Distinct Genomic Landscapes in Early-Onset and Late-Onset Endometrial Cancer

Abstract

PURPOSE

The spectrum of somatic mutations among women with endometrial cancer (EC) younger than 50 years (early-onset EC) remains unknown. We investigated distinct somatic mutation patterns among early-onset and late-onset (age ≥ 50 years) EC patients.

METHODS

This cohort study included individuals age 18+ years diagnosed with pathologically confirmed EC in the American Association of Cancer Research (AACR) Genomics Evidence Neoplasia Information Exchange (GENIE, v9.1) consortium. We explored tumor mutational burden (TMB) and genomic patterns of EC by age at clinical sequencing using multivariable regression models adjusted for race, ethnicity, histology, sequencing assay, sample type, and TMB.

RESULTS

Among 2,425 women with EC, 176 (7.3%) had early-onset EC and 1,923 (79.3%) had nonhypermutated (< 17.78 mutations/Mb) tumors. TMB significantly differed across age and histology groups. Among nonhypermutated ECs, early-onset patients had significantly lower odds of presenting with nonsilent FGFR2 and PIK3R1 somatic mutations compared with late-onset EC patients in adjusted models (FGFR2: odds ratio [OR] = 0.18, 95% CI, 0.04 to 0.76; PIK3R1: OR = 0.54, 95% CI, 0.31 to 0.92). By contrast, early-onset EC patients had increased odds of presenting with nonsilent CTNNB1 and BRCA2 mutations compared with late-onset patients (CTNNB1: OR = 3.32, 95% CI, 2.14 to 5.16; BRCA2: OR = 4.01, 95% CI, 1.55 to 10.38). Subsequent analyses stratified by race, ethnicity, and tumor histology identified distinct patterns of APC, KMT2D, KMT2C, and KRAS by race, ethnicity, and PTEN and APC patterns by histologic subtypes.

CONCLUSION

Early-onset EC harbors a unique genomic landscape compared with late-onset disease. A distinct molecular phenotype of early-onset EC provides novel insights into a unique etiology and may yield clinical implications for developing targeted treatment modalities for younger patients.

INTRODUCTION

Endometrial cancer (EC) is the sixth most commonly diagnosed cancer in women worldwide.^1,2 In contrast to most cancers, EC incidence—particularly in younger patients and in developed countries—has been rapidly increasing in recent years.^3-9 Although the majority of EC patients are diagnosed among women over age 50 years, 2%-14% of patients occur in individuals of reproductive age. The majority of ECs are diagnosed early, with 5-year survival rates of > 95% after surgical resection of the primary tumor.⁴ However, for patients who meet specific criteria, hormonal treatment may be an alternative treatment modality—particularly among young women who are interested in preserving their fertility.^10-14 However, the significant reduction in survival rates among women diagnosed with regional spread or distant disease⁴ illuminates the imminent need to develop therapeutic modalities that target unique tumor features to improve prognostic outcomes within this patient population.⁴

CONTEXT

• Key Objective

• Endometrial cancer (EC) is the sixth most commonly diagnosed cancer in women worldwide. Yet, unique to EC, disease incidence—particularly in younger patients and in developed countries—has been rapidly increasing in recent years. Consequently, there is a timely need to understand potential unique epidemiologic and biologic patterns of this disease to tailor early detection, risk stratification, and clinical intervention strategies for women younger than age 50 years (early-onset) diagnosed with EC.

• Knowledge Generated

• First, we discovered that early-onset EC has a significantly lower tumor mutational burden versus late-onset EC among patients with nonhypermutated tumors. Next, we found that CTNNB1 and BRCA2 showed a significant positive association with early-onset EC, whereas FGFR2 and PIK3R1 showed inverse associations with early-onset EC compared with late-onset disease.

• Relevance

• This study supports future clinical investigations to develop and test personalized therapeutic modalities tailored to young women diagnosed with this malignancy.

Over the past two decades, results from genomic studies have revealed critical insights into the molecular landscape of ECs and their histologic subtypes—a strong predictor of prognostic outcomes in this malignancy.¹⁵ For example, endometrioid ECs (adenocarcinomas) have harbored mutations in PTEN, PIK3CA, PIK3R1, CTNNB1, and ARID1A,^16-21 whereas serous EC present with TP53, FBXW7, PIK3CA, and PPP2R1A mutations.^22-25 Consequently, The Cancer Genome Atlas (TCGA) divided endometrioid and serous ECs into four distinct molecular subgroups with prognostic significance.²⁶ However, to our knowledge, no studies to date have defined the genomic landscape of EC by age. Given the rising incidence of EC among women younger than age 50 years (early-onset EC), an understanding of the unique epidemiology and etiology of this disease is critical to tailor early detection, risk stratification, and clinical intervention strategies for this patient population.

