Abstract B033: Integrating spatial biology and peripheral blood biomarkers to map treatment resistance in pancreatic ductal adenocarcinoma

Abstract

Despite recent advances in RAS-directed therapy, pancreatic ductal adenocarcinoma (PDAC) remains highly lethal. Surgical resection, chemotherapy, and radiotherapy continue to anchor treatment, but therapeutic resistance is pervasive, and most patients succumb to treatment-refractory residual disease. Advances in single-cell and spatial profiling have revealed clinically relevant molecular heterogeneity and tumor architectural changes associated with treatment-refractory PDAC. At the same time, intra- and interpatient heterogeneity has made it difficult to isolate treatment effects from patient-to-patient differences. Seeking to understand the causal drivers of treatment resistance, we performed the first-ever matched spatial profiling analysis of pre- and post-treatment clinical PDAC specimens to map the transcriptional organization of PDAC at unprecedented subcellular spatial and whole-transcriptome molecular resolution. In parallel, we collected serial blood samples from the same patients to identify circulating biomarkers associated with dynamic in situ tumor remodeling via ultrasensitive cell-free RNA analysis. Leveraging our patient-matched samples, we applied a causality-aware machine learning model to resolve distinct modes of treatment response among malignant epithelial cells. We developed novel computational approaches to identify candidate spatial transcriptional programs underlying treatment resistance and map their organization within the tumor. We discovered a novel treatment-refractory epithelial cell population with distinct spatial patterning and a T-cell-rich microenvironment. Circulating biomarkers tracked the abundance of this population and could enable its noninvasive detection. To further characterize the local immune microniche of this population in its full tumor context, we profiled a tumor microarray consisting of untreated resected tumors with state-of-the-art, early-access same slide whole transcriptome spatial transcriptomics and 64-plex spatial proteomics. In the long term, we envision that liquid signatures guided by spatial -omics discovery will provide real-time readouts of patients’ tumor evolution to guide optimal changes in regimen and to manage treatment toxicity breaks. ChatGPT was used to proofread and refine the wording of this abstract; all text was reviewed and edited by the authors.