Clinical and Genomic Landscape of RAS Mutations in Gynecologic Cancers

We retrospectively reviewed all gynecologic cancer patients at our institution who underwent prospective somatic molecular profiling between February 2010 and August 2022 using Clinical Laboratory Improvement Amendments (CLIA)-certified tests. Eligible patients were those whose tumors were tested for RAS mutations. Patients whose molecular profiling was not derived from gynecologic malignancy or did not include RAS were excluded. Clinical and pathologic data were abstracted, including demographic and histopathologic data. All pathology was reviewed by specialized gynecologic oncology pathologists at our institution. Premalignant or borderline histology were excluded. History of endometriosis was collected from surgical pathology or documented patient report. Molecular profiling results were reviewed, and patients were categorized by mutation status of their tumors into RAS-mutated and RAS-non-mutated groups. The RAS-mutated group was further subdivided by KRAS , NRAS , and HRAS mutations. Co-mutation analysis was performed using R. 23 To identify co-mutations in non-RAS genes, cancers with multiple point mutations in the same RAS isoform (n=5) were counted once. Clinical actionability of a mutation was assessed based on literature review by our Precision Oncology Decision Support team using an in-house database of over 31,000 annotated biomarkers. 24 , 25 In brief, a mutation was considered actionable if 1) it or other variants of similar functional effect in the same gene confers increased sensitivity or resistance to clinically available therapies, or 2) there are clinical trials selecting for the variant or variants of similar functional effect. Inhibitors for RAS signaling and MAPK pathway (henceforth, “RAS+MAPK pathway inhibitors”) included those targeting RAS, SHP2, SOS1, BRAF, or MEK. Descriptive statistics were used to summarize patient characteristics. For comparisons between groups, a chi-square test or Fisher exact test was used to detect differences in categorical variables, and Wilcoxon rank-sum test or Kruskal-Wallis test was used to detect differences in continuous variables. Overall survival (OS) was defined as the time from diagnosis to death and was estimated using the Kaplan-Meier method. Times were censored at last contact if death had not been reported. The log-rank test was performed to detect differences in survival. Covariates identified as clinically or statistically significant in univariate analysis were then analyzed in multivariable regression based on a Cox proportional-hazards model to identify predictors of survival. All p values were two-sided, with 0.05 as the cutoff for statistical significance, and 95% confidence intervals (CIs) included where appropriate. Data were managed on the institutional RedCap application. 26 All research activity was approved by the Institutional Review Board at The University of Texas MD Anderson Cancer Center per protocol PA14–0353. All data were deidentified, and due to the minimal risk posed to subjects, informed consent was waived per protocol. Data were generated through clinical sequencing, and raw data are not available. The derived data supporting the findings of this study were generated by the authors, and the deidentified data is available upon request from the corresponding author (Dr David Hong).

In total, 3328 patients with gynecologic cancers had undergone somatic molecular profiling, which composed of 27 unique “next generation sequencing” panels that included RAS from 17 CLIA-certified laboratories ( Supplemental Table 1 ). All patients underwent tumor-based testing except 47 who had liquid-based testing only. 12 had single gene tests for RAS. In this total cohort, 523 (15.7%) had RAS-mutated tumors, and 2805 (84.3%) had RAS-non-mutated tumors ( Table 1 ). Patients with somatic RAS mutations were younger at diagnosis (57 vs 60 years, p<0.0001) and had a higher prevalence of current or prior diagnosis of endometriosis (27.3% vs 16.9%, p<0.0001). The distributions of cancer type and grade were different between the mutation groups, with the RAS-mutated group showing a higher prevalence of uterine cancer (44.0% vs 32.4%, p<0.0001) and lower grades (frequency of grades 1 or 2, 43.2% vs 