Dissecting the impact of tumor-immune spatial interactions in mismatch-repair stratified colorectal cancers - PROJECT SUMMARY/ABSTRACT Predicting tumor progression and response to therapy remains a challenge. In colorectal cancer (CRC), immunotherapy has improved survival in a subset of patients, but has yet to make a significant impact on the ~85% of patients with mismatch-repair proficient (pMMR) tumors. However, there is emerging clinical evidence that immunotherapy has activity in pMMR disease within the right clinical context. I hypothesize that there exists an immunologically distinct subset of pMMR tumors with a unique tumor-immune microenvironment permissive to T cell infiltration, which is not captured by genomics-based precision oncology approaches. In CRC tumors, brisk T cell infiltration has been associated with improved disease-free survival and may indicate tumors primed for immunotherapy response. However, it is not well understood how immune infiltration interplays with the relative importance of somatic mutations, resident immune states, tumor architecture and clinical and demographic factors. Newer spatial profiling technologies, including highly multiplexed immunofluorescence (mIF) and spatial transcriptomics (ST), can discriminate immune subtypes and precisely quantify tumor-immune organization that are not possible to discern through dissociative or bulk approaches. However, the translational relevance of these spatial methods is limited by (i) small sample sizes, (ii) sparse clinical annotation, (iii) and a lack of robust computational tools to integrate across spatial profiling modalities. This proposal aims to advance our understanding of tumor-immune interactions in CRC through analysis of multimodal spatial profiling data on a large cohort (>200) of primary CRC tumors. In Aim 1, I will immunophenotype cells from highly multiplexed IF data and identify recurrent immune networks. I will leverage my experience with non-parametric 2D models for somatic rearrangements to develop a null background model for spatial data. I will then test the significance of immune-immune interactions and their association with T cell infiltration and activation states, stratified by MMR status. In Aim 2, I will use H&E and mIF to test the hypothesis that tumor architecture and tumor-cell intrinsic cell cycle and cohesion programs are significantly associated with physical T cell exclusion. I will then perform gene discovery on the key immune networks and tumor regions identified in Aims 1-2 using an external cohort of whole-slide ST data. In Aim 3, I will combine these insights with an integrated analysis of genomic and clinical data to identify the principal determinants of T cell infiltration and their relative effect on clinical outcomes. I will then extend these findings with an IF-trained deep learning model to test the feasibility of H&E-only models for identifying T infiltration in a large collection of whole-slide H&E images. I anticipate that the computational tools and biological insights developed through this proposal will significantly advance our understanding of CRC biology and serve as a basis for my independent research career in translational oncology and computational multimodal tissue profiling.
