The association between the tumor immune microenvironments and clinical outcome in low-grade, early-stage endometrial cancer patients

Activated immune infiltrates expand opportunities for targeted therapy in p53-abnormal endometrial carcinoma

Tumor protein p53 mutated/abnormal (p53abn) endometrial carcinomas account for over 50% of deaths but comprise only 15% of all endometrial carcinomas. Most patients show limited response to standard-of-care chemotherapy with or without radiotherapy, and only a minority of cases are amenable to targeted therapies like poly-ADP ribose polymerase (PARP) inhibitors and HER2-directed therapies. Recent immunotherapy clinical trials have demonstrated remarkable efficacy, not only in mismatch repair deficient (MMRd) tumors but also in a subset of mismatch repair-proficient (MMRp) tumors. However, the immune microenvironment and its relationship to other therapeutic targets in MMRp endometrial carcinoma remains poorly understood. Here, we characterize the immune microenvironment of p53abn endometrial carcinoma, the most clinically aggressive subtype of MMRp endometrial carcinoma, and correlate antitumor immune signatures with other targetable alterations. We accrued 256 treatment-naïve p53abn endometrial carcinomas and systemically profiled T-cell, B-cell, myeloid, and tumor-cell populations with multiplex immunofluorescence to assess the tissue localization and functional status of immune cells. Shallow whole-genome sequencing was performed on a subset of 126 cases. Patterns of immune infiltration were compared to survival outcomes and mutational signatures. Mixture modeling divided p53abn endometrial carcinoma into tumor-infiltrating lymphocyte (TIL)-rich and TIL-poor subsets. Over 50% of tumors were TIL-rich. TIL-rich cases overexpressed targetable immune evasion molecules and were associated with longer overall and disease-specific survival in multivariate analysis. This effect was particularly pronounced in advanced stage disease and in patients who did not receive adjuvant chemotherapy. TIL did not associate with homologous recombination deficient mutational signatures or HER2 amplification. Our findings demonstrate a biological rationale for immunotherapy in a substantial subset of patients with p53abn endometrial cancer and may help inform combination therapies with immune checkpoint inhibition, PARP inhibitors, and anti-HER2 agents. © 2025 The Author(s). The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

4-1BB agonist targeted to fibroblast activation protein α synergizes with radiotherapy to treat murine breast tumor models

BACKGROUND

Ionizing radiation (IR) is a double-edged sword for immunotherapy as it may have both immunosuppressive and immunostimulatory effects. The biological effects of IR on the tumor microenvironment (TME) are a key factor for this balance. Fibroblast activation protein (FAP) is expressed on the surface of cancer-associated fibroblasts (CAF) in many cancer types and its abundance is associated with the poor immune response to immune-checkpoint-blockade in patients. We hypothesized that IR increases FAP expression in CAFs, therefore the combination of IR with targeted immunomodulators such as an agonistic anti-FAP-4-1BBL fusion protein could enhance the immune-mediated antitumoral effects of these treatments.

METHODS

The murine transplantable TS/A tumor-cell-line co-engrafted with CAFs was used to investigate increases in FAP expression in tumors following irradiation using immunohistochemistry, real-time polymerase chain reaction (RT-PCR) and multiplex tissue immunofluorescence. One lesion of bilateral tumor-bearing mice was only locally irradiated or combined with weekly injections of the bispecific muFAP-4-1BBL fusion protein (a mouse surrogate for RG7826). Tumor sizes were followed over time and TME was assessed by flow cytometry. Selective monoclonal antibody (mAb)-mediated depletions of immune cell populations, neutralizing interferon alpha/beta receptor 1 (IFNAR-I) IFNAR and interferon (IFN)-γ mAbs and gene-modified mice (4-1BB-/-) were used to delineate the immune cell subsets and mechanisms required for efficacy. 67Ga labeled muFAP-4-1BBL tracked by SPECT-CT was used to study biodistribution. In human colorectal carcinoma samples, the inducibility of FAP expression following radiotherapy was explored by multiplex immunofluorescence.

RESULTS

Irradiation of TS/A+CAF tumors in mice showed an increase in FAP levels after local irradiation. A suboptimal radiotherapy regimen in combination with muFAP-4-1BBL attained primary tumor control and measurable abscopal effects. Immune TME landscape analyses showed post-treatment increased infiltration of activated immune cells associated with the combined radioimmunotherapy treatment. Efficacy depended on CD8+ T cells, type I IFN, IFN-γ and ability to express 4-1BB. Biodistribution studies of muFAP-4-1BBL indicated enriched tumor targeting to irradiated tumors. Human colorectal cancer samples pre and post irradiation showed enhanced FAP expression after radiotherapy.

