This retrospective study included 196 patients with EC who underwent surgical resection at our hospital between January 2022 and December 2024. The patients did not receive neoadjuvant chemotherapy or radiotherapy prior to surgery. Formalin-fixed paraffin-embedded (FFPE) tumor samples were retrieved from the Pathology Department Biobank. Two patients were excluded because of sequencing failure or insufficient tumor tissue for DNA extraction, resulting in a final sample size of 194 patients with EC.

Two pathologists performed pathological classification, tumor grading, assessment of myometrial invasion, and lymphovascular space invasion according to the WHO guidelines [5]. Clinical staging was based on the International Federation of Gynecology and Obstetrics staging system (2023) [6]. This study was approved by the Ethics Committee of our hospital, and written informed consent was obtained from all participants.

Each FFPE tissue sample was sectioned into 3.5 µm slices for IHC staining using the LUMATAS automatic pathological staining system. The primary antibodies used were from Fuzhou Maixin Biotech (Anti-MLH1, clone MX063; Anti-PMS2, clone MX073; Anti-MSH2, clone MX061; Anti-MSH6, clone MX056; Anti-p53, clone MX008). All histological and IHC slides were independently reviewed by two pathologists.

Normal protein expression of MLH1, PMS2, MSH2, and MSH6 was defined as normal nuclear expression. Loss of expression was defined as the complete loss of nuclear expression, whereas internal controls (stromal or lymphocytic cells) exhibited strong nuclear staining. Based on the MMR protein expression status, tumors were categorized into two groups: deficient mismatch repair (dMMR) if any of the MMR proteins (MLH1, PMS2, MSH2, and MSH6) were lost, and proficient mismatch repair (pMMR) if all MMR proteins were positively expressed.

p53 expression patterns were classified as overexpression (OE), complete absence (CA), cytoplasmic expression (CY), or wild-type (WT), with OE, CA, and CY constituting abnormal p53 expression patterns. Tumors with p53 abnormalities (p53abn), if ≥75% of tumor cell nuclei exhibited strong or diffuse staining (OE), were completely negative (CA), or showed cytoplasmic staining (CY) [13, 14].

Abnormal subclonal expression was considered to occur in p53 and MMR proteins, if abnormal staining patterns were detected in areas adjacent to the tumor.

Paraffin-embedded tumor samples selected by pathological examination were subjected to DNA extraction using the MagPure FFPE DNA LQ Kit (Magen), according to the manufacturer’s protocol. Genomic DNA concentration was quantified using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific).

POLE gene mutations were detected using Sanger sequencing, as previously described [15]. Exons 9–14 of the POLE gene were amplified using primers provided by Sangon Biotech (Shanghai) Co., Ltd. Only 11 previously identified pathogenic variants (P286R, V411L, S297F, S459F, A456P, F367S, L424I, M295R, P436R, M444K, and D368Y) were targeted [16]. PCR products were confirmed by gel electrophoresis, purified using the MinElute PCR Purification Kit (QIAGEN) according to the manufacturer’s instructions, and subsequently subjected to sequencing using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems). Sequencing reactions were purified and analyzed using a SeqStudio Genetic Analyzer (Applied Biosystems). All POLE mutations were annotated according to the LRG_789 (NM_006231.3) and Human Genome Variation Society (HGVS) guidelines.

Targeted DNA NGS was performed on 131 samples using the clinically validated AmoyDx® Comprehensive Panel (Amoy Diagnostics, Xiamen, China), which comprises 116 cancer-related genes, including 109 genes for single nucleotide variants (SNVs), insertions/deletions (indels), 31 genes for copy number variations (CNVs), and 12 genes for gene fusions. The panel covers mutations in POLE, TP53, PTEN, PIK3CA, ARID1A, KRAS, HER2, BRCA1, BRCA2, and SMARCA4 and includes microsatellite instability (MSI) analysis. Genes and their detection regions are listed in Supplementary Table S1. MSI status was determined based on the percentage of unstable loci out of 55 microsatellite loci: samples with <15% unstable loci were classified as microsatellite stable (MSS), and those with ≥15% were classified as MSI-high (MSI-H). A commercially available targeted AmoyDx EC panel covering POLE, TP53, and MSI was used for 63 samples in this study (Amoy Diagnostics, Xiamen, China).

