Differential impact of proliferation signature on efficacy of neoadjuvant chemoimmunotherapy in sTIL-high and sTIL-low triple-negative breast cancer (TNBC): Biomarker analysis of the NeoPACT trial.

507 Background: TNBCs with enrichment of stromal tumor-infiltrating lymphocytes (sTILs) and/or immune gene expression are more sensitive to neoadjuvant systemic therapy (NAST) and exhibit higher rates of pathologic complete response (pCR). Other biomarkers, including proliferation, are also prognostic in TNBC patients treated with NAST. We aim to investigate the impact of proliferation gene expression on efficacy of NAST in sTIL-high and sTIL-low TNBC. Methods: 110 TNBC patients treated with neoadjuvant chemoimmunotherapy (Carboplatin+Docetaxel+Pembrolizumab) on the phase II NeoPACT trial (NCT03639948) with available whole exome RNA sequencing were included. sTILs were scored in 5% increments by H&E. Tumors were defined as sTIL-high (≥ 20% sTILs) or sTIL-low ( < 20% sTILs). The ImSig Proliferation Signature score (ProlifSig) was computed from RNA sequencing data and samples were classified as ProlifSig-high (≥ median) or ProlifSig-low ( < median). ProlifSig was tested for prediction of pCR in sTIL-high and sTIL-low groups. Logistic regression was used to examine the independent prognostic utility of sTILs and ProlifSig on pCR. Results: 63/110 (57%) patients achieved a pCR. 56/110 (51%) patients were classified as sTIL-high, and 54/110 (49%) classified as sTIL-low. sTILs and ProlifSig as continuous variables were each predictive of pCR (OR = 1.022, 95%CI = 1.009-1.035, P= 0.001 for sTILs; OR = 2.682, 95%CI = 1.23-5.85, P= 0.01 for ProlifSig). In the sTIL-high group, ProlifSig was not associated with pCR either as a continuous score (AUC = 0.56) or when assessed as high/low categories (pCR 78% vs. 67% in ProlifSig-high and ProlifSig-low groups, respectively; OR = 1.79, 95%CI = 0.54-5.89, P= 0.34). In contrast, in the sTIL-low group, ProlifSig was significantly associated with pCR both as continuous score (AUC = 0.74) and when assessed as high/low categories (pCR 57% vs. 29% in ProlifSig-high and ProlifSig-low groups, respectively; OR = 3.18, 95%CI = 1.03-9.86, P= 0.045). On multivariate analysis, sTILs and ProlifSig were independent predictors of pCR (OR = 1.02, 95%CI = 1.01-1.03, P= 0.004 for sTILs; OR = 3.13, 95%CI = 1.44-6.83, P= 0.004 for ProlifSig). Conclusions: In TNBC patients treated with chemoimmunotherapy sTILs and ProlifSig provide complimentary information for prediction of pCR. ProlifSig is positively associated with pCR in sTIL-low tumors but not in sTIL-high tumors. We hypothesize that the therapeutic response in sTIL-high tumors is dominated by lymphocyte-dependent cytotoxic mechanisms, while in sTIL-low tumors, the response may be dominated by proliferation-dependent responses. ProlifSig could identify a subgroup of immune low TNBCs that can achieve substantial rates of pCR with neoadjuvant chemoimmunotherapy.