Morphomolecular Correlation and Clinicopathologic Analysis in Endometrial Carcinoma.

Approval for the study was obtained from the Non-Interventional Research Ethics Committee of Dokuz Eylül University with the decision numbered 2021/06-23, and dated February 22, 2021. Our study included all cases that were diagnosed with endometrial carcinoma in our department between 2016 and 2020 based on the histopathologic examination of hysterectomy specimens, containing sufficient tumor material for POLE mutation analysis to be performed, stained with MLH1, MSH6, MSH2, PMS2, p53 immunohistochemistry of the surgical specimens and with fully accessible follow-up information. A total of 122 cases were included in the study after the cases that had not met the above-mentioned study criteria or could not be evaluated optimally due to tissue follow-up/fixation artifacts were excluded from the study. Before determining the molecular subgroups, histopathologic findings in hematoxylin eosin–stained tissue sections of radical hysterectomy materials of all cases were evaluated by one professor experienced in gynecopathology and one pathology assistant under a light microscope. The guidelines of the “International Immuno-Oncology Biomarker Working Group” were taken into account in the evaluation of tumor-infiltrating lymphocytes (TIL). The presence of TILs between 40 and 76/10 high power field was recorded as mild, and the presence of TILs >76/10 high power field was recorded as distinct 10 . The evaluation of myometrial invasion was based on the article published by Cole et al. 11 Among other histopathologic features; the presence of marked desmoplasia in almost every area of the tumor examined was accepted as distinctive desmoplasia, and the identification of even only one lymphoid follicle in the myometrium, regardless of the tumor was taken into consideration. The absence and presence of cytomorphologic features such as tumor giant cells, coarse chromatin, cherry-red nucleolus, atypical mitosis, and psammomatous calcification were also recorded. Even if only one cell with these features was seen throughout the entire tumor, this feature was accepted as “present.” When pleomorphism filling one high power field is seen, this feature was again accepted as “present.” In addition, as a result of the re-examination of the cases, morphologic features that were not routinely evaluated but thought to be significant were also tried to be determined. To identify the POLE mutant group, genomic DNA was isolated from formalin-fixed paraffin-embedded tumor tissue specimens containing more than 30% tumor and <30% necrotic area retrieved from hysterectomy specimens. Following the quality control and purification procedures, next-generation sequencing was performed according to standard protocols after adding the appropriate primers. Fastq files obtained from Dokuz Eylül University Oncology Institute of the samples amplified with NEXTflexTM Universal Amplicon Prep Kit V2 and sequenced using Illumina MiniSeq system were uploaded to SEQ platform which is a cloud-based variant analysis platform of Genomize Inc., and processed using its analysis version 15.4.3. Analysis of somatic mutations in exons 9, 13, and 14 of the POLE gene was performed using the SEQ analysis platform and manual review. With the SEQ analysis platform, 100× coverage were targeted across all amplicons. Four cases were excluded because not all of the target regions could be read for technical reasons. Cases with any of the 5 mutations identified as hotspots (P286R, V411L, S297F, A456P, and S459F) or other rare pathogenic mutations were identified as POLE mutants. The pathogenic POLE variants identified so far in endometrial carcinomas are summarized in Table 1 12 – 15 . Pathogenic POLE variants in endometrial carcinomas 12 – 15 To identify the MMR-D group, 4 routinely applied immunohistochemical markers MLH1, MSH6, MSH2, and PMS2 were re-evaluated in all cases. Microsatellite instability (MSI) analysis results of real-time polymerase chain reaction 5 mononucleotide panels (BAT25, BAT26, NR21, NR24, and MONO27) of specimens of 63 cases containing tumor and tumor-free normal tissue were evaluated based on the data obtained from the reports. Immunohistochemical staining patterns were classified as “nuclear expression is not lost,” “nuclear expression is completely lost,” “nuclear expression is lost, subclonal”; MSI status was classified as “microsatellite stable,” “microsatellite unstable, low” and “microsatellite unstable, high.” Subclonal MMR staining has been defined as a sudden and sharp transition in tumor cells with a complete loss of nuclear expression with one of the MMR proteins (there must be stromal positivity as an internal control) 16 . After excluding POLE mutant cases, cases with loss of expression in any of the 4 biomarkers or molecular instability in one of the 5 regions mentioned above based on immunohistochemical analysis were included in the MMR-D/MSI hypermutant group. Cases with inconsistent results of MMR immunohistochemistry and MSI were also included in this group. Because MMR-D cases with this atypical phenotype, which have been detected based on the results of both immunohistochemical and molecular analyses in the literature, are frequently seen in noncolorectal carcinoma cancers and associated with genetic syndromes at a higher rate 17 . To identify the p53-abnormal group, immunohistochemically stained slides for p53 in routine practice were re-evaluated in all cases. These cases were classified as “wild type,” “overexpression,” “null,” “cytoplasmic,” and “subclonal” according to their staining patterns. Subclonal p53 staining has been defined as a combination of more than one staining