Combined Inhibition of HRAS and MEK Induces Tumor Regression and Restores Myogenic Differentiation in HRAS-Mutant Rhabdomyosarcoma

Farnesyltransferase and MEK inhibition suppresses ERK reactivation, decreases tumor growth, and promotes myogenesis in HRAS-mutant rhabdomyosarcoma.

Abstract

Hyperactive RAS signaling, induced by mutations in NRAS, HRAS, or KRAS, drives tumorigenesis in most PAX3/7::FOXO1 fusion-negative rhabdomyosarcomas (FN-RMS). Despite the frequency of these mutations, indirect RAS pathway–directed therapies have been ineffective for RAS-driven RMS. Farnesyltransferase (FTase) inhibitors (FTI), such as tipifarnib, inhibit HRAS membrane localization and blunt RAS effector signaling, leading to an antitumor effect in HRAS-mutant FN-RMS preclinical models. However, the effect is not durable. In this study, we investigated the mechanisms of adaptive resistance that limit the activity of FTIs, revealing that response to FTIs was limited by adaptive feedback reactivation of ERK signaling and upregulation of wild-type RAS. The combination of HRAS suppression with FTI and MEK inhibition impaired ERK reactivation and reduced ERK transcriptional output in HRAS-mutant RMS models. Cotargeting FTase and MEK restrained tumor progression and induced terminal myogenic differentiation. These findings highlight an effective combinatorial strategy and support its preclinical translation for patients with HRAS-mutant RMS.

Significance:

Farnesyltransferase and MEK inhibition suppresses ERK reactivation, decreases tumor growth, and promotes myogenesis in HRAS-mutant rhabdomyosarcoma.

Graphical Abstract

Introduction

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood and adolescence (1). RMS is believed to arise from a mesenchymal progenitor cell that despite expression of myogenic markers fails to differentiate into skeletal muscle (2). Patients with low-risk RMS have favorable outcomes with 5-year event-free survival between 70% and 90%. High-risk RMS, however, leads to unacceptable outcomes, with only 20% of patients surviving 5 years from diagnosis (3). Furthermore, multi-agent chemotherapy, surgery, and/or radiation have led to short-term and long-term morbidities while failing to improve survival for these patients (3). Thus, newer and more effective therapies are desperately needed for patients with high-risk RMS.

RAS pathway alterations are found in the majority of PAX3 or PAX7::FOXO1 fusion-negative RMS (FN-RMS), the most common subtype of RMS (4). Of the canonical RAS family gene members HRAS, NRAS, and KRAS (5–7), mutations in HRAS are found in 8% of FN-RMS cases and are enriched in infants, who have an inferior 5-year failure-free survival compared with older patients (4). Hyperactive RAF–MEK–ERK signaling downstream of oncogenic RAS leads to uncontrolled tumor growth and interruption of terminal myogenic differentiation in RAS-mutant RMS (8–10).

Strategies to inhibit RAS-driven FN-RMS have included small-molecule inhibition of constituents of the receptor tyrosine kinase (RTK)–RAS–MEK–ERK signaling axis (9, 11, 12). Included in these are inhibitors of farnesyltransferase (FTase), an enzyme required for posttranslational modification of RAS (prenylation) that allows for membrane localization and RAS effector signaling (13). Tipifarnib is a potent and selective FTase inhibitor (FTI; ref. 13). Clinical trials of tipifarnib in adult (14) and pediatric (15) malignancies showed that tipifarnib was well tolerated but demonstrated only modest clinical activity. One explanation for the lack of clinical efficacy lies in the mechanism of NRAS and KRAS prenylation. KRAS and NRAS can be alternately prenylated by geranylgeranyl transferase (GGTase) when FTase is inhibited (16). HRAS, however, is uniquely dependent on FTase for prenylation and membrane localization (16). Accordingly, FTIs have demonstrated enhanced efficacy in HRAS-mutant preclinical models (17) and in clinical trials of HRAS-mutant head and neck squamous cell carcinoma (HNSCC; ref. 18). In 2021, the FDA granted tipifarnib breakthrough therapy designation for use in patients with HRAS-mutant HNSCC.

We reported the preclinical efficacy of tipifarnib in models of HRAS-mutant FN-RMS. FTI selectively impaired MEK-ERK signaling and diminished tumor growth in HRAS-mutant RMS but not in NRAS-mutant, KRAS-mutant, and RAS-wild-type (WT) xenograft models in vivo. The growth inhibitory effects of FTI, however, were not durable in long-term in vivo models of HRAS-mutant RMS (19). Thus, we set out to uncover mechanisms of adaptive resistance limiting the activity of single-agent FTI. Genetic depletion of HRAS decreased cell proliferation transcriptional signatures and upregulated myogenic differentiation, which was recapitulated with FTI. With prolonged FTI exposure, we observed a rebound in ERK suppression, which was accompanied by a loss of negative regulators of RTK–RAS–MEK–ERK signaling and upregulation of WT RAS. We hypothesized that this signaling adaptation blunted the growth inhibitory effects of FTI. To overcome ERK reactivation, we combined FTI with MEK inhibition (MEKi). Dual blockade of FTase and MEK attenuated ERK rebound and downregulated ERK transcriptional activity. We found synergistic growth inhibition with FTI and MEKi in vitro with tumor regressions seen with the combination in vivo. Additionally, FTI and MEKi induced terminal myogenic differentiation. Taken together, the combination of FTI and MEKi is active in preclinical models and may provide a rational treatment strategy for patients with HRAS-mutant FN-RMS.

Materials and Methods

Cell lines, antibodies, and reagents

Human RMS cell lines SMS-CTR and RD were obtained from the ATCC. RH36 was provided by Dr. David Loeb (The Children’s Hospital at Montefiore, Bronx, NY), and JR-1 was provided by Dr. Marielle Yohe (National Cancer Institute, Frederick, MD). SJRHB000026_X1 (SJRHB26) was provided by Dr. Elizabeth Stewart (St. Jude Children’s Research Hospital, Memphis, TN). The patient-derived RMS cell line JH-ERMS-2 was generated in our laboratory from a biospecimen collected during surgical resection from a pediatric patient with RMS and was previously reported (19). Cell lines were authenticated using short tandem repeat (STR) analysis at the Johns Hopkins University Genetic Resources Core Facility to confirm their identity against published STR profiles, where available. STR profiles are provided in Supplementary Table S1. Cell lines were cultured in Roswell Park Memorial Institute 1640. All growth medium was supplemented with 10% fetal bovine serum and 1% penicillin G (50 U/mL) and streptomycin sulfate (50 μg/mL). Cell lines were maintained in a humidified 37°C incubator with 5% CO₂. All cell lines tested negative for Mycoplasma contamination using MycoAlert Mycoplasma Detection Kit.