Sequencing was performed on an Illumina NextSeq 500 platform (Illumina) with the following quality criteria: Q30 base percentage ≥75%, coverage ≥95%, CoverageRatioUNIQ180 (proportion of hotspot regions with effective sequencing depth ≥180×) ≥95%, cnv_cv (coefficient of variation for sequencing depth within gene regions) <0.4, and cnv_uni (degree of deviation from the panel of normals) <1.5 (see Supplementary Tables S2, S3). The mutation allele frequency (MAF) thresholds for mutation identification by NGS were set as follows: SNVs and insertions/deletions (indels) ≥5% and CNVs ≥4 copies. Sequencing data were analyzed using bioinformatics software, and the therapeutic relevance of the identified genetic variants was interpreted based on publicly available databases, including the NCCN Guidelines and OncoKB [17]. The evidence levels for variant-drug associations were categorized into four tiers (A-D) according to the AMP/ASCO/CAP guidelines [18]: Tier A (FDA-approved or supported by professional clinical guidelines), Tier B (validated by large-scale clinical studies with expert consensus), Tier C (supported by Tier A evidence in other cancer types, used as inclusion criteria in clinical trials, or supported by multiple small-scale studies), and Tier D (preclinical studies or case reports).

Bisulfite conversion of purified DNA was performed using the EpiTect Fast DNA Bisulfite Kit (Qiagen), following the manufacturer’s protocol. PCR products were analyzed using 3% agarose gel electrophoresis to verify successful conversion and amplification (Supplementary Figure S1).

Molecular classification was performed according to the WHO guidelines [5]. Patients with one of the 11 confirmed pathogenic POLE variants detected by Sanger sequencing were categorized as POLEmut. The remaining EC cases were classified based on MMR status, with tumors showing loss of one or more MMR proteins designated as MMRd, and those with intact MMR protein expression classified as pMMR. Subsequently, EC cases were further stratified based on p53 status, where abnormal p53 expression patterns (OE, CA, and CY) were categorized as p53abn, and normal p53 expression (WT) was categorized as NSMP.

For NGS-based classification, patients with one of the 11 confirmed pathogenic POLE variants identified by the NGS panel were grouped as NGS POLEmut. The remaining EC cases were classified based on MSI status, with MSI-H cases categorized as NGS MSI-H/MMRd, and MSS cases further stratified based on TP53 mutation status. Tumors harboring TP53 mutations were designated NGS TP53mut/p53abn, whereas those without TP53 mutations were classified as NGS NSMP/TP53wt.

The consistency between Sanger + IHC and NGS molecular classifications was evaluated using overall accuracy and the kappa coefficient. Associations between molecular subgroups and clinicopathological characteristics were assessed using appropriate statistical tests, including the χ² test for categorical variables and analysis of variance (ANOVA) for continuous variables. All statistical analyses were two-tailed, with a significance set at P < 0.05. Data analysis was performed using the IBM SPSS Statistics software (version 19.0; IBM Corp., Armonk, NY, USA).

A total of 194 endometrial carcinoma specimens were successfully to molecular classification using NGS. The clinical and pathological characteristics of the patients are shown in Table 1. Based on the NGS molecular classification, 15.5% (30/194) of the cases were categorized as NGS POLEmut, 23.2% (45/194) as NGS MSI-H/MMRd, 14.4% (28/194) as NGS TP53mut/p53abn, and 46.9% (91/194) as NGS NSMP/TP53wt. Notably, the NGS classification identified 10.3% (20/194) of cases with multiple molecular features. Among these, ten NGS POLEmut tumors harbored TP53 mutations, with one case exhibiting three distinct molecular features (POLEmut, MSI-H, and TP53mut), and ten NGS MSI-H/MMRd tumors concurrently displayed TP53 mutations.

