Chemotherapy response of neuroblastoma patient-derived xenografts correlated with patient outcome and identified gene expression patterns associated with drug resistance
Introduction
Neuroblastoma (NBL) is the most common extra-cranial solid tumor that develops in the sympathetic nervous system of children. The five years survival rate for high-risk NB patients is less than 60%, and approximately 15% of high-risk neuroblastoma patients develop progressive disease (PD) during induction chemotherapy that includes cyclophosphamide (cyclo) and topotecan (topo). As tumor biopsies are often not obtained at time of progressive disease, identifying underlying genetic/genomic features that contribute to disease progression is challenging. Patient-derived xenografts (PDX) established from NBL cells in bone marrow aspirates or blood samples obtained pre-therapy and at time of progressive disease can be phenotypically and genetically characterized and may identify biomarkers predictive of treatment outcome in high-risk NBL patients.
Methods
Thirty-five high-risk patient tumor, marrow, or blood samples, 16 at diagnosis (DX) and 19 at progressive disease (PD), were received via Children’s Oncology Group (COG) protocol ANBL00B1 and established as PDX models at COG/ALSF repository. PDXs were tested for and free of mouse + human pathogens, characterized by histology, expression of tyrosine hydroxylase, short-tandem repeat assay, and human vs mouse content. PDXs were classified by induction chemotherapy (cyclo (30mg/kg) + topo (0.6 mg/kg) daily for 5 days every 21 days, total of 3 cycles : Non-responders (NR) and responders (R) that could be further divided into partial responders (PR) and complete responders (CR). Gene expression in the PDX models prior to therapy was analyzed by RNA-sequencing at the Human Genome Sequencing Center at Baylor College of Medicine, and the gene expression profiles were analyzed by three different methods including differential analysis (DEGs) via edgeR, binomial regression model and student’s t-test to identify the marker genes for NR group. The markers were validated using real-time RT-PCR in PDX samples.
Results
PDX cyclo/topo response was: non-response (NR, n=7), partial response (PR, n=15), and complete response (CR, n=13). The event-free survival in all treatment groups was significantly extended relative to their control (P < 0.001). Four of seven (57%) of the NR PDXs were established from post-mortem progressive disease (PD-PM) samples while 9 of 13 (69%) of the CR group were established from diagnosis (DX) samples, demonstrating the clinical relevance of the PDX models. The RNA-seq data analysis identified 145 differentially expressed genes across the NR versus response groups. These genes were further filtered by a regression model and t-test significance to define a signature comprising of seven genes for the NR group. The signature genes scored by ssGSEA significantly discern the NR group from responders (P-value < 4.3e-13). Of the 122 common marker genes from DEGs and a regression model, we investigated 14 genes by RT-PCR. RT-PCR correlated with RNAseq expression for 9 of the 14 genes (P-value < 0.05 and Pearson). RT-PCR validation confirmed significant differential expression in association with response to cyclo/topo (NR vs responders) for five of the genes (t-test P-value < 0.05).
Conclusion
Response of PDXs to cyclo/topo was greater for those established pretherapy (DX) compared to those established at death from progressive disease (PD-PM). The gene expression signature of non-responding PDXs can potentially provide a biomarker to identify patients destined to have suboptimal responses to induction chemotherapy. Genes overexpressed in non-responding PDXs provide potential molecular targets to reverse drug resistance. In addition, the well-characterized and clinically relevant panel of PDXs will be valuable in preclinical studies of novel therapeutic approaches for high-risk neuroblastoma patients.