Antibodies

HRAS: Proteintech, 18295-1-AP, RRID: AB_2121046, NRAS: Proteintech, 10724-1-AP, RRID: AB_2154209, KRAS: Lsbio LS-C175665-100, RRID: AB_2920760, MRAS: Abcam, ab176570, RRID: AB_3683691, SHOC2: Cell Signaling Technology, cat. #53600S, RRID: AB_2799440, vinculin: Cell Signaling Technology, cat. #4650S, RRID: AB_10559207, GAPDH: Cell Signaling Technology, cat. #2118L, RRID: AB_561053, phospho-p44/42 mitogen-activated protein kinase (MAPK; Erk1/2; Thr202/Tyr204): Cell Signaling Technology, cat. #4370, RRID: AB_2315112, p44/42 MAPK (Erk1/2): Cell Signaling Technology, cat. #4695, RRID: AB_390779, phospho-MEK 1/2 (Ser217/221): Cell Signaling Technology, cat. #9154, RRID: AB_2138017, MEK 1/2: Cell Signaling Technology, cat. #9122, RRID: AB_823567, MYOG (myogenin): Cell Signaling Technology, cat. #43098, RRID: AB_3717698, myosin heavy chain (MYH1): Invitrogen, MA5-35613, RRID: AB_2849513, Mouse Anti-rabbit IgG [L27A9; horseradish peroxidase (HRP) conjugate]: Cell Signaling Technology, cat. #5127S, RRID: AB_10892860, and Mouse IgG HRP-linked Whole Ab: Cytiva, GENA931, RRID: AB_772193. Antibodies were used for immunoblots at a dilution of 1:1,000.

Tipifarnib was provided by Kura Oncology, Inc., under a Johns Hopkins University institutional-approved Material Transfer Agreement. Trametinib, TNO155, BI-3406, NST-628, avutometinib, and cobimetinib were purchased from Selleck Chemicals. RMC-7977 was purchased from MedChemExpress. Drugs for in vitro studies were dissolved in dimethyl sulfoxide (DMSO) to yield 10 mmol/L stock solutions and stored at −80°C.

Small interfering RNA–mediated knockdown

Commercial small interfering RNAs (siRNA) HRAS, NRAS, KRAS, MRAS, and SHOC2 and nontargeting control (sequences listed in Supplementary Table S2) were purchased from Horizon Discovery Biosciences Limited and transfected into cells using DharmaFECT (Horizon Discovery Biosciences Limited, T-2001-02) following the manufacturer’s instructions.

RNA sequencing

Cells were plated at 2 × 10⁶ cells per well in 10-cm plates and incubated with either compounds or DMSO in triplicate conditions. All cells were harvested by centrifugation, washed with ice-cold phosphate buffered saline (PBS), and pelleted. RNA was isolated from pellets using TRIzol phenol–chloroform, followed by the RNeasy Mini Kit (QIAGEN, cat. #74104) according to the manufacturer’s instructions. RNA concentration was quantified using a UV spectrometer, and the total RNA concentration extracted per sample is provided in Supplementary Table S3. All RNA sequencing (RNA-seq) FastQ reads were aligned with the reference genome (hg38). Libraries were sequenced on an Illumina NovaSeq X instrument in paired-end 2 × 150 bp mode. A read depth of 50 million was targeted for each sample.

RNA-seq data are available on the Gene Expression Omnibus (GEO, GSE312030) and analysis code from Zenodo DOI: 10.5281/zenodo.17779686. Briefly, raw sequencing reads were trimmed using trim-galore-0.6.3 and aligned to the reference genome (hg38). Alignment and gene expression quantification were performed with RSEM v1.3.0, with transcript and gene features from the UCSC hg38 gtf file using the following parameters: rsem-calculate-expression module with the following options: –star –star-gzipped-read-file –calc-ci –star-output-genome-bam –paired-end –forward-prob 0.

Differential expression analysis was performed using a negative binomial test with the R v4.1.2 package DESeq2 v1.46.0 (20) for both genes and transcripts. The analysis focused on the gene-level expression. A cutoff of an absolute value (log₂ fold change) greater than 1 and a Benjamini–Hochberg adjusted P value less than 0.05 was used to define a differentially expressed gene (DEG; Supplementary Table S4). The R v4.1.2 package UpSetR was used to generate UpSet plots of significant genes (21). Gene set enrichment was computed using the Wald statistic estimated in the DESeq2 analysis for gene ranking with the R v4.1.2 package FGSEA v1.32.0 (bioRxiv 2021.02.01.060012) for Hallmark gene sets from the Molecular Signatures Database (MSigDB) v7.5.1 (22), Supplementary Table S5.

Additional visualization and analysis of individual genes were performed on variance-stabilized transform counts (called vst-normalized) by applying the vst function from DESeq2 to gene-level counts from our RSEM normalization. The MAPK Pathway Activity Score (MPAS) was derived from this normalized expression data for ten MAPK-specific genes (PHLDA1, SPRY2, SPRY4, DUSP4, DUSP6, CCND1, EPHA2, EPHA4, ETV4, and ETV5; ref. 23). MPAS was computed as MAPK activity = zi√n, in which z_i is the z-score of each gene’s variance-stabilized transform normalization expression level and n is the number of genes comprising the set (i.e., n = 10). Raw data are shown as Supplementary Tables S6 and S7.

Immunoblotting

Cells were plated at 2 × 10⁶ cells per well in 10-cm plates and incubated with either compounds or DMSO. Cells were harvested by centrifugation, washed with ice-cold PBS, and lysed in NP40 supplemented with phenylmethane sulfonyl fluoride and sodium orthovanadate, or were disrupted on ice in radioimmunoprecipitation assay lysis buffer (#R0278-50, Sigma) supplemented with phenylmethane sulfonyl fluoride and sodium orthovanadate. Protein concentration was determined using Pierce BCA Protein Assay Kit (#23227, Thermo Fisher Scientific) on a microplate reader (SpectraMax M5). Equal amounts of proteins were resolved on 10% or 12% SDS-polyacrylamide gels and transferred to nitrocellulose membranes (#1620112, Bio-Rad). Membranes were probed with primary antibodies and incubated overnight at 4°C. Following overnight incubation, membranes were incubated with secondary HRP-conjugated antibodies for 1 hour at room temperature. Chemiluminescence with the ECL detection reagents, Immobilon Western chemiluminescent HRP substrate (# WBKLS0500, Millipore), or Pierce ECL Western blotting substrate (# 32106, Thermo Fisher Scientific) was determined. The membranes were imaged on the ChemiDoc Touch Imaging System (Bio-Rad). Signal intensity from immunoblots was quantified using densitometry analysis via ImageJ. All experiments shown were replicated at least twice.

Multiplexed inhibitor bead chromatography and mass spectrometry

Multiplexed inhibitor bead chromatography and mass spectrometry (MIB/MS) experiments were performed as described previously (24). SMS-CTR cells were treated with vehicle or tipifarnib (100 nmol/L) for 24 or 48 hours in biological triplicate. Extracts were sonicated 3 × 10 seconds, clarified by centrifugation, and syringe-filtered (0.22 μm) prior to Bradford assay quantitation of concentration. Equal amounts of total protein (0.3 mg) were gravity-flowed over MIB columns in high salt MIB lysis (1 mol/L NaCl). Bound protein was eluted twice with 0.5% SDS, 1% β-mercaptoethanol, and 100 mmol/L Tris–HCl, pH 6.8 for 15 minutes at 100°C. Eluate was treated with DTT (5 mmol/L) for 25 minutes at 60°C and 20 mmol/L iodoacetamide for 30 minutes in the dark. Following spin concentration using Amicon Ultra-4 (10k cutoff) to approximately 100 μL, samples were precipitated by methanol/chloroform, dried in a SpeedVac, and resuspended in 50 mmol/L HEPES (pH 8.0). Tryptic digests were performed overnight at 37°C, and peptides further cleaned using C-18 spin columns according to the manufacturer’s protocol (Pierce). Peptides were resuspended in 2% ACN and 0.1% formic acid. A total of 40% of the final peptide suspension was injected onto a Thermo EASY-Spray 75 μm × 25 cm C-18 column and separated on a 120-minute gradient (5%–40% ACN) using an EASY nLC-1000 coupled to a Thermo Exploris 480 mass spectrometer. Raw files were processed for label-free quantification (LFQ) by MaxQuant LFQ using the UniProt/Swiss-Prot human database. Kinases with fewer than two razor + unique peptides were excluded. Normalized LFQ intensities were imported into Perseus software. In Perseus, LFQ intensities were log₂-transformed, and missing values were imputed by column using default parameters for each sample to enable comparison. Unpaired t tests were performed in R to compare groups, and P < 0.05 was considered significant. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD062989. Raw data are included as Supplementary Table S8.