pattern (classically wild type + one or more mutant staining patterns; or 2 different mutant staining patterns) (each pattern must be present in at least 5% of tumor cells) 18 . All staining patterns except for the wild type staining pattern were evaluated as mutant staining and after excluding POLE ultramutant and MSI hypermutant groups, they were included in the p53-abnormal group. Other cases that did not fall into any of the 3 groups described above were included in the p53 wild type group. The age, follow-up/survival times, and metastatic disease status of the cases were accessed through the hospital information system and these parameters were evaluated. The relationship between molecular subgroups of endometrial carcinomas and histopathologic and clinical prognostic markers was evaluated statistically. Statistical analyses of the data obtained during the study were evaluated using the Statistical Package for the Social Sciences (SPSS) 24.0 package program and summarized with SD, median, minimum-maximum, and percentage distribution. χ 2 test or Fisher exact test was used in the evaluation of numerical data. A logistic regression model was created for multivariate analysis. Kaplan-Meier test was used for survival analysis. Log-rank test was used for survival curves. A Cox regression model was also created in the multiple-life analysis of survival. The level of statistical significance was accepted as P <0.05.

The median age of the patients was 61.5 yrs and 81% of the cases demonstrated endometrioid morphology. Nuclear and histologic grading were evaluated only in cases with endometrioid and mucinous carcinomas, while 54% and 88% of the cases had histologic grade 1 and nuclear grade 2 endometrial carcinomas, respectively. Most of the cases (62%) had stage I disease. Intratumoral and stromal lymphocytes were observed in 54% and 68% of the cases, respectively. Characteristics of 118 patients are summarized in Table 2 . Characteristics of the study population Histologic and nuclear grading was performed only for cases of endometrioid, and mucinous carcinomas. A total of 118 cases with endometrial carcinoma included in the study were classified as POLE mutant (n=4: 3.4%), MMR-D (n=31: 26.3%), p53 mutant (n=22: 18.6%), and NSMP (n=61: 51.7%) groups. Mutations in the POLE gene were detected in 22 of the 118 cases included in the study. Of these 22 patients, 18 carried the nonpathogenic POLE mutation, while only 4 had the most common P286R mutation identified as a pathogen. In addition, 21 unspecified pathogenic mutations were detected in 18 cases, except for P286R. The MMR, MSI, and p53 statuses and nucleotide changes of the cases with pathogenic or nonpathogenic POLE mutations are summarized in Table 3 . MMR, MSI, and p53 status of patients with pathogenic or nonpathogenic POLE mutation MMR indicates mismatch repair; MSI, microsatellite instability; MSI-H, microsatellite instability, high; MSS, microsatellite stable; wt, wild type. Thirty-one of the 118 cases included in the study were in the MMR-D group. MMR immunohistochemistry and MSI status of these cases are summarized in Table 4 . Mismatc repair immunohistochemistry and MSI results of mismatch repair-deficient cases MSI indicates microsatellite instability; MSS, microsatellite stable. Twenty-four (20.3%) of 118 cases in the study had a mutant-type p53 staining pattern. Since 1 of 7 cases with the subclonal staining pattern was in the POLE mutant group and 1 was in the MMR-D group, a total of 22 cases were classified in the p53 mutant group. The data on the p53 staining patterns in the p53 mutant group and the POLE-MMR status of the cases with subclonal p53 staining pattern are summarized in Tables 5 and 6 , respectively. P53 staining patterns in the p53 mutant group Cases demonstrating subclonal staining pattern MMR-D indicates mismatch repair-deficient; POLE, polymerase epsilon. The relationship between molecular subgroups, demographic data, and histopathologic features is summarized in Table 7 . The relationship between molecular subgroups with demographic, and histopathologic features Histologic and nuclear grading was performed only for cases of endometrioid, and mucinous carcinomas (4 cases in the POLE mutant, 28 cases in the MMR-D, 6 cases in the p53 mutant, and 59 cases in the NSMP groups). Macroscopic characteristics of the tumor could not be evaluated optimally in 24 cases. Macroscopic location of the tumor could not be evaluated optimally in 15 cases. MMR-D indicates mismatch repair-deficient; NSMP, no specific molecular profile; POLE, polymerase epsilon. All of the cases in the POLE mutant group consisted of endometrioid histologic type. All cases had intratumoral and stromal lymphocytes. All of the tumors were located in the fundus/corpus. Three of the cases were stage I–II and 1 case was stage IIICI. While lymph node involvement was not observed in 3 of the cases, pelvic lymph node involvement was present in 1 case. Statistical significance was found between the POLE mutant group and lymph node involvement ( P =0.049). Lymphovascular space invasion (LVSI) was present in 3 of the cases. All cases had myometrial invasion in the form of irregular infiltrating glands. The presence of giant cells was observed in 1, pleomorphism in 1, and smudged chromatin in 1 case. The presence of cherry-red nucleoli was observed in 3 cases, and the presence of coarse chromatin in 3 cases. Atypical mitosis was not observed in any case. Cases with mutations other than P286R were not associated with any histopathologic parameter. Due to the low number of cases, no statistical significance was found in other parameters except lymph node involvement in the POLE mutant group. The characteristics of