CONCLUSION

Increased FAP expression in the TME as a result of radiotherapy can be exploited to target agonist 4-1BB immunotherapy to malignant tumor lesions using an FAP-4-1BBL antibody fusion protein.

Impact of Timing the Combination of Radiotherapy and PD-1 Inhibitors on Outcomes in Patients with Hepatocellular Carcinoma

Purpose

The optimal timing for combining radiotherapy with immunotherapy in patients with hepatocellular carcinoma (HCC) remains uncertain and affects treatment efficacy and patient outcomes. This study aimed to evaluate and compare the efficacy and treatment-related adverse events (TRAEs) of synchronously administered radiotherapy and programmed cell death protein (PD)-1 inhibitors and sequential administration in patients with HCC.

Patients and Methods

We retrospectively enrolled 67 patients with HCC who were undergoing liver radiotherapy and PD-1 inhibitor therapy at two medical centers between July 2017 and April 2023. Additionally, we created an experimental tumor model using 6-week-old female mice to validate our findings.

Results

In the concurrent group, the median overall survival was indefinite; however, it was 13 months in the sequential group (95% confidence interval [CI] 6.7-19.3 months, P=0.010). The median progression-free survival was significantly longer in the concurrent group (12 months, 95% CI 9.5-14.5 months) than in the sequential group (7 months, 95% CI 1.3-12.7 months; P=0.043). Grade 3/4 TRAEs occurred in 30.4% (concurrent) and 28.6% (sequential) of patients without any treatment-related deaths. In the mouse model, synchronous treatment significantly inhibited tumor growth compared to sequential treatment (293.4±45.18 mm3 versus 602.7±41.68 mm3; P=0.001). Flow cytometry revealed an increased Tregs/CD3+ T cell ratio and a decreased CD8+/Treg cell ratio post-radiotherapy, suggesting an immunosuppressive tumor microenvironment.

Conclusion

Synchronous treatment demonstrated superior efficacy in treating HCC compared to sequential treatment, with manageable adverse events. The rapid increase in Tregs after radiotherapy may contribute to the reduced efficacy of sequential radiotherapy plus PD-1 inhibitors.

Spatial Cancer-Immune Phenotypes Predict Shorter Recurrence-Free Survival in the No Specific Molecular Profile Molecular Subtype of Endometrial Carcinoma

Compartmentation of the immune response into 3 main spatial cancer-immune phenotypes (SCIs) - inflamed, excluded, and desert - has been proposed as the main predictor of response to immune checkpoint inhibitors in solid tumors. The objective of the study was to define and characterize the SCI in a consecutive series of 213 endometrial carcinomas (ECs) by correlating it with molecular subtypes, clinicopathologic features, and prognosis. Immunohistochemistry (IHC) and next-generation sequencing were used to assign surrogate molecular EC subtypes: POLE mutant (POLE), mismatch repair deficient (MMRd), TP53 mutant (p53abn), and no specific molecular profile (NSMP). Immune cell markers (CD20, CD3, CD8, CD68, PD-L1) were assessed by IHC on whole sections and quantified by digital image analysis to define the 3 SCIs. ECs were stratified into 4 molecular subtypes: 17 (8.0%) POLE, 68 (31.9%) MMRd, 42 (19.7%) p53abn, and 86 (40.4%) NSMP. SCI determination showed 105 (49.3%) inflamed, 62 (29.1%) desert, and 46 (25.6%) excluded tumors. The inflamed phenotype was more prevalent in MMRd (64.7%) and POLE (76.5%) subtypes compared with NSMP (45.3%) and p53abn (21.4%). SCI revealed a strong correlation with disease-free survival in NSMP tumors: inflamed 96.2%, desert 83.2%, and excluded 40.5%. The SCI prognostic impact was also maintained in NSMP cases treated with adjuvant therapy resulting in a significant difference in recurrence between the inflamed and excluded phenotypes. To simplify SCI determination, a subset of immune cell markers was selected as appropriate to define the 3 SCI patterns: high intraepithelial CD8 for the inflamed phenotype; CD68, CD20, and PD-L1 to discriminate between desert and excluded tumors. The integration of SCI into molecular classification could be a promising opportunity to improve the prognostic risk stratification of patients and may guide the therapeutic approach, particularly in the NSMP subtype. Thus, the different patterns of immune response are a new prognostic parameter in the NSMP subtype.

