Introduction

Endometrial carcinoma (EC) is a heterogeneous disease that on a molecular basis can be divided into four subgroups with distinct pathogenesis and prognosis: polymerase ε (POLE) ultramutated, microsatellite instability (MSI) hypermutated, copy-number-low, and copy-number-high tumors [1]. In view of this disease heterogeneity, the development of personalized medicine relies on subclass-specific studies on relevant biomarkers and potential treatment strategies.

Defective DNA mismatch repair (MMRd), leading to instability of microsatellites and high frequency of mutation, is present in 20–40% of ECs [2]. Loss of MMR function results from a biallelic inactivation of the key MMR genes: MLH1, PMS2, MSH2, or MSH6. Sporadic disease is more common, and as regards EC, it generally arises from silencing of MLH1 due to promoter hypermethylation [3, 4]. In some cases, MLH1 methylation reflects a more global promoter hypermethylation pattern (CpG island methylator phenotype, CIMP) [5]. Sporadic tumorigenesis may also be driven by biallelic somatic mutations of the MMR genes [6, 7]. A deleterious germline mutation of one MMR allele defines Lynch syndrome (LS). In a separate and rare mechanism, deletion of the gene coding for epithelial cell adhesion molecule (EPCAM, TACSTD1), leads to constitutional methylation and epigenetic silencing of the adjacent gene, MSH2 [8]. In LS, cancerogenesis is triggered by a “second hit” that inactivates the wild-type allele of the affected MMR gene. Lifetime cancer risk in female patients with LS is 30–43% for colorectal carcinoma (CRC) and 40–60% for EC [9,10,11]. EC represents the sentinel cancer in 35–50% of female LS patients, and it has been estimated that LS accounts for 2–3% of all the ECs [12,13,14]. Hereditary cancer risk is not uniform for all MMR proteins, e.g., female carries of MSH6 germline mutations have a higher risk of EC compared with MLH1 mutation carriers, and their risk for EC is higher than the risk for CRC [15].

Clinical LS screening strategies based on age and family history fail to identify a significant proportion (>40%) of patients with an LS-associated EC [16, 17]. MSI analysis and MMR immunohistochemistry may be used for tumor-based LS screening. These methods identify defective MMR in tumors, but they do not differentiate sporadic defects from LS, which requires germline mutation analysis. In LS-associated tumors, MLH1 promoter methylation is rare, and MLH1 methylation may be used as a surrogate marker for sporadic disease in the screening process [18,19,20].

In CRC, MSI phenotype is considered a positive prognostic factor. Various studies on CRC also suggest that MSI may predict resistance to 5-fluorouracil chemotherapy [21, 22]. The prognostic significance of MMR deficiency in EC has been outlined by the TCGA, but therapeutic implications are unclear. Further, the impact of MLH1 methylation status on prognosis and treatment response needs to be clarified [23,24,25,26]. We classified unselected endometrioid ECs according to MMR protein expression and MLH1 methylation status, and compared MMR subgroups as regards prognosis and the distribution of established clinicopathological risk factors along with various potential molecular biomarkers: estrogen receptor (ER) alpha, progesterone receptor (PR), p53, L1 cell adhesion molecule (L1CAM), AT-rich interactive domain 1A (ARID1A), and an immunotherapy target molecule, programmed death ligand 1 (PD-L1).

Methods

The study was approved by the Institutional Review Board and the National Supervisory Authority for Welfare and Health. A tissue microarray (TMA) was constructed on 842 primary tumor samples from patients who underwent primary surgical treatment for stage I–IV EC at the Department of Obstetrics and Gynecology, Helsinki University Hospital between 2007 and 2012 [27]. The TMA included 745 ECs of the endometrioid subtype. Clinicopathologic data were abstracted from institutional medical and pathology records. Stage was determined according to the FIGO guidelines revised in 2009 [28]. The cutoff for age as a risk factor was based on the finding that age >65 years is an independent poor prognostic factor in endometrial cancer [29]. Disease-specific survival times were calculated as time from surgery to death from EC, and overall survival as time from surgery to death from any cause. Cause of death was mainly based on medical records. Missing data were complemented from death certificates derived from Statistics Finland.

