The Maris Laboratory and Translational Research Program is focused on the childhood cancer neuroblastoma, a malignancy of the developing sympathetic nervous system. This disease continues to exact a significant proportion of overall pediatric cancer morbidity and mortality. The research plan is purposefully comprehensive, with focused investigations spanning from the genetic basis of neuroblastoma to Phase 3 clinical trials. There is a steadfast focus on translational research advances as the main goal of the laboratory efforts are to impact clinical practice, and we approach this challenge with urgency. This summary of our research program is divided into major themes, and each section discusses recent accomplishments in the past five years, as well as major future directions.
Like most human cancers, a subset of neuroblastomas is inherited within families. Pedigrees informative for classic genetic analyses are rare due to the fact that the disease is often lethal prior to reproductive age. We have been collecting these pedigrees and in the past five years achieved a critical mass powered sufficiently for a genome-wide linkage scan. Using a highly informative single nucleotide polymorphism (SNP) microarray, we showed that the majority of neuroblastoma pedigrees were linked to a single chromosomal region at 2p24 (Mosse et al., Nature 2008). The critical region harbored two lead candidates: the MYCN oncogene, that we and others have studied for over a decade (Maris et al., Lancet 2007); and the ALK oncogene, previously shown to be a driver of anaplastic lymphoma tumorigenesis, and more recently by us to be amplified in a small subset of high-risk human neuroblastoma (George et al., PLoS ONE 2007). We then showed that familial neuroblastoma is due to activating mutations in the kinase domain of ALK, that these mutations were indeed oncogenic drivers, that mutations could also be acquired in the tumor cells somatically, and that siRNA-based inhibition of ALK signaling was cytotoxic to tumor cells harboring activated ALK (Mosse et al., Nature 2008).
Since our discovery of ALK mutations as the cause of hereditary neuroblastoma, we have developed a genetic screening assay in our Molecular Diagnostic laboratory for patients with potential hereditary predisposition to neuroblastoma. Because ALK status in tumor tissue may be of clinical utility, we have also developed assays to screen for somatic mutation, amplification and phosphorylation in archived specimens. We have screened a broad array of small molecular inhibitors of the ALK kinase, and selected the most potent for clinical development. A phase 1/2 clinical trial of this inhibitor opened in August 2009 using a novel design that allows for selection of patients with ALK aberrations during the dose finding stage.
Of interest was that fact that each of the families not showing linkage to 2p24 co-segregated other disorders of neural crest-derived tissues such as Hirschsprung disease and/or congenital central hypoventilation syndrome (CCHS). Since triplet repeat expansion mutations in a polyalanine domain of the PHOX2B gene were shown by others to be the cause of CCHS, we tested these families for PHOX2B mutations and showed that each had putative loss-of-function aberrations (Mosse et al., Am J Hum Genet 2004). Subsequently, we showed that loss of PHOX2B function leads to a block in normal sympathetic nervous tissue and neuroblastoma differentiation capabilities (Raabe et al., Oncogene 2008).
Taken together, these studies have identified the two major (and perhaps only) disease genes for familial neuroblastoma. We now have a molecular diagnostic assay to provide genetic counseling to families worldwide, and also have an identified a tractable therapeutic target for exploitation in the clinic.
In order to determine the genetic etiology of the 99% of neuroblastoma cases that do not have a family history of the disease, we are performing a genome-wide association study (GWAS) of 5,000 neuroblastoma cases compared to 10,000 controls. Here we hypothesize that sporadic neuroblastoma is a complex genetic disease, and that susceptibility is related to the interaction of common genetic variations affecting pathways critical to normal sympathetic nervous system development. This is the only pediatric cancer GWAS of which we are aware, and required a decade of preparation in terms of collection of appropriate samples. To date, we have discovered multiple common variations associated with neuroblastoma, and have robustly replicated each of these via European collaborations.
