Abstract PD04-05: Molecular predictors for type of recurrence following conservative treatment for DCIS

Abstract

Background: The identification of molecular predictors of type of recurrence (non-invasive vs invasive) after local therapy for ductal carcinoma in situ (DCIS) remains a challenge. We hypothesized that gene expression profiling using a panel of 32 breast cancer-related genes could identify a signature predictive of type of recurrence. Methods: From a prospectively maintained database of 1873 pts treated with conservative surgery +/− radiation for DCIS (1991–2006), we identified 190 (10%) patients who recurred. Original DCIS archival blocks were available for 108 (57%) of these cases. Freshly cut sections were obtained for microdissection of pure DCIS lesions, RNA was extraction and mRNA transcript quantification was performed with the Nanostring™ nCounter® system. Data were normalized using the nCounter® digital analyzer software for analysis. Results: Among the 108 studied cases, 66 (61%) recurred as DCIS and 42 (39%) recurred as invasive cancer. Median time to recurrence was 40 months (range 7–156 mo) and did not differ by type of recurrence. Similarly, patient age, grade of DCIS, ER status, use of radiation or tamoxifen therapy did not differ. Unsupervised hierarchical clustering using all 32 genes demonstrated two predominant groups, one enriched for recurrent-as-invasive (RI) tumors, and the other enriched for recurrent-as-DCIS (RD) tumors. Refined analysis by selection of 14 genes with significant differential expression between the RI and the RD tumors identified four main groups. The first cluster was dominated by ERBB2 over-expression and is equally composed of RI and RD tumors. The second cluster was composed almost exclusively of RI tumors and showed high level of COX2 and CCND1 and low level of CDKN2A (p16). The clusters 3 and 4 were enriched for RD tumors, with the fourth cluster completely devoid of RI cases. This RD-only group showed the lowest level of CCND1, but highest level of AKT3, EGFR, CDKN2A, and MKI67, thus displaying molecular traits typical of basal-like tumors. Conclusion: In this cohort of patients who recurred following conservative treatment for DCIS, HER2 expression was not predictive, however gene expression analysis demonstrated that high COX2 and CCND1 and low CDKN2a (p16) levels were strongly associated with invasive recurrences (RI), whereas increased AKT3 and MKI67 were associated with non-invasive recurrences (RD). Although the optimal combination of predictive markers requires validation, these data suggest that molecular alterations associated with RI differ from those associated with RD and warrant further investigation. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr PD04-05.