In comparison, the Sanger + IHC molecular classification assigned 11.9% (23/194) of cases to the POLEmut category, 23.7% (46/194) to MMRd, 13.9% (27/194) to p53abn, and 50.5% (98/194) to NSMP. The Sanger + IHC classification identified ten cases with overlapping molecular characteristics: five POLEmut tumors showed abnormal p53 expression or deficient mismatch repair (dMMR), and five MMRd tumors concurrently exhibited abnormal p53 expression.

Figure 1 presents the molecular marker results for both Sanger + IHC and NGS classifications. A comparative analysis revealed that 93.8% (182/194) of the cases showed concordance between the NGS and Sanger + IHC molecular classifications, achieving a kappa coefficient of 0.908 and an overall accuracy of 0.938, (Figures 2, 3). However, there were 12 discordant cases between the two classification methods (Table 2). Specifically, cases 1 and 2 were classified as NSMP by Sanger + IHC but as NGS TP53mut/p53abn, with TP53 variants S260Qfs*3 and G245S detected at a MAF of 2.4% and 15%, respectively. Both cases had tumor cell contents exceeding 50%, suggesting tumor heterogeneity, with a low proportion of TP53-mutated cells. Cases 3 and 4 were classified as MMRd by Sanger + IHC but as NGS NSMP/TP53wt, with no variants detected in the MSH2, MSH6, MLH1, PMS2, or EPCAM genes. Additionally, seven cases with POLE mutations were missed by Sanger sequencing because of their lower sensitivity than NGS, particularly for variants with a MAF <10%. After excluding the seven POLEmut cases that were missed by Sanger sequencing, the concordance between NGS and Sanger + IHC molecular classifications increased to 97.4% (189/194), with a kappa value of 0.962.

As shown in Figure 4, NGS identified 30 POLE gene mutations, including P286R in 15 cases, V411L in 11 cases, A456P in two cases, and single instances of S297F and P436S. In contrast, Sanger sequencing detected only 23 POLE mutations, excluding seven cases with allelic frequencies <10% (Supplementary Table S4). These seven cases, which had low tumor content (20%), underwent tumor region delineation and enrichment based on hematoxylin and eosin (HE) staining prior to DNA extraction and subsequent Sanger sequencing. Remarkably, all seven cases were confirmed to harbor POLE mutations (Figure 5).

IHC for MLH1, MSH2, MSH6, and PMS2 proteins, along with PCR-based MSI analysis, are the standard methods for assessing MSI phenotypes in EC molecular classification. The NGS-based MSI status, determined by analyzing the read distribution across 55 microsatellite loci, was compared with the IHC-based MMR status. The results demonstrated high concordance, with 97.9% (190/194) of cases being consistent between the two methods. Four discordant cases were observed: three IHC-diagnosed dMMR cases were classified as MSS by NGS, and one IHC-diagnosed pMMR case was classified as MSI-H by NGS. These three cases exhibited concurrent loss of MLH1 and PMS2 expression and MLH1 promoter hypermethylation (Supplementary Figure S1). Additionally, two cases displayed partial loss of MMR protein expression: one with a 50% loss of MLH1/PMS2, and the other with a 20% loss of MLH1/PMS2, indicating a subclonal staining pattern (Supplementary Figure S2), and both were positive for MLH1 promoter methylation.

Among the 194 patients, 45 had MSI-H. Owing to tumor-only testing, the germline mutation status remained uncertain. Although the NGS panel also detected MLH1, PMS2, MSH2, and MSH6 gene variants, no pathogenic MMR-related gene mutations were identified in the dMMR tumors. No mutations in EPCAM were detected in the 194 patients. Germline testing for Lynch syndrome is typically recommended. However, no mutations were detected in this cohort.

Both NGS and IHC were used to determine p53 status in all EC cases in this study. Abnormal p53 status was defined by IHC as ≥75% of tumor cell nuclei exhibiting strong or diffuse staining, complete absence of nuclear staining, or cytoplasmic staining. Three cases had p53 expression levels near the threshold; 60%, 65%, and 70% of tumor cells showed moderate positive expression, and were thus classified as p53 wild-type (WT) according to the interpretation criteria.