ENGOT-en11/GOG-3053/KEYNOTE-B21: a randomised, double-blind, phase III study of pembrolizumab or placebo plus adjuvant chemotherapy with or without radiotherapy in patients with newly diagnosed, high-risk endometrial cancer

BACKGROUND

Pembrolizumab plus chemotherapy provides clinically meaningful benefit as first-line therapy for advanced (locoregional extension and residual disease after surgery)/metastatic/recurrent mismatch repair-proficient (pMMR) and mismatch repair-deficient (dMMR) endometrial cancer, with greater magnitude of benefit in the dMMR phenotype. We evaluated the addition of pembrolizumab to adjuvant chemotherapy (with/without radiation therapy) among patients with newly diagnosed, high-risk endometrial cancer without any residual macroscopic disease following curative-intent surgery.

METHODS

We included patients with histologically confirmed high-risk [International Federation of Gynecology and Obstetrics (FIGO) stage I/II of non-endometrioid histology or endometrioid histology with p53/TP53 abnormality, or stage III/IVA of any histology] endometrial cancer following surgery with curative intent and no evidence of disease postoperatively, with no prior radiotherapy or systemic therapy. Patients were randomised to pembrolizumab 200 mg or placebo every 3 weeks (Q3W) for six cycles added to carboplatin-paclitaxel followed by pembrolizumab 400 mg or placebo every 6 weeks (Q6W) for six cycles per treatment assignment. Radiotherapy was at the investigator's discretion. The primary endpoints were investigator-assessed disease-free survival (DFS) and overall survival in the intention-to-treat population.

RESULTS

A total of 1095 patients were randomised (pembrolizumab, n = 545; placebo, n = 550). At this interim analysis (data cut-off, 4 March 2024), 119 (22%) DFS events occurred in the pembrolizumab group and 121 (22%) occurred in the placebo group [hazard ratio 1.02, 95% confidence interval (CI) 0.79-1.32; P = 0.570]. Kaplan-Meier estimates of 2-year DFS rates were 75% and 76% in the pembrolizumab and placebo groups, respectively. The hazard ratio for DFS was 0.31 (95% CI 0.14-0.69) in the dMMR population (n = 281) and 1.20 (95% CI 0.91-1.57) in the pMMR population (n = 814). Grade ≥3 adverse events (AEs) occurred in 386 of 543 (71%) and 348 of 549 (63%) patients in the pembrolizumab and placebo groups, respectively. No treatment-related grade 5 AEs occurred.

CONCLUSIONS

Adjuvant pembrolizumab plus chemotherapy did not improve DFS in patients with newly diagnosed, high-risk, all-comer endometrial cancer. Preplanned subgroup analyses for stratification factors suggest that pembrolizumab plus chemotherapy improved DFS in patients with dMMR tumours. Safety was manageable.

TRIAL REGISTRATION

ClinicalTrials.gov, NCT04634877; EudraCT, 2020-003424-17.

RESEARCH SUPPORT

Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.

MHC class I and II-deficient humanized mice are suitable tools to test the long-term antitumor efficacy of immune checkpoint inhibitors and T-cell engagers

BACKGROUND

Immunodeficient mice engrafted with peripheral blood mononuclear cells (PBMCs) are models to study new cancer immunotherapy agents. However, this approach is associated with xenograft-versus-host disease (xGVHD), which starts early after PBMC transfer and limits the duration and interpretation of experiments. Here, we explore different approaches to overcome xGVHD and better support the development of cancer immunotherapies.

METHODS

Immunodeficient NOD-scid IL2Rgnull (NSG) mice were intravenously transferred with human PBMCs and subcutaneously co-engrafted with HT29 human colon carcinoma cells. Diverse strategies to reduce xGVHD while preserving the antitumor activity of human immune cells were evaluated: (1) ex vivo immune graft modification by depleting CD4+ T cells pre-transfer using magnetic beads, (2) post-transplantation cyclophosphamide administration to eliminate proliferating xenoreactive T-cell clones and (3) using major histocompatibility complex (MHC) class I and II-deficient NSG mice: (Kb Db)null (IA)null (MHC-dKO NSG). Body weight and plasma murine alanine aminotransferase levels were measured as indicators of xGVHD and tumor size was measured every 2-3 days to monitor antitumor activity. The antitumor effects and pharmacodynamics of nivolumab plus ipilimumab and an anti-epithelial cell adhesion molecule (EpCAM)/CD3 T-cell engager (αEpCAM/CD3 bispecific antibody (BsAb)) were evaluated in the model.

