Expression profiling and CGH array as prognostic tools in hepatoblastoma

Hepatoblastoma (HB) is the predominant type of malignant liver tumors in childhood. This rare tumor occurs generally before 3 years of age. The etiology of hepatoblastoma is unknown, but an association with congenital abnormalities is well established. Importantly, deregulation of the Wnt pathway through activating mutations of ß-catenin plays a crucial role in the development of this tumor. HBs are characterized by the proliferation of immature hepatocytes frequently associated to mesenchymal tissue, suggesting that HB is derived from uncommitted hepatic progenitors. In most of the cases, cure of HB patients is achieved by surgery and chemotherapy. However, approximately one fourth of the children does not survive the disease. In this study, we aimed at identifying gene expression signatures and chromosomal alterations related to aggressive HBs. To this end, we analyzed 68 HB samples from 63 patients for which extensive clinical and pathological parameters were collected. Most patients were enrolled in the international clinical trials of the SIOPEL group. DNA and RNA were prepared and their quality was checked by Agilent technology. Gene expression profiles and chromosomal aberrations were assessed on Affymetrix U133A and CGH arrays in a set of 30 tumors. Significant findings were validated by real-time PCR in an independent set of 38 tumors. Unsupervised hierarchical clustering analysis clearly identified two tumor groups: Cluster-1: tumors with predominant fetal histology (p<0.0001), low proliferation rate by Ki67 immunohistochemistry (p=0.04), no evidence of vascular invasion (p=0.007), and relative genetic stability (mean altered chromosomes: 2, p=0.003). Cluster-2: tumors exhibiting less differentiated patterns, high proliferation, blood vessel invasion and higher genetic instability (mean altered chromosomes: 6). Comparison of gene expression profiles between clusters 1 and 2 revealed high expression of genes located on chromosomes 2 and 8 in cluster-2 tumors (p<0.001). Accordingly, in CGH array analysis, gains of chromosomes 2 and 8 were significantly associated to cluster-2. Moreover, the 2-year probability of survival differed significantly between patients with cluster-1 versus cluster-2 tumors (85% vs. 29%, log rank= 0.0022). Eighteen differently expressed genes (p<0.000001) between clusters 1 and 2 were identified as putative prognostic markers and their predictive value was validated in the independent set of 38 HBs. We conclude that gene expression profiles and CGH arrays can provide molecular clues for identifying key events in HB and for designing new therapeutic approaches for patients at greater risk of poor outcome.