Immune Microenvironment in Microsatellite-Instable Endometrial Cancers: Hereditary or Sporadic Origin Matters

Antitumor immune response is thought to be associated with high somatic mutational load, which causes increased production of tumor-specific neoantigens (1). In fact, tumors with high mutation rates, such as melanoma, have proven responsive to immunotherapy. Immunotherapy strategies against tumors with high microsatellite instability (MSI-H) are currently being studied due to the high mutational load inherent to this tumor subtype as well. MSI-H tumors have defects in DNA mismatch repair (MMR), which result in errors in areas of repetitive DNA sequences, known as microsatellites (2). Howitt and colleagues described prediction models that indicate MSI-H endometrial cancer has a 7-fold higher neoantigen load compared with microsatellite stable (MSS) tumors (3). MMR deficiency may occur sporadically due to MLH1 promoter methylation or from germline mutations in MMR genes (MLH1, MSH2, MSH6, PMS2, EPCAM) seen in Lynch syndrome. MSI-H status is seen most frequently in colorectal and endometrial cancers.

Although MSI-H tumors occur in colorectal and endometrial cancers due to either sporadic MLH1 promoter methylation or inherited Lynch syndrome, previous studies have not evaluated potential differences in antitumor response based on the origin of these tumors. Recent studies of the immune microenvironment in an overall cohort of MSI-H endometrial cancer and ultramutated endometrial cancer (related to DNA Polymerase ϵ mutation) have shown increased tumor-infiltrating lymphocytes (TILs) compared with MSS endometrial cancers (3, 4). Colorectal cancer studies have also shown that MSI-H tumors have increased density of TILs and CTLs compared with MSS tumors, but again did not evaluate MLH1-methylated and Lynch syndrome cases separately (5, 6). In addition, elevated immune checkpoint expression of PD-1 and PD-L1 has been demonstrated in the immune microenvironment of MSI-H colorectal tumors (5) and led to an anti-PD-1 clinical trial. A phase II study of PD-1 blockade using pembrolizumab in colorectal cancers showed improved immune-related objective response rates among MSI-H compared with MSS tumors (7).

Previous studies have focused on the role of increased production of neoantigens due to increased mutational load as a driver for antitumor activation. As such, sporadic MLH1-methylated and Lynch syndrome–associated MSI-H tumors were not analyzed separately; however, this assumption has not been tested. In fact, it is well recognized that MSI-H colorectal cancers do not represent a single molecularly homogeneous subgroup. Approximately 70%–80% of MSI-H colorectal cancers can be categorized into a unique subgroup of cases with widespread hypermethylation, called the CpG island methylator phenotype (CIMP; ref. 8). These cases show hypermethylation of CpG islands in the promoter of tumor suppressor genes and inactivation of MLH1 due to aberrant promoter methylation, but CIMP colorectal cancers have different histology, a distinct molecular profile associated with BRAF mutation, and unique clinical characteristics despite some overlapping features with the MSI-H subtype (9–11). The methylator phenotype has not yet been well described in endometrial cancer (12, 13). Again looking to the colorectal cancer literature, the observation that there are distinct molecular subgroups of MSI-H tumors (14, 15) suggests that this assumption of an identical immune response across all MSI-H tumors may be doubtful. The study reported here sought to use robust methodologies to characterize the immune microenvironment of sporadic MLH1-methylated and Lynch syndrome–associated MSI-H endometrial cancer as a means to begin to determine whether microsatellite instability drives equivalent immune activation, regardless of sporadic or hereditary origin. Differences in immune microenvironment between sporadic and Lynch syndrome-associated MSI-H endometrial cancer and MSS endometrial cancer will provide opportunities to identify candidate subgroups for targeted immunotherapeutic regimens.

Over 10,000 women will die from endometrial cancer in the United States this year (16), largely because of resistance to therapy in the advanced/recurrent disease setting. Combination chemotherapy in these circumstances offers response rates (RR) ranging from 30% to 60% and median progression-free survival of 5–14 months (17). In patients that progress despite chemotherapy, available therapies are extremely limited and RR of only 9%–32% are achieved with second-line therapies (17–20). For this reason, it is critical to identify molecular subgroups that are amenable to targeted therapies, including immunotherapy.