The following monoclonal antibodies were used for chromogenic immunohistochemistry on multicore TMA slides: MLH1 (ES05, Dako), MSH2 (G219-1129, BD Biosciences), MSH6 (EPR3945, Abcam), PMS2 (EPR3947, Epitomics), ERa (SP1, Roche/Ventana), PR (16, Novocastra), p53 (DO-7, Dako), ARID1A (HPA005456, Sigma-Aldrich), and L1CAM (SIG-3911, Covance, clone 14.10). TMA slides were scanned with the 3-dimensional Histech Pannoramic 250 Flash II scanner by Fimmic Oy (Helsinki, Finland). Slide images were managed and analyzed with WebMicroscope Software (Fimmic Oy). Virtual slides were scored by a pathologist blinded to clinical data (AP). A second investigator (RB) examined equivocal cases and a consensus was reached. Mismatch repair protein status was considered deficient (MMR-D) when we observed a complete loss of nuclear expression in carcinoma cells of one or more MMR proteins (MLH1, MSH2, MSH6, and PMS2) detected by immunohistochemistry. MMR proteins form heterodimer complexes (MLH1/PMS2 and MSH2/MSH6), and only MLH1 and MSH2 are stable without their dimer partners. Hence, tumors showing loss of MLH1 and PMS2 or MSH2 and MSH6 on immunohistochemistry, were considered MLH1 and MSH2 defective, respectively. Tumors showing isolated loss of PMS2 or MSH6 in IHC were considered to present inactivation of the homonymous gene. We adopted a 10% cutoff for ER/PR positivity based on common guidelines for breast cancer and a previous study on EC [30]. Aberrant p53 staining (p53 abn) was defined as strong and diffuse nuclear staining or completely negative (“null”) staining in carcinoma cells. Weak and heterogeneous staining was classified as wild-type expression. AT-rich interaction domain 1A (ARID1A) staining was classified negative when tumoral cells presented diffuse or clonal-type loss of nuclear expression. As indicated by a previous mutational study, the heterogeneous “checkerboard” pattern of staining was considered positive [31]. Stromal and inflammatory cells served as internal control for MMR, p53, and ARID1A stainings. L1CAM expression was scored as reported earlier, with ≥10% of membranous staining considered positive [27]. Samples with scarce carcinoma cells or completely negative staining of the internal control (when applicable) were discarded. Representative images of chromogenic IHC are depicted in Fig. 1.

Fig. 1: Representative HE and immonohistochemical images of MMR-deficient endometrial carcinoma.
figure 1

G1 endometrioid carcinoma (a–d) with MLH1 loss (b), wild-type p53 (c), and negative L1CAM (d), and G3 endometrioid carcinoma (e–h) with MLH1 loss (f), aberrant p53 (g), and positive L1CAM (h).

Full size image

The fluorescent multiplex immunohistochemistry was carried out as reported previously [32]. PD-L1 expression was defined as partial or complete membranous staining in carcinoma cells, and membranous and/or cytoplasmic staining in immune cells (ICs, i.e., CD3-positive T lymphocytes and CD163-positive macrophages within tumor nests and/or adjacent supporting stroma). The percentage of PD-L1-positive carcinoma cells and ICs was estimated separately and in combination. To calculate the combined positive score (CPS), we divided the total number of PD-L1-positive cells (carcinoma cells, lymphocytes, and macrophages) by the number of viable carcinoma cells, multiplied by 100 (www.agilent.com/cs/library/usermanuals/public/29219_pd-l1-ihc-22C3-pharmdx-gastric-interpretation-manual_us.pdf). Cutoff for positive PD-L1 expression was set at ≥1% for all the scorings. The quantity of CD3-positive tumor-infiltrating T lymphocytes (TILs) was visually estimated on multiplex IHC slides, and was semiquantitatively scored as scarce, moderate, or abundant. For statistical analyses tumors with abundant TILs were considered TIL positive.

For DNA extraction, representative areas of formalin-fixed paraffin-embedded tumor sections were macrodissected as identified by pathologist assessment. DNA was extracted by proteinase K/phenol–chloroform method. POLE exonuclease domain mutation screening of hot spots in exon 9, exon 13, and exon 14 was performed by direct sequencing as described before [32]. Polymerase chain reaction products were sequenced on an ABI3730xl Automatic DNA Sequencer at the Institute for Molecular Medicine Finland, Helsinki. Sequence graphs were analyzed both manually and with Mutation Surveyor (Softgenetics, State College, PA).