the cases in the POLE mutant group are summarized in Figures 1 – 2 and Table 8 . Polymerase epsilon mutant case 22. (A) High-grade ambiguous morphology, grade 3 [hematoxylin eosin (H&E) stain, ×400)]. (B) High-grade, serous like nuclear features, grade 3 (H&E stain, ×200). (C) Prominent tumor-infiltrating and peritumoral lymphocytes (H&E stain, ×200). (D) Tertiary lymphoid structures within the myometrium (H&E stain, ×200). (E) An area containing papillary and solid morphology (H&E stain, ×200). (F) Subclonal p53 staining (wild type and overexpression) in the area pictured in (E) (p53 immunohistochemistry, ×400). Polymerase epsilon mutant case 33. (A) Low-grade endometrioid morphology, grade 1 [hematoxylin eosin (H&E) stain, ×100]. (B) Low-grade nuclear features, grade 2 (H&E stain, ×200). (C) Mild tumor-infiltrating and peritumoral lymphocytes (H&E stain, ×200). POLE mutant case 56. (D) Low-grade endometrioid morphology, grade 2 (H&E stain, ×200). (E) Low-grade nuclear features, grade 2 (H&E stain, ×400). (F) Prominent tumor-infiltrating lymphocytes (H&E stain, ×400). (G) Prominent peritumoral lymphocytes (H&E stain, ×200). Polymerase epsilon mutant case 55. (H) Low-grade endometrioid morphology, grade 1 (H&E stain, ×200). (I) Low-grade nuclear features, grade 2 (H&E stain, ×200). (J) Mild tumor-infiltrating and peritumoral lymphocytes (H&E stain, ×200). Demographic, histopathologic, and clinical characteristics of the cases in the POLE mutant group Ctx indicates chemotherapy; H, histologic; LN, lymph node; LVI, lymphovascular space invasion; Met, metastasis; N, nuclear; Rt, radiotherapy. In the MMR-D group, the most frequently endometrioid histologic subtype (87.1%), and then dedifferentiated carcinoma (6.5%) were seen. Histologic and nuclear grading was evaluated in endometrioid and mucinous carcinomas (28 cases in total) within the MMR-D group. A statistically significant relationship was found between the MMR-D group and the histologic grade ( P =0.019). Generally, cases with histologic grades 1 to 2 were seen, grade 3 cases were observed at a lesser rate. A statistically significant correlation was found between the MMR-D group and the number of intratumoral and stromal lymphocytes ( P <0.001 and <0.001, respectively). Intratumoral lymphocytes were prominent in 17 (54.8%), and mildly apparent in 10 (32.3%) cases; stromal lymphocytes were prominent in 22 (71%) and mildly apparent stromal in 8 (25.8%) cases. A statistical significance was found between the microcystic, elongated and fragmented (MELF) invasion pattern and the MMR-D group ( P =0.007). Among all cases, 7 of the MELF invasion patterns seen in a total of 11 cases were in the MMR-D group. The incidence of pleomorphism and cherry-red nucleoli was at a low level in the MMR-D group, and statistical significance was found between them ( P =0.042 and 0.030). While pleomorphism was not detected in 28 cases (90.3%), there was pleomorphism in 3 cases (9.7%). Cherry-red nucleoli were observed in only 4 cases (12.9%), and there were no cherry-red nucleoli in 27 cases (87.1%). Twelve cases (54.5%) in the p53 mutant group had serous morphology and the relationship with the histologic subtype was statistically significant ( P <0.001). All of the carcinosarcomas were in the p53 mutant group. Histologic and nuclear grading was evaluated in 6 cases with endometrioid carcinomas within the p53 mutant group. Both histologic and nuclear grades were significantly higher in the p53 mutant group with a statistically significant difference between them ( P <0.001 and <0.001, respectively). This group generally consisted of advanced-stage patients, and the relationship between the stage and p53 mutant group was statistically significant ( P =0.003). Of the 4 patients with stage IVB, 3 were in the p53 mutant group. Among the cytomorphologic features, a statistically significant correlation was detected between the p53 mutant group and the presence of tumor giant cells ( P <0.001), pleomorphism ( P <0.001), atypical mitosis ( P <0.001), cherry-red nucleolus ( P <0.001), smudged chromatin ( P <0.001), and coarse chromatin ( P <0.001). These cytomorphologic features were most common in the p53 mutant group. In the NSMP group, 59 cases (96.7%) had endometrioid morphology, and the relationship with histologic subtype was statistically significant ( P <0.001). The NSMP group had a lower histologic and nuclear grade, and a statistically significant correlation was found between these parameters ( P =0.002 and 0.005, respectively). In this group most commonly histologic grade 1, and least frequently histologic grade 3 were observed. Nuclear grade 3 was detected in only 1 patient (1.7%). A statistically significant difference was found between stage I/II tumors and tumors in other stages ( P =0.017). Stage I and II tumors constituted 91.8% of this group. Intratumoral and stromal lymphocytes were observed at a statistically significantly lowest rate in the NSMP group ( P <0.001 and 0.001, respectively). A statistically significant correlation was found between the NSMP group and lymph node involvement ( P =0.034). Only 1 patient (1.6%) had lymph node involvement. A statistically significant relationship was found between the NSMP group and LVSI ( P =0.003). Most of the cases (n=49, 80.3%) did not have LVSI. Among the cytomorphologic features, the presence of giant cells ( P =0.009), pleomorphism ( P <0.001), atypical mitosis ( P =0.005), smudged chromatin ( P <0.001), and coarse chromatin ( P <0.001) were statistically significantly associated with the NSMP group. These cytomorphologic features were less frequently seen in the NSMP group than in the other groups. Herein, the relationships of molecular subgroups