8.1%, p<0.0001 for grade overall) than the non-mutated group. History of a second primary cancer was similar between the groups (n=0.92). OS for the full cohort was 5.3 years (95%CI 5.0–5.7). Patients with RAS-mutated gynecologic cancers had a numerically but not significantly higher OS at 5.8 years (95%CI 4.9–6.9) compared to RAS-non-mutated tumors, with an OS of 5.2 years (95%CI 4.9–5.6; p=0.66). The prevalence of specific KRAS , NRAS , and HRAS mutations were 13.5%, 2.0%, and 0.5%, respectively (ie, 85.9%, 12.8%, and 3.3% of RAS). Comparison of patients with KRAS -mutated tumors to those with KRAS -non-mutated tumors showed similar trends to RAS: younger age at diagnosis (57 vs 59), a higher prevalence of endometriosis (29.0% vs 16.9%), a higher prevalence of uterine cancer (46.6% vs 32.3%), and lower grades (frequency of grades 1 or 2, 43.4% vs 9.1%; all p<0.0001). OS was 5.6 years (95%CI 4.7–6.6) in the KRAS -mutated group and 5.3 years (95%CI 5.0–5.6) in the KRAS -non-mutated group (p=0.74). Clinical characteristics of patients whose tumors harbored NRAS or HRAS mutations are also shown in Table 1 . The NRAS-mutated group had high frequency of ovarian cancer (53.7%) and lower grades (grades 1 or 2, 43.3%), although analysis was limited by sample size. OS durations for NRAS -mutated and HRAS -mutated groups were 7.7 years (95%CI 4.4–9.0) and 8.1 years (95%CI 2.3–15.3), respectively. The histologic distribution of patients with RAS-mutated gynecologic cancer is shown in Supplemental Figure 1 . Of the 523 RAS-mutated cancers, 44.0% were uterine, 43.8% ovarian, 9.4% cervical, and 3.9% vulvovaginal. Among these, the most common histologic types in each cancer included endometrioid endometrial cancer (60.4% of uterine, 26.6% of RAS-mutated), low-grade serous ovarian cancer (34.1% of ovarian, 14.9% of RAS-mutated), cervical adenocarcinoma (59.2% of cervical, 5.5% of RAS-mutated), and vulvar melanoma (66.7% of vulvar, 1.9% of RAS-mutated). Next, histologic groups in the full cohort were condensed and compared by RAS mutation status ( Table 2 ). The histologic distributions significantly differed according to RAS and KRAS status in each cancer type. Compared to their respective non-mutated controls, RAS-mutated and KRAS -mutated tumors included higher frequency of endometrioid endometrial cancer (RAS, 60.4% vs 30.7%; KRAS , 62.2% vs 30.9%; p<0.0001 for both), low-grade serous ovarian cancer (RAS, 34.1% vs 3.7%; KRAS , 30.7% vs 4.8%; p<0.0001 for both), and cervical adenocarcinoma (RAS, 59.2% vs 27.6%; KRAS , 61.4% vs 27.8%; p<0.0001 for both). Moreover, the relative prevalence of RAS mutations in each histologic type was examined ( Supplemental Table 2 , KRAS in Figure 1 ). While only 9 patients had mesonephric-like endometrial cancer, 100% of the tumors had KRAS mutations. Similarly, KRAS mutations were seen in 83.3% (n=6) of mesonephric-like ovarian cancer, 50.0% (n=2) of mesonephric cervical cancer. One vaginal cancer showed mesonephric histology without KRAS mutation. Other histologic types in which KRAS mutations were common included mucinous ovarian cancer (60.4%), low-grade serous ovarian cancer (44.4%), endometrioid ovarian cancer (38.9%), endometrioid endometrial cancer (31.1%), cervical adenosquamous carcinoma (30.8%), and cervical adenocarcinoma (excluding mesonephric histology, 24.8%). NRAS and HRAS mutations were rare ( Supplemental Figure 2 ). Univariate analysis for OS in gynecologic cancer patients is shown in Supplemental Table 3 . The preliminary univariate data at an earlier cutoff was presented as an abstract previously, 27 and the current version represents the updated comprehensive analysis. Factors associated with OS were age at diagnosis, cancer type, and grade. In a multivariable analysis controlling for these factors, somatic RAS mutation was significantly associated with worse OS with a hazard ratio (HR) of 1.3 (95%CI 1.1–1.5, p=0.001; Table 3 , Supplemental Table 3b ). Given the high prevalence of low-grade serous ovarian cancer and the known improved outcomes, analysis was repeated after exclusion of this subset of patients. In this analysis, RAS mutation status was again significantly associated with worse OS with a HR of 1.4 (95%CI 1.1–1.6, p=0.0003; Supplemental Table 3b ). An exploratory multivariable