downstream metabolite of mevalonate, isopentenyl pyrophosphate (IPP), is required for the isopentenylation of the selenocysteine (Sec)-tRNA, which is necessary for the synthesis of selenoproteins like GPX4 [28–29]. Therefore, we hypothesized that statins might promote ferroptosis by inhibiting GPX4 synthesis. To test this, we compared GPX4 protein levels in statin-treated YR1.7 cells to control and found that statins reduce GPX4 abundance (Figure 4A). Densitometric quantification, normalized to GAPDH, confirmed a reproducible decrease following statin treatment (Figure 4B). Next, to determine whether this effect extends in vivo, we assessed GPX4 expression by immunohistochemistry (IHC) in YR1.7 tumors treated with simvastatin and observed a marked reduction, demonstrating that statin treatment suppresses GPX4 expression also in vivo. (Figure 4D–E). Both treated and untreated tumors displayed comparable H&E histology, including poorly differentiated epithelioid and spindled melanocytes with mixed inflammatory infiltrates, mitoses, and apoptotic debris (Figure 4D). In human melanoma, analysis of bulk RNA-sequencing data from The Cancer Genome Atlas (TCGA) showed that GPX4 expression positively correlated with MVK, MVD, and FDPS (Figure S4A). Interestingly, significant negative correlations also existed between GPX4 and HMGCR and FDFT1 (Figure S4B). These results suggest that mevalonate pathway regulation of GPX4 expression may also occur in patient tumors, raising the possibility that restoring mevalonate pathway activity could reverse the effects of statin treatment. Consistent with this, the addition of mevalonate to statin-treated YR1.7 cells rescued cell death, supporting our hypothesis (Figure 4F). Mevalonate also rescued statin-induced cell death in MC38 cells (Figure 4G). Notably, mevalonate failed to rescue RSL3-induced cell death, suggesting that it specifically targets the mechanism of statin-induced cell death rather than acting as a broadly applicable ferroptosis inhibitor (Figure S4C). Several downstream products of the mevalonate pathway beyond IPP have been shown to impact ferroptosis, however, and this did not rule out the possibility that mevalonate rescues statin-induced cell death by increasing them rather than IPP and GPX4. Coenzyme Q₁₀ (CoQ₁₀) is a radical-trapping antioxidant used by ferroptosis suppressor protein 1 (FSP1) to prevent ferroptosis [30–31], squalene has been reported to alter cellular lipid metabolism and protect cells from ferroptosis in a model of anaplastic large cell lymphoma [32], and 7-dehydrocholesterol (7-DHC) was recently discovered to suppress ferroptosis in a Burkitt lymphoma cell line [33]. Inhibiting cholesterol uptake has also been shown to induce ferroptosis in lymphoma cells, and cholesterol itself has been shown to act as an endogenous ferroptosis suppressor [34–36]. Despite these findings, supplementation with CoQ₁₀, squalene, 7DHC, or cholesterol failed to rescue statin-induced cell death (Figure S4D–F). We also observed that statins increase levels of the ferroptosis promoting protein ACSL4 (Figure 4A, 4C). ACSL4 inhibition with the thiazolidinedione pioglitazone, however, was insufficient to prevent statin-induced cell death (Figure S4G).

Figure 4. Mevalonate rescues statin-induced cell death.

A Western blot detection of ACSL4 and GPX4 protein levels in YR1.7 cells after treatment with statins compared to control. B Densitometric analysis of GPX4 and C ACSL4 protein levels from Western Blot. GPX4 and ACLS4 signal intensity was normalized to GAPDH for each condition. D Measurement of GPX4 in YR1.7 tumors treated with simvastatin compared to control. E H-score analysis of GPX4 IHC in YR1.7 tumors treated with simvastatin compared to control alone (n=5, p=0.0018). F Percent viability of YR1.7 and G MC38 cells after treatment with statins alone or in combination with mevalonate compared to control. * p < 0.05, *** p < 0.001, **** p < 0.0001

Statins improve responses to α-PD1 checkpoint blockade and promote ferroptosis in vivo

Multiple studies have demonstrated that increased intratumoral ferroptosis augments the effect of immune checkpoint blockade [37–44]. With this in mind, we speculated that statins might improve responses to immune checkpoint blockade by promoting ferroptosis. To evaluate this, we tested the effects of α-PD1 and simvastatin alone and in combination on YR1.7 and MC38 tumor growth in vivo. On their own, α-PD1 and simvastatin promoted tumor clearance and extended survival in 20% to 60% of mice (Figure 5A–D). When we administered α-PD1 and simvastatin together, however, this result increased to almost 100% of mice (Figure 5A–D). This improved response corresponded with an increase in intratumoral lipid peroxides and MDA suggesting that ferroptosis is the cell death mechanism responsible for the in vivo tumor clearance (Figure 5E–F).

Figure 5. Statins improve responses to α-PD1 checkpoint blockade and promote ferroptosis in vivo.

