Chapter 7: Computational Modeling and Experimental Investigations to Enhance the Successful Response of Anti-PD-1 Cancer Immunotherapies

Immunotherapies with anti-programmed death-1 antibodies (anti-PD-1) have revolutionized cancer treatment. However, the majority of patients still fail to respond and the reasons remain largely unknown. In order to identify tumor characteristics associated with treatment failure, we developed a computational model capable of simulating tumor response to anti-PD-1 immunotherapy, validated with on-site in vitro and in vivo experiments.

The experiments were conducted using non-responsive and responsive murine cell lines, 4T1 mammary carcinoma and CT26 colon carcinoma, respectively. Tumors were induced in athymic nude mice, wild-type mice, and wild-type mice receiving anti-PD-1 immunotherapy to: (1) assess Gompertzian parameters describing intrinsic tumor growth in the absence of T cells, (2) assess the immune-related parameters, and (3) validate model’s predictive ability, respectively. Flow-cytometric analyses of major histocompatibility complex (MHC) class I and PD-1 ligand (PD-L1) expression were performed to assess initial heterogeneity in tumor cell subpopulations. Model parameter sensitivity studies were performed to establish importance of different tumor characteristics on the observed tumor response to anti-PD-1 immunotherapy.

The model correctly predicted the observed trend of response to anti-PD-1 in 4T1, while the predictions in CT26, where dichotomous responses were observed experimentally, were less accurate. Sensitivity studies of measured model parameters and initial conditions suggested that complete responses to anti-PD-1 immunotherapy might be possible only in the case of homogeneous MHC class I positive tumors. On the other hand, heterogeneous tumors containing MHC class I negative tumor cell subpopulations were associated with resistance to anti-PD-1 therapy.