A gene signature of primary tumor identifies metastasized seminoma

Abstract

Background The aim of this study was the prediction of metastatic status in seminoma based on examination of the primary tumor. Methods Total RNA was isolated from metastasized seminoma ( n = 10, T1N1-2M0), non-metastasized seminoma ( n = 21, T1-3N0M0), and corresponding normal tissues. Pooled RNA from 10 biopsies of each tissue type was hybridized on whole genome microarrays for screening purposes. Ninety-two selected gene candidates were quantitatively examined using real-time quantitative polymerase chain reaction (RTQ-PCR). Results Agreement in gene expression was 88% between the whole genome microarrays and RTQ-PCR. Metastasized seminoma showed 1,912 up-regulated and 2,179 down-regulated genes with ≥2-fold differences in gene expression compared non-metastasized seminoma. RTQ-PCR of selected genes showed that mean gene expression values were significantly reduced in metastasized compared with non-metastasized seminoma. The presence of metastases could be predicted based on an 85-gene expression signature by using logistic regression. Sensitivity and accuracy of the 10-fold cross-validation model were 77.8% and 84.2%, respectively. Conclusion A logistic regression model using an 85 gene expression signature allowed identification of metastasized seminoma from the primary tumor with a sensitivity of 77.8%.