This study was conducted with approval from the institutional review board of The University of Texas MD Anderson Cancer Center (Houston, TX). MSI-H endometrial cancer specimens from 2000 to 2015 were identified from archived samples in the Gynecologic Oncology tumor bank and Lynch syndrome patient registry. MSI tumor status was determined clinically using a method developed by the Molecular Diagnostic Laboratory at MD Anderson Cancer Center that has been previously described (21) and results were reported in the medical record. Briefly, MSI testing was performed following extraction of DNA from formalin-fixed paraffin-embedded (FFPE) tumor and normal tissue. A PCR-based method was used for analysis followed by capillary electrophoretic detection of microsatellite markers. Seven microsatellite markers were used in this method. These markers included BAT25, BAT26, BAT40, D2S123, D5S346, D17S250, and TGFBR2. The number of tumor microsatellite repeats for each of the markers was compared with normal tissue from the same case. A tumor was considered MSI-H if three or more of the seven markers demonstrated allelic shift. When all markers were negative for allelic shift, the case was defined as MSS. Expression of MMR proteins (MLH1, MSH2, MSH6, PMS2) was further evaluated in tumors with microsatellite instability using standard clinical pathology procedures. Lack of protein expression of MSH2, MSH6, or PMS2 in the tumor was considered probable Lynch syndrome. For those with MLH1 loss by IHC, MLH1 promoter methylation was analyzed. Those cases showing MLH1 loss by IHC and without MLH1 promoter methylation were classified as probable Lynch syndrome. MSI-H cases were matched approximately 1:2 to MSS endometrial cancer specimens from the tumor bank. Matching was done based on histology, tumor grade, tumor stage, age at diagnosis, and body mass index (BMI). Archived FFPE tissue blocks were cut in 4-μm sections. Clinical data were abstracted from the medical record as available.

Because of the previous reports indicating that polymerase ε (POLE) mutations in endometrial cancer result in an enhanced immune response, MSS cases were further screened for POLE mutations. Cases with POLE mutation were excluded from the analysis. MSS cases that could not be fully sequenced for POLE (resulting in unknown POLE mutation status) were also excluded from analysis. POLE exonuclease domain mutations were assessed in FFPE tissue using PCR-based Sanger sequencing at The University of Texas MD Anderson Cancer Center Sequencing and Microarray Core Facility. DNA was extracted using Pico Pure DNA Extraction kit (Applied Biosystems). Primers and method were used as previously described by Billingsley and colleagues (22).

Fluorescent IHC multiplexing was performed using the Opal Multiplex System (PerkinElmer), and the protocol was adapted from ref. 23. Briefly, antigen retrieval was carried out in a decloaking chamber for 15 minutes with specific antibody conditions as listed separately in Supplementary Table S1. Sequential incubation with each primary antibody for multiplexing was conducted at 4°C in a humidified chamber for 16 hours. After each incubation, slides were washed and species-specific secondary horseradish peroxidase (HRP; Invitrogen) was applied for 10 minutes. Slides were again washed and incubated with fluorescent tyramide signal amplification (TSA) reagent for 10 minutes. This was followed by washing and repeat heating in a decloaking chamber for 15 minutes. The next primary antibody was then applied after blocking. The process continued as previously described until all primary antibodies and corresponding TSA reagents had been applied to the slides. The slides were heated one last time in a decloaking chamber at 96°C for 15 minutes. After cooling and washing, DAPI was applied for 5 minutes and the slides were cover slipped with an aqueous mounting solution (Thermo Scientific).

Three different multiplexing panels were performed with the order, primary antibody and corresponding fluorophore as listed: panel 1: (i) Granzyme B (1:300, clone 11F1, Leica)/Cyanine 3; (ii) CD8 (1:400, clone 4B11, Leica)/FITC; (iii) CD68 (1:500, clone KP1, Biogenex)/Cy5.5; and (iv) PD-L1 (1:1,600, clone E1L3N, Cell Signaling Technology)/Cy3.5; (v) DAPI.