survival analysis was repeated for each cancer type and histology to assess the effect; diminishing sample size in RAS-mutated tumors limits robust analysis ( Table 3 ). After adjusting for age and grade, significantly worse OS was independently seen in uterine (median OS 4.6 years vs 5.0 years, HR 1.3 [95%CI 1.03–1.6], p=0.03) and vulvovaginal cancers (median OS 3.0 years vs 4.5 years, HR 4.1 [95%CI 1.7–9.8], p=0.001). While median OS was improved with RAS mutation in ovarian cancer (median OS 7.5 years vs 5.6 years), multivariable adjustment for age and grade resulted in worse HR 1.2 (95%CI 0.97–1.6), although not statistically significant (p=0.09). In general, most histology showed numerically worse OS with statistical significance shown in endometrioid endometrial cancer (median OS 5.7 years vs 8.7 years, HR 1.6 [95%CI 1.03–2.4], p=0.04), other ovarian cancer (median OS 3.3 years vs 18.8 years, HR 3.4 [95%CI 1.4–8.4], p=0.009; includes granulosa cell, mesonephric-like, seromucinous, adenosquamous, sertoli leydig, struma ovarii, anaplastic, and adenocarcinoma NOS), and vulvovaginal melanoma (median OS 1.6 years vs 4.9 years, HR 3.5 [95%CI 1.5–8.1], p=0.003). Numerically improved OS was seen for low-grade serous, mucinous ovarian, and cervical adenocarcinoma but none were statistically significant. Distribution of histology and grade used in multivariable analyses are shown in Supplemental Table 4 . Next, genomic analysis of RAS mutations in gynecologic cancers was performed. Supplemental Figure 3 and Supplemental Table 5 show specific mutated alleles and codons. The most common mutations were KRAS G12D (28.4% of RAS, 4.7% of total cohort) and KRAS G12V (26.1% of RAS, 4.3% of total cohort). Of interest, KRAS G12C mutation was seen in 5% of RAS-mutated tumors (0.8% of total cohort). Histologic breakdown of these point mutations of interest is shown in Table 2 and Supplemental Table 2 . By cancer type, the prevalence of KRAS G12D mutation was 5.9% in uterine, 4.1% in ovarian, 4.4% in cervical, and 1.6% in vulvovaginal cancers. Histologic types with the highest frequency of KRAS G12D mutation were mesonephric-like ovarian (50.0%, n=3/6), mesonephric-like endometrial (33.3%, n=3/9), mucinous ovarian (20.8%), low-grade serous ovarian (19.5%), endometrioid ovarian (15.3%), and adenosquamous cervical (15.4%). The prevalence of KRAS G12V mutation was 5.3% in uterine, 4.2% in ovarian, 2.9% in cervical, and 0.8% in vulvovaginal cancers. Histologic types with the highest frequency of KRAS G12V mutation were mesonephric-like endometrial (33.3%, n=3/9), low-grade serous ovarian (20.3%), mucinous ovarian (18.8%), mesonephric-like ovarian (16.7%, n=1/6), and endometrioid ovarian (15.3%). Histologic types with the highest frequency of KRAS G12C mutation were mesonephric-like endometrial (11.1%, n=1/9), and mucinous ovarian (6.3%). In NRAS and HRAS , Q61R and G12S were the most common mutations, respectively. Co-mutation analysis of RAS-mutated gynecologic cancers was also performed ( Supplemental Figure 4 ). This revealed PIK3CA (30.9%), PTEN (28.8%), ARID1A (28.0%), and TP53 (27.9%) as the most common co-mutations ( Table 4 ). This trend held when co-mutations were analyzed by cancer type for uterine, ovarian, and combined cervical and vulvovaginal cancers ( Supplemental Figure 5 ). Most mutations were activating in PIK3CA and inactivating in PTEN , ARID1A , and TP53. In all, 64.7% of these alterations were deemed clinically actionable, including 91.1% of PIK3CA , 83.7% of PTEN , and 50.0% of ARID1A mutations. TP53 was not deemed therapeutically actionable in this cohort. Finally, we evaluated the clinical efficacy of RAS+MAPK pathway inhibitors in RAS biomarker-selected gynecologic cancers. A total of 62 of 523 patients had received a RAS+MAPK pathway inhibitor targeting KRAS G12C, SHP2, BRAF, or MEK ( Table 5a ). No patient had received a SOS1 inhibitor. These included 42 ovarian cancers (28 low-grade serous, 14 other histology), 17 uterine cancers (9 endometrioid, 8 other histology), 2 cervical adenocarcinomas, and 1 vaginal melanoma. Compared to patients with RAS-mutated gynecologic cancers who did not receive a RAS+MAPK pathway inhibitor, those who did so showed a significantly improved OS of 8.4 years (95%CI 5.5–12.9) compared to 