Purpose:

The two most common molecular subtypes of endometrial cancers, mismatch repair deficient (MMRd) and p53 wild-type (p53wt) comprise the majority of endometrial cancers and have intermediate prognoses where additional risk stratification biomarkers are needed. Isoform switching of FGFR2 from FGFR2b to FGFR2c (normally expressed in mesenchymal cells), has been reported in other solid carcinomas. The objective of this study was to investigate the role of FGFR2c in risk stratification of endometrial cancer.

Experimental Design:

We have developed and optimized a BaseScope RNA ISH assay to detect FGFR2c. FGFR2c expression was determined in a preliminary screening cohort of 78 endometrial cancers and a clinically and molecularly annotated Vancouver cohort (n = 465). Cox regression model analyses were performed to assess the prognostic value of FGFR2c.

Results:

Univariate and multivariate analyses revealed FGFR2c expression was significantly associated with shorter disease-specific survival (DSS) and progression-free survival (PFS) in endometrioid endometrial cancer (EEC, n = 302). Notably, FGFR2c expression was significantly associated with shorter PFS and DSS in patients with grade 3 EECs (P < 0.003 and P < 0.002) and the European Society Medical Oncology (ESMO) high-risk group (P < 0.0001 and P < 0.002), respectively. Moreover, within the MMRd subtype, FGFR2c expression was significantly associated with shorter PFS (P < 0.048) and DSS (P < 0.001).

Conclusions:

FGFR2c expression appears an independent prognostic biomarker in patients with EEC and further discerns the outcomes within grade 3 tumors, ESMO high-risk groups, as well as within the MMRd and p53wt subtypes. FGFR2c inclusion into future molecular subtyping can further refine risk stratification of EEC.

Translational Relevance

In endometrial cancer, risk stratification has been traditionally performed on the basis of clinicopathologic parameters, but they are not optimal for adjuvant treatment tailoring. New molecular risk stratification has identified molecular subtypes with excellent prognoses (POLE) and very poor prognoses (p53mut). However, approximately 80% of endometrioid endometrial cancers (EEC) belong to the mismatch repair deficient (MMRd) and p53 wild-type (p53wt) subtypes with intermediate prognoses where additional biomarkers for risk stratification are needed. In this study, FGFR2c isoform expression was detected in approximately 50% of EECs, and was associated with shorter progression-free survival and disease-specific survival, especially in patients with grade 3 disease (presumed to have poorer clinical outcomes) and those with high-risk features. Furthermore, FGFR2c expression was associated with poorer outcomes within the MMRd and p53wt molecular subtypes. We showed that integrating FGFR2c into a new molecular classifier significantly improves the prediction of recurrence and endometrial cancer–specific survival, which could potentially reduce under and/or overtreatment of patients with EEC.

Endometrial cancer is the most common gynecologic malignancy in the developed countries with increasing annual rates (1). The majority of patients with endometrial cancer are diagnosed at an early stage and have favorable prognosis (1, 2), however, about 20%–30% of patients with endometrial cancer with “high-risk” features have a high risk of recurrence and poor prognosis. In current practice, patients with endometrial cancer are stratified and assigned to adjuvant radio-chemotherapy based on the traditional clinicopathologic markers, for example, the European Society Medical Oncology (ESMO) risk stratification (3). The drawback for clinicopathologic risk group stratification is most of the clinical parameters are obtained postoperatively and this limits the ability to incorporate prognostic tumor biomarkers during surgical management decisions, for example, avoiding the morbidity associated with full lymph node staging for those patients with low-risk disease.

The Cancer Genomic Atlas (TCGA) networking group originally identified four molecular subtypes with distinct prognoses based upon genomic somatic copy-number alterations, tumor mutational load, and microsatellite instability (MSI) status (4). Subsequently, the McAlpine laboratory has developed, confirmed, and validated the “proactive molecular risk classifier for endometrial cancer (ProMisE)” using surrogate IHC markers and limited sequencing (5–7). Similarly, a Dutch group has validated the prognostic importance of the TCGA molecular subtypes in a large cohort of PORTEC1/2 patients with endometrial cancer with early-stage disease (8, 9) and Cosgrove and colleagues also used a slightly different strategy to validate the molecular subtypes in a large population-based U.S. cohort (10). The four molecular subtypes are described as follows: DNA polymerase epsilon exonuclease domains mutant (POLE EDM) the same as TCGA “POLE/ultramutated” but not identical (excellent prognosis); mismatch repair deficient (MMRd) similar to MSI/“hypermutated” (intermediate prognosis); p53 wild-type (p53wt)/NSMP (no specific molecular profile) similar to TCGA “copy number low” (intermediate prognosis); and p53 mutant/abnormal similar to TCGA “copy number high/serous-like” (worst prognosis). Despite slight differences in how the subtypes were identified in the TCGA, PORTEC, ProMisE, and U.S. patient cohorts, the MMRd and p53wt subtypes comprise approximately 30% and 50% of all diagnosed endometrial cancers and show mostly endometrioid histology (4–7, 9, 10). Identification of additional biomarkers that can further stratify these two most common subtypes into those patients with better and worse prognoses is an unmet clinical need that could guide management decisions of patients with endometrial cancer in clinical practice.

FGFR2 is a member of the FGFR family of receptor tyrosine kinases (RTK). FGFR2 mutations have been identified in 12% and 17% of stage I/II and III/IV, respectively, in a large multi-institutional cohort of endometrial cancer, and are associated significantly with shorter progression-free survival (PFS) and disease-specific survival (DSS) in endometrioid endometrial cancer (EEC) (ref. 11). FGFR2 mutations have been shown to predominantly occur within the MMRd and p53wt molecular subtypes of endometrial cancer (4, 9). FGFR2 is unique among RTKs, in that in addition to the more common mechanisms of receptor activation that drive tumorigenesis (mutations, translocations, and amplification), constitutive receptor activation can also occur through isoform switching (12). FGFR2 has two major isoforms, differentially expressed in epithelial (FGFR2b) and mesenchymal (FGFR2c) cells, which have different FGF ligand binding specificity. FGFR2 signaling is tightly controlled in normal tissues and occurs in a paracrine fashion such that FGF ligands secreted by mesenchymal cells stimulate FGFR2b and FGF ligands produced by epithelial cells stimulate FGFR2c. Isoform switching (inappropriate expression of FGFR2c instead of FGFR2b in epithelial cells) allows the receptor to bind to epithelial-expressed ligands and establish an autocrine loop (12). Several in vitro and in vivo studies reported ectopic expression of FGFR2c contributes to carcinogenesis and aggressive tumor behavior (13–15). In EEC, the most common oncogenic activating mutation of FGFR2 is S252W in the ligand binding domain, and detailed functional studies have shown it phenotypically mimics isoform switching, allowing FGFR2bS252W to bind promiscuously to FGFR2c ligands (16). Despite this, there is a paucity of data regarding FGFR2 isoform switching and the role of FGFR2c expression in progression and prognostication of endometrial cancer, due to a lack of reliable isoform-specific antibodies and the difficulties in distinguishing tumor- and stroma-expressed FGFR2c from RNA-sequencing (RNA-seq) data of bulk tumors. The purpose of this study was to develop a novel chromogenic BaseScope RNA ISH assay to detect FGFR2c in formalin-fixed, paraffin-embedded (FFPE) tissues, assess the frequency of FGFR2 isoform switching in endometrial cancer and its association with aggressive clinicopathologic parameters, and assess the prognostic value of FGFR2c in endometrial cancer using a clinically and molecularly annotated cohort.

Patients

Written informed consent was obtained from all participating patients whenever applicable and the study was conducted according to the Declaration of Helsinki. Two endometrial cancer commercial tissue microarrays (TMA UT 242a and UT 801) were purchased from Biomax after Queensland University of Technology (QUT) Human Research Ethics Committee (HREC) approval (#1800000461). These were used to optimize and validate the BaseScope RNA ISH as well as a screening cohort. Clinical data regarding histologic type, myometrial invasion, FIGO grade, and stage was obtained from Biomax along with TMAs sections.

We also obtained whole sections of endometrial cancer hysterectomy samples from our partner, Mater Hospital (Brisbane, Queensland, Australia) after ethical approval from HREC (Reference # HREC/15/MHS/127) as well as through QUT (Brisbane, Queensland, Australia, 1500000169). FFPE whole sections (4 μm thickness) were obtained from 20 women diagnosed with endometrial cancer and operated on between January 2016 and January 2017 at the Mater Hospital (Brisbane, Queensland, Australia). None of the patients received preoperative neoadjuvant therapy. Histologic type, myometrial invasion, lymphovascular space invasion (LVSI), lymph node metastasis, and histologic grade were assessed by a board-certified pathologist (D.S. Smith). The staging was performed according to the FIGO 2009 staging system (17).

