Abstract 2499: Genomic spatial profiling of archived pre-malignant lesions of the breast epithelium

Abstract

Listen

Translate article

Abstract The widespread adoption of breast cancer screening has increased the detection of pre-malignant lesions such as ductal carcinoma in situ (DCIS) but there are currently no reliable biomarkers to stratify DCIS and avoid over-treatment. Importantly, the identification of novel biomarkers in DCIS is notoriously difficult. The rare and late recurrence to subsequent breast cancer (10% at 10 years) poses challenges in experimental design that only a few established, retrospective studies can solve. Most tissue specimens in the associated registries are small, fixed and archived in paraffin (FFPE), precluding their characterization via standard genomic assays. Besides, pure DCIS often present multi-focal heterogenous histology of unclear mutational make-up. Here, we present an approach combining laser capture microdissection with state-of-the-art genomic library preparation methods to enable molecular profiling from minute amounts of damaged nucleic acids and demonstrate its validity for the spatial molecular characterization of DCIS specimen. Targeted DNA sequencing was performed on limited dilutions of a test DNA from archived breast tissue. Three library preparation strategies were evaluated, included adaptation of classic A-tailing (AT), alternate single-strand (SS) or blunt-end (BE) adapter ligations. The use of an AT protocol resulted less than 10% of targeted base pairs covered at more than 20x (Cov20) from lower than 50ng of DNA and suggested the adapter ligation as initial bottleneck. In contrast, SS or BE adapter ligation from 3ng of DNA reduced the fraction of duplicate reads, increasing the Cov20 to 49% and 100% respectively. Artifactual C to T substitutions, a signature of formalin damage, were restricted to variants of low allelic fraction (<5%). As a result, the use of ensemble variant calling and heuristic filtering to eliminate artifacts and germline variants, led to exome-wide mutational profile of 89% precision and 88% recall from 3ng input, comparable to replicate deep sequencing from a mirrored fresh frozen specimen. The presence of an ERBB2 amplification was validated in all tested conditions. The BE approach for whole exomes was validated on DCIS regions (N=8) and normal epithelium (N=1) dissected from 3 FFPE DCIS blocks. The normal epithelium was unaltered and copy number alterations both consistent and inconsistent within specimen were identified, including FGFR1 and ERBB2 gains or TP53 loss restricted to high-grade lesions. Short variants and indels were rare (~123 per region), mostly shared within each specimen and unlikely to represent driver mutations. The proposed approach can support the spatial mutational profiling of pre-malignant lesions. Compatible with a limited quantity of tissue from archived specimens, the approach sets a new paradigm for the retrospective molecular characterization of specimen with longitudinal follow-up and associated long-term outcome. Citation Format: Daniela Nachmanson, Joseph Steward, Taeyong Kim, Huazhen Yao, Eliza Jeong, Thomas O'Keefe, Gillian L. Hirst, Laura Esserman, Alexander Borowsky, Farnaz Hasteh, Kristen Jepsen, Olivier Harismendy. Genomic spatial profiling of archived pre-malignant lesions of the breast epithelium [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2499.