Abstract 3126: Multimodal genome-wide survey of progressing and non-progressing DCIS

Abstract

Abstract Introduction: Our current understanding of the molecular changes predisposing Ductal carcinoma in situ (DCIS) to progress to invasive breast cancer (IBC) is limited. The aim of this study is to characterize the molecular changes of DCIS to improve DCIS risk assessment. Methods: We obtained paraffin-embedded tissue of 210 DCIS samples with no concurrent or antecedent IBC from 5 cancer institutions, and extracted DNA and RNA. Transcriptome analysis: Illumina’s TruSeq RNA Exome kit was used for library construction, followed by sequencing. Methylome analysis: Bisulfite treated genomic DNA was restored and arrayed using the Illumina 450K methylation chips. DNA Copy number (CNV) analysis: CNV was estimated from the methylation data set using the EpiCopy R package. Results: Samples passing Q/C metrics: Transcriptome: 59 cases, 63 controls. Methylome & CNV: 93 cases, 98 controls. We classified samples into their intrinsic PAM50 subsets. Cases showed an increase in the Her2 subtype, and the controls were enriched in LuminalA (LumA) samples, while Basal and LuminalB (LumB) subtypes were evenly distributed. Compared to the TCGA IBC dataset, proportions of Her2 and LumB subtypes in DCIS were increased while the LumA cohort was decreased. Unsupervised clustering of the transcriptome data resulted in 3 clusters with key differences between PAM50 subtypes: one cluster predominated in Basal and Her2 subtypes, and a second was enriched in hormone-positive samples (LumA and LumB). For the methylome data, the optimal number of clusters was 6. Again, several clusters showed correlation with PAM50 subtype, e.g., one was enriched for hormone-negative subtypes (Basal and Her2), while another was enriched for hormone-positive subtypes (LumA and LumB). For the CNV data, the optimal number of clusters was 4. Here, one CNV cluster appeared predominantly in the Basal subtype, and there were multiple regions showing significant subtype-specific differences with the TCGA IBC data set. While the well-known heterogeneity of DCIS prevents the above broad molecular categories from strongly stratifying DCIS by outcome, we are continuing to analyze our data sets for predictive molecular signatures. Since the long-term outcome of all DCIS in this cohort is known, we are in the unique position to pursue additional questions such as PAM50 subtype differences in times-to-events or lineage fidelity (i.e., whether the subtype of the subsequent IBC matches the preceding DCIS), and it appears that the time to IBC diagnosis is shortest in Basal DCIS, while at the same time, lineage fidelity is lowest in Basal DCIS, as has been previously reported. Conclusions: Our subtype-stratified analyses identified multiple molecular differences both between intrinsic subtypes as well as between DCIS and IBC that suggest subtype-specific characteristics that may be exploitable for risk stratification of DCIS. Citation Format: Marija Debeljak, Soonweng Cho, Bradley Downs, Michael Considine, Brittany Avin-McKelvey, Yongchun Wang, William Grizzle, Katherine Hoadley, Charles Lynch, Brenda Hernandez, Paul van Diest, Wendy Cozen, Ann Hamilton, Debra Hawes, Edward Gabrielson, Ashley Cimino-Mathews, Liliana Florea, Leslie Cope, Christopher B. Umbricht. Multimodal genome-wide survey of progressing and non-progressing DCIS [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3126.