The purpose of this study, composed of 2,425 patients with EC from the international clinicogenomic data-sharing consortium American Association of Cancer Research (AACR)'s Project Genomics Evidence Neoplasia Information Exchange (GENIE) consortium²⁷ and replicated using TCGA data, was to characterize distinct somatic mutation patterns among women diagnosed with early-onset and late-onset EC.

METHODS

Data Sources and Study Population

Next-generation clinical sequencing data from tumor tissues and associated pathology reports have been released by the AACR GENIE project²⁷ from several cancer centers in the United States, Canada, and Europe. This study has been granted data access through Database of Genotypes and Phenotypes (dbGap) project #24541. Somatic mutation and clinical data from EC patients were downloaded from the GENIE project via Synapse (release 9.1).²⁸ This study was exempt from institutional review board approval and informed consent because deidentified GENIE data are publicly available to the entire scientific community.²⁷ We included a total of 2,425 pathologically confirmed EC patients with a unique patient record and matched clinical and sequencing data from January 1, 2011, to December 31, 2019. Clinical and somatic mutation data for 479 patients with EC were also obtained from the TCGA data portal.^26,29

Clinical, Pathologic, and Demographic Features

Clinical and pathologic variables included histologic subtype (adenocarcinoma, adenosquamous or mixed, serous carcinoma, clear cell carcinoma, undifferentiated and/or dedifferentiated, carcinosarcoma, and unspecified and/or others) and sample type (primary tumor or metastatic site). Demographic variables included patient age at sequencing report (a surrogate for age at diagnosis as we have previously described),³⁰ race and ethnicity (non-Hispanic White [NHW], non-Hispanic Black [NHB], and Asian and/or Pacific Islander [API]), and sequencing center. For TCGA data, age at diagnosis, race and/or ethnicity, and pathologic variables including histologic subtype (adenocarcinoma, adenosquamous or mixed, and serous carcinoma) were included in this study.

Somatic Mutations in Putative Cancer Driver Genes

Using clinical-grade targeted gene panel sequencing approaches from different sequencing centers, somatic mutation data for tumor tissues have been previously generated.²⁷ The collected tumor tissues were sequenced to pooled median read depths of 500× for GENIE data and 100× for TCGA data. To ensure consistent somatic variation calling in tumor tissues and to minimize artifacts and germline events, GENIE has applied a stringent filtering pipeline to remove putative germline variants (eg, using pooled blood samples as controls, existing databases of known artifacts, and common germline variants from the 1000 Genomes Project or Exome Sequencing Project with allele frequencies > 0.1%). Tumor mutational burden (TMB) for each study participant was measured by the total number of carrying mutations per Mb. We excluded patients with a sequencing panel covering < 100 kb target regions. Nonhypermutated EC was defined using a conservative cutoff of TMB < 17.78 mutations/Mb. Furthermore, we limited our analysis to nonsilent variants—including missense, frameshift, nonframeshift, splicing, nonsense, and truncating mutations—for the analyses of nonsilent mutation events (eg, bin variable for mutation carrier and non-carrier) and mutation frequencies, similar to our previous studies.^30-32

We performed mutational signature analyses using the computational framework SigProfiler.³³ We first generated mutational matrices of somatic mutations for our cohort samples using the SigProfilerMatrixGenerator tool and then identified mutational signatures using the SigProfilerExtractor tool.

Statistical Analysis

We compared clinical and demographic features between women diagnosed with early-onset EC (age < 50 years) and late-onset EC (age ≥ 50 years) using chi-square tests or Fisher's exact tests (for categorical variables) or t-tests (for continuous variables) as appropriate. To investigate associations between TMB and age group, we used multivariable linear regression models adjusted for race, ethnicity, histologic subtype, sequencing center, and sample type. To investigate associations between mutational signatures and age group, we performed multivariable linear regression with an adjustment of the same abovementioned covariates and TMB. Comparison of somatic mutations for each gene by age group was analyzed using multivariable logistic regression and adjusted for the same covariates and TMB. All covariates were used as fixed effects, and the reference outcome category was women with late-onset EC. In addition, we performed similar analyses stratified by race, ethnicity, and histologic subtypes. To account for multiple testing across analyses, a Bonferroni correction was used to establish P-value thresholds for strong evidence (P < .0018 = 0.05/28 genes) and suggestive evidence (0.0018 < P < .05) for reported associations. All analyses were conducted using R software version 3.3.3 (R Project for Statistical Computing).

