Identification Of Novel Prognostic Signatures In Rhabdomyosarcoma By Whole Transcriptome Expression Profiling: A Discovery And Validation Study

Proceedings: AACR Annual Meeting 2014; April 5-9, 2014; San Diego, CA INTRODUCTION: Pediatric rhabdomyosarcoma (RMS) has varying outcomes, especially in intermediate-risk disease (IR-RMS) due to the inherent inability of clinical staging to accurately risk-stratify a large proportion of patients. This study aimed to identify prognostic signatures in IR-RMS, the clinical subgroup with the most heterogeneous outcomes, which can potentially provide better risk stratification than routine clinicopathologic parameters. Signature performance was validated on an independent set of RMS patients. METHODS: Prospectively-obtained primary tumors from 80 IR-RMS patients on Children's Oncology Group clinical trial protocols formed the training set. Tumors from 19, 15 and 20 patients with low-risk, high-risk and IR-RMS formed the validation set. Annotated and unannotated transcripts were profiled by Affymetrix Human Exon microarrays employing 1,393,765 probe selection regions (PSRs) and used to derive weighted signatures. Potentials of coding and non-coding signatures to predict survival were compared using areas under receiver operating characteristic curves that provided a measure of predictive accuracy. Associated biological processes were analyzed using curated pathway analysis tools. RESULTS: Histologic subtype (p=0.94) and PAX-FKHR fusion status (p=0.66) were unable to predict survival in the training set of IR-RMS. Tumor site was the only clinical predictor of outcome in this set (p=0.041). Cox regression on 17,045 coding transcripts identified a prognostic 30-gene meta-feature (30gMF, p=0.001). Analysis of unannotated transcripts identified a 39-PSR meta-feature (39ncMF) that also predicted survival (p<0.001). Multiple PSRs interrogating the same genomic locus were then replaced by a single PSR that reduced ncMF size to 34 PSRs (34ncMF), which could still predict outcome (p<0.001). Predictive accuracy of 39ncMF was higher than 34gMF (96.4% vs. 70.8%, p<0.001). However, predictive accuracy of the former was comparable to the 34ncMF (96.7%, p=0.54). When applied to the validation set, the 34gMF, 39ncMF and 34ncMF were able to predict outcomes (p=0.022, 0.006, 0.012, respectively). Analysis of biological processes using 34gMF showed enrichment for functions/disorders associated with musculoskeletal development and signaling pathways. Similar analysis of non-coding signatures revealed enrichment for cellular assembly, cell cycle, apoptosis and cancer-associated functions. CONCLUSIONS: A concise non-coding RNA meta-feature was able to better predict outcome in IR-RMS than a coding gene meta-feature, where most standard clinical prognosticators failed. The meta-features were independently validated in IR and non-IR RMS. This suggests that non-coding transcripts can regulate and determine RMS biology and aggressiveness, and be used as novel prognostic indicators. Citation Format: Anirban P. Mitra, Sheetal A. Mitra, Jonathan D. Buckley, Philipp Kapranov, James R. Anderson, Stephen X. Skapek, Douglas S. Hawkins, Timothy J. Triche. Identification of novel prognostic signatures in rhabdomyosarcoma by whole transcriptome expression profiling: A discovery and validation study. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4730. doi:10.1158/1538-7445.AM2014-4730