Integrative machine learning reveals a TLS signature and CCL5-CCR1 axis-associated immune remodeling in breast cancer

The transition from ductal carcinoma in situ (DCIS) to invasive ductal carcinoma (IDC) is a critical but poorly understood step in breast cancer progression. This study characterizes the dynamic remodeling of the tumor microenvironment (TME) during this transition, focusing on tertiary lymphoid structures (TLS) and chemokine signaling.

Using an integrated multi-omics approach-including multiplex immunofluorescence of a clinical cohort, public single-cell and spatial transcriptomics data, and a longitudinal syngeneic mouse model (EO771)-we investigated spatiotemporal TME evolution.

Our findings reveal that TLS reorganization during the DCIS-to-IDC shift is closely associated with the CCL5-CCR1 axis, which is implicated in macrophage-CD8⁺ T cell interactions and correlates with terminal T-cell exhaustion. Longitudinal modeling confirmed CCL5 network dysregulation alongside progressive CD8⁺ T cell dysfunction. Through an integrative machine-learning framework, we developed a robust 21-gene TLS signature that independently predicted patient outcomes in multiple cohorts, even after adjusting for clinical confounders. Finally, molecular docking identified cucurbitacin derivatives as candidate compounds nominated by exploratory in silico analysis targeting the CCL5 network.

This work highlights CCL5-CCR1 axis-related TLS dysregulation as a key spatiotemporal feature of invasion, provides a clinically applicable prognostic signature for early risk stratification, and nominates actionable targets to intercept invasive progression.