Panel 2: (i) CD3 (1:900, clone SP7, Thermo Scientific)/520 nm; (ii) CD4 (1:450, clone 4B12, Thermo Scientific)/540 nm; (iii) PD-L1/620 nm; (iv) CDllc (1:1,000, clone 5D11, Leica)/690 nm; and (v) DAPI.

Panel 3: (i) CD103 (1:5,000, clone EPR4166, Abcam)/520 nm; (ii) CD8/620 nm; and (iii) DAPI.

Because of the batch differences associated with different staining conditions for the three multiplexing panels, comparisons between groups were only performed within an individual panel batch.

Immune cell markers were as follows: CD3 (general T-cell marker), CD4 (CD4⁺ T cells), CD8 (CTLs), granzyme B (activated CTLs or natural killer cells), CD68 (macrophages), PD-L1 (immune checkpoint ligand), CD11c (dendritic cell marker), and CD103 (intraepithelial T-cell or dendritic cell marker). Costaining was evaluated for granzyme B/CD8 (activated CTLs), CD3/CD4 (helper T cells), PD-L1/CD68 (PD-L1⁺ macrophages), PD-L1/CD11c (PD-L1⁺ dendritic cells), and CD8/CD103 (intraepithelial T cells).

Multispectral imaging was performed with the Vectra 2 automated system (PerkinElmer) with the assistance of the Flow Cytometry and Cellular Imaging Facility at MD Anderson Cancer Center (Houston, TX). With this system, an algorithm was created to capture up to 30 random high powered fields of tumor-containing images. Four filters were used to capture images (DAPI, Cy3, FITC, and Cy5), and optimal exposure times were determined for each filter individually.

Following image acquisition, inForm software version, 2.1.5430.24864 (PerkinElmer) was used to create a spectral library from each of the single fluorophore slides to allow for unmixing of the fluorophore images. A sample set of images was used to define areas of tumor epithelium and stroma, and train the inForm software in tissue segment pattern recognition. DAPI staining was used to identify nuclei within each of the tissue compartments, and as a reference to determine cellular cytoplasm and membrane segmentation based on the inForm algorithm. Thresholds and scoring were defined by fluorescent pixel intensity that accurately identified positive staining cells, as determined by manual review of sample images for each of the antibodies. Positive cell staining was also determined on the basis of the target cellular compartment for each of the specific antibodies. Thresholds for each of the antibodies are as follows and values greater than the listed threshold were considered positive staining. Panel 1: granzyme B/Cy3 (membrane score >3), CD8/FITC (membrane score >2), CD68/Cy5.5 (cytoplasm score >1), PD-L1/Cy3.5 (cytoplasm score >5); panel 2: CD3/520 nm (membrane score >0.83), CD4/540 nm (membrane score >0.33), PD-L1/620 nm (cytoplasm score >2.5), and CD11c/690 nm (membrane score >0.2); panel 3: CD103/520 nm (membrane score >2.0) and CD8/620 nm (membrane score >4.5).

An algorithm to determine positive cell counts within the stroma and tumor epithelial compartment was created using the discussed tissue and cell segmentation and fluorescent pixel intensity algorithm using the inForm software. This algorithm was applied to all images. A reviewer blinded to MSI status manually reviewed all images to assess accuracy of the algorithm. Those images that were determined to be inaccurate were added back to the training set to improve segmentation accuracy. Because of the inherent variation of tissue architecture among samples, 100% segmentation accuracy was unable to be obtained and those images with gross segmentation inaccuracy were excluded from the algorithm. Overall there was an average of 16 images analyzed per specimen (range 1–35). In addition, tissue degradation secondary to multiple rounds of heating occurred in some cases and these cases were removed from the analysis.

The number of positive staining cells for each of the antibodies was then calculated per millimeter squared (mm²) within both the tumor epithelium and the tumor-associated stroma for each of the cases using code written in SAS version 9.4.

