Using patient-derived xenografts as sources of 3D tumor sphere cultures to study autologous tumor-infiltrating lymphocytes in metastatic uveal melanoma

Introduction: Uveal melanoma (UM) is a rare form of melanoma, but the most common primary malignancy of the eye. 50% of patients with UM develop metastasis, mainly to the liver, with a median survival of less than a year. Immune checkpoint inhibitors are markedly less effective in patients with the uveal form as compared to the cutaneous form of melanoma. Adoptive cell therapy (ACT) has not been extensively studied clinically or preclinically in UM. There has been a lack of robust ex vivo screening models and very few patient-derived xenograft (PDX) mouse models. Here we develop a 3D ex vivo tumor sphere coculture platform using PDX samples and their autologous tumor-infiltrating lymphocytes (TILs). Material and Methods: UM metastases from ten patients were found to grow when implanted subcutaneously in NOG mice. The PDXs were used to generate tumors from which we cultured tumor spheres, in ultra-low attachment U-bottom wells, and cocultured them with ex vivo expanded autologous or MART1-specific HLA-matched allogenic TILs. The coculture was analysed using ELISA and flow cytometry (FACS) to look at markers of T cell activity. Single-cell RNA sequencing was performed on biopsies and TILs. Immunohistochemistry was used to compare biopsies from PDXs and patients. Using Incucyte, 3D image quantification was performed, showing interactions and effects on tumor spheres after coculture with TILs. Results: UM patient tumors utilized for this study were of choroidal origin, 70% being liver mets and remainder skin mets. All PDXs correlated with the histopathological and molecular characteristics of original patient tumors. All samples had BAP1 mutations, 90% had GNA11/GNAQ with some having additional CYSLTR2 or CDKN2A mutations and 10% had PLCB4 and SF3B1 mutations. Degranulation (CD107a+) and activation (41BB+) with FACS was observed for 30% of all coculture samples, supported by comparative Granzyme B activity with ELISA. Single-cell sequencing of biopsies and TILs, as well as TCR sequencing of the “activated” TILs allowed for a deeper characterization of clonal expansion of tumor-reactive TILs. 3-D image-based analysis showed interactions between tumor spheres and TILs and further quantified differences in flourescently labelled spheres post TILs treatment. Conclusion: These results demonstrate that our tumor sphere platform is effective in studying TILs from metastatic UM samples. The platform is a powerful tool for visualizing and quantifying tumor-TIL interactions, and can aid in identifying immunological mechanisms that could be clinically translatable. It also serves as a valuable platform to screen for means to improve TIL function, i.e., by genetic engineering. To further study autologous TILs therapy in-vivo, we have developed few orthotopic UM liver metastasis NOG-IL2 PDX models, which can be utilised after the tumor sphere experiments. No conflict of interest.