Abstract P6-04-05: Digital Quantification of Post-Neoadjuvant Chemotherapy Breast Tumor RNA Reveals a Basal-Derived Gene Signature Associated with Post-Treatment Ki67

Abstract

Abstract Neoadjuvant chemotherapy (NAC) leads to a pathological complete response (pCR) in -20% of patients with breast cancer. The rest of the patients exhibit residual disease (RD) in the breast after NAC and are more likely to have a metastatic recurrence. Recent data show that tumor cell proliferation, as measured by Ki67 in the residual post-NAC cancer, can inform clinical outcome. We hypothesized that RD after NAC will harbor different gene expression as a function of Ki67 index. A 2nd hypothesis is that identification of key alterations in tumors with post-NAC high Ki67 will identify molecules or networks that can be therapeutically targeted in the immediate post-operative setting to eradicate clinically silent micro-metastases. We utilized Nanostring digital RNA quantification to measure the expression of 350 genes in 49 post-NAC residual breast cancers. The cohort was enriched for triple-negative cancers (24/49). Genes were selected based on relevance to breast cancer, published prognostic signatures, and our own unsupervised class discovery approach from existing microarray data. For the selected gene sets, we included a chemo-resistance signature (CHEMO), a stroma signature (STROMA), and a WNT/metastasis signature (WNT/MET). For the unsupervised analysis, publicly available pre-treatment microarray data from chemotherapy-treated ER-breastcancers that did not achieve pCR were clustered and the two most prominent groups were used to select a 250 gene set (CLUSTER). Ki67 was assessed in the RD for the 49 tumors and correlated to absolute RNA counts of the 350 selected genes. Ki67 was strongly associated with subtype as defined by the PAM50 classifier (P<0.0001). Basal tumors demonstrated the highest Ki67, followed by Luminal B, HER2, and Luminal A tumors (median = 63.5, 27, 22.6, and 5.5%, respectively). The CHEMO signature correlated inversely with Ki67 index (p=2.5e-8): paradoxically, genes positively associated with a higher probability of pCR correlated with a high Ki67 in the post-NAC specimens. Thus, chemotherapy-enriched tumors demonstrate distinct and unexpected gene expression patterns after treatment. The WNT/MET and STROMA gene sets showed significant trends with similar directionality to the CHEMO gene set. Our CLUSTER gene set was highly predictive of Ki67 (p=3.6e-6) and robustly associated with basal subtype (p=2.6e-13). MELK, an anti-apoptotic kinase involved in activation of Bcl-G was positively associated with a high Ki67 independently of tumor subtype (p=0.035 and p=0.045 in luminal and basal, respectively). In luminal tumors, high expression of BRCA1, BRCA2, and RRM2 was associated with high Ki67 (p=0.027, 0.016, and 7e-5, respectively), while reduced DUSP4 expression, a MAPK phosphatase, was predictive of high Ki67 in basal post-NAC tumors (p=6e-4). Existing gene expression predictors built on pre-treatment data did not match expected patterns in chemotherapy-enriched tumors. Breast cancer subtype was strongly associated with a chemotherapy-refractory high Ki67 index in the surgically removed tumors. Further, the assessment of gene expression patterns in RD may be important in ascertaining the molecular underpinnings of aggressive disease. Thus, our defined gene signature may offer insights into molecular targets in patients who do not achieve pCR after NAC. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P6-04-05.