Background Baseline signatures have so far failed to accurately predict response to immune checkpoint inhibitors. A model that mimics the clinical response of tumor immune microenvironment (TiME) upon treatment in culture, would dramatically improve the chances of identifying true responders. Tumor histocultures capture the complex interactions between various tumor microenvironment components effectively. They retain the immune cell repertoire along with spatial context, both of which are adequately represented in Farcast^TM TiME histoculture platform.

Methods Farcast^TM head and neck squamous cell carcinoma histoculture platform was used to evaluate response to treatment with anti-PD1 (Nivolumab) alone or in combination with anti-CTLA4 (Ipilimumab) in culture. Multiple assays including spatial analysis was performed post-treatment.

Results Interferon gamma (INFg) cytokine release, that classically defines T- cell response, was used to stratify samples for response to treatment with Nivolumab (n=45). Histopathological evaluation revealed that 4 of 45 (9%) samples showed an increase in tumor infiltrating immune cells by 1.5 folds, with a concomitant increase in IFNg secretion (Log₂FC ± SD,1.67 ± 0.8), and CD8+GranzymeB+ cells estimated by flow cytometry. Three out of these 4 samples showed an increase in tumoricidal response. To capture the complex spatial interplay between TiME components and the intra-cellular signals that determine response, a 72-protein marker GeoMx DSP (nanoString) panel was used on two samples (one recurrent and a primary) treated with Nivolumab. panCK was used for tumor segmentation. An increased expression of FoxP3 (Treg marker), IDO and phosphorylated tumor markers was observed in the recurrent sample within the tumor nest, suggestive of immune suppressive microenvironment post treatment. Interestingly, the primary sample exhibited increase in checkpoint and T-cell costimulatory molecules within the tumor nest along with a significant decrease in FoxP3 and tumor marker EGFR, suggestive of enhanced T- cell mediated tumor cytotoxicity. To map the relative spatial distribution of CD8+ and FoxP3+ cell populations, multiplex IHC assay was performed, and data analyzed using QuPath. Decrease in distance between CD8+ cells from tumor nest with a concomitant increase in distance between FoxP3+ and CD8+ cells were observed in a high IFNg and GranzymeB secreting sample.

Conclusions In summary, our data suggests that though IFNg secretion was observed in multiple samples post treatment, spatial contexture influenced T-cell response mediated tumor cytotoxicity. Farcast TiME thus provides a unique platform that can combine spatial dynamics of immune cells overlaid with bulk data, providing important insights into understanding underlying mechanisms that counter T-cell response to immune check point inhibitors.

Acknowledgements Technical support from nanoString for performing the GeoMx DSP experiment and analysis is acknowledged.

Ethics Approval All patient samples used in this study was obtained with prior approvals from individual institutional review boards of participating hospitals.