CRISPR library screen

Library virus packaging and multiplicity of infection (MOI) determination was done as described previously (25). The SMS-CTR cell line was seeded in six-well plates at 5 × 10⁵ cells per well, and the next day, virus was added at an MOI of 0.3. Following puromycin selection, a day 2 sample was collected to determine initial library representation, and the remaining cells were seeded in T-225 culture flasks. Each condition (control and tipifarnib treatment for 3 weeks) was done in duplicates. Throughout the treatment, cells were maintained at a minimum of 500× library coverage. At the end of treatment, the cells were collected, genomic DNA was extracted using DNeasy Blood & Tissue Kit (QIAGEN), and single-guide RNA sequences were amplified by library specific PCR primers and an Illumina sequencing adapter with index for each sample. Raw data are shown as Supplementary Table S9.

High-throughput cell viability assays

Matrix 10 × 10 combination studies were used to assess small-molecule synergy in SMS-CTR using the Mechanism Interrogation PlatE (MIPE) v6. Matrix blocks were dispensed using an acoustic dispenser (EDC Biosystems), and 48-hour CellTiterGlo readouts were used to determine cell viability and apoptosis as described previously (26). Data are included as Supplementary Table S10.

Cell growth assay

The IncuCyteTM SX5 Live-Cell Analysis System (Sartorius) was used to track cellular proliferation in real time. Cells were plated at a density of 1,500 to 2,000 cells/well in 96-well plates and treated 16 to 24 hours after plating. Cell confluency was visualized using phase-contrast images with the IncuCyteTM camera; data were collected every 4 hours. Each experiment was conducted at least three times with at least three biological replicates.

Soft agar colony formation assay

A total of 100,000 to 150,000 cells growing in log phase were mixed with 1% agar (Gibco) treated with either DMSO or tipifarnib (10, 30, 100 nmol/L) and plated over a bottom layer of 4% agar in six-well plates. Cells were incubated at 37°C for 3 weeks. Colonies were stained with crystal violet (Sigma-Aldrich) overnight and imaged via ChemiDoc Touch Imaging System (Bio-Rad). The measurements were based on three replicates for each condition. Images captured within a single experiment were taken at the same magnification and exposure time. All experiments shown were replicated at least twice.

Immunofluorescence

SMS-CTR or RH36 cells were plated into six-well plates at 4 × 10⁵ or 3 × 10⁵ cells per well, respectively. To induce differentiation of the RMS cell lines, cultures were treated with 2 nmol/L trametinib, 100 nmol/L tipifarnib, 0.1% vehicle control (DMSO), or the combination of trametinib and tipifarnib for 72 hours, without replenishment. To observe differentiation, after 72 hours, cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, blocked with 1% bovine serum albumin (BSA) in PBS, and probed with primary anti–myosin heavy chain 4 (MYH4) antibody (clone MF20, Thermo Fisher Scientific, #14-6503-37) diluted in blocking buffer overnight. Nuclei were visualized with Hoechst 33342 Nuclear Stain (Thermo Fisher Scientific, #62249). At least five random fields per condition were imaged on a ZEISS Axio Vert.A1 wide-field fluorescent microscope (ZEN software, RRID: SCR_013672). Although investigators were not blinded to the experimental group, random fields were selected in the Hoechst channel to avoid bias associated with MYH4 staining levels. The differentiation index, defined by the total number of nuclei in MYH4-positive cells divided by the total number of nuclei, was calculated for each field. Briefly, total nuclei were calculated automatically by image thresholding, watershed segmentation, and particle counting on Fiji (RRID: SCR_002285). Code is available upon request. Nuclei within MYH4-positive cells were identified manually and quantified using Cell Counter in Fiji (RRID: SCR_002285).

Xenograft studies

NOD SCID gamma (NSG, # 005557) female mice were purchased from The Jackson Laboratory. To establish subcutaneous tumors cells at 80% confluency were trypsinized, resuspended in a 1:1 solution of PBS and Matrigel, and injected into the flanks of 8-week-old mice (5–7.5 million cells per flank). Tumor-bearing mice (defined as having palpable tumors) were randomized into four groups of eight animals (SMS-CTR) or five animals (SJRHB26) by an algorithm that distributes animals based on measured volume to achieve the best-case distribution to ensure that each treatment group has similar mean tumor volume and SD. Sample size determination was accounted on the need for statistical power. Vehicle or inhibitors were administered via oral gavage based on mean group body weight, with a treatment schedule of 5 days on/2 days off. Investigators were blinded to the treatment groups. The endpoint of the experiment for efficacy studies was considered 4 weeks on treatment or longest tumor diameter of 2-cm as per the approved animal protocol, whichever occurred first. Tumors were measured twice weekly by calipers in two dimensions, and tumor volume was calculated by (L × W²) (0.5), in which L is the longest diameter and W is the width. Tumor measurements were obtained by an investigator blinded to treatment allocation. Data are shown as the mean ± SEM. At study termination, mice were euthanized by CO₂ inhalation followed by cervical dislocation. Tumors were excised, weighed, and divided for formalin fixation or snap-freezing in liquid nitrogen for protein extraction. All mouse experiments were approved by the Institutional Animal Care and Use Committee at Johns Hopkins under protocol #MO22M63 and conducted in compliance with the ARRIVE 2.0 guidelines and the NIH Guide for the Care and Use of Laboratory Animals.

Statistical analysis

Statistical analysis was performed for the RNA-seq data in R as described above. For all other assays, statistical analysis was performed with GraphPad Prism 10 (GraphPad software). One-way ANOVA was used to analyze xenograft studies. Analyses were considered statistically significant if P < 0.05.

Results

RAS–MEK–ERK signaling drives tumorigenesis in HRAS-mutant RMS

Oncogenic RAS is essential for tumorigenesis in RAS-mutant RMS, and genetic depletion of RAS decreases tumor growth in RAS-mutant RMS (9). Studies have demonstrated that the MEK–ERK effector pathway is the dominant signaling axis downstream of oncogenic RAS (9, 11) in HRAS-mutant RMS. To deepen our understanding of the transcriptional effects of oncogenic HRAS on RAS effector signaling, we used a panel of RAS-mutant FN-RMS cell lines with varying co-mutations, including PIK3CA, NF1, and TP53 (Fig. 1A). We used siRNA directed against HRAS or NRAS for 48 hours, followed by RNA-seq in three HRAS-mutant RMS cell lines (Supplementary Fig. S1A). A greater number of genes were differentially expressed in HRAS-mutant cell lines using siRNA for HRAS compared with siRNA for NRAS, as expected. HRAS expression was significantly repressed following HRAS knockdown compared with NRAS knockdown (Fig. 1B, left) and HRAS protein and phosphorylated MEK (pMEK) levels decreased with HRAS knockdown (Fig. S1B). Likewise, NRAS knockdown led to decreased NRAS expression and NRAS protein levels (Fig. 1B, middle); Supplementary Fig. S1C), whereas KRAS expression was unchanged with HRAS or NRAS knockdown in all cell lines (Fig. 1B, right).

