P.Depuydt/V.Boeva:shared first authors
G.Schleiermachter/K.De Preter:shared last authors
Prognostic classification of high-risk neuroblastoma patients remains challenging. The aim of this study was to define genomic biomarkers for upfront identification of ultra-high-risk patients that will develop resistance to chemotherapy.
High-resolution array-CGH profiles of diagnostic tumor samples from 570 high-risk neuroblastoma patients were collected from 8 collaborative groups. To correct for possible inter-platform variability, raw data were transformed to a common genome build and segmented using SegAnnDB.
First, datasets from seven European centres were used for training (n=317), while the COG-cohort (n=253) was used for validation of the classifier. We built a classifier using the regions with differential copy numbers (logistic regression analysis) in two extreme training groups, i.e. high-risk patients that die within 18 months and high-risk patients that survive with at least 5 years event-free follow-up. However, this classifier could not discriminate patients with ultra-high-risk in the validation cohort. A possible explanation is that the different cohorts have slightly different population characteristics and treatment protocols.
Second, when we investigated the prognostic power of individual aberrations in both cohorts, we identified regions that are prognostic in one cohort but not in the other, including 1p-deletion and 1q-gain which do not predict poor outcome in the COG cohort. Data from other cohorts will be collected to validate a classifier based on regions that are prognostic across different cohorts.
Indeed, both cohorts also share copy number aberrations that are prognostic, including MYCN-amplification, 4p-deletion and distal 6q-deletion (below -0.3 log relative copy number). Most significant results were observed for 6q-deletion which defines a small subgroup of ultra-high-risk patients with extremely poor outcoma as 92% (12/13) and 100% (9/9) of the training and validation patients with this aberration died of disease.
In conclusion, we identified distal 6q-deletions (more distal to ARID1B previously described in neuroblastoma) that mark neuroblastoma patients for fatal outcome. Further studies are ongoing to identify the 6q-gene/regulatory element that drives the aggressiveness of this disease.