The second cohort of patients was obtained from University of British Colombia (UBC, Vancouver, British Colombia, Canada) after HREC approval at the UBC (Vancouver, British Colombia, Canada, #H09-00939) and QUT (Brisbane, Queensland, Australia, #1800000670), respectively. This cohort was previously used to develop the ProMisE (5) and its subsequent confirmation in an independent dataset (6). Four serial slide sections from four previously constructed TMAs as well as anonymized data including demographic [age and body mass index (BMI)], clinicopathologic (histologic type, grade, myometrial invasion, LVSI, and FIGO stage), molecular subtype (POLE, MMRd, p53wt, and p53abn), and clinical outcome were obtained from the UBC database in collaboration with the McAlpine laboratory. All definitions of the above parameters were reported in previous publications (5, 6). A total of 465 samples were analyzed in this cohort and the REMARK criteria was provided. Nineteen cases had two pairs of cores (four cores) each, while the rest of the cases had a pair of cores analyzed for mRNA FGFR2c, PPIB (peptidylprolyl isomerase, housekeeping gene), and FGFR2 protein IHC.

IHC

FGFR2 protein IHC staining was performed using a standard manual protocol. Primary anti-FGFR2 antibody (catalog no. Ab58201, Abcam) was used at 1:200 dilution and incubated at 4°C overnight. This antibody targets the C-terminal FGFR2 and recognizes both FGFR2b and FGFR2c isoforms. Polymer-HRP–conjugated prediluted secondary antibody (catalog no. K406311-2, Dako, Diagnostic Envision Kit, Agilent Technologies) with diaminobenzidine (catalog no. K3468) detection kit was used and staining was performed according to the manufacturer's guidelines. A negative control of murine skin from a Cre keratin 5 conditional Fgfr2-knockout transgenic mouse (C57BL/6; ref. 18) and a positive control FFPE cell block made from SUM52PE (a breast cancer cell line with an FGFR2 amplification; ref. 19) were used. IHC scoring was performed by two independent observers (A.T. Sengal and P.M. Pollock) using the Histologic (H)-score method as published previously (20).

Chromogenic BaseScope RNA ISH

BaseScope RNA ISH follows a similar workflow and principle to the well-established novel RNAscope RNA ISH technique used for detection of mRNA in situ, with hybridization and amplification of target mRNA with super sensitivity and specificity. The BaseScope RNA ISH assay needs an extra step of amplification and a single Z pair target probe instead of the 20 ZZ pair probes employed in the RNAscope technique. Custom designed, an exon 7 to exon 8 junction-specific probe (BA-Hs-FGFR2-tv2-E7E8) (1ZZ), (gene accession NM_022970.3, 1578-1622 bp) to target epithelial FGFR2b and an exon 7 to exon 9 junction-specific probe (BA-Hs-FGFR2-tv1-E7E9), (NM_000141.4, 1580-1619bp) to target mesenchymal FGFR2c isoform were utilized. These probes are human specific. All probes and BaseScope reagents were purchased from Advanced Cell Diagnostic unless otherwise stated and the FAST RED modified bright field manual kit V1 was used. Four-micron-thick sections from FFPE cell line pellets and endometrial cancer clinical samples (whole sections and TMAs) were baked in a 60°C oven for 1 hour and deparaffinized in three changes of xylene and three changes of absolute alcohol. The sections were blocked with 3% H2O2 (catalog no. 322381) for 10 minutes at room temperature and washed with deionized water. Target retrieval was performed in a (decloaking chamber) pressure cooker (Biocare Medical) with Advanced Cell Diagnostic target retrieval solution (catalog no. 322000) incubated at 104°C for 15 minutes. A permeabilization step was performed using protease III (catalog no. 322380) with incubation in a humidified chamber for 30 minutes for FFPE tissue and 15 minutes for cell pellets in a 40°C oven (HB-1000, Hybridizer, UVP). After washing with deionized water, target probes FGFR2b, FGFR2c, and PPIB, a housekeeping gene positive control, and DabB (dihydrodipicolinate reductase) a bacterial gene negative control were incubated for 2 hours in a humidified chamber in a 40°C oven (HB-1000 Hybridizer, UVP). A subsequent series of amplification steps were performed according to the protocol as follows: amplification (AMP0, 30 minutes at 40°C), AMP1 (15 minutes at 40°C), AMP2 (30 minutes at 40°C), AMP3 (30 minutes at 40°C), AMP4 (15 minutes at 40°C) with the humidified chamber, AMP5 (30 minutes at room temperature), and AMP6 (15 minutes at room temperature). Slides were washed twice for 2 minutes in between each amplification step using diluted washing buffer (catalog no. 310091). Following chromogenic detection utilizing FAST RED dye (catalog no. 322900) incubated for 10 minutes at room temperature, a nuclear counterstaining was performed with Mayer hematoxylin (catalog no. MHS16-500, Sigma Aldrich) incubated for 2 minutes at room temperature, then blued by washing in tap water, and finally slides were covered using aqueous mounting media (Vectamount, catalog no. H-5000, Vector Laboratories).

Scoring of chromogenic RNA ISH (PPIB and FGFR2c)

Chromogenic mRNA ISH for PPIB expression (positive control targeting housekeeping gene) and FGFR2c expression were scored by two independent observers (A.T. Sengal and P.M. Pollock), blind to patient outcome data. Scoring was performed according to the protocol of the RNAscope/BaseScope RNA ISH recommendation (21) as follows: 0, no signal dots or <1 dot signal/10 tumor cells at 40× magnification; 1+, 1–3 signal dots/cell; 2+, 4–10 signal dots/tumor cell; 3+, >10 signal dots/cell with < 10 cluster dots; and 4+, >10 signal dots/cell and ≥10 cluster dots at 20× magnification. Representative images for each score are provided in Fig. 2B. Stromal FGFR2c was not included in scoring giving that stromal cells can express FGFR2c normally. DSS curves for each score showed very little difference (see Results), therefore, final scores were dichotomized as negative if the FGFR2c RNA ISH score was 0 and positive if the FGFR2c RNA ISH score was >1–4. Two independent cores were analyzed for PPIB mRNA (positive control housekeeping gene) and FGFR2c mRNA for each patient. The core with the highest signal was included in the analysis. There was 97% concordance in scoring between the two independent scorers and a final score for discordant cases was reached by consensus. Patients with loss of cores or harboring more than 50% stroma or poor mRNA quality, as defined by a lack of expression of PPIB (RNA ISH score 0), were excluded from all subsequent analyses.

Statistical analysis

All data analysis was performed using IBM, SPSS (version 23). χ2 and ANOVA tests for categorical and continuous variables, respectively, were performed to test for an association between FGFR2c RNA ISH status or FGFR2 protein and demographic and clinicopathologic data, ProMisE molecular subgroups and ESMO risk groups. Fisher exact test was used for dichotomized categorical variables. Time-to-event analyses were calculated as follows: overall survival (OS) from date of surgery or diagnosis to the date of death; DSS from date of surgery to the date of death due to endometrial cancer, and PFS from date of surgery to the date of any recurrence. All patients who were confirmed alive were censored at the last date of follow-up. The Kaplan–Meier method was used to determine survival curves (OS, DSS, and PFS) and P values were calculated using the log-rank test (LRT). Cox proportional hazards models were used to evaluate the prognostic value of each factor. Factors with P < 0.10 were included in a multivariate Cox regression model with a stepwise forward method to include in the final analysis. In the last step, significant factors from the forward selection model (P < 0.05) were included in the final Cox regression model together with established clinicopathologic prognostic factors: age at surgery (<60 vs. ≥60 years), BMI (≤30 vs. >30), myometrial invasion [negative and/or <50% vs. >50%, LSVI (negative vs. positive)], grade (1–2 vs. 3), FIGO stage (I/II vs. III/IV), and ProMisE molecular subgroups (MMRd vs. p53wt, POLE vs. p53wt, and p53abn vs. p53wt). The AUC ROC was used to evaluate the ability of the FGFR2c RNA ISH score 0–4 (five tiers) and FGFR2 protein IRS 0–4 (five tiers) to discriminate outcome (OS, DSS, and PFS) of patients with endometrial cancer in the Vancouver cohort. Furthermore, the discriminatory ability of several models with or without combining FGFR2c RNA ISH score with the established clinicopathologic parameters FIGO grade (three tiers) and FIGO stage (four tiers), ESMO risk group (three tiers), and ProMisE subgroups (four tiers) were performed using an AUC ROC analysis with 95% confidence intervals (CI). The same approach was followed in the TCGA endometrial cancer cohort depending on the available clinicopathologic variables. All models were internally validated using a bootstrap method with 1,000 samples. All reported P values were based on two-tailed tests of significance, where P < 0.05 was considered statistically significant.

A novel chromogenic BaseScope RNA ISH assay to detect FGFR2b and FGFR2c mRNA expression was optimized and validated in endometrial cancer FFPE cell pellets using Ishikawa cells transduced with FGFR2b (22) and AN3CA cells expressing endogenous FGFR2c verified by RT-PCR. The principle, workflow, optimization, and validation are summarized in Fig. 1. To assess the protein expression, anti-FGFR2 antibody targeting the C-terminal that detects both isoforms was used. FGFR2 protein was not detected in murine keratinocytes obtained from a Keratin 5 CRE conditional knockout of FGFR2, validating the specificity of our antibody (Supplementary Fig. S2A and S2B).

Figure 1.