Figure 1.

HRAS knockdown decreases ERK transcriptional output. A, Key genomic alterations in RAS-mutant human RMS cell lines. B, Fold changes (log₂) of HRAS, NRAS, and KRAS RNA expression in RMS cell lines following 48 hours of HRAS siRNA-mediated knockdown or NRAS siRNA-mediated knockdown versus nontargeting control determined by RNA-seq. FDR q < 0.05. C, GSEA of hallmark gene sets in RMS cell lines following 48 hours of HRAS siRNA-mediated knockdown versus nontargeting control, followed by RNA-seq. Blue, most negatively enriched gene sets; red, most positively enriched gene sets. Benjamini–Hochberg adjusted P value. EMT, epithelial–mesenchymal transition. D, RMS cell lines following 48 hours of HRAS siRNA-mediated knockdown or NRAS siRNA-mediated knockdown versus nontargeting control followed by RNA-seq. MPAS was determined by aggregated gene expression data. E, SMS-CTR following 48 hours of HRAS siRNA-mediated knockdown or NRAS siRNA-mediated knockdown versus nontargeting control, followed by RNA-seq. Volcano plots summarizing DEGs identified using DESeq2 analysis for each comparison performed. Dashed lines indicate an absolute value fold change greater than or equal to >1 (vertical) and P adj <0.05 (horizontal). Significantly DEGs from the MPAS gene set are highlighted in red. *, P adj < 0.05; **, P adj < 0.01; ***, P adj < 0.001; ****, P adj < 0.0001.

To examine the impact of HRAS knockdown on the transcriptome, we performed differential expression analysis on the RNA-seq using DESeq2 and then gene set enrichment analysis (GSEA) of the resulting differential expression statistics with pathways in the MSigDB hallmark gene sets (22) in three HRAS-mutant RMS cell lines. HRAS knockdown led to a positive enrichment of the myogenesis gene set in all cell lines (Fig. 1C), supporting the established role of RAS in maintaining the myogenic differentiation block in FN-RMS (8, 9). Cell proliferation opposes myogenic differentiation in RMS, with both processes being regulated by RAS–MEK–ERK signaling (8–10). Consequently, there was a negative enrichment of G₂M checkpoint, E2F target, and MYC target gene sets following HRAS knockdown (Fig. 1C). These changes in transcriptional programs validate the notion that oncogenic HRAS drives cell proliferation and impairs myogenic differentiation in an ERK-dependent manner. The PI3K–AKT–mTOR gene set was not among the most negatively enriched gene sets following HRAS knockdown, even in the cell lines without concomitant hotspot mutations in PIK3CA (SMS-CTR and RH36). In cell lines with oncogenic PIK3CA, such as JH-ERMS-2, we would expect that PI3K–AKT–mTOR signaling would be decoupled from the requirement for activation by RAS and not expect such changes to result from HRAS knockdown. In tumor cells without PI3K oncogenic activation, however, HRAS knockdown also failed to elicit PI3K–AKT signaling downregulation, suggesting that MEK-ERK effector signaling drives tumorigenesis in HRAS-mutant RMS.

Finally, we used the transcriptional MPAS, aggregating variance-stabilized gene expression values from the RNA-seq for ten genes (EPHA4, EPHA2, ETV4, ETV5, PHLDA1, SPRY2, SPRY4, CCND1, DUSP4, and DUSP6) to quantify MAPK activity. MPAS has been shown to predict sensitivity to MAPK signaling inhibition in vitro (23). HRAS knockdown diminished MAPK activity (defined by the MPAS) in HRAS-mutant RMS cell lines compared with NRAS knockdown or nontargeting control (Fig. 1D and E; Supplementary Fig. S1D). Taken together, these data support prior studies that have established the role of oncogenic HRAS in driving cell proliferation and myogenic differentiation block via MEK-ERK effector signaling in HRAS-mutant FN-RMS.

FTase inhibition phenocopies HRAS suppression in HRAS-mutant RMS

To understand the transcriptional response to FTI and uncover therapeutic vulnerabilities, we performed RNA-seq using six RAS-mutant cell lines treated with tipifarnib for 24 hours relative to untreated controls. We inferred treatment changes with differential expression analysis using DESeq2, followed by GSEA of hallmark gene sets (22). This transcriptomic analysis revealed a positive enrichment of myogenesis and a negative enrichment of RAS signaling and cell proliferative transcriptional programs such as G₂M checkpoint, E2F targets, and MYC targets in HRAS-mutant cell lines akin to HRAS knockdown (Fig. 2A). We examined differential gene expression induced by FTI in four HRAS-mutant cell lines to identify consensus DEGs (absolute value of DESeq2 estimated log₂ fold change >1 and P adj <0.05) in HRAS mutants following their inhibition at 24 hours (Fig. 2B and C; Supplementary Fig. S2A). Performing the same in NRAS-mutant cell lines, very few genes were transcriptionally regulated with FTI (Fig. 2B and C; Supplementary Fig. S2A), and as such, the number of DEGs corresponded to the in vitro and in vivo sensitivity to FTI, with effects seen exclusively in cells with oncogenic addiction to mutant HRAS (19). This analysis was also consistent with the selective inhibition of HRAS compared with NRAS or KRAS-mutant lines in response to FTI (19). JH-ERMS-2 stood out from the other three HRAS-mutant lines, with fewer overall DEGs, leading to a consensus gene signature involving all four cell lines of only three genes (gray arrow). To improve the capture of a biologically meaningful gene set, we therefore selected genes that were significantly differentially expressed in at least three of the four HRAS-mutant lines, thereby arriving at a set of 74 genes (gray plus black arrows, Fig. 2B).

Figure 2.

FTI phenocopies genetic suppression of HRAS. A, GSEA of hallmark gene sets in RMS cell lines treated with 24 hours of 100 nmol/L tipifarnib (FTI) vs. DMSO followed by RNA-seq. Blue, most negatively enriched gene sets; red, most positively enriched gene sets. Benjamini–Hochberg adjusted P value: *, P adj < 0.05; **, P adj < 0.01; ***, P adj <0.001; ****, P adj < 0.0001. EMT, epithelial–mesenchymal transition. B, UpSet plot demonstrating statistically significant DEGs (absolute value of log₂ fold change >1 and P adj <0.05) in RMS cell lines treated with 24 hours of 100 nmol/L FTI, followed by RNA-seq. C, RMS cell lines SMS-CTR and JR-1 following 24 hours of FTI vs. DMSO followed by RNA-seq. Volcano plots summarizing DEGs identified using DESeq2 analysis for each comparison performed. Dashed lines indicate an absolute value fold change greater than or equal to >1 (vertical) and P adj < 0.05 (horizontal). Significantly DEGs from the MPAS gene set are highlighted in red. D, Heatmap summarizing fold changes (log₂) in the expression of 74 genes differentially expressed in at least three of four cell lines, including RH36, SJRHB26, and SMS-CTR and JH-ERMS-2, derived from B. E, RMS cell lines treated with 24 hours of 100 nmol/L FTI, followed by RNA-seq. MPAS score determined by aggregated DESeq2 vst-normalized gene expression data. F, Heatmaps summarizing fold changes (log₂) in the expression of MPAS genes in RMS cell lines following 6, 24, and 48 hours of 100 nmol/L FTI vs. DMSO. All conditions were performed in triplicate.

