Association between EC molecular subtypes and survival and clinical features
To explore the relationship between the distinct molecular subtypes of endometrial cancer and the clinical features, we analyzed clinical data of 500 EC patients from TCGA database, including clinical stage, histologic grade, age, history of radiation therapy, diabetes, hypertension, menopause state and BMI (Table 1). The 500 patients were classified into three molecular subgroups: the POLE, MSI and Other subgroup. We discovered that the POLE subgroup has a lower age, BMI, lower fraction of diabetes and hypertension. The tumors in MSI subgroup were identified more often in the early stage, and had a lower age than the Other subgroup (Figure 1A-1B). These results were consistent with the prognosis analysis between patient survival and molecular subtype, age and clinical stage (Figure 1A-1C). But there is no difference in the proportion of histologic grade across the subtypes (Figure 1A). The POLE and MSI subtypes, lower age, grade and clinical stage were all correlated with better patient survival (Figure 1C).
Landscape of somatic mutation files in POLE, MSI, and Other ECs
Tumors with POLE mutation or MSI have been suggested to be hyper or highly mutated. The somatic mutation profiles of 500 EC patients in the TCGA database were obtained from MuTect2” analysis. The top 50 mutated genes were shown in the waterfall plot; around 498 (99.6%) samples in total possessed somatic mutations, and most of the genes had higher mutation frequency in the POLE group (Figure 2A and supplementary table 1). The MSI group had few mutations in TP53, FBXW7, CTNNB1 and PPP2R1A, which was consistent with previously described (Figure 2B) [4]. 6 genes showed more frequent mutations in MSI group, including 4 genes (KRAS, ARID1A, JAK1 and RNF43) [4, 16] that have been previously reported in endometrial cancer and 2 novel genes (KMT2D and SETD1B) (Figure 2B). Except for KRAS with predominantly missense mutation in all groups, other 5 genes showed more frequent frameshift deletions in MSI group than POLE and other groups (Figure 2C). JAK1 and RNF43 with polymerase slippage-associated deletions have been reported previously [16]. KMT2D and SETD1B are chromatin remodeling-related genes, and appear to help predict the degree of myometrial invasion [17].
Enriched immune infiltrates in POLE and MSI ECs than Other ECs
We analyzed the global gene expression profile of 500 EC patients, and identified the differentially expressed genes (DEGs) between the three molecular groups. POLE group and MSI group share a similar expression profile (Figure S1). 1138 genes were upregulated in both POLE and MSI groups, and 792 genes were upregulated in Other group (Supplementary table 1). The gene set enrichment analyses (GSEA) using Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) terms showed that genes up-regulated in POLE and MSI samples were mainly enriched for immune-related functions, such as T cell activation, T cell differentiation, cytokine-cytokine receptor interaction and T cell receptor signaling pathway (Figure 3).
POLE mutation or MSI subtypes in cancers indicate hypermutations, and they are suggested to harbor more tumor-specific neoantigens and higher lymphocytes infiltrates [18]. Thus, we calculated the immune scores of all 500 samples using the ESTIMATE algorithm, and found that immune scores of both POLE and MSI EC groups were significantly higher than the Other group (Figure 4A). To investigate the association between molecular subtypes and specific immune infiltrates in endometrial cancer, we then estimated the percent of 22 immune cell types of all 500 samples by CIBERSORT algorithm and compared immune cell fractions in 3 molecular subtype groups (Figure 4B). Furthermore, the Kaplan-Meier curve with log-rank test was used to analyze the correlation between each immune infiltrate level and the overall survival (OS) time of EC patients. The fractions of 10 immune cell types varied across molecular subtypes (Figure 4C). Importantly, both POLE and MSI groups were found to have more CD8+ T cells, follicular helper T cells, resting NK cells, M1 macrophages, while less activated NK cells, M2 macrophages, activated dendritic cells and mast cells (Figure 4C). POLE groups had more plasma cells (Figure 4C). MSI group had more regulatory T cells (Figure 4C). Among these immune cell types, high infiltrates level of CD8+ T cells, and regulatory T cells were found to be significantly positively related with OS, while high infiltrates level of M2 macrophages and activated dendritic cells were negatively related with OS (Figure 4D). These results were consistent with better clinical outcomes in POLE and MSI EC groups than Other group (Figure 1C).
Cell mapping of immune ecosystem in an MMR-D EC sample
Our previous study [15] also indicated high lymphocytes infiltration in an MMR-D cancer tissue whose MSH6 expression is negative (Figure S2A). The existence of different tumor-infiltrating immune cell types in this sample was also assessed by immunohistochemistry (IHC) staining with CD3, CD8, CD20, CD68 and CD163 antibody (Figure 5A). We further analyzed the cellular composition and immune microenvironment of this sample deeply based on its single-cell RNA–seq data using Seurat R package (Figure 5B). After quality filtering, 4518 cells were used for downstream analysis Graph-based clustering was used to classify cells into groups. We identified 4 major cell types through marker genes: epithelial cells (KRT18), endothelial cells (VWF), fibroblasts (COL6A2), immune cells (PTPRC) in this sample (Figure S2B). The immune cells were further analyzed by identifying subsets in T cells, B cells and myeloid cells (Figure 5C). T cells included naive CD4+ T cells (cluster 5; IL7R+), regulatory T cells (cluster 6; FOXP3+), exhausted CD8+ T cells (cluster 0; PDCD1+) and natural killer T cells (NKT cells) (cluster 3; GNLY+) (Figure 5C-D and Figure S3). Cluster 11 had proliferative cells (MKI67+) of both lymphocyte and myeloid lineages (Figure 5E). Two distinct populations in macrophages (CD163+) were observed: macrophage population 1 (cluster 1) with enriched expression of C1QA, C1QB, C1QC; and macrophage population 2 (cluster 7) with enriched expression of FTL, CXCL8, SPP1 (Figure 5C, 5F and Figure S3). One cluster of monocytes was characterized by enriched expression of LYZ, S100A8 and S100A9 (Figure 5C and Figure S3) [19]. Two subtypes in B cells were further analyzed. Follicular B cells were enriched for the expression of CD19 and MS4A1 [20], while plasma B cells were characterized by the expression of MZB1 and CD38 (Figure 5C, 5G and Figure S2) [19]. These results formed the basis description of the immune subsets in endometrial carcinomas with MSI.
Brooke E et al. reported that the MSI tumors exhibited higher infiltration of CD3+ and CD8+ T cells compared to MSS tumors [21]. However, little is known about the relationship between MSI status and tumor infiltrating B cells. Here, we assessed the infiltration of B cells by IHC staining of MMR-D and MMR-I tumor sections with CD20 antibody (Figure 5H). The MMR-D tumors exhibited a higher trend in the number of B cells than MMR-I tumors, but with no significance (Figure 5I). Patient overall survival analysis from TCGA-UCEC data indicated that higher expression of CD20 (MS4A1) was associated with good prognosis (Figure 5J). The function of B cells in endometrial carcinomas remains to be further explored.