Principles and workflow of bright field chromogenic RNA ISH, optimization and validation of BaseScope assay for detecting alternatively spliced FGFR2 isoforms. A, FGFR2 is encoded by 21 exons and has two isoforms; inclusion of exon 8 gives rise to the FGFR2IIIb isoform (designated as FGFR2b), expressed in normal epithelial cells, whereas inclusion of exon 9 gives rise to the FGFR2IIIc isoform (designated as FGFR2c), expressed in normal mesenchymal cells. FGFR2b and FGFR2c only differ in the second half of the third immunoglobulin-like loop (Ig-III) as indicated with purple and red colors, respectively. FGFR2b and FGFR2c bind to their specific ligands as noted in purple and red colors. FGF1 binds to both isoforms. B, Schematic workflow of bright field BaseScope RNA ISH technique. The first step involves deparaffinization, target retrieval, and permeabilization with a protease, followed by hybridization with target probe (ZZ). Then a series of amplification steps are performed before signal detection with a red chromogen, visualization, and quantification of the signals using bright field microscopy or semiautomated counting of dots using digital platforms. C, Optimization and validation of bright field BaseScope RNA ISH assay for detection of FGFR2 splice isoforms in FFPE cell blocks processed from cell pellets. Ishikawa endometrial cancer cells transduced with FGFR2b (epithelial isoform; top); AN3CA endometrial cancer cells expressing endogenous FGFR2c (mesenchymal isoform; bottom). PPIB, positive control housekeeping gene showing mRNA quality; DabB, technical negative control targeting bacterial gene. Red dots indicate chromogenic mRNA signal product. FGFR2 protein IHC using nonisoform-specific antibody targeting C-terminal of FGFR2.

Figure 1.

Principles and workflow of bright field chromogenic RNA ISH, optimization and validation of BaseScope assay for detecting alternatively spliced FGFR2 isoforms. A, FGFR2 is encoded by 21 exons and has two isoforms; inclusion of exon 8 gives rise to the FGFR2IIIb isoform (designated as FGFR2b), expressed in normal epithelial cells, whereas inclusion of exon 9 gives rise to the FGFR2IIIc isoform (designated as FGFR2c), expressed in normal mesenchymal cells. FGFR2b and FGFR2c only differ in the second half of the third immunoglobulin-like loop (Ig-III) as indicated with purple and red colors, respectively. FGFR2b and FGFR2c bind to their specific ligands as noted in purple and red colors. FGF1 binds to both isoforms. B, Schematic workflow of bright field BaseScope RNA ISH technique. The first step involves deparaffinization, target retrieval, and permeabilization with a protease, followed by hybridization with target probe (ZZ). Then a series of amplification steps are performed before signal detection with a red chromogen, visualization, and quantification of the signals using bright field microscopy or semiautomated counting of dots using digital platforms. C, Optimization and validation of bright field BaseScope RNA ISH assay for detection of FGFR2 splice isoforms in FFPE cell blocks processed from cell pellets. Ishikawa endometrial cancer cells transduced with FGFR2b (epithelial isoform; top); AN3CA endometrial cancer cells expressing endogenous FGFR2c (mesenchymal isoform; bottom). PPIB, positive control housekeeping gene showing mRNA quality; DabB, technical negative control targeting bacterial gene. Red dots indicate chromogenic mRNA signal product. FGFR2 protein IHC using nonisoform-specific antibody targeting C-terminal of FGFR2.

Close modal

Using the optimized exon-junction isoform-specific chromogenic RNA ISH assay, FGFR2c was determined in FFPE clinical patient samples in two clinical cohorts. To evaluate the pattern of FGFR2c expression with respect to grade and stage, FGFR2c RNA ISH was initially analyzed in 20 whole sections of EECs obtained from Mater Pathology and two commercial TMAs (UT 242a) and (UT 801) composed of 20 and 38 endometrial cancer cores in duplicate, respectively. Representative images from grade 1 and grade 3 EECs showing high specificity without background noise are provided (Fig. 2C). In several cases where the whole section was analyzed, FGFR2c expression increased toward the myometrial invasive front of the tumor (Supplementary Fig. S2C and S2D). Eight cases were excluded from the commercial TMA cohort analysis with five cases either missing or containing only stroma and three cases showing poor RNA quality (PPIB housekeeping gene expression was negative). Our unpublished data indicated that FGFR2c expression was not observed across 10 whole sections of normal secretory endometrium in the epithelial/glandular compartment (manuscript in preparation).

Figure 2.

REMARK criteria for FGFR2c analysis in the Vancouver cohort and representative images illustrating RNA ISH scoring, FGFR2 protein, PPIB, and FGFR2c RNA ISH in well and poorly differentiated endometrioid endometrial carcinomas. A, REMARK flow chart for FGFR2c analysis in the whole Canadian cohort of endometrial cancers. B, Representative microphotographs RNA ISH scoring of FGFR2c (0, 1+, 2+, 3+, and 4+) from five different patients. C, Microphotographs showing serial sections of FGFR2 protein IHC, PPIB (RNA ISH score 2+), and FGFR2c (RNA ISH score 3+) from patients with grade 1 (top) and grade 3 (bottom) endometrioid endometrial carcinomas (EECs); PPIB, positive control targeting housekeeping gene; brown color, positive IHC staining; red dots, BaseScope signal product of individual mRNA; and nuclei are counterstained with hematoxylin.

Figure 2.

REMARK criteria for FGFR2c analysis in the Vancouver cohort and representative images illustrating RNA ISH scoring, FGFR2 protein, PPIB, and FGFR2c RNA ISH in well and poorly differentiated endometrioid endometrial carcinomas. A, REMARK flow chart for FGFR2c analysis in the whole Canadian cohort of endometrial cancers. B, Representative microphotographs RNA ISH scoring of FGFR2c (0, 1+, 2+, 3+, and 4+) from five different patients. C, Microphotographs showing serial sections of FGFR2 protein IHC, PPIB (RNA ISH score 2+), and FGFR2c (RNA ISH score 3+) from patients with grade 1 (top) and grade 3 (bottom) endometrioid endometrial carcinomas (EECs); PPIB, positive control targeting housekeeping gene; brown color, positive IHC staining; red dots, BaseScope signal product of individual mRNA; and nuclei are counterstained with hematoxylin.

Close modal

In this preliminary screening cohort, 47% of cases had a positive FGFR2c RNA ISH score (1–4). The association of FGFR2c and FGFR2 protein with various clinicopathologic features is provided in Supplementary Table S1. Univariate analysis showed FGFR2c mRNA expression in EEC was significantly associated with high-grade, deep myometrial invasion, advanced stage, and high ESMO risk group (Supplementary Table S1).

In the second cohort we analyzed expression of FGFR2c and PPIB on several endometrial cancer TMAs that has been used to develop and validate the ProMisE molecular subtype risk classifier (5, 6). The median patients' follow-up was 4.8 years and ranged from 0.019 to 30 years. From a total of 465 cases' samples analyzed, 79 cases were excluded from the statistical analyses for missing cores and/or cores contain only stroma or lack of PPIB expression (Fig. 2A). The majority of the cases (302/386; 78%) were EEC. There was a significant difference between expression of FGFR2c mRNA in EEC and non-EEC (NEEC; P < 0.0001), as well as FGFR2 protein (P < 0.022; Supplementary Fig. S1B–S1D). Roughly, 50% of patients with EEC (150/302) had positive FGFR2c mRNA expression (1–4+), whereas high FGFR2c expression (3/4+) was found in 15% of EECs. FGFR2c expression was less common in NEECs (20/84; 23%). Activating FGFR2 mutations predominantly occur in EECs (23), suggesting these tumors have a different etiology to NEEC. Hence, analysis of FGFR2c was performed independently in EEC and NEECs.

Initially, Cox regression proportional hazard model survival outcome analyses was performed for each pair of RNA ISH score 0 versus 1, 0 versus 2/3, and 0 versus 4 independently and all FGFR2c-positive scores (1–4) were associated with poor outcome (Supplementary Fig. S1E). Given the similar effect of each score on DSS, subsequent analyses were performed by dichotomizing as negative (RNA ISH score 0) and positive (RNA ISH score 1–4). In this large population cohort, positive FGFR2c (RNA ISH scores 1–4) was significantly associated with aggressive tumor characteristics (grade 3, deep myometrial invasion, and positive LVSI), ESMO risk groups, and ProMisE molecular subtypes within EECs (Table 1). In NEEC, FGFR2c was associated with deep myometrial invasion but no other clinicopathologic features (Table 1). FGFR2c expression significantly associated with histologic subtype (P < 0.0001), grade (P < 0.0001), and molecular subtypes (P < 0.002) when analyzed in the whole population (Supplementary Table S2). FGFR2 protein expression was not associated with LVSI, FIGO stage, or ProMisE molecular subtypes in EEC (Supplementary Table S2).

Table 1.

Association of demographic, clinicopathologic biomarkers, ProMisE molecular subtype, and ESMO risk group according with FGFR2c RNA ISH status in endometrioid and NEEC in the Vancouver cohort.