Statistical analysis was performed using GraphPad Prism version 6 (GraphPad Software) and Stata version 14.1 (StataCorp). The Mann–Whitney and χ² tests were used to compare demographic data between the two groups as indicated. For comparison of immune cell markers between MSI-H and matched MSS cases, the Wilcoxon matched-pairs signed rank test was used. Values for each MSI-H case were paired with the matched MSS case. In the case that more than one matched MSS case was available, the average of the matches was used for comparison. The overall number of cases is indicated at the top of each data table; however, when a case was excluded for any reason, the corresponding matched MSS or MSI-H case was also removed from the analysis for that marker. Lynch syndrome MSI-H and sporadic MSI-H cases were compared as a group (not matched) and the Mann–Whitney test was used for this comparison. A P value of <0.05 was used to signify statistical significance. Box plots were created to visualize data for percent positive PD-L1 staining cells. The upper border of the box represents the third quartile, the lower border the first quartile, and the line the median. The whiskers are defined using the Tukey box plot method where whiskers represent 1.5 times the upper and lower interquartile range. Study data were collected and managed using REDCap electronic data capture tools (24) hosted at The University of Texas MD Anderson Cancer. REDCap (Research Electronic Data Capture) is a secure, web-based application designed to support data capture for research studies.

Evaluation of endometrial cancer immune microenvironment was conducted using specimens from the gynecologic oncology archived tumor bank and Lynch syndrome patient registry. In total, 60 MSI-H cases were identified and matched approximately 1:2 to 107 MSS cases. Further testing of MSS cases was performed to exclude POLE-mutant cases (n = 4) or cases with unknown POLE status (n = 7), resulting in 96 matched MSS cases. Of the MSI-H cases, 20 were found to have IHC defects in MMR genes consistent with probable Lynch syndrome (LS) and this group was used for the LS MSI-H subanalysis. A summary of Lynch syndrome testing characteristics for these cases is shown in Supplementary Table S2. One MSI-H case had loss of MSH2 and MSH6 on IHC, but no germline deleterious mutations of DNA MMR genes. Another MSI-H case had unknown specific protein loss, but was positive for microsatellite instability with allelic shift in 5 of 7 microsatellite markers. These last two cases were included only in the overall MSI-H versus MSS analysis. Thirty-eight MSI-H cases demonstrated sporadic promoter methylation of MLH1 and were used for the sporadic MSI-H subanalysis.

There were no significant differences in characteristics used for case matching of MSI-H to MSS (histology, age at diagnosis, BMI, stage, and grade) as shown in Table 1. Some clinical data could not be obtained for a portion of the MSI-H cases, primarily the Lynch syndrome cases. Many LS MSI-H endometrial cancer cases were obtained from a Lynch syndrome patient registry, which included specimens collected from outside institutions and had more limited clinical information available. As shown in Table 1, 11 cases were missing details of depth of myometrial invasion and 12 were missing LVSI information. As comparison of depth of invasion and LVSI were not primary objectives, these cases were not excluded from the cohort. In addition, BMI data were missing for 13 Lynch syndrome cases, 2 cases had unknown grade (both LS MSI-H), and 5 had unknown stage (all LS MSI-H). However, these cases were matched according to the data available and were included in the analysis given the limited number of cases of the LS MSI-H subtype.

Overall, the majority of cases were stage IA in both groups (MSI-H 58.3% vs. MSS 68.8%; P = 0.76) and the majority of cases were grade 2 (MSI-H 68.3% vs. MSS 76.0%; P = 0.42). As most cases were stage IA, only 25.0% had myometrial invasion equal to or greater than 50% in the MSI-H and 27.1% in the MSS cases. Finally, there was a significant difference in lymphovascular space invasion (LVSI) between the two groups with more MSS cases being absent LVSI (69.8%) than that seen in the MSI-H (38.3%) cases (P = 0.01).

