Longitudinal transcriptomic profiling identifies predictors of response to neoadjuvant chemoimmunotherapy in triple-negative breast cancer: results from the NeoTRIPaPDL1 trial.

Background

Triple-negative breast cancer (TNBC) is immunogenic, but only a subset of patients benefits from neoadjuvant immune checkpoint inhibitors combined with chemotherapy. Predictive biomarkers to guide patient selection and treatment adaptation are lacking.

Patients and methods

In the randomized phase III NeoTRIPaPDL1 trial, 280 patients with high-risk TNBC received neoadjuvant carboplatin plus nab-paclitaxel (CT, n = 142) or the same regimen with atezolizumab (CT/A, n = 138). Tumor biopsies were collected at baseline and on the first day of the second cycle (D1C2). Longitudinal RNA sequencing was performed to evaluate hallmark, immune, and ferroptosis-related gene signatures. Associations with pathologic complete response (pCR) were assessed using univariable logistic regression and multivariable XGBoost models.

Results

At baseline, in the CT/A arm, higher proliferation and lower stromal and metabolic programs were associated with increased pCR rate. Absence of tumor cells in early on-treatment biopsies strongly predicted surgical pCR in both treatment arms. Comparison of on-treatment with baseline samples revealed a reduced proliferation and increased immune and stromal signatures. These changes were more pronounced in tumors achieving pCR, particularly in patients receiving chemoimmunotherapy. Immune deconvolution analyses confirmed dynamic remodeling of the tumor microenvironment during treatment, with response-associated enrichment of immune cell populations and reduction of tumor epithelial fractions. Although several metabolic signatures were associated with outcome at baseline, most lost predictive value on-treatment, where immune-related programs became the dominant correlates of response. Exploratory multivariable analyses suggested complementary contributions of on-treatment cytotoxic immune activity and iron metabolism in shaping response to chemoimmunotherapy.

Conclusions

Longitudinal transcriptomic profiling revealed dynamic tumor and microenvironment remodeling during neoadjuvant therapy in TNBC. Baseline tumor-intrinsic features and early treatment-induced immune activation provide complementary information for predicting response to chemoimmunotherapy. Early tumor clearance in on-treatment biopsies represents a strong surrogate of pCR and supports investigation of response-adapted neoadjuvant strategies.