RESULTS

Study Population

A total of 2,425 women diagnosed with EC were identified from the AACR Project GENIE Consortium data during the 9-year study period (Data Supplement). Overall, the mean age at cancer diagnosis was 42.5 years, and approximately one in every 13 women had early-onset EC (7.3%; 176 of 2,425 women). Nearly 83% of women were NHW (2,010 of 2,425), and the proportion of women with early-onset EC differed by race and ethnicity. In particular, API women accounted for a larger proportion of early-onset versus late-onset EC patients. By histologic subtypes, 1,206 patients (49.7%) were diagnosed with endometrial adenocarcinoma, 453 (18.7%) had serous carcinoma, and 284 (11.7%) had carcinosarcoma. Histologic subtypes differed by age group in this cohort; adenocarcinomas comprised a larger proportion of early-onset versus late-onset ECs. By contrast, serous carcinomas and carcinosarcomas accounted for a lower proportion of early-onset versus late-onset EC patients.

Spectrum of Somatic Mutations in EC

The spectrum of somatic mutations in EC is presented in Figure 1. Overall, approximately one in every five endometrial tumors (20.7%, n = 502) were classified as hypermutated. Among hypermutated tumors, we observed that 16%-34% harbored at least a nonsilent mutation in a DNA mismatch repair gene, including POLE (34%), MSH6 (32%), MSH2 (21%), and MLH1 (16%; Fig 1A). Excluding the hypermutated tumors, nonhypermutated tumors had an average of 5.91 mutations/Mb. Subsequent analyses were focused on nonhypermutated EC patients. The nonsilent mutation frequency for the top nine frequently mutated genes (with a mutation frequency > 10%) in nonhypermutated ECs is illustrated in Figure 1B. Lollipop plots featuring the nonsilent mutations for each individual gene are shown in the Data Supplement.

FIG 1.

Genomic spectrum of endometrial cancers by age group. (A) Mutation rate in tumor samples from 2,425 patients. The red dashed line represents 17.78 mutations/Mb, the cutoff used to define hypermutated and nonhypermutated tumors. Mutation frequencies are denoted for mismatch repair genes in hypermutated samples. (B) Spectrum of somatic variations in nonhypermutated endometrial tumors. (C) Mutation frequencies by age group (early-onset and late-onset) for the top 28 frequently altered genes in EC patients. (D) Significant genes identified in the adjusted multivariable logistic regression model. OR, odds ratio.

Among the nonhypermutated tumors, women with early-onset EC had significantly lower TMB compared with late-onset EC patients (P = .003; Data Supplement). In stratified analyses, this association remained robust among NHWs and among those diagnosed with endometrial adenocarcinomas (Data Supplement). Moreover, several histologic subtypes (adenosquamous, serous carcinoma, clear cell carcinoma, carcinosarcoma, and unspecified) showed significant associations with having lower TMB compared with adenocarcinoma cases (Data Supplement). In addition, in mutational signature analyses, women with early-onset EC had significantly higher mutational signature SBS10a compared with late-onset EC patients (P = 3.63 × 10^–5; Data Supplement). All mutational signature figures derived from all raw data are presented in the Data Supplement.

We further investigated the frequency of nonsilent somatic mutations for the top 28 frequently mutated genes (mutation frequency > 3%) by age group (Fig 1C). Seven genes harbored distinct mutation frequencies (difference > 5%) between early-onset and late-onset ECs (TP53: 24.0% [early-onset] v 52.9% [late-onset]; PTEN: 53.6% v 35.1%; PIK3R1: 15.2% v 22.0%; CTNNB1: 39.2% v 13.2%; PPP2R1A: 3.1% v 15.3%; FBXW7: 4.0% v 13.3%; and FGFR2: 1.60% v 6.9%, respectively).

Significant Genes Associated with Early-Onset EC

We examined associations between each of the top 28 commonly mutated genes with age groups (Table 1 and Fig 1D). Among nonhypermutated ECs, women with early-onset EC had significantly lower odds of presenting with nonsilent FGFR2 and PIK3R1 somatic mutations compared with late-onset patients (FGFR2: odds ratio [OR] = 0.18, 95% CI, 0.04 to 0.76, P = .02; PIK3R1: OR = 0.54, 95% CI, 0.31 to 0.92, P = .02). By contrast, early-onset EC patients had three- to four-fold increased odds of presenting with nonsilent CTNNB1 and BRCA2 mutations in adjusted models (CTNNB1: OR = 3.32, 95% CI, 2.14 to 5.16, P = 8.93 × 10^–8; BRCA2: OR = 4.01, 95% CI, 1.55 to 10.38, P = 4.25 × 10^–3). Notably, three (FGFR2, PIK3R1, and CTNNB1) of these four genes presented with concordant associations in both the GENIE and TCGA data sets (Data Supplement).