Using these 74 unique DEGs to define the transcriptional response to FTI in HRAS-mutant RMS, we examined fold changes in gene expression using DESeq2 in four HRAS- and two NRAS-mutant RMS cell lines (Fig. 2D). In HRAS-mutant cell lines FTI selectively decreased the expression of canonical ERK transcriptional targets DUSP6 and SPRY4, consistent with prior reports of ERK transcriptional output changes induced by MEK or BRAF inhibition in BRAF-addicted cells (23). Furthermore, FTI decreased the expression of known activators of RAS–MEK–ERK signaling, including CCL2 (27), CXCL2 (28), and CD271/NGFR (29). Finally, FTI increased the expression of several genes responsible for myogenic differentiation in RMS, including MYOG, MYH8, TNNC2, CAV3, CASQ2, SRL, ARPP21, and KLHL41 (9) in HRAS-mutant cell lines (Fig. 2D). Using this set of 74 significant DEGs, applied to HRAS-mutant cell lines following siRNA-mediated HRAS knockdown, we observed a consistent direction of change in gene expression compared with that elicited by the FTI (Supplementary Fig. S2B). MAPK activity (determined by MPAS) was decreased in all HRAS-mutant cell lines treated with 24 hours of FTI, with no significant changes seen in NRAS-mutant cell lines (Fig. 2C and E; Supplementary Fig. S2A). Among the 10 genes in the MPAS gene set, EPHA4 (encoding Eph receptor A4), a negative regulator of MAPK signaling (30), was increased at 24 and 48 hours, with other genes downregulated over the course of time (Fig. 2F). We therefore concluded that FTI selectively inhibited ERK transcriptional output in HRAS-mutant RMS and phenocopied HRAS knockdown in HRAS-mutant RMS cell lines.

Prolonged FTI exposure leads to ERK reactivation and upregulation of WT-RAS

Growth inhibition with tipifarnib was not durable in long-term in vivo models of HRAS-mutant FN-RMS (19). We surmised that this limitation was due in part to decreased DUSP4, DUSP6, SPRY2, and SPRY4 expression (Fig. 2E and F), which negatively regulates RTK–RAS–MEK–ERK signaling (31). To further elucidate the signaling adaptations to FTI, RMS cell lines were treated with FTI up to 72 hours. HRAS defarnesylation demonstrated by unprocessed HRAS protein (Fig. 3A, red arrow) was maintained at 72 hours, consistent with continued on-target effects of FTI. In RH36 and SMS-CTR, the depth of pMEK suppression increased over time, whereas pERK suppression was most significant at 24 hours and then remained steady or decreased. In the two other HRAS-mutant cell lines, JH-ERMS-2 and SJRHB26, however, the depth of pMEK suppression was the greatest at 24 hours and then diminished. In these cells, pERK inhibition was overall modest to unchanged (Fig. 3A; Supplementary Fig. S3). Overall, incomplete pMEK and pERK responses were observed, despite consistent evidence of target inhibition as detected by a mobility shift in HRAS in all lines (Fig. 3A).

Figure 3.

ERK reactivation limits the activity of single-agent FTI. A, RMS cell lines were treated with 100 nmol/L tipifarnib (FTI) for 24, 48, and 72 hours and subject to immunoblot for the indicated proteins. B, SMS-CTR was treated with 100 nmol/L FTI for 24 and 48 hours, lysed, and subjected to kinome profiling by MIB/MS. Volcano plot shows the fold change (log₂) in MIB binding plotted against the −log₁₀P value. C, RMS cell lines were treated with 100 nmol/L FTI for 24, 48, or 72 hours, and whole-cell lysates were subject to immunoblot for the indicated proteins. D, SMS-CTR was subject to immunoblot for the indicated proteins after 48 hours of nontargeting control (NT), si-NRAS, or si-KRAS in the presence or absence of 100 nmol/L FTI. E, SMS-CTR was subject to immunoblot for the indicated proteins after 48 hours of nontargeting control, si-MRAS, or si-SHOC2 in the presence or absence of 100 nmol/L FTI. F, SMS-CTR was transduced with druggable genome pLentiv2 CRISPR library and treated with tipifarnib for 3 weeks. Single-guide RNA (sgRNA) region was PCR-amplified, and sgRNA representation was determined by next-generation sequencing. Fold change (log₂) for sgRNA read counts between treated and control samples was calculated using MAGeCKFlute. Volcano plot summarizing tipifarnib-sensitizing genes. G, Positively and negatively enriched reactome pathways following CRISPR knockdown. NES, normalized enrichment score.

RTK-driven feedback reactivation of ERK signaling has been identified as a mechanism of resistance to direct KRAS inhibition with allele-specific inhibitors (32). We were therefore interested in how RTK activity may contribute to ERK reactivation following indirect HRAS inhibition. Using MIB/MS kinome profiling, we examined changes to 118 kinases following FTI exposure in SMS-CTR (HRAS Q61K). At 24 hours, binding of MKK4 was decreased, indicating reduced signaling via JNK and p38 pathways (Fig. 3B; ref. 33). By 48 hours, there was decreased MEK1 and ERK1 binding, consistent with blunted RAS–MEK–ERK signaling following FTI (Fig. 3B). Loss of CDK2 binding seen at 48 hours supported the notion that FTI inhibits G₁–S cell-cycle progression. Finally, MKK3 kinase had decreased capture at 48 hours, consistent with continued downregulation of p38 and JNK signaling pathways in response to FTI (34). We found a statistically significant (but modest) increase in the binding of NDR2, a kinase that has been shown to negatively regulate YAP1 activity (35). Otherwise significantly increased RTK binding was not seen at 48 hours using MIB/MS in SMS-CTR.

Prior studies have shown that upregulation of WT RAS and reestablished RAS-MEK-ERK signaling is observed with indirect HRAS inhibition with FTI (17) and following direct KRAS inhibition with allele-specific inhibitors (32). We therefore asked whether the ERK reactivation we observed involved changes in WT RAS isoforms NRAS, KRAS, and MRAS. We treated RMS cell lines with FTI for up to 72 hours. At 48 and 72 hours, NRAS, KRAS, and MRAS protein levels were increased (Fig. 3C). We hypothesized that adaptive MEK–ERK signaling downstream of WT RAS was contributing to ERK rebound. To address this, we asked whether genetic knockdown of WT RAS in the presence of FTI would impair ERK phosphorylation. Using siRNA, we knocked down the expression of NRAS or KRAS in the presence or absence of FTI for 48 hours in SMS-CTR. Phospho-ERK expression was modestly dampened in the presence of FTI at 48 hours; however, NRAS and KRAS knockdown reduced pERK expression compared with FTI alone, supporting the notion that WT NRAS and KRAS contribute to ERK activity (Fig. 3D). Likewise, we knocked down the expression of the RAS family member MRAS and SHOC2, a scaffolding protein that complexes with RAS proteins and protein phosphatase 1 (PP1C) to dephosphorylate RAF kinases and increase MEK–ERK signaling (36). ERK signaling was attenuated with SHOC2 and MRAS knockdown in the presence of FTI in SMS-CTR (Fig. 3E) in accordance with prior reports (37).

