Personalized Combination Treatments and Early Induced Anti-Tumor Immunity in Breast Cancer Zuzana Tatarova1,3, Dylan C Blumberg1, Jessica L Riesterer1, Claudia S Lopez1, Gordon B Mills2, Lisa M Coussens2, Oliver Jonas*3, Joe W Gray*1 1Oregon Health and Science University, Department of Biomedical Engineering, Knight Cancer Institute, OHSU Center for Spatial Systems Biomedicine, Portland, OR, 2Oregon Health and Science University, Department of Cell and Developmental Biology, Knight Cancer Institute, Portland, OR, 3Brigham & Women’s Hospital, Harvard Medical School, Department of Radiology, Boston, MA Identification of effective combinations of immunotherapies with other anti-cancer agents requires an integrated understanding of the effects of the combination on the tumor and associated stromal cells. We have developed an integrated technological platform for this purpose that uses an implantable microdevice for delivery of dozens of drugs or drug combinations into spatially separate regions of a single living tumor and multiplexed histology analyses with probes for 40 proteins to assess the effects of each spatially defined treatment on the tumor and on the tumor microenvironment. We demonstrated the utility of this platform by assessing the effects of seven FDA approved drugs and combinations thereof on two late stage breast cancer mouse models with intact immunity. These studies indicated that the most effective tumor cell kill was produced by the epigenetic modulator, panobinostat, and was linked with upregulation of MHC-I expression on tumor cells, and recruitment of professional antigen presenting macrophages and cytotoxic neutrophils. Our spatial analyses also suggested the density of cancer stem cell and the formation of fibroblasts/ECM barriers as potential resistance mechanisms. Addition of the BCL2 inhibitor, venetoclax, enhanced panobinostat anti-tumor activity, and the local phenotypic cell responses suggested addition of an anti-CD40 agonist antibody to induce maximum therapeutic effect. We demonstrated that the panobinostast/venetoclax/anti-CD40 therapeutic combination was effective in whole animal studies in multiple models of luminal breast cancer and often resulted in complete tumor control. Our findings in this pilot study demonstrate the ability to use effect signatures of locally administered drug microdoses, to predict the most effective, systemically administered combination regimens. Because the assay involves no or minimal toxicity, it opens the possibility to rationally design optimal combination regimens involving immunotherapies for individual patients based on a rapid, intratumor readout. The approach may also enable the systematic identification of biomarkers of response and resistance to novel agents and combinations.