Over the past 4 decades, prognostic stratification of endometrial cancer (EC) has progressively incorporated serous histopathologic findings, lymphatic metastases, lymphovascular space involvement, TP53 variations, ERBB2 expression, mismatch repair and homologous recombination (HR) repair deficiencies, and other variables (1–5). More recently, The Cancer Genome Atlas (TCGA) generated a molecular-based prognostic classification system (6). Despite the evolution of EC risk stratification, corresponding risk-adjusted therapy has minimally affected EC mortality rates. The American Cancer Society 2020 Annual Report concluded that “cancer survival has improved since the mid-1970s for all of the most common cancers except uterine cervix and uterine corpus” and that “stagnant survival rates reflect a lack of major treatment advances” (7). Addressing this lack of major treatment advances requires an enhanced understanding of pivotal tumor-specific molecular drivers to facilitate the discovery of target-specific therapeutics.

TCGA classification of EC identified 4 unique molecular subgroups: 1) POLE ultramutated, characterized by POLE variations; 2) microsatellite instability hypermutated, classified by mismatch repair deficiency and high microsatellite instability (MSI-H); 3) copy-number low, denoted by no specific molecular profile (NSMP); and 4) copy-number high, primarily typified by TP53 variations and serouslike EC (6). Comparative prognoses associated with POLE ultramutated and copy-number high EC are contrasting, in which the prognosis is favorable for POLE ultramutated EC and substantially compromised for copy-number high EC. MSI-H and NSMP ECs have an intermediate prognosis but account for most EC tumors, thereby highly contributing to the annual EC mortality rate (6, 8–10). Therefore, determining the causes of treatment failures in MSI-H and NSMP ECs is paramount. Because MSI-H and NSMP tumors are predominantly low stage, with grade 1 and 2 histopathologic indicators (11), treatment failures may reflect our inability to identify clinicopathologic low-risk cases at high risk for recurrence (i.e., EC harboring occult high-risk disease). Radiotherapy (RT) and/or platinum-based chemotherapy (PbCT) are standard adjuvant treatment modalities for clinicopathologic high-risk MSI-H and NSMP EC. Treatment failures may also result from our inability to predict insensitivity to these standard adjuvant therapeutics (9, 12).

MSI-H EC generally harbors a high tumor mutation burden (TMB), and thus patients with MSI-H tumors are considered potential candidates for immunotherapy (6, 13–16). However, the high associated TMB presumably predicts a greater prevalence of HR gene variations, which portend sensitivity to DNA-damaging agents in cases with clinicopathologic high-risk indicators (17–20). Defining specific vulnerabilities in MSI-H tumors is necessary for decreasing the mortality rate in patients with this molecular profile. Furthermore, because of the absence of identifiable molecularly targetable pathways in NSMP, empiric therapy must be used for the largest EC cohort. Therefore, we used our previously described stratification system based on E2F1 and CCNA2 expression and sequence variations in POLE, PPP2R1A, and FBXW7 (termed ECPPF) to molecularly stratify MSI-H and NSMP ECs (21). The primary aim of our study was to use ECPPF to identify MSI-H and NSMP EC cases with clinicopathologic low-risk indicators that are at risk for occult extrauterine disease and recurrence and those with clinicopathologic high-risk indicators that are insensitive to standard therapeutic regimens.

Endometrioid histopathologic characteristics are present in more than 80% of EC tumors, with most classified as MSI-H or NSMP, low stage, and low grade (6, 11, 29). The observation that MSI-H/NSMP EC with clinicopathologic low-risk indicators accounted for most of the MSI-H/NSMP recurrences denotes major limitations in conventional risk assignment strategies. Improving clinical outcomes requires the identification of molecular vulnerabilities to permit target-specific risk designation and choice of therapeutics. Consequently, we assessed our previously described ECPPF molecular stratification system for its ability to resolve MSI-H/NSMP EC prognoses (21).

