Immune and proliferative programs distinguish anti-PD-1-responsive endometrial cancer

An exploratory analysis of two anti-PD-1-treated endometrial cancers found that the tumor from the patient who responded to therapy contained a broader immune compartment and more proliferating cells than the non-responder’s tumor. The single-cell data also showed stronger regulatory T-cell, cytotoxic T-cell and checkpoint-associated expression in the responder, while spatial transcriptomics revealed a notably higher proliferation signal and lower epithelial and extracellular-matrix signals. These patterns were directionally contextualized in The Cancer Genome Atlas, but the findings remain hypothesis-generating because the treatment comparison included only one patient in each response group.

The study combined three analytical scales: single-cell RNA sequencing and spatial transcriptomics from the Gene Expression Omnibus dataset GSE251923, bulk RNA data from The Cancer Genome Atlas uterine corpus endometrial carcinoma project, and exploratory quantitative reverse-transcription polymerase chain reaction assays. After quality control, the single-cell dataset contained 16,725 cells from 13,390 cells in the responder and 3,335 cells in the non-responder. These cells were assigned to 14 populations on the basis of canonical lineage markers and transcriptional programs. The researchers treated the patient or tissue section, rather than the number of cells or spots, as the biological unit, making the response-related comparisons descriptive.

Compared with the non-responder sample, the responder contained higher proportions of CD8 T cells, regulatory T cells, B cells, plasma cells, CD4 T cells and proliferating cells. The largest percentage-point differences were +7.7 for proliferating cells, +6.7 for B cells, +5.9 for CD4 T cells, +5.4 for plasma cells, +3.3 for regulatory T cells and +3.1 for CD8 T cells. The non-responder was instead dominated by epithelial or tumor cells, which accounted for 74.39% of its annotated cells, compared with 38.56% in the responder. Endothelial cells were also proportionally more common in the non-responder. The responder’s 443 regulatory T cells expressed a range of regulatory-program scores, whereas the non-responder contributed only one such cell, preventing a reliable comparison of their distributions.

At the gene level, the responder sample had higher patient-mean expression of CXCL13, CTLA4, FOXP3, PDCD1, CD8A and MKI67. Functional enrichment connected the responder-associated genes with T-cell receptor signaling, T-helper differentiation, B-cell receptor signaling, cell adhesion, cytokine interactions and PD-1 checkpoint pathways. Some very large effect estimates, including those for low-abundance genes such as CXCL13, likely reflect near-zero expression in the non-responder and should not be interpreted as population-level response effects. The analysis therefore supports a pattern of simultaneous regulatory, cytotoxic and proliferative programs in the profiled responder, not a causal mechanism linking any one program to treatment benefit.

All 8,103 Visium spots passed quality control, including 4,204 from the responder section and 3,899 from the non-responder section. Responder-minus-non-responder differences were +0.324 for the proliferation signature, +0.091 for the regulatory T-cell signature and +0.006 for the CD8 T-cell signature. The epithelial score was 0.412 lower and the extracellular-matrix score 0.664 lower in the responder. Marker patterns identified broad epithelial or tumor-like, stromal-vascular, myeloid-associated, B- or plasma-cell-enriched and proliferation-enriched expression compartments. CD8A and FOXP3 were sparse, and no discrete CD8-T-cell- or regulatory-T-cell-dominant cluster was assigned.

The spatial analysis has important constraints. The source lacked tissue position files, so its expression-space projections could not establish anatomical colocalization, exclusion or neighborhood relationships. Each Visium spot can also contain multiple cells, meaning that the cluster labels describe dominant transcriptional components rather than pure cell identities or formal deconvolution results. Differences in genes detected per spot further indicated that the sections differed in transcript complexity. Although the proliferation contrast was large, the section-level findings characterize two individual tumors and cannot show that a responder generally contains more proliferating cells or that proliferation itself caused the clinical outcome.