Histologic typeEndometrioidNonendometrioid
Clinicopathologic markersFGFR2c negative n (%)FGFR2c positive n (%)PaFGFR2c negative n (%)FGFR2c positive n (%)Pa
Age in years <60 69 (45.4) 48 (32) 0.025 11 (17.2) 4 (20) 0.492 
 ≥60 83 (54.6) 100 (66.7)  53 (82.8) 16 (80)  
 Unknown 0 (0) 2 (1.3)  — —  
BMI Kg/m2 <25 36 (23.7) 46 (30.7) 0.086 28 (43.8) 9 (45) 0.698 
 25–30 31 (20.4) 32 (21.3)  14 (21.9) 6 (30)  
 ≥30 70 (46.1) 55 (36.7)  21 (32.8) 5 (25)  
 Unknown 15 (9.9) 17 (11.3)  1 (1.6) 0 (0)  
Myometrial invasion None or <50% 99 (65.1) 87 (58) 0.034 27 (42.2) 15 (75) 0.015 
 >50% 50 (32.9) 59 (39.3)  34 (53.1) 5 (25)  
 Unknown 3 (9.7) 4 (2.7)  3 (4.7) 0 (0)  
LVSI Negative 95 (62.5) 85 (56.7) 0.021 21 (32.8) 10 (50) 0.424 
 Positive 47 (30.9) 57 (38)  38 (59.4) 9 (45)  
 Unknown 10 (6.6) 8 (5.3)  5 (7.8) 1 (5)  
FIGO grade Grade 1/2 111 (73.) 54 (36) 0.0001 1 (1.6) 1 (5) 0.442 
 Grade 3 39 (25.7) 95 (63.3)  63 (98.4) 19 (95)  
 Unknown 2 (1.3) 1 (0.7)  — —  
FIGO stage 116 (76.3) 106 (70.7) 0.148 23 (35.9) 12 (60) 0.31 
 II 8 (5.3) 16 (10.7)  4 (6.3) 1 (5)  
 III 22 (19.8) 16 (10.7)  23 (35.9) 5 (25)  
 IV 4 (2.6) 8 (5.3)  14 (21.9) 2 (10)  
 Unknown 2 (1.3) 4 (2.7)  — —  
ESMO risk groups Low 77 (66.4) 37 (24.7) 0.0001 — —  
 Intermediate 34 (22.4) 47 (31.3)  — —  
 High 39 (25.7) 65 (43.3)  64 (100) 20 (100)  
 Unknown 2 (1.3) 1 (0.7)  — —  
ProMisE molecular subtype POLE 15 (9.9) 18 (12) 0.009 3 (4.7) 0.293 
 p53wt 89 (58.6) 63 (42)  7 (10.9) 1 (5)  
 MMRd 41 (27) 46 (30.7)  7 (10.9) 3 (15)  
 p53 abn 6 (3.9) 20 (13.3)  40 (62.5) 15 (75)  
 Unknown 1 (0.7) 6 (4)  0 (0) 1 (5)  
Histologic typeEndometrioidNonendometrioid
Clinicopathologic markersFGFR2c negative n (%)FGFR2c positive n (%)PaFGFR2c negative n (%)FGFR2c positive n (%)Pa
Age in years <60 69 (45.4) 48 (32) 0.025 11 (17.2) 4 (20) 0.492 
 ≥60 83 (54.6) 100 (66.7)  53 (82.8) 16 (80)  
 Unknown 0 (0) 2 (1.3)  — —  
BMI Kg/m2 <25 36 (23.7) 46 (30.7) 0.086 28 (43.8) 9 (45) 0.698 
 25–30 31 (20.4) 32 (21.3)  14 (21.9) 6 (30)  
 ≥30 70 (46.1) 55 (36.7)  21 (32.8) 5 (25)  
 Unknown 15 (9.9) 17 (11.3)  1 (1.6) 0 (0)  
Myometrial invasion None or <50% 99 (65.1) 87 (58) 0.034 27 (42.2) 15 (75) 0.015 
 >50% 50 (32.9) 59 (39.3)  34 (53.1) 5 (25)  
 Unknown 3 (9.7) 4 (2.7)  3 (4.7) 0 (0)  
LVSI Negative 95 (62.5) 85 (56.7) 0.021 21 (32.8) 10 (50) 0.424 
 Positive 47 (30.9) 57 (38)  38 (59.4) 9 (45)  
 Unknown 10 (6.6) 8 (5.3)  5 (7.8) 1 (5)  
FIGO grade Grade 1/2 111 (73.) 54 (36) 0.0001 1 (1.6) 1 (5) 0.442 
 Grade 3 39 (25.7) 95 (63.3)  63 (98.4) 19 (95)  
 Unknown 2 (1.3) 1 (0.7)  — —  
FIGO stage 116 (76.3) 106 (70.7) 0.148 23 (35.9) 12 (60) 0.31 
 II 8 (5.3) 16 (10.7)  4 (6.3) 1 (5)  
 III 22 (19.8) 16 (10.7)  23 (35.9) 5 (25)  
 IV 4 (2.6) 8 (5.3)  14 (21.9) 2 (10)  
 Unknown 2 (1.3) 4 (2.7)  — —  
ESMO risk groups Low 77 (66.4) 37 (24.7) 0.0001 — —  
 Intermediate 34 (22.4) 47 (31.3)  — —  
 High 39 (25.7) 65 (43.3)  64 (100) 20 (100)  
 Unknown 2 (1.3) 1 (0.7)  — —  
ProMisE molecular subtype POLE 15 (9.9) 18 (12) 0.009 3 (4.7) 0.293 
 p53wt 89 (58.6) 63 (42)  7 (10.9) 1 (5)  
 MMRd 41 (27) 46 (30.7)  7 (10.9) 3 (15)  
 p53 abn 6 (3.9) 20 (13.3)  40 (62.5) 15 (75)  
 Unknown 1 (0.7) 6 (4)  0 (0) 1 (5)  

Note: P < 0.05 is indicated in bold.

Abbreviations: FGFR2c negative, FGFR2c RNA ISH score (0); FGFR2c positive, FGFR2c RNA ISH score (1–4); p53, tumor protein 53; p53 abn, null/missense p53 mutation.

aP value was determined using χ2 test

Univariate Cox proportional hazards model and Kaplan–Meier curve survival analyses within EEC revealed FGFR2c expression (RNA ISH scores 1–4) was significantly associated with decreased OS (P < 0.003), DSS (P < 0.001), and PFS (P <0.009) (Table 2; Fig. 3A). This was not evident in all cases of the cohort (EEC and NEEC combined; Supplementary Fig. S3A–S3C) or the NEEC subpopulation (Supplementary Fig. S3D–S3F). Multivariate Cox proportional hazards model analyses adjusting for age, BMI, myometrial invasion, LVSI, histologic grade, FIGO stage, ESMO risk group, and molecular subtype showed that FGFR2c expression was an independent prognostic biomarker with shorter DSS and PFS (Table 3). Notably, patients with EEC with FGFR2c expression have higher risk of endometrial cancer–specific death in univariate (HR, 2.5; 95% CI, 1.409–4.353; P < 0.001) and multivariate analyses (HR, 2.3; 95% CI, 1.27–4.496; P < 0.022).

Table 2.

Univariate Cox regression survival outcome analyses according to FGFR2c expression and clinicopathologic profiles, ProMisE molecular subtypes, and ESMO risk groups in endometrioid endometrial carcinoma in Vancouver cohort.