Differences in immune cell markers between all MSI-H and MSS endometrial cancers was examined using three fluorescent IHC multiplexing panels. A representative image for each multiplexing panel is demonstrated in Supplementary Fig. S1. The number of positive staining cells for each of the markers in both the stromal and tumor epithelial compartments in all MSI-H versus MSS endometrial cancer is shown in Table 2. Among all MSI-H tumors, the median number of granzyme B⁺, PD-L1⁺, CD3⁺, and CD4⁺ staining cells were significantly higher within the tumor-associated stroma compared with MSS endometrial cancer. Tumor epithelial CD8⁺ and granzyme B⁺ staining was also significantly higher in all MSI-H compared with MSS endometrial cancer, while CD68⁺ staining was significantly lower in the tumor epithelial compartment of MSI-H cases. The percentage of PD-L1⁺ cells within the two compartments was also assessed, as this value has been used as an indicator for response to immune checkpoint–blocking therapies in some clinical trials. In this study, the median percentage of PD-L1⁺ staining stromal cells was significantly increased in the MSI-H cases compared with MSS cases (60.5% vs. 48.3%; P < 0.01). However, in the tumor epithelium, there was no significant difference in median percentage of PD-L1 expression between the two groups (MSI-H 2.0% vs. MSS 1.5%; P = 0.46; Fig. 1A). No other statistically significant difference in individual staining was seen within the stromal or tumor epithelial compartment for the two groups.

Costaining of markers was assessed within the groups and among each of the tissue compartments as shown in Table 2. Costaining of CD8⁺ and granzyme B⁺ cells represents the number of activated CTLs, and was significantly higher in MSI-H compared with MSS cases in both the stromal and tumor epithelial compartments. Th cells were evaluated via costaining of CD3/CD4. Stromal CD3⁺CD4⁺ cells were increased in MSI-H compared with MSS but showed no statistically significant differences within the tumor. Costaining for PD-L1⁺ macrophages (PD-L1⁺ and CD68⁺) was also evaluated, and showed no significant differences. Additional staining was performed to evaluate PD-L1⁺ dendritic cells (PD-L1⁺ and CD11c⁺) and CD103/CD8 (intraepithelial CD8⁺ cells). To streamline data presentation, results for costaining of PD-L1⁺ dendritic cells and CD8⁺CD103⁺ immune cell populations, as well as separate CD11c and CD103 staining are not shown. There were no statistically significant differences for these markers in any of the group comparisons (Tables 2–5).

To evaluate differences in immune markers of sporadic MSI-H compared with matched MSS cases, a secondary subanalysis was then performed excluding LS MSI-H cases. This analysis included only MSI-H endometrial cancer with MLH1 promoter methylation compared with matched MSS endometrial cancer. Within this subgroup, the number of granzyme B⁺, CD3⁺, and CD4⁺ cells within tumor-associated stroma was higher among the sporadic MSI-H versus MSS cases (Table 3). PD-L1⁺ cells were increased within the stroma and tumor of sporadic MSI-H versus MSS cases. The median percentage of PD-L1⁺ staining cells (62.5% vs. 46.0%; P < 0.0001) was also significantly higher in the stromal compartment for sporadic MSI-H versus MSS cases, similar to that seen in all MSI-H cases (Fig. 1B). The percentage of PD-L1⁺ cells within the tumor epithelial compartment was also significantly different in sporadic MSI-H compared with MSS cases (2.5% vs. 1.0%; P = 0.009; Fig. 1B).

As in the evaluation of all MSI-H cases, costaining was evaluated and a comparison of the median number of positive cells is shown in Table 3. Sporadic MSI-H cases demonstrated an increase in the median number of PD-L1⁺ macrophages (costaining of PD-L1 and CD68) in both stroma and tumor compared with MSS. Interestingly, this difference was not present in the overall MSI-H cohort analysis. T helper cells (costaining of CD3 and CD4) were increased in the stroma compared with matched MSS cases.

A subanalysis of immune markers in LS MSI-H compared with matched MSS endometrial cancer was conducted (Table 4). LS MSI-H endometrial cancer demonstrated a significantly higher median number of CD8⁺-staining cells in both the stromal and tumor epithelial compartments when compared with MSS endometrial cancer. The number of PD-L1⁺ cells in the tumor was significantly lower in LS MSI-H cases versus MSS cases. The median number of CD68⁺ staining cells was found to be significantly reduced in the tumor epithelial compartment in LS MSI-H cases compared with MSS tumors. In regards to the median percentage of PD-L1⁺ cells, there was no difference between LS MSI-H and MSS cases in the stroma (LS MSI-H 52.0% vs. MSS 64.0%; P = 0.62); however, Lynch syndrome cases showed a significant decrease in the percentage of PD-L1⁺ cells in the tumor epithelial compartment (LS MSI-H 1.0% vs. MSS 3.5%; P = 0.03; Fig. 1C).

