Clinical Cancer Research vol:18 issue:3 pages:826-838
PURPOSE: The importance of KIT and PDGFRA mutations in the oncogenesis of gastrointestinal stromal tumors (GISTs) is well established, but the genetic basis of GIST metastasis is poorly understood. We recently published a 67 genes-expression prognostic signature related to genome complexity (CINSARC for Complexity INdex in SARComas) and asked whether it could predict outcome in GISTs.Experimental design: We performed genome and expression profiling on 67 primary untreated GISTs. RESULTS: We show and validate here that it can predict metastasis in a new dataset of 67 primary untreated GISTs. The gene whose expression was most strongly associated with metastasis was AURKA, but the AURKA locus was not amplified. Instead, we identified deletion of the p16 (CDKN2A) and retinoblastoma (RB1) genes as likely causal events leading to increased AURKA and CINSARC gene expression, to chromosome rearrangement and ultimately to metastasis. Based on these findings we established a Genomic Index that integrates the number and type of DNA copy number alterations. This index is a strong prognostic factor in GISTs. We show that CINSARC class, AURKA expression and Genomic Index all outperform the AFIP grading system in determining the prognosis of patients with GISTs. Interestingly, these signatures can identify poor prognosis patients in the group classified as intermediate-risk by the AFIP classification.CONCLUSIONS: We propose that a high Genomic Index determined by Comparative Genomic Hybridisation (CGH) from formalin-fixed paraffin-embedded samples could be used to identify AFIP intermediate-risk patients who would benefit from imatinib therapy.