Abstract / Summary
To explore the biological and potential prognostic significance of B and T lymphocyte attenuator ( BTLA ) in high-grade serous ovarian cancer (HGSOC), including its association with neoadjuvant chemotherapy (NACT) groups and the tumor immune microenvironment (TIME). Single-cell RNA sequencing (scRNA-seq) and integration with public transcriptomic datasets were used to characterize BTLA expression and its cross-sectional associations with PN–TN treatment-group status (PN, pre-neoadjuvant chemotherapy; TN, post-neoadjuvant chemotherapy) and the tumor immune microenvironment. Exploratory binary machine-learning models using the vital status recorded at the last available follow-up were applied solely for candidate feature prioritization; these models did not incorporate follow-up duration or right-censoring information and were not interpreted as time-to-event survival-prediction models. Time-to-event associations were evaluated separately using LASSO-Cox and Cox proportional-hazards regression, which incorporated overall-survival time and censoring status. Pseudotime trajectory analysis was performed to explore the distribution of BTLA across inferred transcriptional states. All computational analyses were considered exploratory and hypothesis-generating. Exploratory cross-sectional analyses suggested differences in BTLA -detectable cells between the PN and TN groups. However, treatment group was confounded with patient identity and sampling site, and only one patient contributed samples at both time points; therefore, these findings cannot be interpreted as treatment-induced changes. Seventeen prognostic genes were identified. In a subsequent stratified multivariable analysis of 321 TCGA-OV patients, higher BTLA expression was associated with a lower mortality risk after accounting for relevant clinical and immune-composition variables (adjusted HR per one-standard-deviation increase = 0.76, 95% CI 0.66–0.88; P < 0.001). In an exploratory cross-cohort comparison, bulk-tissue BTLA expression was higher in TCGA-OV samples than in GTEx normal-ovary samples. However, disease status was completely confounded with cohort and data source, and the observed difference could not be separated from differences in tissue composition or immune-cell infiltration. BTLA expression was elevated in tumor tissues and was associated with immune infiltration and correlated with transcription factors such as STAT3 and IRF1. Pseudotime analysis showed heterogeneous BTLA expression along the inferred trajectory, which was interpreted as exploratory rather than chronological. BTLA showed an exploratory association with ovarian-tumor tissue datasets and the immune context of HGSOC; however, the available bulk-transcriptomic comparisons could not distinguish tumor-specific regulation from differences in immune-cell abundance, tissue composition, or data source. Prognostic associations were inconsistent across cohorts. The pre-/post-NACT, cross-cohort expression, and survival findings are exploratory and hypothesis-generating; larger independent cohorts, particularly with paired pre-/post-NACT sampling and orthogonal validation of BTLA expression, are required before BTLA can be considered a validated prognostic biomarker in HGSOC.