Because sensitivity to RT and chemotherapeutics are in part predicated on attenuated DNA-damage repair pathway elements (18, 30, 31), we integrated HR variant gene status into our stratification system. Our findings show that ECPPF prognostically stratifies MSI-H and NSMP ECs and that integration of HR variant gene status may facilitate development of new molecular-based therapeutic strategies. The ECPPF-stratified MHR group identified cases at high risk for recurrence in those conventionally classified as clinicopathologic low risk. ECPPF stratification also identified NSMP EC with variant CTNNB1 as a subgroup with a high risk of recurrence and the MHR group with wild-type HR genes as insensitive to standard adjuvant PbCT and/or RT. In contrast, the ECPPF-stratified MLR group predicted sensitivity to standard therapies (PbCT and/or RT) in EC with clinicopathologic high-risk indicators and/or HR variant genes.

The high TMB associated with MSI-H EC has understandably led to embracing immunotherapy in persistent or recurrent disease (13–16). High TMB also predicts an increased prevalence of HR variant genes and most likely sensitivity to RT and DNA-damaging agents (17). Consistent with this precept, recent studies reported improved clinical outcomes with RT in patients with MSI-H EC (32, 33). In the current study, ECPPF effectively stratified MSI-H cases, in which patients in the MHR group with early-stage, low-grade tumors, who are rarely candidates for adjuvant therapy, had a poor DFS rate. Moreover, the ECPPF-stratified MLR and MHR groups with HR variant genes and clinicopathologic high-risk indicators had higher DFS rates than did the MHR group with wild-type HR genes and clinicopathologic high-risk indicators.

These observations advocate for strategizing treatment options for MSI-H and NSMP EC tumors with clinicopathologic low-risk indicators according to ECPPF-stratified MLR and MHR groups. Thus, pending HR gene status, conservative surveillance is recommended for the MLR group, and early therapeutic intervention (innovative clinical trials) is recommended for the MHR group. Furthermore, a therapeutic strategy for MSI-H and NSMP EC with clinicopathologic high-risk indicators would consist of treatment with contemporary PbCT and/or RT for patients with ECPPF-stratified MLR tumors and MHR tumors with HR variant genes but innovative clinical trials for those with ECPPF-stratified MHR tumors with wild-type HR genes.

Although NSMP EC is primarily early stage and low grade, it is associated with an intermediate prognosis (6, 8–10). This paradox suggests that its high recurrence rate is linked to unrecognized high-risk disease in conventionally classified low-risk EC. ECPPF appeared to effectively stratify NSMP EC cases, particularly for early-stage, low-risk tumors. The MLR group had a distinctly favorable DFS rate over that of the MHR group with occult high-risk disease. CTNNB1 variations in low-grade NSMP EC are reportedly associated with poor patient outcomes (18, 34, 35). ECPPF efficiently stratified NSMP tumors with variant CTNNB1, and all documented recurrences occurred in the MHR group. ECPPF appears to provide a molecular profile for TCGA-designated copy-number low tumors by identifying occult high-risk disease within conventional low-risk cases and high-risk disease that is insensitive to standard adjuvant PbCT and/or RT.

The recently published version of the National Comprehensive Cancer Network guidelines encourages molecular assessment of EC and states that probing for POLE hotspot variations and performing immunohistochemical analysis of mismatch repair deficiency and wild-type TP53 expression may identify MSI-H and NSMP, respectively (36). However, without specific profiling of adverse molecular drivers, targetable vulnerabilities remain unrecognized. Therefore, ECPPF classification provides new profiling insights to refine prognoses and improve therapeutic efficacy.