The results of a CRISPR/Cas9 screen targeting the druggable genome in SMS-CTR further pointed to WT RAS activity as an adaptation to FTI. We performed a screen in the presence of sublethal (GI₅₀) doses of FTI and identified genes whose loss decreased or increased sensitivity to FTI. The top FTI-sensitizing genes included RCE1, FNTA, ZDHHC9, SHOC2, and PGGT1B (Fig. 3F), genes that encode enzymes needed to modify and activate proteins including RAS. Reactome pathway analysis of the CRISPR screen results revealed that genes involved in mitochondrial translation were significantly enriched among those that conferred resistance, whereas genes involved in RAS processing were significantly enriched among those that conferred sensitivity to FTI (Fig. 3G). FNTA encodes a common α subunit of FTase and GGTase, whereas PGGT1B encodes the β subunit of GGTase (38). When FTase function is blocked by FTI, NRAS and KRAS proteins require GGTase for prenylation, membrane localization, and activation (16). In concert with Ras converting enzyme 1 (Rce1), a membrane protease that cleaves carboxyl-terminal amino acids, and ZDHCC9, which palmitoylates HRAS and NRAS (39), these enzymes function to modify KRAS and NRAS proteins allowing for their membrane localization and MEK–ERK effector signaling (40). We surmised that inhibition of enzymes required for RAS activation sensitized the cells to FTI via interruption of NRAS and KRAS processing and the growth stimulating effects of downstream ERK signaling. Suppression of SHOC2 also sensitized SMS-CTR to FTI, consistent with our knockdown studies (Fig. 3E). We therefore concluded that WT RAS, in conjunction with SHOC2, contributed to ERK reactivation following FTI.

Combination therapies targeting RAS–MEK–ERK signaling are synergistic in HRAS-mutant RMS

To address adaptive resistance to single-agent FTI, we compared the antiproliferative effects of small-molecule inhibitor combinations cotargeting nodes within the RAS–MEK–ERK signaling pathway. We utilized the MIPE 6.0 small-molecule library of approved and investigational drugs (26). The high-throughput compound screen included 28 small-molecule inhibitors combined with FTI (tipifarnib) in SMS-CTR (Fig. 4A). Synergistic growth inhibition was identified with RAS–MEK–ERK inhibitors, such as the multiselective inhibitor of GTP-bound WT and mutant RAS RMC-6236/daraxonrasib, supporting the notion that inhibition of WT RAS can improve the efficacy of FTI. Additionally, synergistic effects were seen with the pan-RAF/MEK molecular glue NST-628 and the MEK inhibitor cobimetinib. Cobimetinib additionally synergized with FTIs tipifarnib and lonafarnib (Supplementary Fig. S4). These results affirmed that the dual blockade of HRAS and MEK–ERK is a rational potential strategy in HRAS-mutant RMS. Inhibitors of SHP2 (RMC-4550, TNO155/batoprotafib), a scaffold and phosphatase required for activation of RAS, did not synergize with FTI. This result was in accordance with our data that suggested a minimal contribution of RTKs to the signaling adaptions to FTI in SMS-CTR (Fig. 3B).

Figure 4.

Combination therapies targeting RAS–MEK–ERK signaling are synergistic in HRAS-mutant RMS. A, Excess HSA score versus rank representing 28 synergy scores derived from a 10 × 10 matrix screen in SMS-CTR with tipifarnib (FTI). RAS–RAF–MEK pathway inhibitors are indicated in bold. B, SMS-CTR was treated with the indicated inhibitors in the presence or absence of 100 nmol/L FTI for 48 hours. Cells were lysed and subject to immunoblot for the indicated proteins. C, RMS cell lines were treated with the indicated inhibitors in the presence or absence of 100 nmol/L FTI for the indicated time points. Phase confluence was measured by IncuCyte real-time imaging.

We examined ERK signaling and cell growth perturbations in the presence of additive combinations identified through the drug screen. We expanded our investigation to include the MEK inhibitor trametinib and RAF–MEK molecular “clamp” avutometinib, as well as inhibitors upstream of RAS such as SOS1i and SHP2i. As expected, SOS1i and SHP2i did not have single-agent activity in SMS-CTR. When combined with FTI, several RAF–MEK inhibitors more profoundly downregulated ERK phosphorylation than either agent alone (Fig. 4B). We measured the impact of combination therapies on long-term cell growth using real-time cell confluence monitoring using the IncuCyte Live-Cell Analysis System. FTI and MEKi decreased RMS cell growth up to 7 days compared with FTI combined with inhibitors acting upstream of RAS to target SHP2 and SOS1 in RMS cell lines (Fig. 4C). Thus, FTI and inhibitors of RAS–RAF–MEK synergize to decrease ERK phosphorylation and impede cell growth in HRAS-mutant RMS.

Combined FTI plus MEKi inhibits tumor progression

Numerous studies have evaluated the utility of RAS-ERK pathway inhibition (using allosteric inhibitors of MEK1/2) in RAS pathway–altered pediatric cancers. Of the inhibitors included in our combination studies, trametinib has undergone several clinical evaluations in RAS-driven pediatric cancers (41, 42), has a manageable safety profile, and has an FDA approval for treatment of BRAF V600E pediatric low-grade glioma in combination with dabrafenib (41). With this in mind, we chose the combination of FTI and trametinib (MEKi) for additional studies.

In HRAS-mutant FN-RMS, MEKi impeded ERK signaling but alone did not control tumor growth with prolonged treatment (9). A mechanism of resistance to MEKi also involves ERK reactivation (9, 11); therefore, we evaluated the effects of the combination on MEK-ERK signaling. When we combined FTI and MEKi, the combination effects on pERK in SMS-CTR and SJRHB26 were greater than either MEK or FTI alone, but in RH36 and JH-ERMS-2, the combination induced no greater effect on pERK at late time points than MEKi alone (Fig. 5A). We assessed the effects of the combination on the MPAS after 48 hours of drug exposure. For three of the HRAS-mutant RMS cell lines, dual inhibition downregulated MAPK transcriptional activity to a greater extent compared with FTI or MEKi alone with the greatest effect seen in SMS-CTR (Fig. 5B).

Figure 5.

Combined FTI plus MEKi inhibits tumor progression. A, RMS cell lines were treated with DMSO, 100 nmol/L of FTI, 2 nmol/L of MEKi, or the combination for 24, 48, and 72 hours and subject to immunoblot from whole-cell lysate. B, RMS cell lines treated with 48 hours of DMSO, 100 nmol/L of FTI, 2 nmol/L of MEKi, or the combination exposure, followed by RNA-seq. MPAS determined by vst-normalized aggregated gene expression data. C, RMS cell lines were treated with DMSO, 10 nmol/L of FTI, 2 nmol/L of MEKi, or the combination and grown in soft agar for 3 weeks. Representative images (left) and quantification (right). Scale bars, 10 mm. D and E, NSG mice bearing SMS-CTR (D) and SJRHB26 (E) xenografts were treated with vehicle, tipifarnib at 20 mg/kg twice daily (FTI), trametinib 0.15 mg/kg daily (MEKi), or the combination (5 days on/2 days off) for 30 days. Tumor volumes were calculated twice weekly. The average tumor volume is graphed as a function of days on treatment. Error bars, mean ± SEM. One-way ANOVA was used to calculate statistical differences in treatments groups vs. vehicle. *, P adj < 0.05; **, P adj < 0.01; ns, nonsignificant.

