In the past decade, the development of high-throughput sequencing technologies and computational algorithms has facilitated the understanding of cancer genomics. In endometrial cancer (EC), the establishment of molecular subtypes by the Cancer Genome Atlas (TCGA) consortium (1), on the one hand, has affected patient stratification, promoting individualized clinical management. In 2021, the National Comprehensive Cancer Network (NCCN) guidelines for uterine neoplasms recommended molecular subtyping in EC diagnosis (2). In addition, the European Society of Gynecologic Oncology (ESGO)/European Society for Radiotherapy and Oncology (ESTRO)/European Society of Pathology (ESP) guideline for EC further incorporated molecular subtypes into the risk stratification system for guiding postoperative adjuvant therapies (3).

On the other hand, molecular subtypes, to some extent, also indicated possibly different routes of EC development and differences in cancer microenvironmental features. Specifically, immune components in the cancer microenvironment have attracted increasing attention in recent years due to their potential roles in predicting patients’ prognosis and guiding immune checkpoint blockade therapies (4, 5). Improvements in methodologies, including single-cell and spatial transcriptomics, immune deconvolution algorithms (6) and multiplex immunofluorescence assays (7), have significantly promoted research in cancer immune microenvironment. In 2018, European researchers, for the first time, established an immune risk score based on tumor-infiltrating T cells in colon cancer tissue and validated its effectiveness in predicting recurrence in a large retrospective cohort (8). In EC, previous findings have indicated the prognostic value of immune-related gene signatures (9–11). However, most previous studies on the immune microenvironmental features of EC were only based on data of next-generation sequencing. Furthermore, extensive data regarding the association of tumor-infiltrating immune cells with patients’ molecular features are still needed to establish incorporated risk stratification systems for clinical applications.

In this study, we aimed to analyze the tumor immune microenvironmental features in different molecular subtypes of EC using multiplex immunofluorescence method, so as to provide a better understanding of the mechanisms underlying the differences in prognosis and immunotherapeutic efficacy among different EC subtypes.

EC is one of the most common gynecologic malignancies worldwide (26). Personalized treatment strategies for EC are essential both for better precision care and for reducing treatment-related health economic burdens. With the emerging trend of applying immunotherapies in EC treatment, and the increasing evidence indicating the potential role of immunological features in predicting treatment responses (27–29), understanding tumor immune microenvironmental features and the associations with molecular features of cancer is necessary to guide immunotherapy design and predict patients’ prognosis. In this study, we compared the clinicopathological features of different molecular subtypes of primary EC. Additionally we systemically analyzed the association of EC molecular subtypes with tumor immune microenvironmental features using experimental approaches. The information provided here could be informative for the design of relevant basic and clinical studies in the future.

Our data revealed that TP53 mutant EC was associated with non-endometrioid histology, advanced stage, and multiple negative prognostic factors, indicating compromised survival outcomes. These results were in accordance with previous studies (20, 30–33). The following analysis of the tumor immune microenvironment could, to some extent, provide explanations for this. Indeed, in recent years, studies based on cancer genomics have indicated that immunological subtypes could be used to predict patient prognosis (5). There are also available models for predicting clinical response to immune checkpoint blockade therapies (34, 35). Nevertheless, current models are still not sufficiently accurate, and further explorations are warranted.

Previous studies have indicated that increased infiltration of intratumoral CD8⁺ T cells is associated with a better prognosis in different cancer types (36–38), and a recent study based on multiplex immunofluorescent assays further supported the prognostic value of tumor infiltrating T cells in early-stage endometrial cancer (39). In this study, we also observed relatively high percentages of CD8⁺ T cells in samples of TMB-H EC, which is believed to have a favorable survival outcome (23). However, studies have shown that the functional status of CD8⁺ T cells changes with tumor progression, and different stages of dysfunctional T cells, characterized by the expression of specific immune checkpoint molecules, are thought to be associated with distinct response rates to immune checkpoint inhibitors (40). In this study, we noticed that TP53 mutant EC showed abundant infiltration of CD8⁺ PD-1⁺ T cells in the tumor parenchyma and a high proportion of CD8⁺ T cells with PD-1 expression in the tumor mesenchyme, indicating unfavorable clinical outcomes of TP53 mutant EC.