Methylation-specific multiplex ligation-dependent probe amplification (MS-MLPA) was performed on MLH1-deficient tumors to evaluate MLH1 promoter methylation levels in Deng promoter regions C and D. We used the SALSA MMR MS-MLPA Kit ME011 (MRC-Holland) on 250 ng of DNA from each sample. All MS-MLPA reactions, analyses, and calculations of methylation dosage ratios were done according to the manufacturer’s instructions. MS-MLPA products were separated by capillary electrophoresis (on ABI 3730 Automatic DNA sequencer, Applied Biosystems), and analyzed using GeneMapper 5.0 genotyping software (Applied Biosystems). To calculate the methylation ratio, each peak height from HhaI-digested tumor DNA was divided by its corresponding peak height from the undigested tumor DNA. To compensate for differences in PCR efficiency of the individual samples, each peak height (digested and undigested) was normalized, dividing the probe amplification product by the average value of the control probes without a HhaI enzyme site. According to the manufacturer’s recommendations, the hypermethylation threshold is defined as the mean methylation dosage ratio in reference samples (from healthy patients) plus 2 standard deviations. Since our reference samples did not present methylation in the above-mentioned regions, the technical threshold of 0.15 was used as a cutoff [33]. Tumors with a methylation ratio >0.15 (corresponding to 15% of methylated DNA) in region C and/or region D were considered hypermethylated.

Other variables selected for statistical analyses were FIGO stage, age, grade of differentiation, depth of myometrial invasion, lymphovascular invasion (LVI), tumor size, peritoneal cytology, quantity of TILs, expression of PD-L1 (tumor cells, ICs, and CPS), ER, PR, p53, L1CAM, ARID1A, and type of adjuvant therapy. Chi-squared test and Fisher exact test (two-sided) were used for comparison of categorical variables. Survival curves were calculated by the Kaplan–Meier method. A log-rank test was used to test for survival differences. Simple and multivariable analyses for prognostic factors were conducted by the Cox proportional hazard model. Risk factors that were identified as potential predictors in unadjusted analysis were included in the multivariable model. Statistical significance was set at P < 0.05. Data were analyzed using Statistical Package for Social Sciences version 25 software (IBM Corp., Armonk, NY).

Results

Pertinent clinicopathologic characteristics of the study cohort are illustrated in Table 1. MMR immunohistochemistry provided conclusive results for 683/745 cases of endometrioid EC (Fig. 2). Conforming to the original TCGA algorithm, we discarded from the MMRd group 1 tumor that was known to harbor concomitant MLH1 methylation and POLE mutation [1]. Loss of MMR protein expression was observed in 35.8% of the tumors as follows: MLH1 + PMS2 loss in 203 (29.8%), isolated PMS2 loss in six (0.9%), MSH2 + MSH6 loss in nine (1.3%), and isolated MSH6 loss in 19 (2.8%). Concomitant loss of MLH1 + PMS2 and MSH6 was present in seven (1.0%) tumors. Due to uncertain classification, we discarded these seven cases with multiple abnormalities from the methylation status-based analyses.

Table 1 Clinicopathologic data (n = 682a).
Full size table
Fig. 2: MMR immunohistochemistry and MLH1 methylation analysis.
figure 2

MMR mismatch repair, IHC immunohistochemistry, Met methylated, NonMet nonmethylated. Methylation-linked (light gray) and nonmethylated (dark gray) MMR-deficient subgroups.

Full size image

MS-MLPA was successfully carried out on 157/203 MLH1-/PMS2-negative tumor samples. Cases with unsuccessful methylation analysis (46/203, 22.7%) due to unsuccessful DNA extraction or low-quality DNA, were excluded from the methylation-based analyses. Methylation states of Deng regions C and D were mostly congruent: 132 of 157 samples were methylated, and 14 unmethylated in both regions (P < 0.0001). Methylation of only one of the regions was observed in eleven (7.0%) tumors: six in region C and five in region D. Four of the five cases with isolated hypermethylation in region D, displayed near-cutoff levels of methylation in region C. In total, hypermethylation (methylation ratio >0.15) was observed in 91.1% of the MLH1-deficient tumors. For the purposes of this study, the cases with conclusive IHC and methylation results were divided into subgroups: MMR-proficient tumors (MMRp, n = 438) and MMR-deficient tumors (MMRd, n = 244). MMRd cases with conclusive methylation results were further subclassified as MLH1-methylated (MMRd Met, n = 143) and -nonmethylated (presumably mutated) tumors (MMRd NonMet, n = 48). Considering the methylation rate in this cohort, an estimated 76% of all the MMRd cases and 27% of all the 682 samples presented methylation-linked MMR deficiency. Of all the samples, an estimated 7.6% were MMRd and nonmethylated (presumably mutated).

