Pembrolizumab Therapy Exacerbates EBV-Induced Infections and Tumors in a Long-Term Fully Humanized Mouse Model

INTRODUCTION: A promising rich pipeline of combination therapies targeting checkpoint molecules expressed on T cells and/or tumor cells is currently being developed to abrogate tumor-induced immunosuppression. Novel in vivo models suitable for validating these immunotherapies and predict safety issues are warranted to accelerate their translation to patients.

AIM: Epstein Barr virus (EBV) is a type 1 carcinogen that is directly associated with the development of human B cell neoplasms. We modelled EBV infection and tumor progression in long-term humanized mice and investigated the activation of T cells with PD-1 expression. Further, we performed studies evaluating the effects of an anti-PD-1 antibody (pembrolizumab/ keytruda) in on EBV infections and/or tumor growth.

METHODS: Humanized mice transplanted with human cord-blood CD34⁺ stem cells and showing long-term (15 weeks) human T cell reconstitution were infected with an oncogenic recombinant Epstein Barr Virus (EBV), encoding enhanced firefly luciferase (fLuc) and green fluorescent protein (GFP). EBV infections were monitored by optical imaging analyses and PCR. CD8⁺ and CD4⁺ T cell subtypes (PD-1⁺, naïve, central memory, effector memory and terminal effector) were sequentially monitored in blood by longitudinal flow cytometry analyses and in organs at experimental endpoint. Histopathological analyses were performed to characterize EBV infection (EBER⁺) and PD-1⁺ T cell-rich infiltrates in tissues and tumors. We used the model to evaluate the effects of pembrolizumab administered after EBV challenge at low dose (first dose 1.65mg/kg and then 3.30 mg/kg, every other week, n=3) or high dose (first dose 5.00 mg/kg and then 10.00 mg/kg every other week, n=3) in respect to EBV infected controls (n=2).

RESULTS: EBV-fLuc was detectable one week after infection by non-invasive optical imaging in the spleen, from where it spread rapidly and systemically. Among the EBV-infected mice, 8/18 (=44%) developed macroscopically visible tumors in the spleen. For further analyses of the data, we then compared EBV-infected mice with ("EBV-Tumor") or without ("EBV") macroscopic tumors. At 6 weeks post-infection, the relative CD8⁺ T cell frequencies increased significantly and constantly (control Vs. EBV p=0.0021, control Vs. EBV-Tumor p=<0.0001, EBV Vs. EBV-Tumor p=0.0072). For absolute cell counts in tissues, CD8⁺ T cell increases were more dramatic in mice infected with EBV and developing tumors. These differences amounted to approximately tenfold relative to controls and 3-fold relative to mice not developing tumors. Mice infected with EBV showed 90-100% of the CD4⁺ and CD8⁺ T cells in lymphatic tissues expressing PD-1. Mice with EBV-tumors showed twice as many PD-1⁺ CD4⁺ and three times as many PD-1⁺ CD8⁺ T cells as infected mice without tumors. Histopathology combined with EBER in situ hybridization, showed foci of EBV infected cells in close association with PD-1⁺ infiltrating lymphocytes, often in perivascular regions. This model was then used to evaluate dose-dependent effects of pembrolizumab. The check-point inhibitor controlled EBV-fLUC spread for 2 weeks, but later prompted increased levels of infections. At endpoint analyses, mice receiving pembrolizumab showed larger dissemination of tumors.

CONCLUSIONS: We are currently performing additional experiments in order to elucidate this mechanism of EBV rebound. This humanized mouse model contributes to risk assessment prior to clinical trials of the use of checkpoint inhibitors in patients after transplantations at high risk of EBV infections.