Innate immune cells, including macrophages and NK cells, could also modulate the tumor immune microenvironment and regulate antitumor responses. Macrophages can be further divided into M1 macrophages and M2 macrophages based on cell surface markers and functions. M1 macrophages mainly show antitumoral functions, while M2 macrophages promote tumor progression via stimulating tumor cell proliferation, angiogenesis, and epithelial-mesenchymal transitions (41). According to our data, a trend of higher fractions of M2 macrophages in TP53 mutant EC relative to the other molecular subtypes was observed. This could possibly help explain the higher rate of advanced-stage diseases in this subtype as a result of tumor progression. Besides, NK cells could also be divided into two types, CD56^dim and CD56^bright (CD56 brightly stained NK cells). CD56^bright NK cells are mainly responsible for secreting cytokines, while CD56^dim NK cells show more potent cytotoxic effects (42). One recent study indicated that low NK cell infiltration in the tumor was associated with worse survival (43). However, based on our data, the fraction of CD56^dim NK cells was not significantly different among different EC molecular subtypes, possibly due to the limited sample size.

T_reg cells are another vital cell type with prognostic significance. Evidence has shown that high T_reg cell infiltration level is associated with compromised survival and hyperprogression of disease following immune checkpoint blockade therapy (44, 45). But interestingly enough, in this study, a higher proportion of T_reg cells was observed in the tumor mesenchyme of the TMB-H subtype, instead of the TP53 mutant subtype, which are commonly thought to have poor survival. According to previous studies, T_reg cells responsible for modulating immune responses typically express the same transcription factors, which are also expressed in the cells that they regulate; moreover, in antitumor immune responses, T_reg cells could be triggered by the same chemotaxis molecules that also recruit CD8⁺ T cells to the tumor site (46). Therefore, the higher proportion of CD8⁺ T cells in the tumor mesenchyme in TMB-H tumors may have contributed to the higher proportion of T_reg cells. Furthermore, the similar ratio of T_reg cells to CD8⁺ T cells in the three molecular subtypes also supported the above hypothesis.

In summary, EC patients with high TMB showed abundant tumor-infiltrating CD8⁺ T cells and relatively high levels of PD-L1 expression in tumor cells, which is consistent with data from previous studies on POLE mutant and MSI-H EC (47–49). The above indicates that there are strong antitumor immune responses in TMB-H tumors and that this subtype is potentially suitable for immune checkpoint blockade therapies. In the NSMP subtype, the TMB, the proportions of multiple tumor-infiltrating immune cells and the expression levels of immune checkpoint molecules were low, indicating a lack of effective antitumor immune responses. In the TP53 mutant subtype, the TMB level was low. However, the proportions of T_reg cells, M2 macrophages, PD-L1⁺ CD68⁺ macrophages and CD8⁺ PD-1⁺ T cells were relatively high, indicating a strong immune suppressive microenvironment in this molecular subtype. Based on the immune microenvironmental features analyzed above, we summarize the immune phenotype of the three molecular subtypes as normal immune response, absence of immune infiltration, and suppressed immune response.

The distribution pattern of immune cells in the tumor tissue could also indicate differences in their functions. Recently, Keren et al. (50) proposed a model for describing immune infiltrative patterns in the tumor: compartmentalized pattern, which suggested that tumor cells and immune cells form relatively independent regions; cold pattern, which indicated low levels of tumor infiltrating immune cells; and mixed pattern, which implied the highly mixed distribution of tumor and immune cells. Among these, patients with the compartmentalized pattern showed better survival than those with the mixed pattern (50). In this study, we also analyzed the distribution of immune cells in the tumor parenchyma and mesenchyme. Our study showed that CD8⁺ T cells were distributed in both tumor parenchymal and mesenchymal regions in TMB-H EC. The fraction of CD8⁺ T cells in the tumor mesenchyme was the highest in the TMB-H subtype, which means that a part of T cells could form a relatively isolated region adjacent to the tumor cells. Among the three molecular subtypes, the NSMP subtype displayed the lowest degree of CD8⁺ T cell infiltration. In the TP53 mutant subtype, CD8⁺ T cells were mostly distributed in tumor parenchyma, with a relatively low median level of infiltration in tumor mesenchyme. The features of CD8⁺ T cell distribution in the three molecular subtypes, to some extent, resemble that of the model mentioned above (50). Besides, TP53 mutant EC typically shows the worst survival (1, 20, 30–33) across all subtypes, which is also consistent with the survival features revealed by the above model (50). Another recent study analyzed the interactions of cellular neighborhoods in colorectal cancer with distinct immune infiltrative features (51). In brief, in tumors with numerous tertiary lymphoid structures, T cell exchange between the T cell cluster and tumor invasive front could help enhance the antitumor immune response; while in tumors with diffuse inflammatory infiltration, the immune suppressive macrophage cluster showed strong contact with the tumor invasive front and inhibited effective immune responses (51). The mechanisms described above might also explain the distinct survival features in EC of different molecular subtypes and immune infiltrative patterns.