Costaining comparison of immune markers for LS MSI-H versus MSS endometrial cancer is also shown in Table 4. There was a significantly higher median number of activated CTLs (costaining with CD8 and granzyme B) in the stroma and tumor of LS MSI-H versus MSS cases. The median number of PD-L1⁺ macrophages was decreased in Lynch syndrome cases compared with MSS, in contrast to the observation in the sporadic cohort. No other significant differences were noted from costaining analysis.

Given the intriguing differences shown between each subgroup of MSI-H endometrial cancer relative to MSS endometrial cancer, we then directly compared LS MSI-H and sporadic MSI-H cases to determine whether the origin of microsatellite instability influences the immune microenvironment. As shown in Table 5, LS MSI-H endometrial cancer had increased CD8⁺ cells in the stroma and reduced numbers of CD68⁺ macrophages in the stromal and tumor compartments compared with sporadic MSI-H endometrial cancer. Costaining revealed increased activated CTLs in the stroma of LS MSI-H endometrial cancer. Furthermore, PD-L1⁺ macrophages were increased in the stroma and tumor of sporadic MSI-H endometrial cancer compared with Lynch syndrome cases. There were no significant differences in median percentage of PD-L1⁺ cells when comparing sporadic MSI-H to LS MSI-H cases in either the stroma (sporadic MSI-H 62.5% vs. 55.0% LS MSI-H; P = 0.19) or tumor epithelial (sporadic MSI-H 2.0% vs. 1.0% LS MSI-H; P = 0.20) compartments (Fig. 1D). This comparison was unmatched and differences between the two groups did exist based on the known etiology of the disease; Lynch syndrome MSI-H patients were younger than sporadic MSI-H patients (51.9 years vs. 64.2 years; P < 0.001) and had a different distribution of histology and grade (see Supplementary Table S3), consistent with previous reports on sporadic versus Lynch syndrome–associated endometrial cancer. BMI was unknown for 13 of 20 Lynch syndrome cases, making it difficult to reliably compare BMI in this cohort.

A detailed and robust multiplexed fluorescent IHC examination of immune cell populations in MSI-H and MSS endometrial cancer specimens was initially undertaken to identify specific differences in the immune microenvironment, and this analysis showed that antitumor immune response is elevated in MSI-H endometrial cancer. By separately evaluating both sporadic and LS MSI-H endometrial cancer cases, we were able to identify the distinct differences relative to MSS endometrial cancer. This prompted further analysis of a direct comparison between the two MSI-H subtypes, which identified differences in specific T-cell and macrophage populations. While the different clinicopathologic characteristics of Lynch syndrome–associated endometrial cancer precluded a matched analysis to sporadic MSI-H endometrial cancer, the altered immune cell infiltrates identified in the analysis for both subgroups compared with matched MSS endometrial cancer provide further support to this observation that origin of microsatellite instability influences immune microenvironment.

In the overall analysis including all MSI-H cases, MSI-H endometrial cancers had increased granzyme B⁺ cells and CTL activation within both the tumor epithelium and tumor-associated stroma, suggesting increased immune-mediated antitumor response in MSI-H endometrial cancers. Increased CD8⁺ T cells and decreased macrophages were observed for the overall MSI-H endometrial cancer cohort in the tumoral compartment. Stromal PD-L1 expression was also significantly increased in MSI-H endometrial cancer and reflects immune response exhaustion and suppression.

Subanalysis of sporadic MSI-H endometrial cancer also demonstrated a difference in the immune microenvironment compared with MSS endometrial cancer. Sporadic MSI-H endometrial cancer had increased granzyme B⁺ in the stroma, with increased PD-L1⁺ cells in both tumor and stroma compared with MSS endometrial cancer, reflecting increased antitumor response in sporadic MSI-H endometrial cancer. Costaining further revealed increased stromal helper T cells, as well as increased stromal and tumoral PD-L1⁺ macrophages in sporadic MSI-H endometrial cancer compared with MSS endometrial cancer.

