[1669] Discovery of Molecular Subtypes in Leiomyosarcoma through Integrative Molecular Profiling

AH Beck, CH Lee, DM Witten, S Zhou, K Montgomery, R Tibshirani, T Hastie, RB West, M Van de Rijn. Stanford University Medical Center, Palo Alto, CA; Vancouver General Hospital, Vancouver, BC, Canada; Stanford University, Palo Alto, CA

Background: The molecular pathogenesis and heterogeneity of soft tissue leiomyosarcoma (LMS) are poorly understood. We employed integrative molecular profiling to characterize molecular subtypes of LMS.
Design: Gene expression profiling was performed on 51 LMS cases using 44k cDNA arrays. Unsupervised clustering was performed to identify subclasses of LMS, and the protein expression of 6 differentially expressed genes was assessed on 3 LMS tissue microarrays (n=425). Array comparative genomic hybridization (aCGH) was performed on 44k arrays for 20 LMS cases. Sparse canonical correlation analysis was performed to identify sets of copy number changes that show significant correlation with sets of gene expression changes. Connectivity map analysis (CMAP) was performed to identify drugs that induce patterns of gene expression inversely correlated with that seen in the LMS subtypes.
Results: Unsupervised clustering demonstrates 3 LMS clusters, which show a high degree of reproducibility (Median In-Group Proportion=1 and p = 0.05, following 100 iterations of the clusterRepro algorithm). The clusters show differential expression of multiple gene classes, including genes involved in smooth muscle differentiation. Cases from the muscle-enriched cluster show increased copy number alterations (p=0.03), with the majority of cases showing gain at Xp22, 7q31, 3q13, and loss at 1p36 and 16q24 (FDR= 0.016). Sparse canonical correlation analysis performed by chromosome shows that copy number changes in chromosomes 2, 12, and 15 each show significant correlation (p0.05) with sets of gene expression changes. Immunohistochemistry is performed for 6 markers that were highly expressed in the muscle-enriched cluster (ACTG2, CASQ2, SLMAP,ELF1,MYLK,PRUNE2) and demonstrates strong expression of the majority of the proteins in 25% of cases, no strong expression of any of the markers in 35% of cases, and intermediate expression in 30% of cases. The CMAP demonstrates that the 3 LMS clusters defined by gene expression analysis show significant negative connectivity with distinct classes of chemotherapeutic agents.
Conclusions: This integrative analysis demonstrates distinct molecular LMS subtypes, provides insight into their pathogenesis, and suggests that they may respond to distinct classes of chemotherapeutic agents.
Category: Special Category for 2009 - Pan-genomic/Pan-proteomic approaches to Cancer

Monday, March 9, 2009 9:30 AM

Poster Session I Stowell-Orbison/Autopsy Award # 238, Monday Morning