5.5 years (95%CI 4.4–6.6, HR 0.67, p=0.031). However, after adjustment for age at diagnosis, histology, and stage in ovarian and uterine cancer cohorts separately, the use of a RAS+MAPK pathway inhibitor was not associated with OS (p>0.05 for both cancer types, Table 5b ). Histologic breakdown and sub-analyses are presented in Supplemental Table 6 . Five of the 25 eligible patients whose tumors showed a KRAS G12C mutation received KRAS G12C inhibitors: 3 had ovarian cancer (high-grade serous, low-grade serous, clear cell) and 2 had endometrial cancer (mixed endometrioid, clear cell; mixed endometrioid, serous, mucinous) ( Supplemental Table 7 ). The best responses were 3 partial responses (−43%, −74%, and −86% by RECIST) and 2 stable diseases (−7% and −17%); the partial responses included a duration of response of 22.0 months and ongoing responses at 5 and 16 months. Clinical benefit was seen in all patients. Co-mutations in tumors with KRAS G12C mutation are shown in Supplemental Table 8 .

Gynecologic cancers with RAS mutations have a distinct histopathologic distribution, including a high prevalence in mesonephric-like cancers, mucinous ovarian cancer, low-grade serous ovarian cancer, endometrioid ovarian cancer, and cervical adenocarcinoma. Although characterized by younger age at diagnosis, uterine cancer type, lower grade, and endometriosis, patients with RAS-mutated gynecologic cancers may have worse survival compared to RAS non-mutated cancers after adjustment for confounders. The most commonly actionable co-alterations are PIK3CA, PTEN , and ARID1A , which may aid in combination therapy design. Patients whose tumor harbors a RAS mutation should be considered for RAS pathway targeting in trial setting. In prior series utilizing immunohistochemistry or polymerase chain reaction (PCR) methods, KRAS had been associated with low-grade serous ovarian cancer, 28 mucinous ovarian cancer, 15 and endometrioid endometrial cancer. 13 , 14 In endometrial cancer, the prevalence of KRAS mutation has been reported to be around 18%–26%, 13 , 14 which was consistent with both The Cancer Genome Atlas analysis showing a rate of 25% 29 and our data. Considering the differing timing of testing in these studies, the similarity in prevalence supports the early occurrence of this mutation in carcinogenesis. 15 A clinically novel histologic finding of our study is the high frequency of RAS mutations in mesonephric-like cancers, endometrioid ovarian cancer, and cervical adenocarcinoma. This was partially comparable to a recent study of MAPK pathway mutations using the American Association of Cancer Research Genomics Evidence of Neoplasia Information Exchange database. 20 Moreover, we were able to statistically analyze clinical and survival characteristics of RAS mutations in a single-institutional cohort of gynecologic cancer patients undergoing somatic molecular profiling. Centralized specialist pathology review is a strength of this study. The identification of rare but aggressive histology such as mesonephric-like cancer 30 may have future therapeutic implications. Importantly, most historic studies have reported no prognostic association of RAS mutations in gynecologic cancers, including in endometrial cancer, 31 , 32 while others reported a survival correlation in subgroups. 33 In a large meta-analysis of ovarian cancer cohorts in Denmark, KRAS mutation was associated with worse prognosis, which had not been observed in prior smaller studies. 34 Similar findings were suggested in cervical cancer cohorts. 35 In our large data set, after controlling for age, cancer type, and grade, RAS mutation was significantly associated with worse survival overall with independent significance in uterine and vulvovaginal cancers, trend in ovarian cancer, and notable effect in endometrioid endometrial and vulvovaginal melanoma histology. Recent studies in low-grade serous ovarian cancer had shown improved OS with MAPK pathway mutation after controlling for age, stage, prior diagnosis of borderline tumor, prior therapy, 16 platinum sensitivity 17 and with the use of whole-exome sequencing. 18 While the cause of discrepancy in this histology is unclear, the association between RAS mutation and worse outcomes has been demonstrated in colorectal, 36 , 37 lung, 38 pancreatic, 39 thyroid, 40 and hematologic cancers. 