Lactylation-mediated remodelling of the breast cancer microenvironment: single-cell multidimensional analysis and prognostic model construction.

Background: The breast cancer tumour microenvironment (TME) exhibits marked cellular and metabolic heterogeneity that contributes to disease progression and therapeutic resistance. Lactylation, a lactate-derived post-translational modification, has emerged as a potential link between metabolic reprogramming and tumour-associated transcriptional and immune changes. However, its cell-type distribution and clinical relevance in breast cancer remain incompletely defined. Methods: We integrated 26 scRNA-seq samples spanning ER+, HER2+, and triple-negative breast cancer (TNBC), comprising 98,572 cells. A literature-curated lactylation-related transcriptional module score was calculated using AddModuleScore, and epithelial cells were stratified into high- and low-score states for downstream analyses. Candidate genes were prioritised by integrating single-cell differential expression, a curated lactylation-related gene pool, and tumour-associated expression changes in TCGA-BRCA. A 14-gene prognostic model was developed using LASSO-Cox regression in TCGA-BRCA and externally validated in GSE20685 and METABRIC. Additional analyses evaluated associations with immune infiltration, somatic alterations, and predicted drug sensitivity. Functional relevance was explored through CALR knockdown in breast cancer cells and xenograft assays. Results: Single-cell analysis identified six major cell populations and revealed marked subtype-related heterogeneity. Lactylation-associated transcriptional activity was highest in epithelial cells, particularly in TNBC, and was associated with pathways related to immune response, cell cycle, and metabolic adaptation. High-score epithelial states showed enhanced epithelial-fibroblast-myeloid communication and were linked to an immune-modulatory microenvironment with partial immune-suppressive features. The 14-gene signature stratified patients into significantly different prognostic groups in TCGA-BRCA and retained prognostic value in GSE20685 and METABRIC. Internal resampling suggested modest but reproducible discrimination, although measurable optimism was observed in the training cohort. High-risk tumours were associated with distinct immune/stromal patterns, whereas mutation-frequency differences and tumour mutation burden did not show robust group differences after statistical correction. CALR knockdown suppressed proliferation, induced G1-phase arrest and apoptosis, and reduced xenograft growth. Conclusions: This study defines a lactylation-associated transcriptional programme in breast cancer and links it to epithelial-state heterogeneity, microenvironmental remodelling, and patient prognosis. The proposed 14-gene signature may provide a transcriptome-based framework for risk stratification, but the findings should be interpreted cautiously because the lactylation score is an indirect surrogate rather than a direct measurement of lactylation itself. Further mechanistic and clinical validation will be required.