The predominant challenges for treating MSI-H and NSMP ECs are distinguishing occult high-risk disease in conventionally characterized histomorphologic low-risk EC and predicting insensitivity of EC with clinicopathologic high-risk indicators to contemporary therapies (12). The 43.8% 3-year DFS rate of the ECPPF-stratified MHR group with clinicopathologic low-risk indicators is the first evidence that molecular stratification can distinguish high-risk disease in conventionally classified low-risk MSI-H/NSMP EC. Importantly, 81% of documented recurrences occurred in the ECPPF-stratified MHR group with wild-type HR genes which accounted for only 28% of all TCGA MSI-H/NSMP EC cases. Furthermore, our analysis of MSI-H/NSMP EC with clinicopathologic high-risk indicators showed that the 3-year DFS rate of the MLR group and MHR group with HR variant genes was considerably disparate from that of the MHR group with wild-type HR genes. This finding is consistent with treatment sensitivity to standard therapeutic regimens in the MLR group and MHR group with HR variant genes and treatment insensitivity in the MHR group with wild-type HR genes. These observations introduce a potential molecular-based paradigm shift for precision therapeutics in MSI-H and NSMP ECs. On the basis of our previously reported molecular schematic of disrupted EC signaling pathways (21), reasonable therapeutic options may include modulators of the CCNA2-E2F1-CIP2A axis or PP2A activity and FBXW7 substrates alone or in combination with PbCT or PARP inhibitors (27, 36–40).

Strengths of our analyses include the size of the patient population, centralized pathology review, and availability of robust genomic and transcriptomic databases. Assessing the primary clinical outcome of DFS rather than overall survival is an additional strength of the study that afforded analysis of limitations in contemporary staging, prognostication, and therapeutic efficacy. Limitations of this study included the variability and length of follow-up periods. Because the upper quartile of time to recurrence exceeded the lower quartile of follow-up time, the extent of recurrences in our study population may be underestimated. Recent meta-analyses of prognostic indicators independent of the TCGA signature suggest that lymphovascular space involvement (LVSI) has prognostic value for EC and that deep MI may be a predictor of recurrence (41, 42). LVSI information was not accessible in the TCGA study population and warrants inclusion in subsequent ECPPF stratification studies. In previously reported multivariate analyses incorporating the ECPPF elements with numerous traditional risk factors, deep MI was not an independently significant risk factor (21, 43). Moreover, definitive information in TCGA regarding treatment strategies for the patients was not available. However, because specimens were submitted chiefly from comprehensive cancer centers, it is reasonable to assume that the therapies were consistent with contemporary established guidelines. Such therapies include surgical staging and adjuvant RT and/or chemotherapy in patients with clinically identified at-risk disease; as such, 98% of patients receiving chemotherapy received PbCT (6). Knowledge of the sites of recurrence (e.g., local, regional, or distant) would have also been preferable for this study, particularly for tumors with predominantly clinicopathologic low-risk indicators.

In summary, ECPPF identifies occult high-risk disease in MSI-H/NSMP EC with clinicopathologic low-risk indicators and appears to efficiently triage treatment sensitivity of MSI-H/NSMP EC with clinicopathologic high-risk indicators to contemporary adjuvant therapies. Consequently, ECPPF stratification confers a new therapeutic paradigm for MSI-H/NSMP EC: 1) observation only for MLR cases with clinicopathologic low-risk indicators; 2) PbCT and/or RT for MHR cases with HR variant genes and clinicopathologic low-risk indicators, MLR cases with clinicopathologic high-risk indicators, and MHR cases with HR variant genes; and 3) innovative clinical trials for MHR cases with wild-type HR genes and clinicopathologic low-risk or high-risk indicators.

CCNA2, cyclin A2; CTNNB1, catenin beta 1; E2F1, E2F transcription factor 1; FBXW7, F-box and WD repeat domain-containing protein 7; POLE, DNA polymerase epsilon, catalytic subunit; PPP2R1A, protein phosphatase 2 scaffold subunit alpha.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article.

Conceptualization: JG-B and KP; Data curation: JG-B and KP; Formal analysis: JG-B, JB-G, BK, KH, and KP; Funding acquisition: SD, AF, and KP; Investigation: JG-B, SW, JB-G, AW, MM, SD, AF, BK, KH, SY, FC, and KP; Methodology: JG-B, BK, KH, and KP; Project administration: KP; Resources: JG-B and KP; Software: JG-B, AW, and MM; Supervision: KP; Pathology: BK, KH, and FC; Writing – original draft: JG-B, JB-G, AW, MM, BK, KH, SY, and KP; Writing – review and editing: JG-B, SW, JB-G, AW, MM, SD, AF, BK, KH, SY, FC, and KP. All authors contributed to the article and approved the submitted version.