Proportions of clinicopathologic risk factors and biomarker expression in the molecular subgroups are shown in Table 2. MMR deficiency was associated with older age, high grade of differentiation (G3), advanced stage (II–IV), larger tumor size, abundant TILs, PD-L1 positivity in ICs and CPS, wild-type p53, negative L1CAM, ARID1A loss, and type of adjuvant therapy. Within the MMRd cases, methylation correlated significantly with older age and larger tumor size.

Table 2 Clinicopathologic characteristics according to MMR subgroups.
Full size table

Median follow-up time was 81.3 months (range 1–136) in the whole cohort and 75.7 months (range 2–136) in the MMRd cohort. Disease-specific mortality occurred in 94 (13.8%) of all the 682 patients, including 47/244 (19.3%) of MMRd cases and 47/438 (10.7%) of MMRp cases. Five-year disease-specific survival and overall survival rates were 83.2 and 71.3% for the patients with a MMRd-Met tumor, and 91.7 and 83.3% for the patients with a MMR-NonMet tumor. In the Kaplan–Meier survival analysis based on MMR and methylation status, MMRd-Met phenotype predicted lower disease-specific survival in the whole cohort (P = 0.007, Fig. 3). The negative effect of methylation status retained its significance after controlling for confounders (Table 3). Kaplan–Meier survival analyses were also performed within subgroups of patients receiving radiotherapy, chemotherapy, or both. In these treatment-specific subgroup analyses, we found no statistically significant correlation between MMR phenotype and survival (data not shown). In the simple proportional hazards model performed exclusively on MMRd cases, disease-specific survival was associated with FIGO stage, grade 3 of differentiation, deep myometrial invasion, LVI, peritoneal cytology, tumor diameter, ER negativity, L1CAM positivity, and type of adjuvant therapy, but not methylation (Table 4). Multivariable analysis confirmed the independent effect of LVI (Table 4).

Fig. 3
figure 3

Disease-specific survival according to the MMR phenotype.

Full size image
Table 3 Simple and multivariable Cox regression models of disease-specific survival in endometrioid endometrial carcinoma.
Full size table
Table 4 Simple and multivariable Cox regression models of disease-specific survival in MMR-deficient endometrioid endometrial carcinoma.
Full size table

Discussion

EC is a pathogenetically heterogeneous group of sporadic and hereditary tumors. We classified unselected endometrioid ECs according to their MMR IHC profile and MLH1 methylation status in order to explore clinicopathologic differences between MMR subclasses of EC.

The prevalence of specific MMR protein losses was similar to that reported by Watkins et al. [34]. Compared to other studies, the relative frequency of MLH1 loss was high and the frequency of MSH2 loss was low, possibly reflecting geographical differences in the prevalence of various mutations [34,35,36]. In the group of MLH1-deficient cases, 91.1% exhibited hypermethylation, which is in line with previous studies (74–93%) [37, 38]. Methylation of the promoter area is not always complete, and the results of various methylation studies may vary according to the specific promoter regions that have been analyzed [39]. Further, partial methylation of the MLH1 promoter does not invariably lead to gene silencing [39, 40]. To increase specificity of the methylation analysis, we investigated MLH1 methylation exclusively in promoter regions that are confirmedly associated with MLH1 gene silencing [41, 42].

MMRd has been found to correlate with several negative prognostic factors of EC including advanced stage, high grade, LVI, and myometrial invasion, findings corroborated by the present study [24, 26, 43, 44]. However, in the pre-TCGA studies, where dichotomous comparisons between MMRd and MMRp subgroups were made, MMRd was generally not associated with poor survival [45]. In the TCGA study with four molecular subgroups, MMRd along with the copy-number-low ECs, presented an intermediate prognosis compared with the indolent POLE mutated and the aggressive copy-number-high ECs [1]. As regards the prognostic significance of methylation status in EC, we found an association between the MLH1-methylated phenotype and lower disease-specific survival as compared with MMRp EC. Similarly, in the study by Cosgrove et al., poorer recurrence-free survival was observed in the subgroup of methylated MMRd EC as compared with the nonmethylated phenotypes [26]. Shikama et al. reported poor overall survival in patients with sporadic EC compared with nonmethylated (presumed LS) MMRd tumors [25]. It is noteworthy that when measuring overall survival (as opposed to disease-specific survival), differences between presumed hereditary and sporadic disease may reflect the divergent age distribution in these groups. In our study this effect was mitigated by the adoption of disease-specific survival as the measure of prognosis.