Notably, in this study, both the TMB-H subtype and TP53 mutant subtype showed high levels of immune checkpoint molecule expression. However, in the TMB-H subtype, PD-L1 was mainly expressed in tumor cells, while in the TP53 mutant subtype, there were high levels of PD-L1 expression in macrophages and high levels of PD-1 expression in T cells. These results indicate that the cellular distribution of immune checkpoint molecules could also provide information on the immune response status of the patient and differences in patients’ responses to immune checkpoint blockade therapies. Further studies with larger sample sizes are needed to confirm the findings.

In this study, we quantitively analyzed tumor immune microenvironmental features in different EC molecular subtypes and provided information about the spatial distributions of multiple immune cell types in the tumor tissue. A key strength is that we adopted experimental methods for in situ visualization of the cells, which showed direct evidence for tumor immune infiltrations. In this regard, this study could provide a vital supplement to previous studies based on bulk tissue sequencing and computational deconvolution. By systemically analyzing the TMB, tumor infiltrating immune cells and immune checkpoint molecules, we summarized the immunophenotypes of different EC molecular subtypes, thus providing clues for understanding their distinct survival features and treatment responses. However, there are also some limitations. First, the sample size of the study was relatively small, which limited the statistical power in some analyses. Studies with larger sample sizes are needed to further validate our findings. Second, since most cases in this study were treated recently and follow-ups are on-going, survival information is still lacking. Long-term follow-up is necessary to analyze the associations of tumor immune microenvironmental features with patients’ recurrence, survival and responses to multiple treatment modalities. Finally, as an explorative study, the panel we used for testing tumor-infiltrating immune cells included multiple cellular markers. Further explorations are needed to refine the testing strategies and develop clinically feasible panels for better practical applications.

Currently, with the deep clinical influence of TCGA molecular subtyping and its surrogate methods (1, 20, 31), the diagnosis and treatment of EC are becoming increasingly more comprehensive and individualized. Molecular markers provide vital information and rationality for applying targeted drugs and immune checkpoint blockade therapies in specific patient groups (52). Nevertheless, heterogeneity in prognosis and treatment response could still be seen even within the same pathological and molecular subtype, which urges further refinement of the current risk stratification system (3). During this effort, barriers still exist that the cellular architectures of the tumor tissues and the biological behaviors of malignant and surrounding stromal and immune cells are far less understood in EC. As indicated above (50, 51), this information may also be highly associated with clinical outcomes. The results from this study, on the one hand, established the connection between molecular subtypes and the immune microenvironmental features of EC. On the other hand, it paved the way for further designing related studies. We encourage more efforts using multiplex imaging methods to establish a prognostic or treatment-related classification system based on immunological markers. Based on these efforts, incorporating effective immunological features into current EC patients’ risk stratification systems would be another vital step for better individualized treatment.

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (53) in National Genomics Data Center (54), China National Center for Bioinformation / Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA003458) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human. The raw data of multiplex immunofluorescence assays have been deposited in the OMIX, China National Center for Bioinformation / Beijing Institute of Genomics, Chinese Academy of Sciences (https://ngdc.cncb.ac.cn/omix: accession no. PRJCA013221).

The studies involving human participants were reviewed and approved by Institutional Review Board of Peking University People’s Hospital. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

YD and ZW contributed to study conception and design. Development of methodology was performed by YD, DH, LC, LZ and ZW. Acquisition of data was conducted by YD, LZ, XZ, NK, LQ, LL and HL. Data analysis was performed by YD, DH and LZ. The first draft of the manuscript was written by YD, and the manuscript was reviewed and revised by LC, JW and ZW. All authors contributed to the article and approved the submitted version.

This work was supported by the National Key Technology Research and Developmental Program of China (2022YFC2704300, 2022YFC2704303), the National Natural Science Foundation of China (81972426) and the Capital’s Funds for Health Improvement and Research (2022-2Z-4086, 2022-1-4081).

We would like to thank Dr. Xiaochen Zhao from 3D Medicines Inc., for her help in the development of methodology and data interpretation, and Dr. Ting Bei, who is from 3D Medicines Inc., for her help in the review of this manuscript.

Authors DH and LC were employed by 3D Medicines Inc.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.