TABLE 1.

Frequency and Differential Pattern of Nonsilent Somatic Mutations Between Patients With Early-Onset and Late-Onset EC

Heterogeneity by Race/Ethnicity and EC Histologic Subtypes

Given pronounced racial and ethnic disparities in EC,^34-39 we next performed stratified analyses of nonhypermutated ECs to explore distinct genomic patterns of early-onset disease across racial and ethnic groups. Of interest, we discovered several additional genes significantly associated with early-onset EC. Among NHWs, women with early-onset EC had significantly decreased odds of presenting with nonsilent KRAS mutations compared with NHW women with late-onset EC (KRAS: OR = 0.54; 95% CI, 0.29 to 0.996; P = .049, Data Supplement). This finding is in addition to the associations observed in our entire cohort with PIK3R1 and FGFR2. Moreover, early-onset EC patients among NHWs had 2.68-fold increased odds of presenting with nonsilent APC mutations; this is in addition to the association observed above in our entire cohort with CTNNB1 and BRCA2 (APC: OR = 2.68; 95% CI, 1.05 to 6.86; P = .04). Among NHB women, early-onset EC patients showed suggestive evidence of having increased odds of presenting with KMT2D or KMT2C nonsilent mutations compared with late-onset patients; this is in addition to the observed significance of CTNNB1 (Data Supplement). For API women, in addition to the observed significance of CTNNB1, strong evidence emerged that younger patients had increased odds of presenting with KMT2D mutations compared with women with late-onset EC (Data Supplement). However, ORs and 95% CIs were unstable for these genes because of the small sample size of the NHB and API groups in this data set.

By tumor histology, several additional genes were found to be significantly associated with early-onset EC. As expected, results for adenocarcinoma patients were similar to our findings in the overall cohort, with the exception of BRCA2. Early-onset endometrial adenocarcinoma patients with nonhypermutated tumors had 81% and 48% decreased odds of presenting with FGFR2 and PIK3R1 variations, respectively, compared with their late-onset counterparts (on the basis of suggestive evidence). By contrast, patients with early-onset endometrial adenocarcinoma had increased odds of presenting with nonsilent CTNNB1 mutations with strong evidence (Data Supplement). For serous carcinomas, early-onset patients had higher odds of presenting with nonsilent PTEN variations versus the late-onset group, on the basis of suggestive evidence. For carcinosarcomas, patients with early-onset EC had increased odds of presenting with nonsilent mutations in APC when compared with late-onset patients (on the basis of suggestive evidence) although as with racial and ethnic subgroups, it should be noted that ORs and 95% CIs were unstable because of the small sample size (Data Supplement).

Moreover, we also examined whether these genes of interest had heterogeneity across racial, ethnic, and histologic subgroups. Of interest, KMT2D showed significant heterogeneity by race and ethnicity in our cohort (P for heterogeneity = .01, Data Supplement). Although the APC gene did not show a significant P for heterogeneity in the overall cohort, heterogeneity did reach statistical significance for APC mutations across histologic subtypes (P for heterogeneity = .02).

Replication of Results in TCGA

In our study, replication of our GENIE data analysis and findings was performed with TCGA data.²⁶ Baseline characteristics of patients with EC from the TCGA cohort are also presented in the Data Supplement. Concordant with our findings in the GENIE cohort, analysis of TMB significantly differed across age groups and histologic subtypes in TCGA (Data Supplement). Notably, the direction of association (beta coefficients) was consistent for most significantly identified genes across data sets (ie, 75% [entire cohort] and 100% [adenocarcinoma]). However, most genes, which were reported to reach statistical significance in the GENIE cohort, did not reach significance in TCGA, which can be potentially attributed to 80% fewer EC patients in TCGA compared with GENIE (479 v 2,425 women with EC, Data Supplement).

DISCUSSION

This study investigated the spectrum of somatic mutations among women with EC by age groups using two well-annotated patient cohorts. First, we discovered that early-onset EC has a significantly lower TMB and higher mutational signatures SBS10a versus late-onset EC among patients with nonhypermutated tumors. Next, we found that CTNNB1 and BRCA2 showed a significant positive association with early-onset EC, whereas FGFR2 and PIK3R1 showed inverse associations with early-onset EC compared with late-onset disease. Stratified analyses by race, ethnicity, and histologic subtype also revealed distinct mutation patterns, including significant positive associations with early-onset EC for APC among NHW women, KMT2D among NHB and API populations, KMT2C among NHBs, PTEN in serous carcinomas, and APC in carcinosarcoma histologic subtypes. Conversely, among NHWs, KRAS mutations harbored an inverse association with the early-onset EC.