Comparison of the LS MSI-H subgroup versus matched MSS cases also showed increased immune activation in Lynch syndrome cases. This was demonstrated by increased CD8⁺ cells and activated CTLs in the tumor and stroma of LS MSI-H versus MSS endometrial cancer. The number of macrophages was decreased in the tumor epithelial compartment in LS MSI-H versus MSS endometrial cancer. In contrast to the overall analysis of MSI-H cases and the sporadic MSI-H versus MSS analyses, LS MSI-H cases demonstrated no difference in PD-L1⁺ staining in the stroma compared with MSS endometrial cancer and a significant reduction in PD-L1⁺ staining in the tumor epithelial compartment.

Our study is particularly novel in that we were able to directly compare LS MSI-H and sporadic MSI-H endometrial cancer cases to evaluate distinct changes in the immune microenvironment that could be shaped by either the hereditary MMR defect or sporadic MLH1 hypermethylation origin of microsatellite instability. This comparison revealed increased CD8⁺ cells and activated CTLs in the stroma of Lynch syndrome compared with sporadic MSI-H cases. The number of CD68⁺ macrophages was decreased in the stroma and tumor of Lynch syndrome cases. PD-L1⁺ macrophages were increased in sporadic MSI-H endometrial cancer in both the stroma and tumor. While a matched analysis was not possible due to the unique features of these two subgroups of endometrial cancer, it is important to note that these characteristics are representative of sporadic and hereditary MSI-H endometrial cancer cases overall. A limitation to this study is the number of statistical comparisons for each analysis. We report unadjusted P values for this hypothesis-generating study, and must acknowledge the possibility of false positives and false negatives in this analysis. Importantly, the results are overall consistent with the different alterations identified when each MSI-H subgroup was compared with the matched MSS cases. While it is not yet clear how these differences could impact response to immunotherapies, these results indicate that it will be essential to evaluate response separately in both sporadic MSI-H and LS MSI-H endometrial cancer. Additional studies are necessary to confirm this observation in an independent series and further probe the mechanisms underlying this heterogeneity in the MSI-H immune microenvironment.

The findings from our study provide a comprehensive view of the immune microenvironment of MSI-H endometrial cancer. Previous studies of MSI-H and POLE mutant endometrial cancers have found elevated levels of CD3⁺ and CD8⁺ infiltrating lymphocytes, which suggests that these tumors elicit a strong immune response (3). The overall MSI-H endometrial cancer analysis in our study revealed increased CD8⁺ cells in the tumor epithelial compartment and increased CD3⁺ cells in the stroma; there was also a significant difference in granzyme B⁺ cells and activated CTLs in all MSI-H. This provides further evidence of a more active immune microenvironment in MSI-H endometrial cancer, and suggests that the CTLs that are present are activated and capable of mounting an antitumor immune response.

Tumor-mediated immune evasion in patients with MSI-H endometrial cancer was seen in our study, as evidenced by elevated PD-L1 expression among stromal cells in the overall cohort of MSI-H endometrial cancers and in the tumor and stroma of sporadic MSI-H endometrial cancers compared with matched MSS cases. In addition, our study found low overall expression of tumor PD-L1⁺ cells. PD-L1 expression on tumor versus stromal cells may have important implications for therapeutic response to anti-PD-L1 therapies, but expression varies among tumor types (25, 26). Although many studies have focused on tumor PD-L1 expression and its ability to predict response to anti-PD-1 and anti-PD-L1 therapy, its role as a biomarker remains unclear (25, 26). In some studies, tumor PD-L1 expression has been associated with response to anti-PD-1 therapy (26). In other studies however, expression of PD-L1 on infiltrating immune cells is associated with response to anti-PD-L1 therapy rather than tumor cells (25). In our study, the increased PD-L1 expression highlights the variability of PD-L1 in the tumor microenvironment and suggests that patients with MSI-H endometrial cancer may respond favorably to immune checkpoint blocking agents. In addition, we found PD-L1⁺ macrophages to be significantly increased in sporadic MSI-H compared with both MSS endometrial cancer and LS MSI-H endometrial cancer. The exact role of this subpopulation of cells in the tumor microenvironment cannot be established from our study; however, we speculate that these cells play a role in suppression of the immune response in these tumors. Additional studies are needed to characterize the role of PD-L1⁺ immune cells in MSI-H endometrial cancers.