with mortality rate, presence of metastases, and overall and disease-free survival were examined. The correlation between mortality rate and molecular subgroups was statistically significant ( P =0.038). No death was observed in any of the cases in the POLE mutant group. The presence of the tumor in the p53 mutant group ( P =0.012) increased the mortality rate. There was no significant difference between the POLE mutant and MMR-D groups and the presence of metastasis. The presence of the tumor in the p53 mutant group was statistically significantly associated with the risk of metastasis as well as the mortality rate ( P =0.005). The presence of the tumor in the NSMP group was a condition that reduced the risk of metastasis, and statistical significance was found between them ( P =0.02). The relationship of molecular subgroups with survival and metastasis is summarized in the table below (Table 9 ). The relationship of molecular subgroups with survival rates and metastases MMR-D indicates mismatch repair-deficient; NSMP, no specific molecular profile; POLE, polymerase epsilon. In survival analyses, statistical significance was found between molecular groups and survival times (log rank, P =0.005). In our series, none of the 4 cases in the POLE mutant group died. The expected survival time of the MMR-D group was 70.88±3.38 mo, and the expected survival time of the NSMP group was 71.59±2.63 mo. There was no significant difference between NSMP, MMR-D, and POLE mutant groups in terms of survival times. The expected survival time was statistically lower in the p53 mutant group (72.54 vs. 57.44 mo, log rank, P <0.001). In addition, no death was observed in any of the cases with subclonal p53 staining. The survival of these cases was better than the cases showing other abnormal staining patterns (overexpression, null or cytoplasmic) or even wild-type staining, and a statistically significant difference was found between them (log-rank, P <0.001) (Fig. 3 ). Kaplan-Meier survival curves. (A) Overall survival time in molecular groups. (B) Overall survival time in the p53 mutant group. (C) Overall survival time of the cases according to p53 staining patterns. A statistically significant difference was found between disease-free survival and molecular groups (log-rank, P <0.001). Due to the small number of cases, the disease-free survival time of the POLE mutant group was 49.50 mo, and 71.82 mo in the MMR-D group. There was no significant difference between the POLE mutant and MMR-D groups in terms of disease-free survival times. The expected disease-free survival time was statistically lower in the p53 mutant group (75.40 vs. 58.67 mo, log-rank, P <0.001). In the p53 wild type (NSMP) group, the expected disease-free survival time was statistically longer (76.90 vs. 67.85 mo, log-rank, P =0.012) (Fig. 4 ). Kaplan-Meier disease-free survival curves. (A) Disease-free survival time in molecular groups. (B) Disease-free survival time in the p53 mutant group. (C) Disease-free survival time in the no specific molecular profile (NSMP) group. In the logistic regression model created to predict the cases in the MMR-D group; the absence of the cherry-red nucleoli in the tumor was an independent indicator of this group (odds ratio: 0.07, 95% CI: 0.005–0.984, P =0.049). In the logistic regression model created to predict the cases in the NSMP group; the absence of distinct stromal lymphocytes in the tumor was found to be an independent predictor alone in predicting this group (odds ratio: 0.54, 95% CI: 0.003–0.865, P =0.039). Any independent determinant was not detected in multivariate analysis to predict the cases in the POLE mutant and p53 mutant groups. In our study, we categorized 118 patients with endometrial carcinoma according to the ESGO/ESTRO/ESP guidelines for therapeutic purposes based on their stage, histologic subtype, grade, and the presence/absence of lymphovascular invasion, regardless of their molecular classification, into low (n=43), intermediate (n=28), high-intermediate (n=28), high (n=22) risk, and advanced metastatic (n=4) groups. We reclassified all cases according to the molecular subtypes of endometrial carcinoma into risk groups, only the risk group of 3 (2.5%) cases changed. Among these 3 cases, the detection of a p53 mutation resulted in an increase in the risk group for 2 cases; in the remaining case, the presence of POLE mutation reduced the risk group. In addition, the risk group of 2 cases could not be determined according to molecular classification. Because, in the current ESGO/ESTRO/ESP guidelines, insufficient data are available to allocate patients with stage III–IVA POLE mutant endometrial carcinoma and stage I–IVA MMR-D or NSMP clear cell carcinoma with myometrial invasion into a prognostic risk group based on the molecular classification. The change between the 2 risk groups and the characteristics of these 5 cases are summarized in Tables 10 and 11 . According to the ESGO/ESTRO/ESP guidelines the change between 2 risk groups Characteristics of the cases that changed the classification of risk group LVSI indicates lymphovascular space invasion; NSMP, no specific molecular profile; POLE, polymerase epsilon. In addition, we narrowed down the study population by identifying patients for whom molecular classification could alter risk groups. Because of the molecular classification would not contribute to the management of patients in the FIGO stage III and IV categories, we excluded 20 patients classified in these categories. In addition, patients with serous histologic subtype and carcinosarcoma of any stage do not likely benefit from p53 and MMR testing, as they will be treated aggressively regardless of the molecular subtypes