OSDSSPFS
Variables (Ref)Events/totalHR (95% CI)LRT PEvents/totalHR (95% CI)LRT PEvents/totalHR (95% CI)LRT P
Age >60 (age <60) 107/299 2.36 (1.51–3.70) 0.0001 53/269 1.78 (0.97–3.24) 0.061 47/278 1.049 (0.59–1.88) 0.87 
BMI >30 (BMI <30) 88/269 1.19 (0.77–1.84) 0.43 44/244 1.24 (0.69–2.24) 0.48 41/246 1.42 (0.77–2.63) 0.27 
MI ≥50% (no MI or <50%) 105/294 2.17 (1.47–3.20) 0.0001 51/264 4.50 (2.49–8.15) 0.0001 45/274 4.15 (2.25–7.64) 0.0001 
LVSI positive (LSVI negative) 99/283 1.96 (1.31–2.94) 0.001 47/264 3.67 (2.02–6.67) 0.00001 40/263 3.34 (1.76–6.34) 0.0001 
Grade 3 (grade 1–2) 108/298 1.99 (1.32–3.006) 0.001 54/268 3.699 (1.97–6.94) 0.0001 45/275 3.614 (1.821–7.17) 0.0001 
FIGO stage III/IV (FIGO stage I/II) 106/295 2.32 (1.46–3.70) 0.0001 52/265 3.97 (2.28–6.92) 0.00001 45/274 4.488 (2.46–8.20) 0.0001 
FGFR2c score (1–4) (FGFR2c score 0) 109/301 1.80 (1.22–2.64) 0.003 55/271 2.48 (1.41–4.35) 0.001 47/278 1.93 (1.07–3.48) 0.009 
Molecular subtype 107/297 — 0.006 55/268  0.004 47/274  0.039 
MMRd (p53wt)  1.81 (1.16–2.83) 0.009  2.163 (1.20–3.90) 0.01  1.868 (1.00–3.49) 0.048 
POLE (p53wt)  0.55 (0.24–1.285) 0.17  0.318 (0.08–1.35) 0.121  0.191 (0.026–1.43) 0.107 
p53 abn (p53wt)  1.74 (0.931–3.255) 0.083  2.494 (1.095–5.681) 0.03  1.933 (0.77–4.85) 0.160 
ESMO risk group 108/298 — 0.0001 54/268 — 0.0001 45/275 — 0.0001 
ESMO risk intermediate (low)  1.20 (0.714–2.149) 0.447  1.949 (0.736–5.16) 0.179  1.128 (0.34–3.737) 0.844 
ESMO risk high (low)  2.40 (1.50–3.83) 0.0001  6.47 (2.85–14.68) 0.0001  7.67 (3.18–18.48) 0.0001 
OSDSSPFS
Variables (Ref)Events/totalHR (95% CI)LRT PEvents/totalHR (95% CI)LRT PEvents/totalHR (95% CI)LRT P
Age >60 (age <60) 107/299 2.36 (1.51–3.70) 0.0001 53/269 1.78 (0.97–3.24) 0.061 47/278 1.049 (0.59–1.88) 0.87 
BMI >30 (BMI <30) 88/269 1.19 (0.77–1.84) 0.43 44/244 1.24 (0.69–2.24) 0.48 41/246 1.42 (0.77–2.63) 0.27 
MI ≥50% (no MI or <50%) 105/294 2.17 (1.47–3.20) 0.0001 51/264 4.50 (2.49–8.15) 0.0001 45/274 4.15 (2.25–7.64) 0.0001 
LVSI positive (LSVI negative) 99/283 1.96 (1.31–2.94) 0.001 47/264 3.67 (2.02–6.67) 0.00001 40/263 3.34 (1.76–6.34) 0.0001 
Grade 3 (grade 1–2) 108/298 1.99 (1.32–3.006) 0.001 54/268 3.699 (1.97–6.94) 0.0001 45/275 3.614 (1.821–7.17) 0.0001 
FIGO stage III/IV (FIGO stage I/II) 106/295 2.32 (1.46–3.70) 0.0001 52/265 3.97 (2.28–6.92) 0.00001 45/274 4.488 (2.46–8.20) 0.0001 
FGFR2c score (1–4) (FGFR2c score 0) 109/301 1.80 (1.22–2.64) 0.003 55/271 2.48 (1.41–4.35) 0.001 47/278 1.93 (1.07–3.48) 0.009 
Molecular subtype 107/297 — 0.006 55/268  0.004 47/274  0.039 
MMRd (p53wt)  1.81 (1.16–2.83) 0.009  2.163 (1.20–3.90) 0.01  1.868 (1.00–3.49) 0.048 
POLE (p53wt)  0.55 (0.24–1.285) 0.17  0.318 (0.08–1.35) 0.121  0.191 (0.026–1.43) 0.107 
p53 abn (p53wt)  1.74 (0.931–3.255) 0.083  2.494 (1.095–5.681) 0.03  1.933 (0.77–4.85) 0.160 
ESMO risk group 108/298 — 0.0001 54/268 — 0.0001 45/275 — 0.0001 
ESMO risk intermediate (low)  1.20 (0.714–2.149) 0.447  1.949 (0.736–5.16) 0.179  1.128 (0.34–3.737) 0.844 
ESMO risk high (low)  2.40 (1.50–3.83) 0.0001  6.47 (2.85–14.68) 0.0001  7.67 (3.18–18.48) 0.0001 

Note: P < 0.05 is indicated in bold.

Abbreviations: LRTP, LRT probability; MI, myometrial invasion; p53, tumor protein 53; p53 abn, null/missense p53 mutation; Ref, reference.

Figure 3.

Kaplan–Meier curve survival outcome analyses according to the FGFR2c status stratified by histologic type and ProMisE molecular subtypes in Vancouver cohort. Analysis restricted to patients with EEC (A), MMRd molecular subtype (B), and p53wt ProMisE molecular subtype (C). Negative, FGFR2c RNA ISH score (0) and positive, FGFR2c RNA score (1–4) FGFR2c, FGFR 2c isoform; LRTP, LRT P value.

Figure 3.

Kaplan–Meier curve survival outcome analyses according to the FGFR2c status stratified by histologic type and ProMisE molecular subtypes in Vancouver cohort. Analysis restricted to patients with EEC (A), MMRd molecular subtype (B), and p53wt ProMisE molecular subtype (C). Negative, FGFR2c RNA ISH score (0) and positive, FGFR2c RNA score (1–4) FGFR2c, FGFR 2c isoform; LRTP, LRT P value.

Close modal
Table 3.

Multivariate Cox regression survival outcome analyses of FGFR2c adjusted for cofounding clinicopathologic parameters and ProMisE molecular subtypes in EEC in the Vancouver cohort.

OS in EEC, events (80/252)
Variables (Ref)HR (95% CI)LRT P
Age >60 (age <60) 2.25 (1.402–3.6) 0.001 
LVSI positive (LSVI negative) 1.72 (1.12–2.66) 0.014 
Myo invasion >50% (no Myo invasion or <50%) 1.67 (1.084–2.57) 0.02 
FGFR2c RNA ISH score (1–4) (FGFR2c RNA ISH score 0) 1.43 (0.94–2.17) 0.092 
Molecular subtype  0.031 
MMRd (p53wt) 1.44 (0.867–2.384 0.159 
POLE (p53wt) 0.30 (0.09–0.99) 0.048 
P53 abn (p53wt) 1.79 (0.82–3.88) 0.143 
DSS in EES, events (45/248) 
Myometrial invasion >50% (no Myo invasion or <50%) 3.23 (1.65–6.33) 0.001 
Grade 3 (grade 1–2) 2.317 (1.11–4.77) 0.023 
FIGO stage III/IV (I/II) 2.625 (1.38–5.03) 0.004 
FGFR2c RNA ISH score (1–4) (FGFR2c RNA ISH score 0) 2.25 (1.13–4.50) 0.022 
Molecular subtype  0.02 
MMRd (p53wt) 1.90 (0.95–3.78) 0.068 
POLE (p53wt) 0.23 (0.051–1.01) 0.052 
p53 abn (p53wt) 0.98 (0.319–3.01) 0.972 
PFS in EEC events (40/258) 
Myo invasion >50% (no Myo invasion or <50%) 3.31 (1.65–6.63) 0.001 
FIGO stage III/IV (I/II) 2.69 (1.34–5.39) 0.005 
FGFR2c RNA ISH score (1–4) (FGFR2c RNA ISH score 0) 2.01 (1.05–3.83) 0.035 
OS in EEC, events (80/252)
Variables (Ref)HR (95% CI)LRT P
Age >60 (age <60) 2.25 (1.402–3.6) 0.001 
LVSI positive (LSVI negative) 1.72 (1.12–2.66) 0.014 
Myo invasion >50% (no Myo invasion or <50%) 1.67 (1.084–2.57) 0.02 
FGFR2c RNA ISH score (1–4) (FGFR2c RNA ISH score 0) 1.43 (0.94–2.17) 0.092 
Molecular subtype  0.031 
MMRd (p53wt) 1.44 (0.867–2.384 0.159 
POLE (p53wt) 0.30 (0.09–0.99) 0.048 
P53 abn (p53wt) 1.79 (0.82–3.88) 0.143 
DSS in EES, events (45/248) 
Myometrial invasion >50% (no Myo invasion or <50%) 3.23 (1.65–6.33) 0.001 
Grade 3 (grade 1–2) 2.317 (1.11–4.77) 0.023 
FIGO stage III/IV (I/II) 2.625 (1.38–5.03) 0.004 
FGFR2c RNA ISH score (1–4) (FGFR2c RNA ISH score 0) 2.25 (1.13–4.50) 0.022 
Molecular subtype  0.02 
MMRd (p53wt) 1.90 (0.95–3.78) 0.068 
POLE (p53wt) 0.23 (0.051–1.01) 0.052 
p53 abn (p53wt) 0.98 (0.319–3.01) 0.972 
PFS in EEC events (40/258) 
Myo invasion >50% (no Myo invasion or <50%) 3.31 (1.65–6.63) 0.001 
FIGO stage III/IV (I/II) 2.69 (1.34–5.39) 0.005 
FGFR2c RNA ISH score (1–4) (FGFR2c RNA ISH score 0) 2.01 (1.05–3.83) 0.035 

Note: P < 0.05 is indicated in bold.

Abbreviations: LRT P, LRT probability; Myo, myometrial; p53, tumor protein 53; p53 abn, null/missense TP53 mutation; Ref, reference.

Cox regression survival outcome analyses were also performed by further stratifying the EECs by FIGO grade (grade 1/2 vs. grade 3) and ESMO risk (low/intermediate vs. high) to assess the prognostic significance of FGFR2c (Fig. 4). Notably, FGFR2c expression was significantly associated with shorter PFS (P < 0.0017) and DSS (P < 0.0023) within the FIGO grade 3 EEC. While FGFR2c expression in grade 1/grade 2 EECs was significantly associated with shorter PFS (P < 0.045), this was not statistically significant for DSS. The prognostic significance of FGFR2c was most evident in the ESMO high-risk patients where FGFR2c-positive tumors were associated with shorter PFS (P < 0.001) and DSS (P < 0.002). In the low- and intermediate-risk patients, FGFR2c showed a trend toward shorter PFS and DSS. When similar analyses were performed in the total cohort without stratifying by histologic type, FGFR2c lost its prognostic power (Supplementary Tables S3 and S4), revealing FGFR2c is prognostic only in EECs.

Figure 4.

