Since the last reporting period, I completed the evaluation of the intrinsic cellular sensitivity to 4-hydroperoxycyclophosphamide (4HC), an activated form of cyclophosphamide and to a metabolite of cyclophosphamide thought to produce the antitumor effect, phosphoramide mustard cyclohexylamine salt (PM) for 431 HapMap cell lines from patients of African, Caucasian and African-American ancestry. We have increased the number of cell lines phenotyped from 265 during the NCE period to increase our statistical power to detect associations. Cyclophosphamide is a prodrug that requires hepatic cytochrome P450 activation and therefore requires pre-activation in LCLs, which lack P450 expression. 4HC spontaneously degrades to acrolein, the urothelial toxin, and phosphoramide mustard, the active alkylating metabolite. We hope to elucidate genomic signatures associated with both acrolein- and phosphoramide mustard-related sensitivity and toxicity through studying the parent drug and a metabolite important in antitumor activity. I have exposed our cell lines to clinically relevant concentrations of both of these compounds and have used IC50 – the concentration that produces a 50% cell growth inhibition – as a phenotype for my genome wide association study.

The Dolan lab has pioneered an integrative approach to GWAS that incorporates baseline mRNA exon array expression data for each cell line.  A second association study, between genotype and gene expression, helps to inform the functional role of single nucleotide polymorphisms throughout the genome in terms of their impact on baseline expression of nearby (cis-) and distant (trans-) genes. This data is incorporated into a “Triangle Method” of pharmacogenomic discovery whereby SNPs associated with drug IC50 are narrowed down to include only those SNPs that are associated with the expression of one or more genes, so called expression quantitative trait loci (eQTLs).  Further refinement is achieved by including eQTL target genes that are also associated with the drug phenotype.

In a meta-analysis of all three populations of LCLs analyzed, I have identified 261 SNPs associated with 4HC sensitivity and 307 SNPs associated with PM sensitivity at a p-value cutoff of 1 x 10^-4.  12 4HC SNPs and 22 PM SNPs are associated with the expression of one or more genes at a p-value cutoff of 3 x 10^-6.  7 eQTL target genes are associated with 4HC IC50 and 4 eQTL target genes are associated with PM IC50. One gene, ORMDL3, was common to both the 4HC and PM triangle analysis. These findings will be the basis of my further studies at the University of Chicago under the continued mentorship of Dr. Dolan.

In addition, top SNPs from the 4HC and PM GWAS will be evaluated against germline genotypes of 1,400 children with high-risk neuroblastoma who received cyclophosphamide-based chemotherapy. We hypothesize that a subset of the variants associated with in vitro resistance to cyclophosphamide will also associate with poorer event-free survival. Those variants that associate with both in vitro resistance and treatment failure will be the basis of further investigations to better understand the biology of cyclophosphamide resistance.

Directions for Future Translational Research

1. Validation of cyclophosphamide sensitivity-related SNPs in clinical cohorts

Despite Epstein-Barr virus transformation and genetic differences between LCLs and normal human cells, previous Dolan lab experiments using the “Triangle Method” have proven to be a clinically relevant discovery mechanism for germline genetic variants associated with both chemotherapy response and toxicity in patients. One important example of the success of this approach is with the Australian Ovarian Cancer Study (AOCS). Patients (approx. 400) were followed for a minimum of 5 years, past the expected median progression free survival (PFS) for stage III ovarian cancer of 17.1 months. A primary aim of the trial was to evaluate the association between SNPs identified from the Dolan Lab cell-based model on PFS in ovarian cancer patients who are treated with a taxane and carboplatin. In the initial GWAS, 342 SNPs were associated with carboplatin-induced cytotoxicity, of which 18 unique SNPs were retained after assessing their association with gene expression. One SNP (rs1649942) was replicated in an independent LCL set (p-value_{Bonferroni adjusted}=9×10^-3). It was found to be significantly associated with decreased PFS in phase-1 AOCS patients (P_per-allele=2×10^-2), with a stronger effect in the subset of women with optimally debulked tumours (P_per-allele=4×10^-3) (RS Huang et al., Clin Cancer Res. 2011 Jun 24, epub ahead of print). Secondly, using data from a CALGB-sponsored genome-wide analysis of paclitaxel-induced peripheral neuropathy, there was a statistically significant enrichment of SNPs discovered in the cell-based model after exposure to paclitaxel in the top SNPs associated with peripheral neuropathy in patients on the clinical trial (unpublished data, H. Wheeler et al.). These findings suggest that the genome wide analyses of human LCLs can serve as a powerful discovery tool in cancer pharmacogenomics.

Similar to the studies described above, I plan to validate my cell based findings in a large clinical cohort. In a recent analysis published in the Journal of Clinical Oncology that I helped complete, Drs. Tara Henderson and Susan Cohn among others discovered that African-American children with neuroblastoma had a statistically significant increase in late relapses (>2 years off therapy) when compared to Caucasian children, suggesting that African-American children may have a higher incidence of minimal chemotherapy-resistant disease. I hypothesize that African-American children will have higher mean allelic frequency of common genetic variants associated with chemotherapy resistance in the LCL model. To test this hypothesis, I am analyzing germline genotypes from over 1,400 patients with high-risk neuroblastoma obtained from the lab of Dr. John Maris and clinical outcome data obtained from the Children’s Oncology Group (COG), to look for an enrichment of the top SNPs discovered in the cyclophosphamide project in the subset of patients who do not respond to cyclophosphamide-based chemotherapy. In addition, I will perform a comprehensive analysis of SNPs associated with sensitivity to cisplatin, carboplatin and etoposide discovered in the LCL model for their role in non-response to neuroblastoma therapy.

This study offers the potential to discover variants associated with non-response to cyclophosphamide therapy and to personalize treatment to improve response and reduce toxicity associated with this commonly used drug.

2. Understand the role ofORMDL3in cellular sensitivity to cyclophosphamide

The single gene identified in both the 4HC and PM triangle analyses was ORMDL3. ORMDL3, a member of the ORM family of endoplasmic reticulum membrane-anchored proteins, and is encoded by a gene located on chromosome 17q21. This gene recently gained attention when SNPs that were associated with ORMDL3 transcript abundance were implicated in several genome-wide association studies of asthma (Moffatt MF et al. Nature 448, 470-473, 2007). Relatively little is known about the function of ORMDL3, although preliminary studies show that it appears to be involved in cellular calcium and sphingolipid homeostasis and may regulate the unfolded protein response. In our studies, increased expression of ORMDL3 was associated with relative cellular resistance to 4HC and PM. Interestingly, this gene is located in an area of chromosome 17 that is often amplified in clinically aggressive subtypes of neuroblastoma and breast cancer.

I am planning a series of experiments to further understand the role of ORMDL3 in cellular sensitivity to cyclophosphamide. First, I plan to knock down the gene transcript using siRNA in human LCLs. I hypothesize that the reduced ORMDL3 expression induced by siRNA knockdown will lead to increased sensitivity to both 4HC and PM. I also plan to knock down this gene transcript in both neuroblastoma and breast cancer cell lines and measure its effect on cyclophosphamide sensitivity. If these experiments prove successful, I plan to proceed with small molecule library screening to look for chemical inhibitors of ORMDL3 and to measure their effect on cyclophosphamide sensitivity.

Given ORMDL3’s putative role in the unfolded protein response as well as calcium and sphingolipid homeostasis, future directions of this project could include global assessment of these factors in normal and neoplastic cell lines with and without ORMDL3 knockdown after exposure to 4HC and PM. Manipulations of these systems could prove to be important in increasing cellular sensitivity to cyclophosphamide.