Next, we examined alterations in anchorage-independent growth from combined FTI + MEKi. Using soft agar colony formation assays, we evaluated the effects of FTI and MEKi on RMS cell growth in the absence of extracellular matrix contact, a hallmark of carcinogenesis (43). FTI and MEKi inhibited anchorage-independent cell growth and prevented cell colony formation at low doses of FTI and MEKi (10 and 2 nmol/L, respectively; Fig. 5C). The growth inhibitory effects of FTI plus MEKi were validated in cell line–derived xenograft and patient-derived xenograft models of HRAS-mutant RMS (SMS-CTR and SJRHB26). Concurrent inhibition of FTI + MEKi completely suppressed tumor formation, with all tumors regressing on therapy through 4 weeks of treatment in SMS-CTR (Fig. 5D). In SJRHB26, the combination slowed tumor growth compared with control, but an additive treatment effect was not seen with FTI and MEKi (Fig. 5E). FTI plus MEKi was well tolerated in NSG mice, with stable body weights observed in all treatment arms (Supplementary Fig. S5). The combination therefore overcame adaptive resistance mechanisms to both single-agent MEKi and FTI, leading to a robust antitumor response in SMS-CTR.

Combined FTI + MEKi induces myogenic differentiation

RMS is thought to arise from a mesenchymal progenitor cell that despite expression of myogenic transcription factors fails to undergo terminal muscle differentiation (2). In RAS-driven FN-RMS, hyperactive ERK signaling impedes myogenic differentiation via various mechanisms, including inhibitory phosphorylation of RNA polymerase II and binding and phosphorylating MYOD1 to repress MYOG expression (9). Interruption of ERK signaling weakens the stability of MYC and decreases expression of cell-cycle regulators and allows for transcription of MYOG, promoting myogenic differentiation and ultimately slowing tumor growth (9). We therefore assessed the effect of FTI plus MEKi on MYOG protein expression. Concurrent inhibition potently induced MYOG across all cell lines, most notably at 72 hours (Fig. 6A). Increased MYOG protein levels were also observed with combination therapy in our in vivo models (Supplementary Fig. S6A). Additionally, markers of terminal myogenic differentiation (MYH4 and MYH1) were also increased with both agents via immunoblot (Fig. 6A) and via immunofluorescent staining (Fig. 6B and C), supporting the notion that ERK suppression permits myogenesis but that this effect is seen only after several days.

Figure 6.

Combined FTI and MEKi induce terminal myogenic differentiation. A, RMS cell lines were treated with DMSO, 100 nmol/L of FTI, 2 nmol/L of MEKi, or the combination for 24, 48, and 72 hours and subject to immunoblot from whole-cell lysate for the indicated proteins. B, RH36 or SMS-CTR were treated DMSO, 100 nmol/L of FTI, 2 nmol/L of MEKi, or the combination for 72 hours and subject to immunofluorescent staining for MYH4 (green) or Hoechst nuclear stain (blue). Representative images show areas of similar cell density. C, Differentiation index, defined by the total number of nuclei in MYH4-positive cells divided by the total number of nuclei, was calculated for each field. Brown-Forsythe and Welch ANOVA with Dunnett multiple comparison post-test was used to calculate statistical differences in treatment groups vs. vehicle.*, P adj < 0.05; ****, P adj < 0.0001; ns, nonsignificant. D, SMS-CTR was treated with DMSO, 100 nmol/L of tipifarnib (FTI), 2 nmol/L of trametinib (MEKi), or the combination for 48 hours, followed by RNA-seq. Normalized enrichment scores for RMS cell state metaprograms (proliferative, differentiated, and progenitor). E, SMS-CTR was treated with DMSO, 100 nmol/L of tipifarnib (FTI), 2 nmol/L of trametinib (MEKi), or the combination for 48 hours, followed by RNA-seq. Volcano plots summarizing the number of DEGs identified using DESeq2 analysis for each comparison performed. Dashed lines indicate an absolute value fold change greater than or equal to >1 (vertical) and P adj <0.05 (horizontal). Red, genes in the progenitor signature; green, genes in the proliferative signature; blue, genes in the differentiated signature, corresponding to cell state metaprograms shown in D. NES, normalized enrichment score.

Recent studies have sought to classify different cell states found in RMS tumors, defining three RMS cell state metaprograms (progenitor, proliferative, and differentiated) that have discrete transcriptional signatures (44). Proliferative cells are characterized by the expression of cell-cycle genes, whereas differentiated cells express muscle lineage markers. In FN-RMS, progenitor cells resemble bipotent skeletal mesenchymal cells and are treatment-resistant (44). We were interested to learn how FTI plus MEKi may modulate these cell states. We applied the RMS cell state transcriptional signatures (44) to RMS cell lines treated with FTI, MEKi, or both (at 48 hours) using RNA-seq. The combination demonstrated a greater transcriptional response and downregulated the proliferative signature to a greater degree than either agent individually or HRAS knockdown in SMS-CTR (Fig. 6D and E). FTI plus MEKi led to a positive enrichment for the differentiated signature and a negative enrichment for the progenitor signature comparable with HRAS knockdown or MEKi (Fig. 6D and E). Likewise, the combination downregulated the proliferative signature genes and upregulated differentiated signature genes compared with either agent alone in SJRHB26 (Supplementary Fig. S6B). Together, these data demonstrate that ERK suppression via FTI plus MEKi impedes cell proliferation, decreases treatment-refractory cells, and promotes myogenic differentiation in HRAS-mutant FN-RMS cells.

Discussion

RAS oncogenes are recurrently altered in one third of FN-RMS cases, with HRAS mutations disproportionately found in infants (4). RAS proteins undergo a series of posttranslational modifications that include the addition of a prenyl group by FTase or GGTase (45), which is required for plasma membrane localization and activation (45). Mutations in HRAS hinder GTP hydrolysis, locking the protein in a constitutively active state, allowing for uncontrolled signaling through RAS effectors, including RAF and PI3K (46). Hyperactive RAS–MEK–ERK signaling drives cell proliferation and impairs myogenic differentiation allowing for tumor formation (9,10). Small-molecule inhibition of RAS or nodes of the RAS–MEK–ERK signaling axis have been clinically unsuccessful for FN-RMS to date.

We explored the utility of indirect HRAS suppression using FTI, a strategy that prevents HRAS membrane localization and effector signaling (13). HRAS knockdown decreases ERK transcriptional output, represses cell proliferation programs, and upregulates myogenic differentiation programs in HRAS-mutant RMS cell lines. HRAS inhibition using FTI phenocopies HRAS knockdown, but this single-agent approach was not durable in HRAS-mutant FN-RMS murine models (19). We set out to delineate adaptations to single-agent FTI and identified processes converging on ERK reactivation.

At the transcriptional level, FTI-mediated MEK-ERK inhibition resulted in decreased expression of ERK transcriptional targets responsible for cell proliferation (ETV4 and ETV5) and negative regulation of RTK–RAS–MEK–ERK signaling (SPRY2, SPRY4, and DUSP6; refs. 31, 47). ERK additionally directly phosphorylates and decreases CRAF activity (47). Consequently, several nodes of the RAS–RAF–MEK–ERK signaling axis can contribute to ERK reactivation following HRAS inhibition with FTI.