RESULTS

CD4+ T-cell depletion attenuates xGVHD but also abrogates the antitumor activity. Cyclophosphamide limits the antitumor response and does not substantially prevent xGVHD. In contrast, xGVHD was significantly attenuated in MHC-dKO NSG recipients, while the antitumor effect of human PBMCs was preserved. Furthermore, the administration of nivolumab plus ipilimumab caused exacerbated xGVHD in conventional NSG mice, thereby precluding the observation of their antitumor effects. Severe xGVHD did not occur in MHC-dKO NSG mice thus enabling the study of complete and durable tumor rejections. Similarly, NSG mice treated with an αEpCAM/CD3 BsAb showed complete tumor regressions, but died due to xGVHD. In contrast, MHC-dKO NSG mice on treatment with the αEpCAM/CD3 BsAb achieved complete tumor responses without severe xGVHD. A significant proportion of mice rendered tumor-free showed tumor rejection on rechallenge with HT29 cells without further treatment. Finally, tumor-infiltrating CD8+ T-cell number increase, activation and CD137 upregulation were observed on αEpCAM/CD3 BsAb treatment.

CONCLUSION

Humanized MHC-dKO immunodeficient mice allow and refine the preclinical testing of immunotherapy agents for which experimentation is precluded in conventional immunodeficient mice due to severe xGVHD.

Correlation of BUB1 and BUB1B with the development and prognosis of endometrial cancer

This study aimed to evaluate the expression and clinical significance of budding uninhibited by benzimidazole 1 (BUB1) and BUB1 mitotic checkpoint serine/threonine kinase B (BUB1B) in endometrial carcinoma (EC). BUB1 and BUBIB expressions were evaluated by bioinformatics. Protein expression, clinical features, prognosis and immune cell infiltration were explored in 20 EC tumors. siRNA was used to evaluate BUB1 and BUBIB function in EC cells. BUB1 and BUBIB were highly expressed in 26 cancers. BUB1 was associated with overall survival (OS) in eight cancers and disease-free survival in ten; BUB1B was associated with OS in nine cancers and DFS in eleven. BUB1 and BUBIB exhibited high frequencies of gene changes (mainly mutations, > 5%) in cancer. BUB1 was negatively correlated and BUB1B was positively correlated with cancer-associated fibroblasts and endothelial cell infiltration. BUB1 and BUBIB knockdown decreased migration and invasion in EC cells. High BUB1 expression correlated with tumor malignant phenotypes (P < 0.05). High BUB1 mRNA expression reduced OS (P = 0.00036) and recurrence-free survival (P = 0.0011). High BUB1B mRNA expression reduced OS (P = 0.0024). BUB1/BUB1B correlated with activated CD8 + T and CD4 + T cell infiltration. BUB1 and BUBIB are highly expressed and correlated with clinicopathological characteristics in EC. BUB1 and BUBIB are potential prognosis markers and immunotherapy targets.

Endometrioid Endometrial RNA Index Predicts Recurrence in Stage I Patients

PURPOSE

Risk prediction with genomic and transcriptomic data has the potential to improve patient outcomes by enabling clinicians to identify patients requiring adjuvant treatment approaches, while sparing low-risk patients from unnecessary interventions. Endometrioid endometrial carcinoma (EEC) is the most common cancer in women in developed countries, and rates of endometrial cancer are increasing.

EXPERIMENTAL DESIGN

We collected a 105-patient case-control cohort of stage I EEC comprising 45 patients who experienced recurrence less than 6 years after excision, and 60 Fédération Internationale de Gynécologie et d'Obstétrique grade-matched controls without recurrence. We first utilized two RNA-based, previously validated machine learning approaches, namely, EcoTyper and Complexity Index in Sarcoma (CINSARC). We developed Endometrioid Endometrial RNA Index (EERI), which uses RNA expression data from 46 genes to generate a personalized risk score for each patient. EERI was trained on our 105-patient cohort and tested on a publicly available cohort of 263 patients with stage I EEC.

