Combinatorial Immunotherapy and Associated Tumor Metabolic Stress in Pre-clinical HNSCC Mmodels

Purpose/Objective(s) The combination of conventionally fractionated radiation (conRT) and anti-PD1/PD-L1 has underperformed, speaking to a lack of understanding in the interaction between conRT and the immune response. To begin to address this issue, we have undertaken a comprehensive evaluation of conRT and PD-1/PD-L1 in pre-clinical models. Materials/Methods We utilized MOC2 cell lines (+/- CRISPR knock out (KO) of PD-L1) to generate flank tumors in either immunodeficient (Nu/nu) or immunocompetent (C57Bl/6) mice (n=6-10/group) which were treated with 2 Gy x 9d +/- anti-PD1 mAb (250 ug/week x 2) and evaluated for effects on tumor response, measured as tumor volume on day 14 post-conRT. One way ANOVA and with post-hoc comparisons were performed on these values. Additionally, select tumors were analyzed via multi-channel flow cytometry, bulk RNA-Seq, and reverse phase protein array (RPPA). For RNA Seq data, differences between each individual group were analyzed using a moderated t statistic, with adjustment for multiple testing performed by fitting a beta uniform mixture (BUM) model. For RPPA data, two-way ANOVA with post-hoc comparisons were performed. Additional analysis of RNA Seq data to predict immune infiltrate was performed utilizing TIMER2.0, which provides data from 6 models of predicting immune infiltrate. Ingenuity pathway analysis (IPA) was performed on RNA-Seq and RPPA data. Results Interestingly, a modest repression of tumor grow due to PD-L1 KO was observed in Nu/Nu mice (p=0.04), while no effect was seen in the C57Bl/6 model (p=0.3). However, taken as a whole, the combination of conRT and PD-L1 KO in the C57Bl/6 model led to the most numerically meaningful reduction in TGD (p=3e-2 vs. PD-L1 KO; p=3e-3 vs. conRT), with a similar magnitude of effect when conRT was combined with PD1 mAB in a follow-up experiment (p=0.005). Between group comparison of RNA-Seq and RPPA data from tumors collected ∼14-20 days post conRT, followed by IPA analysis of differentially expressed targets identified some of the most significantly affected pathways in the conRT + PD-L1 KO groups being repression of oxidative phosphorylation (p=4.2e-4) from the RNA seq data, accompanied by repression of PI3k/Akt (p=3.9e-20) and an activation of AMPK signaling (p=3.8e-11) in the RPPA. Finally, analysis of predicted immune infiltrate depended upon the methodology used, however a significant reduction in several macrophage lineages was observed in multiple models (CIBERSORT, CIBERSORT-ABS, QUANTISEQ and xCELL). Analysis of flow cytometry data is ongoing. Conclusion Tumor analysis reveals a potential long-term reduction in oxidative phosphorylation with a concordant upregulation of AMPK signaling in PD-L1 KO tumors treated with conRT pointing to metabolic targeting as a potential therapeutic sensitization strategy.