Upregulation of WT RAS has been described as an adaptation to indirect HRAS inhibition (17) or direct KRAS inhibition (32) that can be overcome through concurrent RTK inhibition. In our study, increased NRAS and KRAS protein did not coincide with increased RTK activity in SMS-CTR. Accordingly, SHP2i and SOS1i did not meaningfully enhance the effects of FTI alone. Among the WT RAS isoforms upregulated in response to FTI, we observed increased MRAS protein. MRAS in complex with SHOC2 and PP1C has been implicated in driving ERK reactivation via dephosphorylation of RAF (48). MRAS, but also H/N/KRAS, can bind to SHOC2-PP1C to dephosphorylate RAF (48). In our study, MRAS and SHOC2 knockdown resulted in reduced ERK phosphorylation in the presence of FTI in SMS-CTR. Additionally, MRAS is enriched in differentiating RMS cells (44). Therefore, in this context, MRAS upregulation may have also resulted from cells that have undergone myogenic differentiation in response to single-agent FTI in addition to an adaptation driving ERK reactivation.

We sought to address ERK reactivation via concurrent inhibition of FTI and MEK signaling. We evaluated several small-molecule inhibitors and found that the combination of FTI and MEKi had synergistic growth-inhibitory effects. As a single agent, FTI only induced a modest and variable reduction in pERK. Likewise, MEKi-induced pERK inhibition has been shown to be short-lived in FN-RMS (9), with rebound around 24 hours, which was also observed in our study. The combination of FTI and MEKi improved pERK suppression at 72 hours in SJRHB26 and SMS-CTR. Mechanistically, FTI and MEKi regulated RAS-MEK-ERK transcriptional activity to a greater extent than either agent alone. These transcriptional changes were associated with tumor growth suppression via reduced ERK-mediated cell-cycle progression and release of myogenic differentiation block. This finding was true across our panel of genomically diverse cell lines. Cell lines with concomitant PIK3CA hotspot mutations were sensitive to the combination although JH-ERMS-2 (PIK3CA p.Q546E) was less sensitive than SJRHB26 (PIK3CA H1047R). However, SJRHB26 was much less sensitive in vivo compared with SMS-CTR. The antiproliferative and differentiating effects of the combination highlight the interplay of ERK signaling and MYC stability and activity. In RMS, MEKi leads to c-MYC dephosphorylation, downregulation of cell cyclins, and induction of myogenic differentiation (9), a notion that was supported by our findings.

The present study underscores the necessity of durable ERK inhibition to elicit effective combination approaches. Although we found an additive effect of the FTI plus MEKi combination on shifting RMS cell state metaprograms, we acknowledge that in leveraging analyses of RNA-seq data to infer the types of molecular and cellular pathways altered by therapy, it is possible that these genes may have no causal impact on mediating the therapeutic response. Furthermore, enhancement of MYOG and MYH1 were seen by immunoblot at 72 hours, suggesting that the modest effect on differentiation programs observed in the RNA-seq could be more significant if measured at a later time point. Another limitation of the study using analysis of bulk RNA-seq data is that gene expression differences may also be due to changes in the total number of cells and changes in cell-cycle genes in proliferating cells. We mitigated the former statistically with negative binomial tests and note that future work evaluating these changes using single-cell RNA-seq could offer additional insight. Our findings are also limited by the use of a cell line–derived xenograft, which may not fully recapitulate intratumoral heterogeneity or the tumor-immune microenvironment. Longer in vivo studies will also be needed to assess durability and evaluate treatment-emergent resistance. As newer RAS–MEK–ERK inhibitors, including RAS tricomplex inhibitors and RAF–MEK molecular glues and clamps with novel mechanisms of action, progress through clinical evaluation, their utility in combination with FTI will also need to be further assessed.

In conclusion, these results reinforce the role of aberrant ERK signaling in driving tumor growth and impairing myogenic differentiation in HRAS-mutant FN-RMS. We uncovered perturbations to the RAS signaling network following FTI that converge on ERK reactivation. We propose a combinatorial approach targeting HRAS and MEK using two clinically relevant drug classes. The combination of FTI and MEKi may extend clinical responses to either agent and improve outcomes for patients with HRAS-mutant RMS.

Supplementary Material

Supplementary Table 1

Cell line STR profiles

Supplementary Table 2

siRNA sequences

Supplementary Table 3

RNA isolation

Supplementary Table 4

DESeq2 log2 fold change

Supplementary Table 5

Hallmark GSEA

Supplementary Table 6

MPAS score

Supplementary Table 7

VST

Supplementary Table 8

MIB.MS

Supplementary Table 9

CRISPR library screen

Supplementary Table 10

MIPE synergy screen

Figure S1

HRAS knockdown decreases ERK transcriptional output.

Figure S2

FTI phenocopies genetic suppression of HRAS.

Figure S3

ERK reactivation limits the activity of single-agent FTI.

Figure S4

Combination therapies targeting RAS-MEK-ERK signaling are synergistic in HRAS-mutant RMS.

Figure S5

Combined FTI and MEKi suppresses tumor growth.

Figure S6

Combined FTI and MEKi restores terminal myogenic differentiation.

Data Availability

The RNA-seq data generated in this study are publicly available in GEO accession GSE312030. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD062989. All other raw data are available upon request from the corresponding author.

Authors’ Disclosures

P. Odeniyide reports grants from Children’s Cancer Foundation, Rally Foundation for Childhood Cancer Research & Mighty Millie Foundation, Tap Cancer Out, and Summer’s Way & Friends of TJ during the conduct of the study. M.E. Yohe reports grants from Kura Oncology, Inc., outside the submitted work. E.J. Fertig reports grants from NIH during the conduct of the study, as well as grants from Break Through Cancer, Lustgarten Foundation, Roche/Genentech, AbbVie Inc., and NFCR and personal fees from Mestag Therapeutics and Resistance Bio/Viosera Therapeutics outside the submitted work. C.A. Pratilas reports grants from Kura Oncology, Inc., V Foundation Translational Research Grant, Rally Foundation for Childhood Cancer Research & Mighty Millie Foundation, during the conduct of the study; grants from Novartis Institute for Biomedical Research, and personal fees from Day One Biopharmaceuticals outside the submitted work; and a patent for US7812143B2 issued and US20240238284A1 issued. No disclosures were reported by the other authors.

Authors’ Contributions

P. Odeniyide: Conceptualization, data curation, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. A. Skaist: Data curation, formal analysis, writing–review and editing. E. Fenner: Data curation, investigation. H. Amirkhanian: Data curation, investigation. A. Baker: Data curation, investigation. A. Lisok: Data curation, investigation. L. Zhang: Data curation, investigation L.B. Fridman: Data curation, investigation. R.I. Rojas: Data curation, investigation. K.E. Hebron: Data curation, investigation. C. Davis: Data curation, investigation. X. Zhang: Data curation, investigation. G. Feldman: Investigation. S.P. Angus: Data curation, supervision, investigation. C.J. Thomas: Data curation, supervision, investigation. A.V. Vaseva: Data curation, supervision, investigation, writing–review and editing. M.E. Yohe: Data curation, supervision, investigation, writing–review and editing. E.J. Fertig: Data curation, supervision, investigation, writing-review and editing. C.A. Pratilas: Conceptualization, supervision, funding acquisition, investigation, visualization, project administration, writing–review and editing.