Kaplan–Meier curve survival outcome analyses according to the FGFR2c isoform expression stratified by FIGO grade and ESMO risk group in EEC. Survival outcome analyses in patients with EEC grade 1/2 (A), grade 3 (B), low–intermediate ESMO risk group (C), and high ESMO risk group (D).

Figure 4.

Kaplan–Meier curve survival outcome analyses according to the FGFR2c isoform expression stratified by FIGO grade and ESMO risk group in EEC. Survival outcome analyses in patients with EEC grade 1/2 (A), grade 3 (B), low–intermediate ESMO risk group (C), and high ESMO risk group (D).

Close modal

Notably, FGFR2c expression was an independent prognostic biomarker that stratifies those with poorer outcomes within the MMRd and p53wt subtypes (Fig. 3B and C). Kaplan–Meier curve survival analyses indicated patients with endometrial cancer with FGFR2c expression had reduced OS (P < 0.049), DSS (P < 0.001), and PFS (P < 0.047) within the MMRd molecular subtype (Fig. 3B). Similarly, patients with p53wt endometrial cancer with FGFR2c expression had shorter OS (P < 0.046) and DSS (P < 0.049) and showed a trend toward lower PFS (P < 0.050; Fig. 3C). We also noted that patients with EEC within the p53-mutant subtype with negative FGFR2c (n = 6/23) had a favorable prognosis with 100% DSS, P < 0.038 and PFS, P < 0.07 (Supplementary Fig. S4G and S4H).

To compare the discriminatory power of FGFR2c expression to the traditional clinicopathologic prognostic markers (grade, FIGO stage, and ESMO risk group) and determine whether discriminatory prognostic outcome increased when combined, we assessed the AUC ROC. When analyses were restricted to EEC the AUC ROC for FGFR2c was OS, 0.603, 95% CI, 0.536–0.670, P < 0.003; DSS, 0.618, 95% CI, 0.546–0.70, P < 0.007; and PFS, 0.591, 95% CI, 0.502–0.681, P < 0.048 (Supplementary Fig. S5A–S5D). FGFR2c did not have any discriminatory prognostic power when the whole cohort was analyzed (Supplementary Figs. S3A–S3C and S5E) or in the NEEC subpopulation (Supplementary Fig. S3D–S3F). AUC ROC analyses showed that FGFR2c had similar predictive power to FIGO grade, FIGO stage, and ESMO risk groups to discriminate OS, DSS, and PFS and integration of FGFR2c with grade, stage, and ESMO risk groups improved the discriminatory prognostic outcome (Supplementary Fig. S5D). Notably, FGFR2 protein did not show any prognostic power highlighting the benefit of analyzing specific mRNA isoforms (Supplementary Fig. S5A–S5C, green line). Finally, we propose a schematic flow chart on how FGFR2c RNA ISH could be incorporated into the ProMisE algorithm in clinical practice (Supplementary Fig. S5F).

Finally, we have analyzed the publicly available TCGA RNA-seq data to determine FGFR2c expression in 386 EEC samples using computational exon–exon junctions counts following STAR method approach (see Supplementary Materials and Methods; ref. 24) to validate the above results. As RNA-seq contains reads from both the tumor and stroma, we used a conservative cutoff of 1.5× more junction reads mapping to exon 9 (FGFR2c) versus exon 8 (FGFR2b). With this cutoff, only 7% of the cases were found to predominantly express FGFR2c, significantly lower than the frequency documented in the screening and Vancouver cohorts. This difference indicates RNA-seq is not as sensitive as BaseScope RNA ISH for determining FGFR2c expression. Nevertheless, both univariate and multivariate Cox regression analyses showed that patients with EEC expressing high levels of FGFR2c had shorter PFS (HR, 3.458; 95% CI, 1.75–6.82; LRT P < 0.015) and OS (HR, 2.9; 95% CI, 1.12–7.62, LRT P < 0.029), respectively (Supplementary Table S6; Supplementary Fig. S6). FGFR2c expression was independent of FGFR2 mutation (Supplementary Table S5) and similar to the Vancouver cohort was significantly associated with FIGO grade 3 EEC.

Endometrial cancer is a heterogeneous and complex disease with wide spectrum of morphologic and molecular types that have different prognostic outcomes. In current practice, risk stratification is carried out on the basis of traditional pathologic parameters including histologic types, LVSI, grade, and stage. These parameters are subjected to interobserver variability. The lack of reproducibility in pathologic risk stratification potentially results in endometrial cancer patients' undertreatment or overtreatment and contributes to large disparities in treatment outcomes (25). The new TCGA molecular subtyping that has been amended and validated by multiple independent groups hold promise in stratifying patients with endometrial cancer. Although, the new molecular subtyping identifies POLE EDM (with excellent prognosis) and the p53 abnormal/serous-like (with worst prognosis), approximately 80% of all endometrial cancers, predominantly endometrioid histotypes, fall into the MSI/MMRd and p53wt/NSMP cohorts (with intermediate prognoses). These two common molecular subgroups need further stratification with more reliable biomarkers.

In this study, we investigated the prognostic value of the mesenchymal FGFR2c isoform in the two Vancouver cohorts previously used for establishing the “ProMisE” molecular subtypes. Our findings have demonstrated that FGFR2c expression is significantly associated with reduced PFS and DSS in EECs in both univariate and multivariate analyses. This landmark finding reveals FGFR2c expression discerns the outcomes within grade 3 EECs (presumably to have poor clinical outcome) and those classified as ESMO high-risk group (Fig. 4). Our findings are consistent with the finding that FGFR2-activating mutations predominantly occur and are associated with poor outcome in the most common endometrioid histologic subtype (11, 23). Detailed FGFR2 sequencing is not available for the ProMisE cohort, however, computational analyses of RNA-seq data from the endometrial cancer TCGA cohort demonstrated FGFR2 mutations and isoform switching are mutually exclusive (Supplementary Table S5). This indicates FGFR2 can be activated in EEC either by mutation or isoform switching and supports the evaluation of targeting FGFR2 in patients with EEC with FGFR2 activation by either mechanism.

When we stratified by molecular subtypes, FGFR2c showed significant prognostic power in the MMRd subtype (Fig. 3B) and to a lesser extent in the p53wt subtype of endometrial cancers. While previous investigations have reported L1-cell adhesion molecule (L1CAM) expression in MMRd endometrial cancers was associated with poor outcome (26), L1CAM expression occurs in only 7% of women in this subtype. In contrast, we observed FGFR2c expression in approximately 50% of MMRd cases showing FGFR2c expression may be a more informative risk stratification biomarker. Within the PORTEC1/2 patient cohort, β-catenin mutation/nuclear β-catenin expression and 1q32.1 amplification lack prognostic significance in the MMRd subtype, however, they were significantly associated with shorter relapse-free survival and cancer-specific survival in the p53wt/NSMP subtype (intermediate risk group; refs. 27, 28). Several studies reported β-catenin mutations are associated with shorter PFS in FIGO grade 1/2 early-stage EEC (28–30). In this study, patients with grade 1/2 or low–intermediate risk EEC with FGFR2c expression showed a trend toward reduced PFS and DSS. As the Vancouver cohort does not have β-catenin mutation data for all patients, the relation between Wnt/β-catenin pathway activation and FGFR2c expression was not assessed. Although FGFR2 mutations have been reported to occur significantly less often in endometrial cancers with β-catenin mutations (28), we have not seen this in our previous study (31). Hence, the functional connection between β-catenin mutations and FGFR2c expression needs further investigation.

This is the first reliable report that showed FGFR2c expression was significantly associated with shorter PFS and DSS revealing this biomarker could be used to identify those patients with poorer outcomes in the MSI/MMRd subtype. Assessment of FGFR2c expression in the high-risk PORTEC3 cohort where multiple different prognostic biomarkers are also being evaluated will allow us to assess whether a combination of FGFR2c, L1CAM, β-catenin mutation/nuclear β-catenin expression, and 1q32.1 amplification can further stratify these MMRd and p53wt molecular subtypes.

We noted that FGFR2c mRNA is less expressed and has no significant prognostic value in the NEECs characterized by p53 mutations, suggesting FGFR2 alternative splicing does not contribute to the etiology of these tumors. Similarly, FGFR2c showed no prognostic significance in the POLE EDM despite 50% in this subtype expressing FGFR2c. We suggest that the high mutational burden and neoantigen expression in the POLE EDM subtype and resulting immunogenicity compensates for any poor prognosis associated with expression of FGFR2c.

In other cancer types, in vitro, in vivo, and clinical studies have reported isoform switching from FGFR2b to FGFR2c in tumorigenesis and cancer progression (14, 15, 32). In vitro studies have shown that ectopic expression of FGFR2c in several different cell lines increases cell proliferation (33) as well as migration and invasion (34). FGFR2c was reported to drive epithelial–mesenchymal transition (EMT) in breast cancer cells (35) and recently, in epithelial keratinocytes (34) and also inhibits cell differentiation (36). Overexpression of FGFR2 assessed by IHC using nonisoform-specific antibodies has been associated with poor prognosis in several cancers including breast (37), gastric (38), and pancreatic (39). This is the first reliable report using novel chromogenic BaseScope RNA ISH that the FGFR2c isoform is associated with unfavorable tumor characteristics (grade 3, deep myometrial invasion, and LSVI) and poor outcomes in EEC. This is possibly explained by the fact that activation of FGFR2c drives pathologic type III EMT (34). Our findings contrast with those previously reported in a small panel of 47 endometrial cancers using IHC who reported FGFR2c is more common in grade 1/2 EECs (40). This study used a supposed FGFR2c isoform–specific antibody, however, the specificity of this antibody for FGFR2c has only ever been provided through Western blotting of denatured proteins and data did not exclude cross-reactivity with FGFR1c.

