Transcript signatures that predict outcome and identify targetable pathways in MYCN-amplified neuroblastoma

Abstract

Background: In the pediatric cancer neuroblastoma (NB), patients are stratified into low, intermediate or high-risk subsets based in part on MYCN amplification status. While MYCN amplification in general predicts unfavorable outcome, no clinical or genomic factors have been identified that predict outcome within these cohorts of high-risk patients. In particular, it is currently not possible at diagnosis to determine which high-risk neuroblastoma patients will ultimately fail upfront therapy.

Experimental design: We analyzed the prognostic potential of most published gene expression signatures for NB and developed a new prognostic signature to predict outcome for patients with MYCN amplification. Network and pathway analyses identified candidate therapeutic targets for this MYCN-amplified patient subset with poor outcome.

Results: Most signatures have a high capacity to predict outcome of unselected NB patients. However, the majority of published signatures, as well as most randomly generated signatures, are highly confounded by MYCN amplification, and fail to predict outcome in subpopulations of high-risk patients with MYCN-amplified NB. We identify a MYCN module signature that predicts patient outcome for those with MYCN-amplified tumors, that also predicts potential tractable therapeutic signaling pathways and targets including the DNA repair enzyme Poly [ADP-ribose] polymerase 1 (PARP1).

Conclusion: Many prognostic signatures for NB are confounded by MYCN amplification and fail to predict outcome for the subset of high-risk patients with MYCN amplification. We report a MYCN module signature that is associated with distinct patient outcomes, and predicts candidate therapeutic targets in DNA repair pathways, including PARP1 in MYCN-amplified NB.

Keywords: Biomarkers; Confounding variables; Gene signatures; MYCN; Networks; Neuroblastoma.

Figure 1

Assessment of NBsigDB and randomly generated signatures. A) P‐value scores of the signatures comprising NBsigDB, signatures are ranked in order of significance. B) p‐value scores for 10,000 randomly generated for k‐length signatures (k = 5, 10, 25, 50). C) Histogram of p‐values for 10,000 randomly generated 25 gene signatures. D) Proportion of significant and non‐significant signatures among 5‐, 10‐, 25‐, and 50‐gene signatures is similar among each tested signature size.

Figure 2

NBsigDB signatures identify a difficult to classify phenotype among NB patients. A) Blue and red indicate the accuracy of prediction. Patients are divided based on survival/death and ordered based on ease of prediction. Gray boxes highlight the patients that are generally miss‐classified by the NB signatures.

Figure 3

Identification of factors that confound NBsigDB signatures. A) Among good outcome patients MNA tumors are significantly more difficult to classify using the 11 NB signature panel (*p < 0.0001, Wilcoxon rank sum test). B) Among poor outcome patients, MNA tumors are significantly easier to classify using the 11 NB signature panel (*p < 0.0001, Wilcoxon rank sum test). C) Incorrectly classified good outcome tumors are enriched in MYCN amplified tumors (*p < 0.0001, Fisher's exact test). D) Correctly classified poor outcome tumors are enriched in MYCN amplified tumors (*p < 0.0001, Fisher's exact test). E) The CORE signatures is a better predictor of MYCN amplification than patient outcome (*p = 0.0081, Fisher's exact test). F) p‐value scores for NBsigDB signatures in NA NB. All signatures are significant (hashed line indicates p = 0.05). G) p‐value scores for NBsigDB signatures in MNA NB. None of the signatures are significant (hashed line indicates p = 0.05).

Figure 4

Identification of a module based signature that predicts outcome of MNA NB patients. A) AUC analysis to identify outcome associated genes in MNA NB (AUC > 0.7 used as cut‐off). B) Modules identified from network analysis of identified predictive genes. C) AUC values for each of the individual predictive modules, as well as a combination module. The combination module is the best predictor of outcome. D) Kaplan–Meier survival analysis of each of the MYCN modules in the validation cohort (i–viii: Modules 0–6, combination: *p = 0.01, p = 0.0001, p = 0.07, p = 0.003, p = 0.1, p = 0.003 p = 0.01, p = 0.0004, log‐rank test). High Risk: High molecular risk. Low Risk: Low molecular risk.

Figure 5

PARP1 inhibitors specifically target MNA NB cell lines. A) Pathway analysis of module 6 genes highlights enrichment in double stranded DNA repair pathways (R: Reactome, K: KEGG, N: NCI Pathways). B) Western blotting reveals high and low MNA NB cell lines. C) Dose response experiments reveal that the PARP1 inhibitor BYK204165 more potently suppresses the growth of MNA vs. NA NB cell lines. D&E) BYK204165 selectively induces the expression of the apoptotic markers cleaved‐PARP and cleaved‐Caspase3 (CASP3) in MNA NB cell lines.