UC San Diego

Precancer can represent a benign condition in some and a potent precursor to lethal cancer in others. As cancer screening and early detection of cancer improves, precancer diagnoses are becoming quite common, yet our poor understanding of precancer leaves clinicians unclear on how to treat them. Precancerous lesions are often tiny and archived, formalin-fixed, and paraffin-embedded (FFPE), which degrades and damages the small amount of nucleic acid available. This makes the genomic and transcriptomic analysis of precancers quite challenging as many of the lesions, especially the smallest, are often incompatible with standard DNA or RNA sequencing assays. My work reports the optimization of both experimental and computational methods to perform genetic profiling of precancer lesions - including acquired (somatic) and inherited (germline) variation - and the application of these methods to improve our understanding of breast precancer, ductal carcinoma in situ (DCIS).

A low-input FFPE DNA optimized exome sequencing workflow was paired with RNA sequencing, histological analysis, and immune microenvironment spatial profiling to generate a multimodal breast precancer atlas across 85 histological regions microdissected from the lesions of 39 patients diagnosed with pure DCIS. This allowed the measurement of relationships between molecular and phenotypic features, phylogenetic analysis, and identification of immune states revealed by stromal and epithelial-specific profiling. Further, a low-coverage whole-genome sequencing with reference-based imputation method, which provided reliable germline genotyping directly from archival FFPE tissues, was leveraged to evaluate the role of germline genetics in long-term precancer progression. I applied this methodology to samples from 36 patients with DCIS and identified that increased established breast polygenic risk scores were predictive of adverse outcomes.

The improved mapping provided by the breast precancer atlas, and suggestive evidence that germline variants contribute to DCIS outcome represent valuable resources in the study of the natural history of breast precancer. In particular, the findings emphasize the importance of including contextual information - spatial, microenvironmental, and inherited genetic factors - to fully characterize lesions. By only using archival tissue, I set the stage for the large retrospective studies that will be needed to guide the clinical management of precancer.