revealed. Therefore, we excluded 8 more patients with serous histologic subtype (n=7) and carcinosarcoma (n=1) from the remaining 98 patients. We statistically analyzed the morphologic parameters of the patients in the p53 mutant (n=6) and MMRD groups (n=25) among 90 patients. However, since detection of POLE mutation would not change the molecular classification of the cases already in the low-risk group, we reallocated the remaining 98 patients into risk groups and excluded low-risk patients (n=43) from the categorization. As stated in the ESGO/ESTRO/ESP guidelines, all stage I and II cases with POLE mutant endometrial carcinoma were in the low-risk group, so we did not exclude patients with serous subtype and carcinosarcoma. We found that only 1 out of a total of 55 patients in the intermediate, high-intermediate, and high-risk groups had POLE mutation. The new algorithm created by narrowing the patient population is summarized in Figure 5 . New algorithm for identifying patients for whom molecular classification could alter risk groups. (A) Algorithm to determine the patient population to which p53 and mismatch repair immunohistochemistry will be applied. (B) Algorithm to determine the patient population to which polymerase epsilon mutation analysis will be applied. We reanalyzed morphologic parameters in the MMR-D/MSI group (n=25) and, compared our new results with the characteristics of the MMR-D group (n=31) which was before excluding patients. Histologic grade, intratumoral/stromal TIL, and MELF invasion pattern, which were previously found to be statistically significant for this group (see the section MMR-D/MSI group), continued to maintain their significance after cases with serous histologic subtype, carcinosarcoma and advanced stage of the disease were excluded. On the other hand, no statistical significance was found between the parameters that were previously found to be statistically significant, such as the absence of the cherry-red nucleoli and pleomorphism. After excluding cases with the advanced-stage disease, serous subtype, and carcinosarcoma, the largest decrease in the number of cases was seen in the p53 mutant group (n=6). Histologic and nuclear grade, presence of tumor giant cells, pleomorphism, atypical mitosis, cherry-red nucleoli, smudged chromatin, and coarse chromatin which were previously found to be statistically significant for p53 mutant group and mentioned under the section p53 Mutant group, continued to maintain their significance. After excluding cases with serous subtype and carcinosarcoma, any statistical significance was not detected between the histologic subtype and p53 mutant group, and the presence of any other morphologic parameter did not reach the level of statistical significance. To narrow down the group of patients who could benefit from POLE mutation analysis, we excluded patients in the low-risk group, in addition to patients in the advanced stage. Since only 1 of the remaining 55 cases had a POLE mutation, statistical analysis was not performed for the POLE mutant group. After narrowing down the patient population, the logistic regression model created to foresee the cases in the molecular groups could not identify any independent predictive marker. In addition to investigating morphologic parameters that can replace molecular groups, we also investigated the relationship between morphologic features and prognosis in the histologic subtypes of endometrioid, mucinous, and dedifferentiated carcinoma, in the remaining 90 cases. The relationship between the MELF pattern of invasion and mortality rate was statistically significant ( P =0.008). Fifty percent of the deceased patients had the MELF invasion pattern. A statistically significant relationship was found between the presence of metastasis and histologic grade ( P =0.032). There were no metastatic cases among those with a histologic grade 1. The presence of LVSI was found to be associated with the presence of metastasis ( P =0.011). All cases with metastatic carcinoma had LVSI. The relationship between the MELF pattern of invasion and metastasis was also statistically significant ( P =0.02). MELF pattern of invasion was observed in 66.7% of the patients with metastatic carcinoma. In the survival analyses, the overall survival time of the cases where the MELF pattern of invasion was observed was statistically significantly shorter than that of the cases without this invasion pattern, (51.74 vs. 76.26 mo, log-rank, P <0.001). Besides, a statistically significant relationship was found between the disease-free survival time and MELF invasion pattern. The disease-free survival of the cases where the MELF pattern of invasion was observed was lower than that of the cases where it was not observed, (60.48 vs. 78.21 mo, log-rank, P <0.001) (Fig. 6 ). Cases with LVSI had a statistically significantly shorter disease-free survival time than those without (log-rank, P =0.001). Kaplan-Meier survival curves. (A) Overall survival curve for microcystic, elongated and fragmented (MELF) invasion pattern. (D) Disease-free survival curve for MELF invasion pattern. In the Cox regression model created for multivariate survival analysis, the presence of the MELF pattern of invasion was identified as an independent prognostic factor that reduces overall survival time (hazard ratio: 16.395, 95% CI: 2.140–125.606, P =0.007). For disease-free survival, any independent prognostic factor could not be identified.

