Background: Expression-based signatures have been shown to predict neoadjuvant therapy response; but further studies are needed to deconvolve the contribution of different immune cell types. The I-SPY 2 TRIAL is a standing neoadjuvant platform trial which evaluates experimental agents/combinations when added to standard chemotherapy. In this study, we compared published T/B cell-related signatures at 3 different levels of resolution as predictors of response in the I-SPY 2 TRIAL: (1) a combined T/B-cell co-expression module, correlated with general lymphocytic infiltrate, (2) individual T-cell and a B-cell specific signatures derived from purified immune cells and refined using tumor expression, and (3) 9 T cell subpopulation-specific signatures, including a CD8+ T resident memory phenotype (TRM) and a CD8+ T effector memory subset (TEM), generated from microdroplet-based single cell (sc) RNA sequencing of over 6000 tumor associated CD3+ T cells.

Methods: Expression data from 989 I-SPY 2 patients randomized to one of 9 possible experimental arms or the standard chemotherapy control were available for analysis. Pre-treatment biopsies were assayed using Agilent gene expression arrays. All I-SPY 2 biomarker analyses follow a pre-specified analysis plan. We used logistic modeling to assess each signature as a predictor of pCR within each arm (likelihood ratio test p<0.05). This analysis is also performed adjusting for HR/HER2 status, and within receptor subsets. Our sample size for each arm is small; and our statistics are descriptive rather than inferential. Our analysis is exploratory and does not adjust for multiplicities of other biomarkers outside this study.

Results: In the population as a whole, immune signatures predict response across multiple classes of agents (8/10 arms), including the checkpoint inhibitor Pembrolizumab (Pembro). However, the cell-type and subpopulation-specific signatures most predictive of response vary by subtype and agent. For instance, the T/B-cell co-expression module associates with response to Pembro and the Angiopoetin-1/-2 inhibitor AMG-386 in both HR-HER2- and HR+ERHHER2- subtypes. However, in the HR-HER2- subtype, the T-cell signature and the sc-derived CD8-TRM signature are most predictive; whereas in the HR+HER2- subtype, it is the B-cell, CD8-TRM and a novel CD4 signature that are most strongly associated with response. In the HER2+ subtype, the T/B-cell module and B-cell signature is associated with response to the AKT-inhibitor MK2206. Interestingly, among the sc-derived signatures, it is the CD8-TEM and multiple CD4 population-specific signatures, rather than CD8-TRM, that associate with response.

Conclusion: Our exploratory study suggests that immune signatures are associated with response to multiple I-SPY 2 experimental agents and implicates different immune cell types as response-predictive within breast cancer subtypes. Single cell sequencing derived population specific signatures may help further de-convolute how different immune cell types contribute to therapy responsiveness.

Citation Format: Yau C, Wolf DM, Campbell M, Savas P, Lin S, Brown-Swigart L, Hirst G, Asare S, Zhu Z, I-SPY 2 TRIAL Consortium, Loi S, DeMichele A, Yee D, Berry D, Esserman L, van 't Veer L. Expression-based immune signatures as predictors of neoadjuvant targeted-/chemo-therapy response: Experience from the I-SPY 2 TRIAL of ˜1000 patients across 10 therapies [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 P3-10-06.