A previous study showed that tumors from patients with MSI-H colorectal cancer demonstrate significant immune cell expression of PD-L1 and very little tumor PD-L1 expression (5), similar to the results reported here. These findings in colon cancer have led to clinical trials of anti-PD-1 therapy in colorectal cancer that showed improved immune-related objective response rates in patients with MSI-H tumors compared with MSS tumors (7). An additional arm of this study also included patients with any MSI-H tumors, two of which were sporadic endometrial cancer patients. Both of these patients showed a response to anti-PD-1 therapy with one complete response and one partial response. Although this observation is limited to only two patients, the presence of increased stromal PD-L1 expression seen in our study further supports the use of immune checkpoint–blocking agents in this population.

Unlike prior studies investigating the immune microenvironment of MSI-H tumors, our study included a unique subgroup of Lynch syndrome cases. The observation of an altered MSI-H immune microenvironment held true for analysis of both subgroups, sporadic (MLH1-methylated) and Lynch syndrome–associated MSI-H endometrial cancer, but with different specific immune cell population changes in each group. In the Lynch syndrome MSI-H subanalysis, CD8⁺ cells and activated CTLs were increased. Yet there was a significant decrease in tumoral PD-L1 expression and no difference in stromal PD-L1 expression, in contrast to what was observed in the overall analysis and sporadic MSS endometrial cancers. This suggests that LS MSI-H endometrial cancers may have lesser response to single-agent immune checkpoint-blocking agents. Instead, combination of an anti-PD-1 or anti-PD-L1 agent with other immune checkpoint–blocking agents may be necessary. These could include combination with anti-CTLA-4 or newer agents that are under development such as anti-TIM-3. While there are important clinical implications for the observed disparity between the immune microenvironment of LS MSI-H and sporadic MSI-H endometrial cancer, further study of the mechanisms underlying these differences will be key to identifying novel immunotherapeutic strategies and identifying appropriate candidates for immunotherapy.

This study is the first to specifically address the immune microenvironment of Lynch syndrome–related endometrial cancer and the largest cohort to date characterizing the immune microenvironment of MSI-H compared with MSS endometrial cancer. Overall, our study demonstrates increased activity in the tumor immune microenvironment of MSI-H compared with MSS endometrial cancer tumors. In addition, this study suggests that the mechanism responsible for microsatellite instability impacts the immune microenvironment of these tumors, highlighting an important area for further study. Clinical trials investigating single and combination immune checkpoint–blocking agents are needed to determine the role of immunotherapy in patients with MSI-H endometrial cancer and outcomes for sporadic and Lynch syndrome cases should be evaluated separately.

Y. Yuan is a consultant/advisory board member for Agenus. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.B. Pakish, S.C. Mok, R.R. Broaddus, K.H. Lu, M.S. Yates

Development of methodology: J.B. Pakish, Q. Zhang, M.S. Yates

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.B. Pakish, Q. Zhang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.B. Pakish, Z. Chen, H. Liang, G. Chisholm, Y. Yuan, S.C. Mok, R.R. Broaddus, M.S. Yates

Writing, review, and/or revision of the manuscript: J.B. Pakish, Z. Chen, H. Liang, Y. Yuan, R.R. Broaddus, K.H. Lu, M.S. Yates

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.B. Pakish, G. Chisholm, S.C. Mok

Study supervision: K.H. Lu, M.S. Yates

The authors would like to acknowledge Dr. Jared Burks (Co-Director of the Flow Cytometry and Cellular Imaging Core Facility) for his expert assistance with multispectral imaging and quantitation systems.

This work was supported by the MD Anderson Uterine Cancer SPORE (P50CA098258; to K.H. Lu and R.R. Broaddus), a T32 training grant for gynecologic oncology (CA101642; to K.H. Lu), and by the MD Anderson Cancer Center Support Grant (CA016672) that supports the Flow Cytometry and Cellular Imaging Facility and Sequencing and Microarray Core Facility.

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.