Molecular data are increasing day by day for endometrial carcinomas, which are currently classified mainly on the basis of their morphologic features. As a result of studies to increase the clinical applicability of molecular classification, 4 groups reflecting TCGA subgroups have been defined: POLE mutant, MMR-D, p53-abnormal, and NSMP. In ongoing studies, it has been determined and confirmed that these groups show both histopathologic and prognostic differences 19 – 21 . Consequently, as a reflection of the increasing data, new treatment regimens have been started to be recommended, so we performed this study in order to determine the histopathologic and prognostic features of the cases in our series according to the ProMisE classification. The incidence of the POLE mutant group in our series was 3.4%, which was below that reported in the literature (3.9%–12.9%) 2 , 4 , 14 , 21 – 28 . Detection of only the most frequently seen P286R protein change in the POLE mutant group in our series was associated with the lower number of cases. We noted that P489S, Q453R, and M487T, among 21 nonpathogenic or unknown protein changes detected other than P286R, accompanied the MMR-D group. When the cases with MMR-D and nonpathogenic POLE mutations were examined in the literature, it was revealed that the prognosis of this group was similar to that of the MMR-D group. As detected in one of our cases, the nonpathogenic Q453R mutation accompanying the MMR-D group was described by León-Castillo et al. 12 . However, the relationship between P489S and M487T and the MMR-D group has not yet been found in the literature. Apart from this, among the other protein changes detected, D414N and L283F have been defined as nonpathogenic mutants in the literature 12 , 22 . In addition, although there are studies in the literature reporting that variants of unknown significance in DNA polymerase genes in different tumors are also effective on mutation burden and prognosis 29 , we did not find a relationship between the histopathologic and prognostic features of the eighteen POLE mutations of unknown significance in our series. One case among our POLE mutant cases showed subclonal p53 staining as well as POLE mutation, and this case had a histologic grade of 3, consistent with the literature data 30 . This case, the youngest patient in our series, had obturator lymph node metastasis (stage IIICI). Although stage III cases have been described in the literature, it has been emphasized in studies that the incidence rate of these cases is quite low 31 . However, in a study examining POLE mutant cases with abnormal p53 staining, 6 of 31 cases were in the stage III/IV category 30 . In addition, a metastatic lesion was detected in the liver 2 mo later during the follow-up of the case and classified in the high-risk category before its molecular classification was known according to ESGO/ESTRO/ESP recommendations; even with the identification of the POLE mutation, it is still unclear in which risk group this case will be classified. The current ESMO guidelines published in 2022 indicated that stage III POLE mutant cases can be considered as low risk, however, there is currently no data on the safety of skipping adjuvant therapy 32 . As a result, this rare POLE mutant case with abnormal p53 staining that received advanced chemo, and radiotherapy is very important in terms of its contribution to the literature. The histopathologic features of the POLE mutant group in the literature are roughly as follows: high nuclear grade, rich in TILs and peritumoral lymphocytes, and presence of tumor giant cells 33 – 37 . Despite a small number of our cases, the nuclear atypia level of the POLE group was low in our series, similar to the MMR-D and NSMP groups. In addition, the only case with a nuclear atypia level of 3 showed subclonal (wild type + overexpression) p53 staining and each area had endometrioid morphology. As a result, it was determined that the cellular atypia levels of these tumors were not very high, unlike the literature. This issue requires further investigation and research. In addition to these findings, due to the low number of cases in our series, we did not find any significance in histopathologic or prognostic parameters except lymph node involvement. Data in the literature has indicated that the presence of POLE mutation has a protective effect against lymph node metastases 19 , 38 . In our study, the presence of lymph node metastasis was found to be statistically significant in the POLE mutant group. In the cohort of Imboden et al. 22 , statistical significance was not found between the POLE mutant group and stage, which supports our findings. Therefore, the presence of lymph node metastases is an important factor in determining the treatment regimen in these cases, especially in populations with a low frequency of POLE mutations. “Tertiary lymphoid structures,” a special type of ectopic lymphoid formations, have been defined in endometrial carcinomas, and it has been reported that these structures are frequently encountered in the POLE mutant and MMR-D groups 39 , 40 . These tertiary lymphoid structures may form due to immunogenic tumor neoantigens that are more likely to be found in cancers with a high mutational burden. In our study, 1 of the lymphoid follicle structures detected in a total of 4 cases were in the POLE mutant, 2 in the MMR-D, and 1 in the NSMP group, and the relationship of this histopathologic parameter with molecular subgroups was statistically significant. The absence of this morphologic finding in the p53 mutant group is both a valuable finding in terms of prognosis and supports the thesis that it is present in immunogenic tumors. In support of our findings, Dessources et al. 41 reported that the lowest CD8+ T cell level was in the high copy number group among all molecular subtypes. In conclusion, this finding, which is not evaluated in routine practice, maybe a significant parameter for the identification