Along with the traditional clinicopathological factors and methylation status, we investigated the expression of various biomarkers in the MMR subgroups of EC. MMRd carcinomas rarely displayed aberrant p53 or L1CAM positivity, both considered molecular markers of aggressive EC [1, 27, 46, 47]. Mutation of ARID1A, a chromatin-remodeling protein, is frequent in endometrioid ECs [48]. Compared with MMRp, we observed ARID1A protein loss more frequently in MMRd tumors, with a particularly high frequency in MMRd-Met tumors. The expression of ER, PR, p53, L1CAM, or ARID1A did not correlate independently with survival in the MMRd subgroup.

Based on abounding data on CRC, detection of MMR deficiency may predict the efficacy of adjuvant therapies. The relationship between MMR phenotype and response to adjuvant treatment has not been fully established for EC [24, 25, 49]. Preclinical studies have provided evidence suggesting that impaired MMR confers resistance to certain chemotherapeutic agents, including platinum-based compounds, which are commonly used in the treatment of gynecological cancers [50]. By contrast, defective MMR has predicted better response to adjuvant radiotherapy in EC [51, 52]. Our study could not demonstrate a significant correlation between MMR status and response to adjuvant therapies. Subclass-specific studies on treatment response generally suffer from low power, which could be increased by meta-analyses conducted on larger amounts of subclass data. Interestingly, defective MMR appears to predict favorable response to antiPD1/PD-L1 immunotherapy [53]. In fact, MMRd carcinomas characteristically present higher T-cell counts and more frequent PD-L1 positivity (T-cell inflamed PD-L1 positive phenotype) [32, 54]. It is not known whether these findings are universal across various subgroups of MMRd carcinomas (sporadic vs. hereditary, methylated vs. mutated). Previous studies suggest that ECs with mutation-associated MMR deficiency present higher quantities of TILs compared with ECs with epigenetic MMRd, but data on PD-L1 expression are contrasting [54,55,56]. In our cohort MMRd tumors presented higher quantities of TILs and more frequent PD-L1 positivity on ICs and CPS compared with MMRp tumors, but we could not demonstrate significant immunological differences between methylation-based phenotypes of MMRd ECs.

As an advantage, our single center study provides sufficiently long follow-up times and accurate information on causes of death. We chose to investigate MMR deficiency with immunohistochemistry instead of MSI analysis because of the additional information it offers on the specific MMR proteins. Four-antibody immunohistochemistry and MSI analysis predict mutations with comparable sensitivities. Rare cases with mutations producing nonfunctional proteins with retained immunohistochemical antigenicity, may be misinterpreted by IHC [57]. By contrast, MSH6 mutations are more frequently detected by IHC [58, 59]. We strictly reported MMR deficiency only for cases showing adequate staining of the internal control. TMA-based methodology allows to perform numerous immunohistochemical stainings on a vast number of tumor samples. On the other hand, uneven distribution of protein expression may lead to false negativity in a TMA study. Previous studies have shown that TMAs with 2–3 core biopsies per tumor adequately represent the tumor phenotype, even for heterogeneous antigens such as ER, PR, p53, and Her-2 [60, 61]. To increase sensitivity, we included four cores from each tumor in our TMA. We previously demonstrated a high concordance between the same TMA and whole section stainings of L1CAM, a highly heterogeneous antigen [27]. In addition, our prevalence of MMR deficiency was at the upper end of the expected values. We thus consider our TMA adequately representative of the MMR status in our study cohort. Lack of germline data may lead to erroneous classification into sporadic and hereditary disease. In fact, a deleterious germline mutation is detected in only half of the EC patients classified as “probable LS” (MMR deficiency not associated with MLH1 methylation) [62]. Possible pathogenetic mechanisms of this “Lynch-like syndrome” include currently unidentified germline mutations, somatic cell mosaicism, or biallelic somatic mutation [7, 34, 35, 63]. Acknowledging this unsolved genetic scene, we based our study on protein expression and methylation phenotypes without further assumptions about sporadic or hereditary origin of the disease.

In conclusion, MMR protein expression and MLH1 methylation profiles define distinctive phenotypes that correlate with prognostic factors and immunologic features of EC. Methylation-linked MMRd phenotype predicts poor survival in endometrioid EC. Further studies are necessary to investigate the predictive value of MMR subclasses for both traditional adjuvant therapy and immunotherapy.