The TCGA-UCEC analysis provided broader cohort context. It included 553 primary tumors and 35 normal tissues after removal of one recurrent tumor sample. Tumors displayed a strong proliferation program, with enrichment of cell-cycle, mitotic-checkpoint, DNA-replication and DNA-repair pathways; 7,471 genes met the study’s differential-expression criteria in the tumor-versus-normal comparison. Among 377 tumors with non-missing survival data, including 57 deaths, a high regulatory T-cell signature was associated with longer overall survival in a median-split analysis (log-rank p = 0.0472; hazard ratio 0.58, 95% confidence interval 0.34–1.00). By contrast, a high proliferation signature was associated with shorter overall survival (hazard ratio 1.72, 95% confidence interval 1.01–2.94; p = 0.046 for the Cox analysis).

The survival findings do not imply that regulatory T cells are protective or that proliferation is causal. In TCGA tumors, regulatory T-cell, CD8 T-cell and T-cell-exhaustion signatures were strongly correlated, with Spearman coefficients of 0.80 to 0.90, while the proliferation signature showed little correlation with the immune signatures. A regulatory T-cell score may therefore mark a broadly immune-infiltrated microenvironment rather than the activity of a single cell type. The survival models were univariable and divided continuous scores at their medians, leaving room for residual confounding and loss of information. The TCGA cohort also lacked treatment annotations and was not used as a direct test of anti-PD-1 response.

Across datasets, 15,842 genes were shared between the single-cell and TCGA analyses. The descriptive single-cell screen identified 6,463 genes meeting its mean-expression and effect-size thresholds, while 4,717 TCGA genes met the false-discovery-rate criterion in the MSI-H-versus-MSS comparison. Among 1,659 genes shared by these sets, 1,365 had concordant directions, corresponding to 82.3% concordance. Regulatory T-cell, cytotoxic T-cell, T-cell-exhaustion and proliferation signatures were directionally higher in both the responder-versus-non-responder comparison and the MSI-H-versus-MSS comparison, whereas the myeloid-derived suppressor-cell signature was not concordant. These contrasts nevertheless addressed different questions: clinical response in two treated patients versus molecular subgroup status in a larger observational cohort, so microsatellite instability-high status was not treated as a surrogate for anti-PD-1 response.

Exploratory cell and neutrophil experiments supplied limited mechanistic context without directly modeling checkpoint blockade. Under 1% oxygen, the Ishikawa and HEC-1A endometrial cancer cell lines had higher transcript levels of MKI67, FOXP3, CXCL13, PDCD1 and CTLA4 than under normoxia, with approximate increases ranging from 2.3-fold to 4.1-fold. Primary neutrophils from three donors exposed to endometrial-cancer ascites-conditioned medium showed PAD4 mRNA increased 3.8-fold and ELANE mRNA increased 2.9-fold relative to control medium. These assays demonstrated condition-associated transcription, but tumor-cell hypoxia does not establish immune-cell abundance or clinical response, and increased PAD4 and ELANE transcripts do not demonstrate functional neutrophil extracellular trap formation. Confirming whether the candidate programs are predictive, prognostic or simply markers of molecular subtype will require independent, multicenter cohorts with treatment annotations, paired samples and functional spatial and immune assays.

Subject of Research: Immune and proliferation programs associated with anti-PD-1 response in endometrial cancer

Article Title: Treg and immune proliferative programs characterize anti-PD-1-responsive endometrial cancer: an integrative single-cell, spatial transcriptomic, and TCGA-UCEC reanalysis

Article References: Zheng, Y., Chen, J., Wang, Q., & Li, Q. (2026). Treg and immune proliferative programs characterize anti-PD-1-responsive endometrial cancer: an integrative single-cell, spatial transcriptomic, and TCGA-UCEC reanalysis. Frontiers in Genetics, 17, Article 1942724. https://doi.org/10.3389/fgene.2026.1942724

Image Credits: AI Generated

DOI: 10.3389/fgene.2026.1942724

Keywords: endometrial cancer, anti-PD-1 therapy, regulatory T cells, tumor proliferation, single-cell RNA sequencing, spatial transcriptomics, TCGA-UCEC, microsatellite instability, neutrophil extracellular traps, Treg, immune, proliferative