TMB is currently being evaluated as a predictive biomarker for immune checkpoint inhibitors in several cancer types, as tumors with a higher number of mutations may be more likely to respond to immunotherapies.⁴⁰ For hypermutated ECs, immune checkpoint inhibitors can now be administered if standard treatment fails in metastatic cancer.^3,15 Although MSS and/or nonhypermutated tumors still account for the majority of EC patients, we do not yet have biomarkers to predict which tumors will be more likely to respond to therapies. In our study, we sought to identify genomic patterns of early-onset and late-onset EC specifically among patients with nonhypermutated tumors. Interestingly, herein, we observed significant differences in TMB across age groups, as early-onset ECs (including both hypermutated and nonhypermutated tumors) had a higher TMB compared with late-onset patients (data not shown). However, specifically, among nonhypermutated tumors, TMB was lower among early-onset EC versus late-onset EC patients. This is consistent with previous findings that mutations accumulate with age in nonhypermutated tumors.⁴¹ These findings suggest that fewer patients may benefit from immunotherapies and further emphasize the clinical need for timely development of personalized therapeutic modalities.

To date, few studies have been conducted to examine somatic mutation patterns by age among EC patients. One previous study, using 31 samples of normal control endometrial tissue and 81 samples of EC tissue, examined DNA methylation patterns for tumor-related genes.⁴² Consistent with our present findings, CTNNB1 and genes involved in Wnt signaling were reported to be frequently mutated among EC patients diagnosed at age 40 years and younger, whereas genetic and epigenetic alterations of fibroblast growth factor signaling genes were more common in tumors of women diagnosed at older age.⁴² In another study, using gene expression profile clustering for 271 endometrioid ECs, they identified four transcriptomic clusters with distinct gene signatures.⁴³ In their clusters, cluster II—significantly enriched with CTNNB1 mutations—was mostly composed of younger age patients.⁴⁴ Moreover, the work of Polymeros and colleagues previously demonstrated that among British South Asian women, patients with somatic CTNNB1 mutations tended to be younger age at diagnosis although these findings only reached marginal significance.⁴⁵ Nevertheless, as pronounced racial and ethnic disparities persist in EC,⁴⁶ these findings may shed light on molecular features underscoring these disease-specific patterns in future clinical investigations. Moreover, molecular classification of EC has been shown to be associated with progression-free survival, prognosis, and clinical outcomes.^44,47 Although biomarker-directed systemic therapies are still limited in a small portion of patients,⁴⁸ if studies on molecular stratification in the setting of early-onset ECs are accumulated along with pathologic findings, they can be used to inform practice, potentially delineating treatment options and enhancing the future practice of cancer medicine.

The use of data from the GENIE consortium and TCGA is a strength of this study as it allowed for a relatively large number of pathologically verified EC patients with tumor sequencing data to be included in our cohort. However, there are some limitations to this study. First, the sample size across racial, ethnic, and tumor histology groups was limited. Second, GENIE does not provide information about disease stage, grade, treatment history, or overall survival, nor about detailed individual-level factors (eg, germline genetic features, family history of cancer, lifestyle-related factors, obesity, and hormonal-related factors) of relevance in EC. Consequently, we were unable to explore whether differences in these factors were associated with distinct genomic patterns of early-onset EC. Third, as a limitation to targeted gene panel sequencing, we were only able to analyze a set of known cancer-related genes, which does not permit for the discovery of novel putative cancer driver genes related to early-onset EC at present. Notwithstanding these limitations, findings from this study posit potential biologic factors contributing to a distinct etiology of early-onset EC and biologic contributors to racial and ethnic disparities among early-onset EC patients.

In conclusion, this study observed a distinct spectrum of somatic variations among early-onset EC patients compared with late-onset EC patients. The unique molecular phenotype of early-onset EC may offer insights into disease-specific etiology and lay the foundation for developing targeted treatment modalities for younger patients.

DATA SHARING STATEMENT

The data supporting findings from this study are available from the corresponding authors upon request.

AUTHOR CONTRIBUTIONS

Conception and design: Jungyoon Choi, Andreana N. Holowatyj, Xingyi Guo

Financial support: Andreana N. Holowatyj, Xingyi Guo

Collection and assembly of data: Jungyoon Choi, Xingyi Guo

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).