Predictive gene expression patterns of response to adriamycin and cyclophosphamide (AC) in human breast cancers

Abstract

527 Background: Adriamycin and cyclophosphamide (AC) and docetaxel (D) are widely used in the treatment of breast cancer. We conducted a prospective, randomized, multicenter trial to discover predictive markers of AC and D and hypothesized that gene expression profiles can select appropriate patients who may respond to AC, and that these patterns are different from our published docetaxel (D) profiles Methods: One hundred and twenty patients were randomized to either 4 cycles of AC (60/600 mg/m2) or D (100mg/m2) prior to surgery. Core biopsies from 60 patients were obtained before treatment with neoadjuvant AC. Pathologic responses were assessed after AC. Gene expression patterns were determined using Affymetrix U133A GeneChips. Differential genes for AC response were then validated by QRT-PCR in an independent cohort of 33 patients treated with AC. Results: The median age was 48 yrs (range 30–72), clinical response rates were 57% (34/60), and pathological complete response (pCR) or near pCR (npCR) was observed in 22% (12/60) in AC arm. Differential expression between sensitive and resistant tumors with a low false discovery rate (FDR 5–10%) was obtained. Of these 82 differentially expressed genes, pathways up-regulated in sensitive tumors included TOP2A, metabolism (LYZ), survival (CFLAR, CASP3), cell cycle (MKI67), cytokines and other inflammatory genes. This molecular portrait for AC was not predictive of docetaxel response. By QRT-PCR of 4 genes (LYZ, CFLAR, MKI67 and TOP2A) in the independent tumor set, LYZ was predictive of AC pathologic complete response. Additional genes will be validated in the second cohort. Conclusions: The molecular profile for AC is different from the docetaxel expression profile. This potential predictive test may allow selection of the most appropriate chemotherapy schedule for women with breast cancer. No significant financial relationships to disclose.