NGS detected at least one potentially pathogenic TP53 mutation in 48 of the 194 cases, including variants such as V73Wfs*50, C141Y, R181C, R213*, C238Y, S241F, G245S, R248W, R248Q, T253A, L257R, ES260Qfs*3, R267W, R273C, R273H, P278S, R282W, and E285Gfs*20. Additionally, two novel TP53 variants (R202_V203del and V157_M160del) were identified using NGS, both of which were associated with mutant p53 expression levels.

Figure 4 highlights that 18 cases with TP53 mutations detected by NGS were classified as p53 wild-type by IHC, resulting in a moderate concordance with a kappa coefficient of 0.688 and an overall concordance rate of 89.7% (174/194 cases). When POLEmut and MSI-H/MMRd cases were excluded, in which TP53 mutations were often considered passenger mutations that did not influence the classification, the consistency between IHC-based p53 staining and NGS-based TP53 mutation detection improved significantly. Among the remaining 119 cases, only three showed discordances, leading to an overall concordance rate of 97.5% (116/119 cases), as presented in Table 3.

Subclonal p53 expression (defined as mutant-pattern staining confined to discrete tumor foci occupying <10% of the total tumor area) was observed in 10/194 (5.2%) cases (Supplementary Figure S3). TP53 mutations were detected in 5/10 cases, all of which met the NGS detection threshold (MAF≥5%). Among these 10 tumors, five were molecularly classified as either mismatch repair deficient (MMRd, n = 1) or polymerase epsilon-mutated (POLEmut, n = 4). Notably, all five MMRd/POLEmut tumors predominantly exhibited wild-type p53 expression patterns with only focal mutant-pattern staining. TP53 mutations were identified in 4/5 cases, suggesting potential discordance between molecular and immunohistochemical (IHC) results due to either low tumor cellularity in sampled regions or subclonal mutations below the IHC detection threshold.

Forty-two mutations associated with targeted therapy trials were identified in 131 samples (Figure 6). The most frequently mutated genes were PTEN (76.3%), PIK3CA (50.4%), and ARID1A (35.9%). Beyond the primary molecular markers POLE, TP53, and MSI, additional mutations with frequencies exceeding 10% included KRAS (25.2%) and CTNNB1 (17.6%).

The distribution of specific gene mutations varied significantly across the FIGO stages (Table 4). For example, PTEN mutations decreased from 86.8% in FIGO G1 to 59.3% in FIGO G3 (P < 0.05), whereas PIK3CA mutations decreased from 60.5% in G1 to 29.6% in G3 (P < 0.05). Conversely, TP53 mutations increased markedly from 2.6% in G1 to 59.3% in G3 (P < 0.001) and HER2 mutations increased from 0% in G1 to 11.4% in G3 (P < 0.05).

Age-related differences were also notable; patients aged ≥58 years exhibited significantly higher frequencies of TP53 mutations (28.8% vs. 15.5%, P > 0.05), CTNNB1 mutations (20.5% vs13.8%, P > 0.05), and HER2 mutations (4.1% vs. 1.7%, P > 0.05) than those aged <58 years.

In advanced EC (FIGO Stage III-IV), mutations in PTEN (60.9% vs. 79.8% in early stages, P < 0.05), PIK3CA (43.5% vs. 51.9%, P > 0.05), ARID1A (21.7% vs. 39.4%, P > 0.05), CTNNB1 (8.7% vs. 20.2%, P > 0.05), FBXW7 (0% vs. 12.5%, P > 0.05), and FGFR2 (4.3% vs. 11.5%, P > 0.05) were significantly less frequent than in early-stage disease. Conversely, HER2 mutations were significantly more frequent in patients with advanced EC (13.0% vs. 1.0%, P < 0.01).

Histological subtype-specific findings revealed that all four clear-cell EC cases harbored TP53 mutations, as did both serous EC cases, with one also exhibiting HER2 amplification, and the other presenting a BRCA1 mutation alongside MYC amplification. Among the four dedifferentiated EC cases, one had POLEmut, one had dMMR, one had p53abn, and one had NSMP with concurrent HER2 and NF1 mutations. Additionally, one case of mixed carcinoma (serous and sarcoma) was classified as a dMMR.