41 Accordingly, our study suggests the generally poor prognostic implications of RAS mutation in gynecologic cancers. Several downstream inhibitors of RAS have been tested in clinical trials in gynecologic cancer patients. In general, the response rate to RAS/MAPK pathway monotherapy targeting RAF or MEK has been disappointing at 6%–16%. 4 – 7 An important consideration in these studies is biomarker selection. While selumetinib did not meet pre-trial specifications for efficacy in all-comer recurrent or persistent endometrial cancer patients, the authors hypothesized the potential benefit of biomarker selection. 4 In low-grade serous ovarian cancer, exploratory analysis of selumetinib in a small cohort of patients with KRAS- mutated or BRAF -mutated cancers did not show a predictive association, 5 but a post hoc analysis of binimetinib compared to investigator’s choice chemotherapy again associated KRAS mutation with improved response to therapy. 6 In a recent phase 2/3 trial of trametinib versus standard of care in low-grade serous ovarian cancer, the median progression-free survival was significantly improved in the trametinib arm, reaching 13.0 months compared to 7.2 months. 8 The incidence of KRAS mutation was 11%–12% in each arm, and the presence of KRAS , BRAF , or NRAS mutation was associated with a marked increase in progression-free survival (13.2 vs 7.3 months, HR 0.41) and objective response rate (ORR, 50.0% vs 8.3%, HR 15.1). The authors suggest this mutation profile may be predictive of ORR (p=0.11). In our study, patients with RAS-mutated cancers who received a RAS+MAPK pathway inhibitor compared to those who did not had a significantly improved OS of 8.4 years compared to 5.5 years (HR 0.67, p=0.031), although the difference was not statistically significant in multivariable analysis. The inclusion of a heterogeneous, predominantly downstream treatment regimen in a small sample size of patients limited our analysis. Combining the poor prognostic implications of RAS mutation and the available data in prospective trials, we recommend consideration of RAS pathway inhibitor trials in patients with RAS-mutated gynecologic cancers. Owing to the complexity of RAS pathway signaling, including various downstream compensatory mechanisms, combination therapy with RAS pathway inhibitors is actively being explored. Preclinical studies have demonstrated activation of the PI3K pathway in response to MEK inhibition; 42 , 43 accordingly, dual inhibition of RAS and PIK3CA pathways has shown synergistic efficacy in vitro and in vivo . 44 – 46 Unfortunately, studies attempting this combination therapy have been fraught with unacceptable toxicity. 47 – 49 In our data, PIK3CA and PTEN were frequently altered, regardless of cancer type. Discovery of novel agents and therapeutic regimens to minimize toxicity remains crucial. ARID1A was found to be an additional candidate. Furthermore, KRAS G12C inhibitors showed robust efficacy in gynecologic cancer patients. Hence, the development of direct inhibitors of further upstream targets may offer promising results. Such inhibitors in clinical trial include those for KRAS-off, RAS-on, SHP2, and SOS1. There are several limitations of this study. While this study includes the most comprehensive clinicogenomic, survival, and therapeutic data of RAS-mutated gynecologic cancers to date, it is also limited by the inclusion of heterogeneous cancer types, genomic platforms, and therapeutic regimens. Statistical adjustment was performed to account for some of these factors. While the study spans 12 years, the approved biomarker-directed therapies in gynecologic cancers during this time would not have significantly favored panel-based somatic molecular profiling for a specific disease type. However, because somatic molecular profiling usually occurs in the recurrent setting, this selection criteria likely biased the patient population toward those with more high-risk disease. In fact, data on the lines of treatment is unavailable but assumed to be in the recurrent setting, which is consistent with the median time from diagnosis to molecular testing of 24 months. In this regard, there may be a left truncation bias on the timing of sequencing itself, although