Although, RNA-seq has revolutionized our understanding of cancer, it is most often performed in bulk tumor specimens where the temporal and spatial expression of FGFR2c cannot be concluded given normal expression of FGFR2c in the periglandular stroma. While TCGA RNA-seq data analyses confirmed our findings, a very low number of cases met our conservative cutoff showing the superiority of the BaseScope RNA ISH assay, specifically that it allows the semiquantification of mRNA FGFR2c within the in situ morphologic context. Chromogenic BaseScope RNA ISH in addition to being a robust, sensitive, and specific assay, which can detect alternatively spliced isoforms (41, 42), has the advantage of detecting small fragments of RNA (as small as 20 bps). This feature makes BaseScope RNA ISH suitable for detection of FGFR2c mRNA in old archival FFPE tissues where mRNA degradation has potentially occurred.

On the basis of these findings, FGFR2c could be integrated into the new ProMisE molecular classification decision tree algorithm (ref. 6; Supplementary Fig. S5F). We propose that after initial screening of patients with endometrial cancer for MMRd, FGFR2c RNA ISH should be performed to separate MMRd patients into favorable (FGFR2c negative) and unfavorable (FGFR2c positive). Following p53 mutation screening, we are proposing FGFR2c could be performed in those tumors identified as p53wt. In this way, FGFR2c determination could be avoided in POLE EDM and p53-mutant subtypes as it had no role in risk stratification in these subtypes. It should be noted that within this ProMisE cohort there was a small number of EECs with p53 mutations and the six cases that were shown to be FGFR2c negative had no evidence of recurrence/progression. This finding warrants further validation in a larger cohort of p53 abnormal EEC to see whether this holds true with larger numbers.

In the era of precision medicine, molecular risk stratification in endometrial cancer is a promising approach in the clinic. The Dutch PORTEC1/2 and Vancouver cohort studies demonstrated that incorporating the new molecular subtyping in risk stratification improved the ESMO risk grouping (5, 6, 9). Similarly, we have incorporated FGFR2c to the traditional clinicopathologic markers/ESMO risk grouping to assess whether FGFR2c improves risk stratification and better discriminates prognostic outcome (OS, DSS, and PFS) in EEC. Interestingly, FGFR2c expression when combined with grade, stage, and ESMO risk group showed an improvement in discriminating outcome. This suggests integration of FGFR2c with the clinicopathologic parameters and/or molecular classifiers offers more effective risk stratification of patients with endometrial cancer.

The new FGFR2c incorporated algorithm of molecular risk stratification could potentially be used to determine the optimal adjuvant therapy. For example, it could be used to identify those higher risk patients in the intermediate risk group to treat with pelvic external beam radiotherapy versus vaginal brachytherapy or radio-chemotherapy versus radiotherapy alone in the high-risk group. Another possible application of these risk stratification biomarkers is to assess patients with endometrial cancer on preoperative diagnostic biopsies to identify FGFR2c-positive tumors with the MMRd and p53wt subtypes which could undergo either sentinel lymph node staging or more extensive lymph node staging in settings where guidelines to perform full surgical staging are not in place. Preclinical in vitro and in vivo studies are underway to confirm FGFR2c expression as a therapeutic target. We hope that the increased proportion of endometrial cancer cases that could benefit from FGFR inhibition may drive interest in FGFR combination trials combining immune checkpoint inhibition and FGFR inhibition in MMRd patients with FGFR2 activation.

In conclusion, expression of the mesenchymal FGFR2c isoform appears to be an independent prognostic biomarker that is significantly associated with shorter PFS and DSS in EEC. Specifically, FGFR2c expression identifies a subset of patients with poorer outcomes within grade 3 EECs, and those classified as ESMO high-risk patients. FGFR2c expression also discerns the outcomes within the MMRd and p53wt molecular subtypes. Addition of FGFR2c to emerging molecular classification strategies could significantly improve risk stratification of EEC. Future studies include evaluation of FGFR2c expression as a predictive biomarker in clinical trials evaluating FGFR inhibition alone or in combination with other anticancer agents both in the metastatic and adjuvant settings.

C.E. Snell is an employee/paid consultant for Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.T. Sengal, E.D. Williams, P.M. Pollock

Development of methodology: A.T. Sengal, E.D. Williams

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.T. Sengal, C.E. Snell, D.S. Smith, S. Leung, P.M. Pollock, A. Talhouk, J.N. McAlpine, P.M. Pollock

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.T. Sengal, A.-M. Patch, D.S. Smith, S. Leung, E.D. Williams, P.M. Pollock

Writing, review, and/or revision of the manuscript: A.T. Sengal, A.-M. Patch, C.E. Snell, D.S. Smith, A. Talhouk, E.D. Williams, J.N. McAlpine, P.M. Pollock

