Abstract 4350: Translating diagnostic gene expression profiles for pediatric solid tumors

Abstract

Abstract BACKGROUND: Features of cancer genomics including gene expression levels can be employed to create biomarker profiles that can predict prior to therapy the diagnosis and prognosis of an individual patient. We aim to translate robust and reproducible diagnostic and prognostic profiles for pediatric solid tumors into clinical tools applicable to all tumor specimens, fresh frozen (FF) or formalin fixed (FFPE). RESULTS: We used Affymetrix GeneChip Human Exon 1.0 ST (HuEx) arrays on 24 cell lines or tumors (4 fusion-positive rhabdomyosarcoma [RMSpos], 5 fusion-negative RMS [RMSneg], 6 Ewing sarcoma family of tumors [ESFT], 5 neuroblastoma [NB], and 4 osteosarcoma [OS]) to identify a 41-feature diagnostic metagene that clearly distinguished both the original test samples and a set of validation samples. We then incorporated the 41-feature metagene into a 48-gene panel. To translate this diagnostic signature, we compared the performance of HuEx arrays with several mid-plex assay platforms including Fluidigm's quantitative reverse-transcriptase PCR (q-RT-PCR) and NanoString's nCounter™ Digital Analyzer by measuring gene expression on an 18-cell line panel (4 RMSpos, 4 RMSneg, 5 ESFT, 1 NB, and 4 OS). With q-RT-PCR, we observed r2 values of 0.703 ± 0.085 (RMSpos), 0.558 ± 0.138 (RMSneg), 0.689 ± 0.070 (ESFT), and 0.607 ± 0.121 (OS). nCounter reported r2 values of 0.672 ± 0.171 (RMSpos), 0.620 ± 0.143 (RMSneg), 0.675 ± 0.094 (ESFT), and 0.583 ± 0.129 (OS). Hierarchical clustering with all platforms was able to distinguish the RMSpos, ESFT, and NB cell lines; however, the RMSneg and OS cell lines were less clearly distinguished because some of the OS cell lines lacked high expression of genes characteristic of OS tumors. Cross-platform comparisons yielded good correlation between nCounter and Q-RT-PCR (average r2 = 0.710 ± 0.183). An 18-sample dilution series revealed that EWS-FLI1 type I or PAX3-FKHR transcripts were consistently detected by q-RT-PCR or nCounter even at 1:1000 RNA dilution. Among biological duplicates, q-RT-PCR reported an average r2 = 0.916 ± 0.039 while nCounter obtained average r2 = 0.983 ± 0.011 including raw cell lysates. In addition, we also compared Fluidigm's q-RT-PCR with Applied Biosystems’ Taqman Low-Density Arrays (TLDA) on five NB patient bone marrow samples to detect residual tumor cells. We observed an average r2 = 0.929 ± 0.119 across five NB genes and one housekeeping gene. Finally, we have also successfully profiled FF vs. FFPE tumors on HuEx and are currently applying these mid-plex technologies to paired FF and FFPE samples in order to create diagnostic profiles applicable to routine FFPE material. CONCLUSIONS: Mid-plex platform can reliably distinguish bone and soft tissue sarcoma cell lines and can be used to translate the application of diagnostic signatures. These data warrant further studies to analyze FFPE samples, to compare additional mid-plex platforms, and to test potential prognostic signatures. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 4350. doi:10.1158/1538-7445.AM2011-4350