Abstract 1682: Imaging mass cytometry reveals key spatial features among immune cells in hepatocellular carcinomas treated with neoadjuvant cabozantinib and nivolumab

Abstract

Abstract Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer and carries a poor prognosis. Most HCC is unresectable at presentation, and until recently, the use of perioperative systemic therapy has been hampered by a lack of any effective therapies. We recently conducted a single-arm trial of neoadjuvant cabozantinib followed by nivolumab for borderline resectable or locally advanced HCC (NCT03299946), through which secondary resectability was obtained in a subset of patients. Of 15 patients enrolled, 12/15 patients achieved margin-negative resection, and 5/12 resected patients experienced a major or complete pathologic response. Here we characterize changes in the tumor microenvironment (TME) induced by neoadjuvant therapy. We constructed a tissue microarray containing 37 core liver biopsies (15 from responders, 22 from nonresponders), stained with a 27-marker panel, and used ilastik and CellProfiler to segment the resulting images, producing a single-cell dataset comprising 59,453 cells. We then used FlowSOM to perform unbiased clustering of cells, which we annotated into 17 cell types. Next, we performed spatial analysis using Voronoi diagrams and top neighbors mapping. We generated a minimum spanning tree using shortest Euclidean distances to model the simplest spatial relationships among all cell types and ranked their importance using random forest models. Grossly, responder cores were characterized by the presence of tertiary lymphoid aggregates, as well as a higher percent abundance of several immune cell types, including CD4 T (p < 0.05) and CD8 T cells (p < 0.005). In responders, Voronoi diagrams revealed denser packing of most immune cell types, particularly B cells (p < 0.005), and top neighbors analysis indicated higher numbers of lymphoid-lymphoid, myeloid-myeloid, and lymphoid-myeloid neighbors. This suggests that response is characterized by immune infiltration of the TME. Exploring this further, a minimum spanning tree showed that in nonresponders, CD8 T cells were flanked by CD163+ macrophages, whereas in responders, HCC cells were closely linked to lymphoid cells. Importance plots from random forest models for B, CD4 T, and CD8 T cells revealed that top predictors of responder status were higher minimum distance from CD163+ Arg1+ macrophages and lower minimum distance from CD163+ Ki67+ macrophages, which express higher levels of PD-L1. This suggests that proximity of B and T cells to macrophages that exert immunosuppression via Arg1 is a critical feature of resistance to cabozantinib plus nivolumab, whereas proximity to proliferative macrophages that express higher levels of PD-L1 is a key feature of response. In conclusion, cabozantinib and nivolumab can effectively promote antitumor immunity by altering both the abundance and spatial organization of macrophages, B cells, and T cells in the HCC TME. Citation Format: Shu Zhang, Qingfeng Zhu, Nicole Gross, Soren Charmsaz, Atul Deshpande, Stephanie Xavier, Aditya Mohan, James Leatherman, Guanglan Mo, Jennifer Durham, Aleksandra Popovic, Brad Wilt, Dongxia Lin, Derek Quong, Robert Anders, Elana Fertig, Elizabeth M. Jaffee, Mark Yarchoan, Won J. Ho. Imaging mass cytometry reveals key spatial features among immune cells in hepatocellular carcinomas treated with neoadjuvant cabozantinib and nivolumab [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1682.