of molecular groups and maybe a guide in terms of both treatment and prognosis. Therefore, the inclusion of this parameter in future pathology reports may be considered. The MMR-D group was found at a rate of 26.3%, close to its frequency in TCGA. Low histologic grade, prominent intratumoral and stromal lymphocytes, presence of myometrial invasion in the MELF pattern and absence of pleomorphism and cherry-red nucleoli, which are cytomorphologic features, were found to be significant in the MMR-D group in our series, consistent with the literature. In our study, both intratumoral and peritumoral lymphocytes were observed in all of the POLE mutant groups, but statistical significance could not be obtained due to the small number of cases. In the MMR-D group, the rate of intratumoral and stromal lymphocytes was found to be much higher than in non–MMR-D tumors. As previously reported, the presence of significant TIL can be used as a predictor for POLE mutant cases after excluding the MMR-D group 42 . The p53 mutant group remained below the frequency of TCGA with 18.6% in our series. The features of this group, which has the highest mean age (68±8.38 yr) in our series, include the presence of serous histologic subtype, high histologic and nuclear grade, and advanced stage, and among cytomorphologic features in consistent with literature findings 43 , giant cells, pleomorphism, atypical mitosis, cherry-red nucleoli, smudged chromatin and coarse chromatin were observed. When we examined its relationship with molecular groups, we found that these cytologic parameters were most frequently observed in the p53 mutant group, and they had the lowest incidence in the NSMP group. In our case series, the p53 mutant group has the worst prognosis among all molecular groups, and the presence of the tumor in this group increases the mortality rate, the risk of metastasis, and reduced overall and disease-free survival times. However, this group alone could not be identified as an independent prognostic determinant in determining the risk of mortality and metastasis, or overall and disease-free survival times. Subclonal p53 expression in endometrial carcinomas has been defined as the presence of more than one pattern of p53 expression in the tumor, and studies on this subject are relatively few. After technical errors are excluded, this staining pattern, in endometrioid carcinomas, is often explained by the acquisition of a TP53 mutation in a subclone of the tumor due to tumor progression in cases of underlying POLE mutation or MMR-D 18 , 44 . The importance of the subclonal p53 pattern has not yet been clarified in tumors in which POLE mutations are not observed or MMR proteins are preserved 45 . However, when the current literature is reviewed, 2 studies have drawn attention in this context: (1) in the study conducted by Singh et al. 18 , in which the correlation of p53 immunohistochemistry and TP53 mutation was examined, it was stated that the subclonal staining pattern in MMR-proficient/POLE wild type endometrial carcinomas was correlated with the TP53 mutation. (2) In a study published by Vermij et al. 46 TP53 mutation was found in both of the 2 MMR-proficient/POLE wild type cases with subclonal p53 staining. Therefore, in our study, overexpression, null, and cytoplasmic staining as well as subclonal staining patterns were also included in mutant p53 staining patterns. Subclonal staining patterns was observed in 7 of 118 cases in our study. Subclonal staining pattern in each of 2 cases was accompanied separately with POLE, and MMR-D groups, respectively (Table 6 ). When we examined only the p53 staining patterns among all cases, we found a remarkable result regarding the subclonal staining pattern. When the survival time of the cases was analyzed statistically according to the p53 staining patterns, we found that the cases with subclonal staining had a longer survival time than the cases with wild- type staining (log-rank, P <0.001). Compared to cases with other abnormal staining patterns, the rate of metastasis was lower in cases with subclonal p53 staining without any case of mortality. Considering the prognostic differences, subclonal staining in MMR-proficient/POLE wild type endometrial carcinomas should not be evaluated in the p53 mutant group, but in the p53 wild type group which may be more accurate in terms of making an accurate molecular classification and predicting the clinical course. The NSMP molecular group is the least molecularly and prognostically characterized ProMisE group and is in the good-intermediate group in terms of prognosis 3 , 20 , 47 . Therefore, the treatment of patients in this group receives minimal benefit from the molecular-based classification system. In our case series, this group mostly consisted of the endometrioid histologic subtype and did not contain a significant number of intratumoral and stromal lymphocytes. The last incidence of LVSI and lymph node involvement was seen in this group. Cytomorphologic features such as giant cells, pleomorphism, atypical mitosis, smudged, and coarse chromatin were less common in the NSMP group than in the other groups. In addition, the presence of the tumor in this group both reduced the risk of metastasis and increased disease-free survival. Among the 4 molecular subgroups, this group had the longest disease-free survival times. Determining the risk groups of the cases through molecular classification to provide appropriate treatment to the patients was also the ultimate goal of our study. However, detecting a change in the risk group of only 3 patients (2.5%) brought 3 more questions to our minds: (1) Can we narrow down the population and reduce the number of immunohistochemical/molecular