this is mitigated by the requirement for testing in both comparator groups. Similarly, immortal time bias is possible. Overall survival analysis for each histology is exploratory in nature and limited by sample size. We acknowledge the possibility that RAS may be differentially prognostic based on histology, as demonstrated in low-grade serous ovarian cancer. Despite the inclusion of RAS+MAPK pathway inhibitors in both standard-of-care and clinical trial settings, the number of patients who received therapy was modest, limiting robust statistical analysis. Furthermore, there is again a left truncation survival bias for patients receiving RAS+MAPK pathway inhibitors, which we attempted to correct in multivariable analysis. Minor limitations include lack of germline test results, possible reporter bias for history of endometriosis, and limited information on combination therapy in patients who received RAS+MAPK pathway inhibitors in ongoing trials. Atypical RAS mutations may be explored in future studies. Despite these limitations, this study represents a large single-institution data set of granular clinical and molecular characterization of gynecologic cancers with RAS mutations, including new insight into survival and therapeutic options. RAS mutation in gynecologic cancers is a relatively prevalent, clinically unique entity with a potential impact on survival. Developmental efforts should focus on RAS pathway-targeted therapy in this population, particularly in combination approaches.

RAS genes ( KRAS , NRAS , and HRAS ) are among the earliest described oncogenes and are the most frequently mutated gene family in human cancers. 1 The RAS family of G proteins are master regulators of the mitogen-activated protein kinase (MAPK) pathway, defined by the RAF-MEK-ERK signaling axis, and ultimately control cell cycle progression, cell survival, growth, metabolism, motility, and migration. 2 Mutations in RAS disrupt the guanine exchange cycle, constitutively activating it in the GTP-bound state. RAS-driven oncogenesis is an obvious target for targeted therapy and has been somewhat of a holy grail in developmental therapeutics. Despite decades of developmental efforts, RAS has eluded targeting and was termed “undruggable.” 3 Various strategies for downstream inhibition of the RAS/MAPK pathway in non-biomarker-selected gynecologic cancers have also only shown modest response (objective response rate, ORR 6–16%; 4 – 7 notable 26% with trametinib 8 ). However, the recent advent of allele-specific direct RAS inhibition has been a breakthrough, reigniting interest and accelerating developmental efforts in this space. In non-small cell lung cancers with specific KRAS mutation at codon G12C, both sotorasib and adagrasib have shown meaningful response (ORR 32.2% 9 and 42.9%, 10 respectively), leading to accelerated approval by the U.S. Food and Drug Administration. In the ongoing KRYSTAL-1 trial, adagrasib monotherapy showed ORR up to 57.1% in the gynecologic cancer cohort. 11 Novel downstream combinatory approaches such as RAF/MEK and FAK inhibitors (avutometinib plus defactinib) are also showing promise, especially in KRAS -mutated tumors. 12 Other interesting early data are reported in smaller cohorts. An understanding of the landscape of RAS mutations is crucial to determine the applicability of therapeutic RAS inhibition in gynecologic cancers. Historical reports of RAS mutations in gynecologic cancers were mainly based on smaller pathologic studies, including those identifying an association with endometrioid endometrial cancer 13 , 14 and mucinous ovarian cancer. 15 More recently, larger genomic analyses were performed in low-grade serous ovarian cancer 16 – 18 and pan-cancer data. 19 , 20 KRAS mutation has been associated with endometriosis in preclinical models, 21 particularly in malignant transformation, 22 although the clinical association in gynecologic cancer patients is unknown. Moreover, characterization of RAS mutations in other gynecologic cancers, comparison to non-mutated counterparts, and the relevant therapeutic experience and survival outcomes are lacking. Therefore, we aimed to describe RAS mutations in gynecologic cancers as they relate to clinicopathologic and genomic features, survival, and implications for therapy.