A Kaplan-Meier survival curve of YR1.7 and B MC38 tumors treated with α-PD1 and simvastatin alone or in combination compared to control. C Spider plots of tumor volume for YR1.7 and D MC38 tumors treated with vehicle, α-PD1, simvastatin, or α-PD1 and simvastatin. E Measurements of intratumoral lipid peroxides and F MVA in YR1.7 tumors treated with α-PD1 and simvastatin alone or in combination compared to control. * p < 0.05, ** p < 0.01., *** p < 0.001, **** p < 0.0001

Discussion

The therapeutic potential of statins in cancer has been documented across a variety of models, and several mechanisms have been attributed to it [15]. We began by showing that statins induce cell death in patient-derived melanoma cell lines and that lower expression of mevalonate pathway genes correlates with improved outcomes in human melanoma. Building on these findings, we combined functional and phenotypic cell death assessments, transcriptomic data, and syngeneic tumor models in murine melanoma and colorectal cancer cell lines to demonstrate that lipophilic statins uniformly promote ferroptosis by depleting GPX4 and that statin-induced ferroptosis augments antitumor immune responses following α-PD1 checkpoint blockade.

In this study, we found that statin-induced cell death in YR1.7 and MC38 exhibited key hallmarks of ferroptosis, including downregulation of GPX4 and could be rescued by ferroptosis inhibitors (ferrostatin-1 and liproxstatin-1) for the majority of lipophilic statins. These results are consistent with prior studies showing that statins inhibit selenoprotein synthesis and statin-induced cell death has features of ferroptosis and may be rescued by ferroptosis inhibitors although this is universally consistent for all statins [11, 45–47]. In line with our findings, additional accounts of statin-induced cell death in vitro are primarily restricted to lipophilic statins, and the effect has been attributed to reactive oxygen species generation and depletion of metabolites like mevalonate and IPP, which are critical for isopentenylation and maintenance of GPX4 abundance [48–52]. Additional reports show that statins both induce and inhibit apoptosis and pyroptosis in a variety of settings [53–56]. Therefore, it remains likely that the cell death modality engaged by statins is primarily ferroptosis but may also be context dependent. This is apparent in other areas of biology like ischemic kidney injury, for example, where ferroptosis, necroptosis, and pyroptosis appear to play overlapping roles in tissue injury and inflammation [57–58].

Links between cholesterol synthesis, statins, and ferroptosis are not limited to studies of statin-induced cell death. As discussed earlier, the downstream mevalonate metabolites CoQ₁₀, squalene, 7-DHC, and cholesterol have all been shown to prevent ferroptosis in various contexts [29–35]. Statins can also inhibit synthesis of CoQ₁₀, an inhibitor of ferroptosis, but our studies did not show rescue of statin-induced cell death by CoQ₁₀ administration. Furthermore, in patient-derived colorectal cancer cells, cholesterol synthesis was identified as a regulator of stemness and drug resistance [59]. The same study demonstrated that loss of either HMGCR or FDPS leads to upregulation of arachidonic acid metabolism, which promotes ferroptosis, and our transcriptomic analysis replicated this finding in YR1.7 and MC38 [60]. Repression of the mevalonate pathway and induction of ferroptosis have also been independently identified as mechanisms of p53-mediated tumor suppression [61–62]. While these two mechanisms have not yet been linked, it is possible that they cooperate to restrict tumorigenesis. Additionally, HMG-CoA increases following statin treatment and can directly modify the active site of fatty acid synthase, which synthesizes long chain fatty acids [62]. Whether this alteration influences cellular lipid dynamics and sensitivity to ferroptosis has not yet been evaluated.

Some studies have demonstrated a role for tumor ferroptosis in immunotherapy responses [36–43]. Here, we show that simvastatin promotes ferroptosis to stimulate antitumor immune responses alone and in combination with α-PD1. This builds on prior work showing that simvastatin synergizes with immunotherapy in head and neck squamous cell carcinoma and ovarian clear cell carcinoma and lovastatin does the same in non-small cell lung cancer [56, 63–65] although these studies did not elucidate the role of ferroptosis in this anti-tumor immune response. Cholesterol has been shown to limit the proliferation of, induce exhaustion, and blunt memory formation in CD8+ T cells [66–68]. Statins do not significantly lower serum cholesterol in mice, however, so a reduction in cholesterol likely cannot explain the antitumor effects that we observed. Beyond primary tumors, two groups have also shown that statins suppress cancer metastasis on their own and in combination with other agents [49, 69]. Although neither specifically assessed ferroptosis as a mechanism of this effect, it represents a likely contributor and another interesting area for future investigation given that dysregulated cholesterol homeostasis has been shown to promote ferroptosis resistance and metastasis in multiple breast cancer and melanoma cell lines [70].

Finally, our patient-derived melanoma cell line and gene expression data underscore the therapeutic potential of targeting the mevalonate pathway in melanoma patients. Although a few retrospective studies have shown an overall survival benefit associated with statin use in melanoma [71–73], and over 95 clinical trials are currently investigating the anticancer effects of statins, very few published prospective clinical data are available. A phase II clinical trial is currently active (NCT06157099, estimated completion 2029-09-01) to evaluate whether atorvastatin administered to resected high-risk stage IIA, IIB or IIIA melanoma can prevent rates of metastasis since statins are also inhibitors of new lymphatic vessel formation. However, to the best of our knowledge no studies have examined the impact of statin use on survival specifically for melanoma patients receiving immunotherapy [71–73], which are now the standard of care and most promising treatment option for many cancers. Indeed, trial NCT05636592 is an active (not currently recruiting) prospective observational study designed the evaluate the safety and efficacy of immune checkpoint inhibitors in combination with statins in advanced non-small cell lung cancer patients. Since many patients do not respond at all to immune checkpoint inhibitors or relapse after an initial response, continuing to identify and test treatment options, such as statins, with the potential to enhance responses to immunotherapy remains a critical area for future research.