RESULTS

EERI was able to predict recurrences with 94% accuracy in the training set and 81% accuracy in the test set. In the test set, patients assigned as EERI high-risk were significantly more likely to experience recurrence (30%) than the EERI low-risk group (1%) with a hazard ratio of 9.9 (95% CI, 4.1-23.8; P < 0.001).

CONCLUSIONS

Tumors with high-risk genetic features may require additional treatment or closer monitoring and are not readily identified using traditional clinicopathologic and molecular features. EERI performs with high sensitivity and modest specificity, which may benefit from further optimization and validation in larger independent cohorts.

New insights into immune cells in cancer immunotherapy: from epigenetic modification, metabolic modulation to cell communication

Cancer is one of the leading causes of death worldwide, and more effective ways of attacking cancer are being sought. Cancer immunotherapy is a new and effective therapeutic method after surgery, radiotherapy, chemotherapy, and targeted therapy. Cancer immunotherapy aims to kill tumor cells by stimulating or rebuilding the body's immune system, with specific efficiency and high safety. However, only few tumor patients respond to immunotherapy and due to the complex and variable characters of cancer immune escape, the behavior and regulatory mechanisms of immune cells need to be deeply explored from more dimensions. Epigenetic modifications, metabolic modulation, and cell-to-cell communication are key factors in immune cell adaptation and response to the complex tumor microenvironment. They collectively determine the state and function of immune cells through modulating gene expression, changing in energy and nutrient demands. In addition, immune cells engage in complex communication networks with other immune components, which are mediated by exosomes, cytokines, and chemokines, and are pivotal in shaping the tumor progression and therapeutic response. Understanding the interactions and combined effects of such multidimensions mechanisms in immune cell modulation is important for revealing the mechanisms of immunotherapy failure and developing new therapeutic targets and strategies.

The Clinical and Pathological Characteristics of POLE-Mutated Endometrial Cancer: A Comprehensive Review

Endometrial cancer shows high histological and molecular heterogeneity. The POLE mutation is a significant molecular alteration in endometrial cancer, leading to the identification of a specific subtype known as POLE-mutated endometrial cancer. This subtype exhibits a high tumor mutation burden, abundant lymphocyte infiltration, and a favorable prognosis, making it a promising candidate for immune checkpoint inhibitor therapy. This paper presents a comprehensive review of the clinical and pathological characteristics, outcomes, treatment advancements, pathogenic POLE gene detection, and alternative testing methods for POLE-mutated endometrial cancer.

Tumor-infiltrating immune cell score as an independent prognostic predictor for endometrial carcinoma: Insights from a comprehensive analysis of the immune landscape

BACKGROUND

Immune cells are crucial components in the tumor microenvironment and have a significant impact on the outcomes of patients.

AIMS

Here, we aimed to establish a prognostic score based on different types of tumor-infiltrating immune cells for Endometrial Carcinoma (EC).

METHODS AND RESULTS

We enrolled and analyzed 516 EC patients from The Cancer Genome Atlas. The relative abundance of 22 immune cells were estimated by using the CIBERSORTx algorithm. Cox regression was performed to identify potential prognostic immune cells, which were used to develop a Tumor-infiltrating Immune Cell Score (TICS). The prognostic and incremental value of TICS for overall survival were compared with traditional prognostic factors using the C-index and decision curves. Clustering analysis using all immune cells identified three immune landscape subtypes, which had weak correlation with survival. A TICS was constructed using CD8T cells, resting memory CD4 T cells, activated NK and activated DCs, and classified patients as low-, moderate- and high-risk subgroups. The low-risk subgroup had higher tumor mutation burden and activation of IL2/STAT5, IL2/STAT3 and IFN-gamma response pathways. Conversely, the high-risk subgroup was associated with DNA copy number variation, hypoxia and EMT process. The TICS subgroups significantly predicted overall survival, which was independent of patient age, tumor stage, grade and molecular classification. Moreover, we developed a nomogram incorporating TICS and clinicopathologic factors, which significantly improved the predictive accuracy compared to the clinicopathologic model alone.

CONCLUSION

The TICS is an effective and independent prognostic predictor for EC patients and may serve as a useful supplement to clinicopathological factors and molecular subtyping.

A novel LINC00478 serves as a tumor suppressor in endometrial carcinoma progression

BACKGROUND

Long noncoding RNAs (lncRNAs) are involved in the pathogenesis and progression of various cancers, but their roles in endometrial cancer (EC) are largely unknown.

