High throughput quantification of absolute gene copy numbers in human tumors from microarray data

Abstract

2935

The genomic microarray technology provides information about gene copy numbers in cancers. Data from clinical samples are not only influenced by the copy numbers but also by the normal tissue content of the biopsy and the total DNA content (ploidy) of the tumor cells. These factors vary among tumors, leading to unreliable results when data from different patients are compared. We have developed a method to extract absolute gene copy numbers from the relative values obtained by traditional analysis, using a formula that includes the copy numbers, ploidy, tumor cell fraction, and an experimental constant describing the dynamics of the relative values.

The method was applied on genomic BAC array data from 100 lymphomas and 52 cervical carcinomas. The formula was verified by fluorescence-in situ-hybridization (FISH) of selected genes in the lymphomas, for which the ploidy and tumor cell fraction were measured by flow cytometry and used as known variables. In cervical carcinomas, where the ploidy but not the tumor cell fraction was known, the calculated absolute values were also in agreement with FISH data. We used the method for gene copy number mapping of chromosome 17q in the cervical carcinomas, relating the copy numbers in this region to treatment outcome and expression of corresponding genes, as determined from gene expression microarray analysis. Increased copy numbers within 17q21.32 - q25.1 were associated with poor progression free survival (p < 0.05). Similar relationships to survival were seen at the expression level for six genes within this region: NME1, MRPS23, THRAP1, GK001, GRB2, and ATP5H. A significant correlation between the expressions and copy numbers was found for each of these genes, suggesting that they are possible target genes for the 17q gain in cervical carcinomas. The method was also used to identify intratumor heterogeneity in gene copy numbers, which indicates pronounced genomic instability. The patients with heterogeneous tumors showed lower survival probability than the others (p < 0.001, lymphomas; not significant, cervical carcinomas). This work suggests that increased reliability in genomic microarray studies and novel information about the intratumor genomic heterogeneity in human tumors can be achieved by the use of our method.