Large-scale serial analysis of gene expression reveals genes differentially expressed in ovarian cancer.

ABSTRACT

Difficulties in the detection, diagnosis, and treatment of ovarian cancer result in an overall low survival rate of women with this disease. A better understanding of the pathways involved in ovarian tumorigenesis will likely provide new targets for early and effective intervention. Here, we have used serial analysis of gene expression (SAGE) to generate global gene expression profiles from various ovarian cell lines and tissues, including primary cancers, ovarian surface epithelia cells, and cystadenoma cells. The profiles were used to compare overall patterns of gene expression and to identify differentially expressed genes. We have sequenced a total of 385,000 tags, yielding >56,000 genes expressed in 10 different libraries derived from ovarian tissues. In general, ovarian cancer cell lines showed relatively high levels of similarity to libraries from other cancer cell lines, regardless of the tissue of origin (ovarian or colon), indicating that these lines had lost many of their tissue-specific expression patterns. In contrast, immortalized ovarian surface epithelia and ovarian cystadenoma cells showed much higher similarity to primary ovarian carcinomas than to primary colon carcinomas. Primary tissue specimens therefore appeared to be a better model for gene expression analyses. Using the expression profiles described above and stringent selection criteria, we have identified a number of genes highly differentially expressed between nontransformed ovarian epithelia and ovarian carcinomas. Some of the genes identified are already known to be overexpressed in ovarian cancer, but several represent novel candidates. Many of the genes up-regulated in ovarian cancer represent surface or secreted proteins such as claudin-3 and -4, HE4, mucin-1, epithelial cellular adhesion molecule, and mesothelin. Interestingly, both apolipoprotein E (ApoE) and ApoJ, two proteins involved in lipid homeostasis, are among the genes highly up-regulated in ovarian cancer. Selected serial analysis of gene expression results were further validated through immunohistochemical analysis of ApoJ, claudin-3, claudin-4, and epithelial cellular adhesion molecule in archival material. These experiments provided additional evidence of the relevance of our findings in vivo. The publicly available expression data reported here should stimulate and aid further research in the field of ovarian cancer.