Gene expression-based classification of nonseminomatous male germ cell tumors

Male adult germ cell tumors (GCTs) comprise two major histologic groups: seminomas and nonseminomas. Nonseminomatous GCTs (NSGCTs) can be further divided into embryonal carcinoma (EC), teratoma (T), yolk sac tumor (YS), and choriocarcinoma (CC) on the basis of the lineage differentiation that they exhibit. NSGCTs frequently present as mixed tumors consisting of two or more histological subtypes, often limiting correlative studies of clinical and molecular features to histology. We sought to develop a molecular classifier that could predict the predominant histologic subtype within mixed NSGCT tumor samples. The expression profiles of 84 NSGCTs (42 pure and 42 mixed) and normal age-matched testes were obtained using Affymetrix microarrays. Using prediction analysis for microarrays, we identified 146 transcripts that classified the histology of pure NSGCTs samples with 93% accuracy. When applied to mixed NSGCTs, the classifier predicted a histology that was consistent with one of the reported components in 93% of cases. Among the predictive transcripts were CGB (high in CC), LCN2 (high in T), BMP2 (high in YS), and POU5F1 (high in EC). Thus, the expression-based classifier accurately assigned a single predominant histology to mixed NSGCTs, and identified transcripts differentially expressed between histologic components with relevance to NSGCT differentiation.