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.T. Sengal

Study supervision: C.E. Snell, E.D. Williams, P.M. Pollock

We would like to acknowledge the funding body Cancer Australia (1087165) and Queensland University of Technology (QUT) in supporting A.T. Sengal through QUTPRA scholarship, as well as QUT/School of Biomedical Sciences to P.M. Pollock. Authors would like to thank all the patients who provided their samples. We are also grateful for professor Sabine Werner's laboratory for providing the murine skin blocks for anti-FGFR2 antibody optimization as well as the IHBI/QUT histology core and Translational Research Institute of Australia microscopy facility for whole slide scanning. The Translational Research Institute is supported by a grant (APP108382) from the Australian Government.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2018
.
CA Cancer J Clin
2018
;
66
:
7
30
.
2.
American Cancer Society
. 
Cancer facts and figures 2016
.
Atlanta, GA
:
American Cancer Society
; 
2017
.
3.
Colombo
N
,
Preti
E
,
Landoni
F
,
Carinelli
S
,
Colombo
A
,
Marini
C
, et al
Endometrial cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up
.
Ann Oncol
2013
;
24
:
vi33
8
.
4.
The Cancer Genome Atlas
. 
Integrated genomic characterization of endometrial carcinoma
.
Nature
2013
;
497
:
67
73
.
5.
Talhouk
A
,
McConechy
MK
,
Leung
S
,
Li-Chang
HH
,
Kwon
JS
,
Melnyk
N
, et al
A clinically applicable molecular-based classification for endometrial cancers
.
Br J Cancer
2015
;
113
:
299
310
.
6.
Talhouk
A
,
McConechy
MK
,
Leung
S
,
Yang
W
,
Lum
A
,
Senz
J
, et al
Confirmation of ProMisE: a simple, genomics-based clinical classifier for endometrial cancer
.
Cancer
2017
;
123
:
802
13
.
7.
Kommoss
S
,
McConechy
MK
,
Kommoss
F
,
Leung
S
,
Bunz
A
,
Magrill
J
, et al
Final validation of the ProMisE molecular classifier for endometrial carcinoma in a large population-based case series
.
Ann Oncol
2018
;
29
:
1180
8
.
8.
Stelloo
E
,
Bosse
T
,
Nout
RA
,
MacKay
HJ
,
Church
DN
,
Nijman
HW
, et al
Refining prognosis and identifying targetable pathways for high-risk endometrial cancer; a TransPORTEC initiative
.
Mod Pathol
2015
;
28
:
836
44
.
9.
Stelloo
E
,
Nout
RA
,
Osse
EM
,
Jurgenliemk-Schulz
IJ
,
Jobsen
JJ
,
Lutgens
LC
, et al
Improved risk assessment by integrating molecular and clinicopathological factors in early-stage endometrial cancer-combined analysis of the PORTEC cohorts
.
Clin Cancer Res
2016
;
22
:
4215
24
.
10.
Cosgrove
CM
,
Tritchler
DL
,
Cohn
DE
,
Mutch
DG
,
Rush
CM
,
Lankes
HA
, et al
An NRG Oncology/GOG study of molecular classification for risk prediction in endometrioid endometrial cancer
.
Gynecol Oncol
2018
;
148
:
174
80
.
11.
Jeske
YW
,
Ali
S
,
Byron
SA
,
Gao
F
,
Mannel
RS
,
Ghebre
RG
, et al
FGFR2 mutations are associated with poor outcomes in endometrioid endometrial cancer: an NRG oncology/gynecologic oncology group study
.
Gynecol Oncol
2017
;
145
:
366
73
.
12.
Turner
N
,
Grose
R
. 
Fibroblast growth factor signalling: from development to cancer
.
Nat Rev Cancer
2010
;
10
:
116
29
.
13.
Oltean
S
,
Sorg
BS
,
Albrecht
T
,
Bonano
VI
,
Brazas
RM
,
Dewhirst
MW
, et al
Alternative inclusion of fibroblast growth factor receptor 2 exon IIIc in Dunning prostate tumors reveals unexpected epithelial mesenchymal plasticity
.
Proc Natl Acad Sci U S A
2006
;
103
:
14116
21
.
14.
Kawase
R
,
Ishiwata
T
,
Matsuda
Y
,
Onda
M
,
Kudo
M
,
Takeshita
T
, et al
Expression of fibroblast growth factor receptor 2 IIIc in human uterine cervical intraepithelial neoplasia and cervical cancer
.
Int J Oncol
2010
;
36
:
331
40
.
15.
Matsuda
Y
,
Ueda
J
,
Ishiwata
T
. 
Fibroblast growth factor receptor 2: expression, roles, and potential as a novel molecular target for colorectal cancer
.
Patholog Res Int
2012
;
2012
:
574768
.
16.
Yu
K
,
Herr
AB
,
Waksman
G
,
Ornitz
DM
. 
Loss of fibroblast growth factor receptor 2 ligand-binding specificity in Apert syndrome
.
Proc Natl Acad Sci U S A
2000
;
97
:
14536
41
.
17.
FIGO Committee on Gynaecologic Oncology
. 
Revised FIGO staging for carcinoma of endometrium
.
Int J Gynaecol Obstet
2009
;
105
:
103
4
.
18.
Meyer
M
,
Muller
AK
,
Yang
J
,
Moik
D
,
Ponzio
G
,
Ornitz
DM
, et al
FGF receptors 1 and 2 are key regulators of keratinocyte migration in vitro and in wounded skin
.
J Cell Sci
2012
;
125
:
5690
701
.
19.
Tannheimer
SL
,
Rehemtulla
A
,
Ethier
SP
. 
Characterization of fibroblast growth factor receptor 2 overexpression in the human breast cancer cell line SUM-52PE
.
Breast Cancer Res
2000
;
2
:
311
20
.
20.
Gatius
S
,
Velasco
A
,
Azueta
A
,
Santacana
M
,
Pallares
J
,
Valls
J
, et al
FGFR2 alterations in endometrial carcinoma
.
Mod Pathol
2011
;
24
:
1500
10
.
21.
Hongwei
W
,
Nan
S
,
Li-Chong
W
,
Xingyong
W
,
Son
B
,
Allissa
N
, et al
Quantitative ultrasensitive bright-field RNA in situ hybridization with RNAscope
. In:
Walker
JM
, editor.
In situ hybridization protocols. Methods in molecular biology (methods and protocols)
.
New York, NY
:
Humana Press
; 
2014
. p.
201
12
.
22.
Stehbens
SJ
,
Ju
RJ
,
Adams
MN
,
Perry
SR
,
Haass
NK
,
Bryant
DM
, et al
FGFR2-activating mutations disrupt cell polarity to potentiate migration and invasion in endometrial cancer cell models
.
J Cell Sci
2018
;
131
:
jcs213678
.
23.
Pollock
PM
,
Gartside
MG
,
Dejeza
LC
,
Powell
MA
,
Mallon
MA
,
Davies
H
, et al
Frequent activating FGFR2 mutations in endometrial carcinomas parallel germline mutations associated with craniosynostosis and skeletal dysplasia syndromes
.
Oncogene
2007
;
26
:
7158
62
.
24.
Dobin
A
,
Davis
CA
,
Schlesinger
F
,
Drenkow
J
,
Zaleski
C
,
Jha
S
, et al
STAR: ultrafast universal RNA-seq aligner
.
Bioinformatics
2012
;
29
:
15
21
.
25.
Nout
RA
,
Smit
VT
,
Putter
H
,
Jurgenliemk-Schulz
IM
,
Jobsen
JJ
,
Lutgens
LC
, et al
Vaginal brachytherapy versus pelvic external beam radiotherapy for patients with endometrial cancer of high-intermediate risk (PORTEC-2): an open-label, non-inferiority, randomised trial
.
Lancet
2010
;
375
:
816
23
.
26.
Karnezis
AN
,
Leung
S
,
Magrill
J
,
McConechy
MK
,
Yang
W
,
Chow
C
, et al
Evaluation of endometrial carcinoma prognostic immunohistochemistry markers in the context of molecular classification
.
J Pathol Clin Res
2017
;
3
:
279
93
.
27.
Depreeuw
J
,
Stelloo
E
,
Osse
EM
,
Creutzberg
CL
,
Nout
RA
,
Moisse
M
, et al
Amplification of 1q32.1 refines the molecular classification of endometrial carcinoma
.
Clin Cancer Res
2017
;
23
:
7232
41
.
28.
Kurnit
KC
,
Kim
GN
,
Fellman
BM
,
Urbauer
DL
,
Mills
GB
,
Zhang
W
, et al
CTNNB1 (beta-catenin) mutation identifies low grade, early stage endometrial cancer patients at increased risk of recurrence
.
Mod Pathol
2017
;
30
:
1032
41
.
29.
Myers
A
,
Barry
WT
,
Hirsch
MS
,
Matulonis
U
,
Lee
L
. 
β-Catenin mutations in recurrent FIGO IA grade I endometrioid endometrial cancers
.
Gynecol Oncol
2014
;
134
:
426
7
.
30.
Imboden
S
,
Tapia
C
,
Scheiwiller
N
,
Kocbek
V
,
Altermatt
HJ
,
Janzen
J
, et al
Early-stage endometrial cancer, CTNNB1 mutations, and the relation between lymphovascular space invasion and recurrence
.
Acta Obstet Gynecol Scand
2020
;
99
:
196
203
.
31.
Byron
SA
,
Gartside
M
,
Powell
MA
,
Wellens
CL
,
Gao
F
,
Mutch
DG
, et al
Fgfr2 point mutations in 466 endometrioid endometrial tumors: relationship with MSI, KRAS, PIK3CA, CTNNB1 mutations and clinicopathological features
.
PLoS One
2012
;
7
:
e30801
.
32.
Zhao
Q
,
Caballero
OL
,
Davis
ID
,
Jonasch
E
,
Tamboli
P
,
Yung
WK
, et al
Tumor-specific isoform switch of the fibroblast growth factor receptor 2 underlies the mesenchymal and malignant phenotypes of clear cell renal cell carcinomas
.
Clin Cancer Res
2013
;
19
:
2460
72
.
33.
Ishiwata
T
,
Matsuda
Y
,
Yamamoto
T
,
Uchida
E
,
Korc
M
,
Naito
Z
. 
Enhanced expression of fibroblast growth factor receptor 2 IIIc promotes human pancreatic cancer cell proliferation
.
Am J Pathol
2012
;
180
:
1928
41
.
34.
Ranieri
D
,
Rosato
B
,
Nanni
M
,
Magenta
A
,
Belleudi
F
,
Torrisi
MR
. 
Expression of the FGFR2 mesenchymal splicing variant in epithelial cells drives epithelial-mesenchymal transition
.
Oncotarget
2016
;
7
:
5440
60
.
35.
Warzecha
CC
,
Sato
TK
,
Nabet
B
,
Hogenesch
JB
,
Carstens
RP
. 
ESRP1 and ESRP2 are epithelial cell-type-specific regulators of FGFR2 splicing
.
Mol Cell
2009
;
33
:
591
601
.
36.
Ranieri
D
,
Rosato
B
,
Nanni
M
,
Belleudi
F
,
Torrisi
MR
. 
Expression of the FGFR2c mesenchymal splicing variant in human keratinocytes inhibits differentiation and promotes invasion
.
Mol Carcinog
2018
;
57
:
272
83
.
37.
Sun
S
,
Jiang
Y
,
Zhang
G
,
Song
H
,
Zhang
X
,
Zhang
Y
, et al
Increased expression of fibroblastic growth factor receptor 2 is correlated with poor prognosis in patients with breast cancer
.
J Surg Oncol
2012
;
105
:
773
9
.
38.
Ahn
S
,
Lee
J
,
Hong
M
,
Kim
ST
,
Park
SH
,
Choi
MG
, et al
FGFR2 in gastric cancer: protein overexpression predicts gene amplification and high H-index predicts poor survival
.
Mod Pathol
2016
;
29
:
1095
103
.
39.
Nomura
S
,
Yoshitomi
H
,
Takano
S
,
Shida
T
,
Kobayashi
S
,
Ohtsuka
M
, et al
FGF10/FGFR2 signal induces cell migration and invasion in pancreatic cancer
.
Br J Cancer
2008
;
99
:
305
13
.
40.
Peng
WX
,
Kudo
M
,
Fujii
T
,
Teduka
K
,
Naito
Z
. 
Altered expression of fibroblast growth factor receptor 2 isoform IIIc: relevance to endometrioid adenocarcinoma carcinogenesis and histological differentiation
.
Int J Clin Exp Pathol
2014
;
7
:
1069
76
.
41.
Harms
U
,
Li
N
,
Rouault
M
,
Wilkens
K
,
Monroe
R
,
Ma
X-J
, et al
In-situ detection and characterization of EGFR and EGFRvIII expression heterogeneity in glioblastoma FFPE tissues
.
Ann Oncol
2017
;
28
:
mdx511.010
.
42.
Zhu
Y
,
Sharp
A
,
Anderson
CM
,
Silberstein
JL
,
Taylor
M
,
Lu
C
, et al
Novel junction-specific and quantifiable in situ detection of AR-V7 and its clinical correlates in metastatic castration-resistant prostate cancer
.
Eur Urol
2018
;
73
:
727
35
.