Treatment-associated remodeling of a mitochondrial translation-related gene set in paired pretreatment and residual triple-negative breast tumors

Background Longitudinal transcriptional remodeling can reveal treatment-associated tumor states. We examined how a fixed 117-gene nuclear-encoded mitochondrial-translation-related set (MT117) changes across treatment in triple-negative breast cancer (TNBC) and whether its change is isolated or accompanied by other transcriptional programs. Methods We analyzed five public longitudinal datasets. The primary analysis used 19 paired pretreatment and residual TNBC tumors from GSE260693. External or auxiliary analyses used six strict TNBC and 27 all-subtype pairs from GSE240671, seven TNBC pairs from GSE43816, malignant-cell pseudobulk from GSE246613, and chemotherapy-persistent PDX/cell-line systems from GSE164715. Patients or model systems were the inferential units. Results MT117 decreased in GSE260693 (mean paired change, -0.754; 95% CI, -1.051 to -0.456; paired t P=4.68 × 10^-5), with decreases in 18 of 19 patients, alongside decreases in multiple contextual programs. In GSE240671, the TNBC-minus-non-TNBC difference was 0.099 (95% CI, -0.247 to 0.445; P=0.575), providing no evidence of TNBC specificity. In the distinct pembrolizumab-plus-radiotherapy GSE246613 cohort, 11 of 13 malignant-cell pairs decreased (mean, -0.131; 95% CI, -0.267 to 0.004; P=0.0567), but matched-module and covariate-adjusted analyses did not establish pathway-specific independence. Both external paired TNBC cohorts had confidence intervals crossing zero, and only one of four persistent experimental systems followed the primary human direction. Conclusions MT117 is best interpreted as an annotation-defined treatment-associated transcriptional signal that co-occurs with other transcriptional changes. Mitochondrial protein synthesis was not directly measured. The findings do not establish selective mitochondrial-translation inhibition, TNBC specificity, a universal persistence mechanism, or a validated biomarker.