analyses? (2) After narrowing down the patient population, can we find morphologic parameters that will represent the molecular groups? (3) Identify prognostic morphologic parameters in early-stage endometrioid, mucinous, and dedifferentiated carcinomas, independent of molecular groups? Determination of POLE status cannot be routinely performed in today’s practice due to the low accessibility to molecular methods, the need for qualified personnel to evaluate and interpret the results, and its high cost. Especially in populations with limited economic opportunities, such as our country, alternative options should also be considered. There are studies presenting new algorithms based on histopathologic parameters for the more effective use of tests 48 – 50 . In response to our first question, we have realized that instead of investigating 118 patients, by applying p53 and MMR tests to 90 patients, we could decrease application rate of immunohistochemical tests to 76.3%; and by performing POLE analysis on 55 patients, we could further decrease it to 46.6%. In our statistical analysis, we could not identify any new morphologic parameters that could represent molecular groups, other than the statistically significant morphologic parameters previously indicated. Among all cases (n=118), the absence of cherry red nucleoli was identified as an independent predictor for the MMR-D subgroup, while the absence of prominent stromal TIL was an independent predictor for the NSMP subgroup. However, when the study population was made more homogeneous (n=90), no independent predictor was found for the molecular subgroups. According to the guidelines released by ISGyP and ESGO/ESTRO/ESP, traditional pathologic characteristics like histotype, grade, myometrial invasion, and LVSI remain significant prognostic indicators. These features accurately assess risk in ~60% of the cases 51 , 52 . Therefore, when parameters related to the stage and histologic subtype of the tumor were excluded, we encountered a surprising result in our statistical analyses aimed at finding a prognostic indicator. The MELF pattern of invasion was associated with the mortality rate, presence of metastasis, overall, and disease-free survival. Moreover, it was an independent prognostic factor that reduced the overall survival time. The critical importance of MELF pattern of invasion should be emphasized. In fact, overlooking this pattern can change the depth of myometrial invasion and consequently the stage of the disease, and the condition can be also confused with LVSI. The incidence of the MELF pattern of invasion was 9.3% among all our cases, with the majority of these being in the MMR-D group. The relationship between the MELF pattern of invasion and the MMR-D group has been significantly identified in a way that corroborates the literature data. At present, the existence of MELF pattern of invasion has not any impact on patient care. However, along with other invasion patterns, it continues to be a central topic in research studies 53 , 54 . In the current literature, although the MELF pattern of invasion has been associated with an increased rate of nodal involvement and a shorter overall survival time, its significance as an independent risk factor is uncertain 55 – 57 . However, the main important point we revealed in our study is that when parameters that are already known to have poor prognostic value are excluded, the MELF pattern of invasion reduces overall survival by 25 mo and disease-free survival by 18 mo. Furthermore, in our study, the MELF pattern of invasion was found to be an independent prognostic factor for the first time that shortens the overall survival time. So, considering its prognostic significance, we think that the MELF pattern of invasion should be specified in pathology reports to alert clinicians to its negative impact on prognosis. One of the limitations of our study is the small sample size of the POLE mutant group. Another is that we could not determine the pathogenicity of total of 18 POLE exonuclease domain mutations of unknown significance which we detected in 15 cases. Based on the POLE scoring system published by León-Castillo et al. 12 for the determination of pathogenicity, based on the incidence rates of mutation types, tumor mutation load of 5 hotspot POLE mutations, and variant recurrence, more detailed interpretation of these mutations can be made in a different study. The third limitation of our study is that we could not perform TP53 mutation analysis in POLE wild type/MMR-proficient cases with subclonal p53 expression. Also, it should be considered that our study lacks a validation cohort to confirm the clinical applicability of algorithm that we have used to identify prognostic and predictive morphologic parameters. In conclusion, while aiming to identify predictive and prognostic morphologic markers for centers with limited resources that could not financially afford the application of molecular tests, we achieved 2 significant results: (1) using the algorithm we designed, and by identifying patients for whom molecular classification could alter categorization of their risk groups, we observed that molecular tests can be utilized more efficiently in populations with limited economic resources, such as in our country. (2) MELF pattern of invasion is a marker that shortens disease-free survival time, and it is also an independent prognostic factor that reduces overall survival. Even though we could not identify a marker to replace molecular classification, we have demonstrated that the number of tests required for risk stratification can be reduced, and thus we have discovered a new morphologic marker of prognostic significance.