METHODS

The expressions of LINC00478 and PTBP1 in EC tissues were determined by RT-qPCR. Cell counting kit-8, flow cytometry and Transwell assays were executed for detecting the roles of LINC00478 in EC cells proliferation, migration and invasion. The mouse-xenograft models were established by subcutaneous injection in vivo. The interaction between LINC00478 and PTBP1 was confirmed by RNA pull-down assay and RNA-binding protein immunoprecipitation assay.

RESULTS

LINC00478 was significantly down-regulated in EC tissues while compared to that in their paracancerous samples, and a higher expression level of LINC00478 was negatively correlated with clinical progress of EC patients. Functional experiments in vivo and in vitro revealed that LINC00478 overexpression could dramatically retard the proliferation of EC cells, decrease the rate of colony formation, suppress the migration and invasion abilities of EC cells in vitro and inhibit tumor growth in vivo. Mechanistically, LINC00478 regulated the expression of PTBP1, a key factor in the Warburg effect, and affected the metabolic process of EC cells.

CONCLUSIONS

LINC00478 acts as a tumor suppressor in EC by negatively controlling PTBP1 expression and influencing the Warburg effect, providing a potential biomarker and therapeutic target for patients with EC.

Novel cell subtypes of SPP1 + S100P+, MS4A1-SPP1 + S100P+ were key subpopulations in intrahepatic cholangiocarcinoma

BACKGROUND

In this study, we integrated single-cell RNA sequencing (scRNA-seq) data to investigate cell heterogeneity and utilized MSigDB and CIBERSORTx to explore the pathways of major cell types and the relationships between different cell subtypes. Subsequently, we explored the correlation of cell subtypes with survival and used Gene Set Enrichment Analysis (GSEA) analyses to assess the pathways associated with the infiltration of specific cell subtypes. Finally, multiplex immunohistochemistry in tissue microarray cohort were performed to validate differences in protein level and their correlation with survival.

RESULTS

iCCA presented a unique immune ecosystem, with increased proportions of Epi (epithelial)-SPP1-2, Epi-S100P-1, Epi-DN (double negative for SPP1 and S100P expression)-1, Epi-DN-2, Epi-DP (double positive for SPP1 and S100P expression)-1, Plasma B-3, Plasma B-2, B-HSPA1A-1, B-HSPA1A-2 cells, and decreased proportions of B-MS4A1. High level of Epi-DN-2, Epi-SPP1-1, Epi-SPP1-2, B-MS4A1, and low level of Epi-DB-1, Epi-S100P-1, and Epi-S100P-2 was significantly associated with longer overall survival (OS), and high level of B-MS4A1_Low_Epi-DN-2_Low was associated with the shortest OS. Moreover, the results of MsigDB and GSEA suggest that bile acid metabolism is a crucial process in iCCA. Finally, we found that S100P+, SPP1+, SPP1 + S100P+, and MS4A1-SPP1 + S100P+ were highly expressed, whereas MS4A1 was lowly expressed in iCCA, and patients with high level of S100P+, SPP1 + S100P+, and MS4A1-SPP1 + S100P+ exhibited shorter survival.

CONCLUSIONS

We identified the cell heterogeneity of iCCA, found that iCCA is a unique immune ecosystem with many cell subtypes, and showed that the novel cell subtypes of SPP1 + S100P+ and MS4A1-SPP1 + S100P+ were key subpopulations in iCCA.

Weakly supervised deep learning to predict recurrence in low-grade endometrial cancer from multiplexed immunofluorescence images

Predicting recurrence in low-grade, early-stage endometrial cancer (EC) is both challenging and clinically relevant. We present a weakly-supervised deep learning framework, NaroNet, that can learn, without manual expert annotation, the complex tumor-immune interrelations at three levels: local phenotypes, cellular neighborhoods, and tissue areas. It uses multiplexed immunofluorescence for the simultaneous visualization and quantification of CD68 + macrophages, CD8 + T cells, FOXP3 + regulatory T cells, PD-L1/PD-1 protein expression, and tumor cells. We used 489 tumor cores from 250 patients to train a multilevel deep-learning model to predict tumor recurrence. Using a tenfold cross-validation strategy, our model achieved an area under the curve of 0.90 with a 95% confidence interval of 0.83-0.95. Our model predictions resulted in concordance for 96,8% of cases (κ = 0.88). This method could accurately assess the risk of recurrence in EC, outperforming current prognostic factors, including molecular subtyping.