Abstract Background In the GeparNuevo trial, the PD-L1 inhibitor durvalumab increased the rate of pathologic complete response (pCR; ypT0 ypN0) in triple-negative breast cancer if treatment started in a two-week window before neoadjuvant taxane/anthracycline chemotherapy (61 % pCR vs. 41%; p = 0.048; Loibl et al. ASCO 2018). Overall, pCR rates increased only numerically from 53 % to 44 % (p = 0.281). Herein, we aimed to evaluate the predictive value of PD-L1 immunohistochemistry in pre-therapeutic core biopsies. In addition, we identified dynamics in gene expression using repeated biopsies. Patients and Methods 174 patients were randomized to receive durvalumab or placebo with neoadjuvant chemotherapy. In the window part, 117 patients received a single dose of durvalumab (or placebo) before chemotherapy. Core biopsies were taken at three times: pre-treatment (“A”; N=174), after the window part (“B”; N=88) and after 12 weeks of nab-Paclitaxel (“C”; N=33). PD-L1 immunohistochemistry in A-biopsies (Ventana SP263 Assay) was recorded as percentage of cells with membranous staining in tumor cells and lymphocytes (TILs). We defined a tumor as PD-L1 high if ≥ 25 % of either compartment was stained. Ki-67 was stained on all available A, B and C biopsies (MIB-1, Dako, 1:100) and recorded as the percentage of tumor cells with nuclear staining. We profiled all available biopsies with targeted RNASeq using the HTG EdgeSeq platform (Oncology Biomarker panel, 2560 genes). Sequencing (IonTorrent S5) was successful in 162 A-, 79 B- and 31 C-biopsies. Results PD-L1 expression was high in 24 % of A-biopsies and was predictive for pCR in the complete cohort (OR 2.561; 1.183-5.554; p = 0.017). PD-L1 status of the TILs, but not of the tumor cells, was predictive (OR 1.313; 1.040-1.656; P= 0.022). The effect was not specific for durvalumab treatment. Higher levels of Ki-67 were predictive for pCR in B- biopsies in all patients (OR 1.399; 1.053-1.858; P =0.021) and in the placebo arm, but not in the durvalumab arm. Ki-67 levels in C-biopsies were not predictive; neither was the change in Ki-67 between pre-treatment and later time points (B vs. A or C vs. A). In a differential mRNA expression analysis (A vs. B), we found seven differentially expressed genes after one dose of durvalumab. We observed strong effects on gene expression after taxane treatment (A vs. C), but no significant difference according to treatment. These genes were associated with biological processes involved in therapy response. The pre-treatment levels of 12 of 69 markedly differentially expressed genes were associated with worse response to chemotherapy. Conclusion In A-biopsies, PD-L1 in TILs was predictive for response, and in B-biopsies, Ki-67 was predictive, but neither marker could specifically predict response to durvalumab. We observed limited effects of a single half-dose of durvalumab on global gene expression, but could identify substantial differential expression after taxane treatment. The evaluation of gene expression dynamic offers a promising approach for the identification of resistance-associated markers. The study was financially supported by AstraZeneca and Celgene Citation Format: Sinn BV, Loibl S, Karn T, Untch M, Kunze CA, Weber KE, Treue D, Wagner K, Hanusch CA, Klauschen F, Fasching PA, Huober J, Zahm D-M, Jackisch C, Thomalla J, Blohmer J-U, van Mackelenbergh M, Rhiem K, Felder B, von Minckwitz G, Burchardi N, Schneeweiss A, Denkert C. Pre-therapeutic PD-L1 expression and dynamics of Ki-67 and gene expression during neoadjuvant immune-checkpoint blockade and chemotherapy to predict response within the GeparNuevo trial [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr PD5-05.