5438

Background: The enhancer of zeste homologue 2 (EZH2) is a member of the polycomb group of genes and is involved in cell cycle regulation. Strong EZH2 protein expression has been documented as a marker of invasion and proliferation in a variety of malignant tumors. In colorectal cancer (CRC), however, the status of EZH2 is not clear. In this pilot study, for the first time the immunohistochemical expression of EZH2 protein in invasive colorectal carcinomas has been evaluated.

Design: We generated a tissue microarray (TMA) with 85 specimens from 38 cases that included CRC and adjacent histologically normal epithelium. Immunohistochemical staining wasperformed using a DAKO immunostainer with the rabbit anti-EZH2 antibody (18-7395; 1:200, Zymed). The manual semiquantitative analysis was based on the evaluation of the intensity of nuclearreaction for EZH2 and the percentage of tumor cells stained positive and categorized in 4-tier system as negative (0), weakly positive (1+), moderatelypositive (2+) or strongly positive (3+). Two quantitative types of scores were generated using Automated Cellular Imaging System (ACIS, Clarient): an automated score (ACIS-Auto) and an ACIS-assisted score. Aperio ScanScope and TMALab software for nuclear algorithm were applied as an additional type of quantification. Automatically generated measurements were then translated into the 4-tier system to compare with semiquantitative analysis. Scores were compared between CRC and normal epithelium using sign-rank test. Receivers operating characteristic (ROC) analyses were performed to quantify the accuracy of 4 methods to discriminate tumor and normal tissue.

Results: Normal colorectal epithelium was moderately and weakly positive for EZH2 expression in 61% (22/36) and 33% (12/36) specimens, respectively. No strongly positive staining has been detected. In contrast, CRC showed strong EZH2 immunostaining in 55% (21/38) of cases and moderate expression in 42% (16/38) cases. Moreover, higher EZH2 expression in CRC specimens versus normal epithelium was significant by all 4 scoring methods (p≤0.0001). According to ROC analysis, semiquantitative and ACIS-assisted methods were the most accurate.

Conclusions: Significant increase in EZH2 expression in CRC versus to adjacent normal colonic epithelium detected in our study suggests the involvement of EZH2 in CRC tumor progression. Interactions between the EZH2 and otherpredictive and prognostic markers will be investigated on larger cohort. The most effective way for the evaluation of novel biomarkers in cancer on TMAs is the combination of computer-assisted methods with pathologist-based verification of each image.

98th AACR Annual Meeting-- Apr 14-18, 2007; Los Angeles, CA