Improved outcome prediction of children with neuroblastoma using a multigene expression signature

Few tumours have engendered as much fascination and frustration for clinicians and scientists as neuroblastoma (NB). Although not frequent in absolute terms, NB represents an important clinical challenge, as it is fatal in almost half of the patients diagnosed; as such it is responsible for approximately 15% of childhood cancer mortality. Current therapeutic stratification is useful, but sometimes patients with identical clinico-pathological parameters, receiving the same treatment, can have markedly different clinical courses. The challenge remains to identify additional tumour-specific prognostic markers. As gene expression deregulation is a key trait of cancer cells, such an analysis of NB tumours might contribute to better outcome prediction and management of the patients. Amongst the various methods available to measure gene expression, the reverse transcription quantitative polymerase chain reaction (RT-qPCR) is the most rapid, sensitive, and reproducible method. In the first part of this thesis, we developed and optimized a workflow for accurate gene expression analysis using qPCR. We validated different methods for the assessment of RNA integrity and purity using only minute amounts of RNA and studied the impact of RNA quality on gene expression results. This analysis clearly demonstrates a measurable impact of RNA quality on the variation of the reference genes, on the significance of differential expression of prognostic marker genes between two cancer patient risk groups, and on risk classification performance using a multigene signature. Next, a sample pre-amplification method was extensively evaluated in order to generate sufficient material enabling large- scale qPCR gene expression studies using limited amounts of sample material. The results obtained from this study show a constant yield of pre-amplified cDNA independent of the amount of input RNA; preservation of differential gene expression after pre-amplification without introduction of substantial bias; no co-amplification of potential contaminating genomic DNA; no necessity to purify the pre-amplified material; and finally the importance of good RNA quality for successful pre-amplification. Finally, we introduced and validated a strategy that employs universally applicable, quantifiable external oligonucleotide standards to detect and correct inter-experimental variation in order to enable comparison of data and large multicentre studies. Using the proposed standards and data-analysis procedure, we obtained a perfect concordance between expression values from eight different genes in 366 patient samples measured on three different qPCR instruments and matching software, reagents, plates and seals, demonstrating the power of this approach to detect and correct inter-run variation and to enable exchange of data between different laboratories, even when not using the same qPCR platform. In the second part of this thesis, we applied the above described strategies on a large panel of NB samples and aimed to develop and validate a prognostic multigene signature for better outcome prediction of NB patients at the time of diagnosis. We first created the Neuroblastoma Gene Server (NBGS) database, an intuitive online tool to compare a user’s own list of genes with a collection of curated lists of differentially expressed genes in NB, derived from published high-throughput neuroblastoma gene expression studies and annotated by the specific group comparison or perturbation that resulted in differential gene expression. Next, we developed an innovative data-mining strategy based on a re- analysis of published gene expression studies in order to prioritize robust prognostic marker genes. Hence, a classifier was built, the strength of which lies on its small number of genes and on its cross-platform validity in which it outperforms other published prognostic signatures. We extended this study by a re-analysis of three additional published gene expression studies and an extensive literature screening for single candidate prognostic genes. In total, we composed a list of 59 prognostic genes that were profiled in the largest NB patient series (n=579) to date using qPCR starting from only 20 ng of RNA. The signature classifies patients accurately with respect to overall and progression-free survival (p