A major underlying hypothesis for our GWAS was that due to the presumed lesser impact of environmental factors on pediatric cancer etiology, the effect size of common variations on neuroblastoma would be more robust than those typically seen in other complex diseases. This proved true when we took an early look at our GWAS data after only 1,000 cases were studied and discovered a robust signal at 6p22 that was replicated in three independent case-series (Maris et al., N Engl J Med 2008). This signal fell within a novel gene that remains of unknown function, but subset analyses showed that not only were particular SNP alleles at 6p22 associated with neuroblastoma, but specifically to the aggressive subset of cases. These data were the first genetic evidence that the various phenotypes observed in patients, from benign to highly malignant, may be different in terms of their genetic origins.
Based on these observations, we performed a second GWAS focused only on the “high-risk” group of patients. Not only was the 6p22 signal enriched, but we also discovered a new association signal of multiple SNP alleles within the BARD1 gene at 2q35 that was also robustly replicated in independent case-series (Capasso et al., Nat Genet 2009). While BARD1 has been speculated to have to be a tumor suppressor, we have subsequently shown that the presence of the risk alleles (SNP genotype associated with neuroblastoma) was highly correlated with the expression of an oncogenic isoform of BARD1 in both developing neural tissues and in neuroblastomas. Ongoing work in the lab is focused on further understanding the biological meaning of the BARD1 association signal, and preliminary data show that high penetrance mutations occur somatically as well.
Having shown that SNP variations are associated with predisposition to develop neuroblastoma, we next turned our attention to the other major cause of human genetic diversity—copy number variation (CNV). We first needed to develop new computational tools to accurately detect CNVs from SNP-array data (Diskin et al., Nucleic Acids Res 2008), and then performed a CNV-based GWAS, resulting in the first validated CNV associated with predisposition to cancer (Diskin et al., Nature 2009). The CNV identified in this study resides in a complex region of segmental duplication at 1q21, but we were able to clone a novel member of the neuroblastoma breakpoint family of genes (NBPF23), so named because the founding member NBPF1 was discovered at the breakpoint of a constitutional translocation in a child with neuroblastoma. We showed that the CNV was directly related to NBPF23 expression in both developing sympathetic nervous tissues and neuroblastoma, with the homozygous deletion or single copy NBPF23 status (low expression) being associated with neuroblastoma.
As of September 2009, the GWAS is currently approximately 50% complete in terms of samples genotyped, and while the coming year will focus on completion of all genotyping and discovering additional variants with the goal of uncovering all of the genetic risk to develop neuroblastoma, we are now focused on understanding the biology beneath the signals. In addition, we have discovered several other SNP and CNV associations that have been validated and either have been submitted for publication or will be soon. Particularly intriguing is a new association signal in the LMO1 oncogene, and we showed that the risk genotypes are associated with increased expression and somatic gain of the locus in tumor tissues (Wang et al., Submitted.). We have grants pending to perform a complete resequencing of all identified and replicated SNP and CNV signals in both germline and paired tumor tissues, to explore the hypothesis that these GWAS signals are actually markers for important genes and pathways, and that rarer mutations will also occur in these regions. We already have proof for this at the BARD1 locus, and these mutations give us much more tractable tools to understand how the genes are involved in normal sympathetic nervous system development, and how common variations and mutations may influence tumorigenesis.
In parallel to the above work focused on germline DNA and cancer predisposition, my laboratory also focuses on somatic cancer genomics. Neuroblastoma is an incredibly diverse neoplasm, with the highest percentage of cases that will undergo a complete spontaneous regression without cytotoxic therapy of all human malignancies. However, on the other end of the spectrum, 50% of human neuroblastomas are widely disseminated by hematogenous and lymphatic metastases at the time of diagnoses, and survival probability for these Stage 4 cases remains poor. Thus, our efforts in translational genomics for neuroblastoma are several-fold. First, we seek to precisely define neuroblastoma phenotypes at the molecular level, allowing for a genomic classification of the disease. Second, we hope to utilize these data as predictive biomarkers of response to therapy and outcome, and to implement genomic classifiers in treatment stratification systems. Finally, we seek to discover the key oncogenic drivers of the high-risk phenotype, with a particular emphasis on those that can be leveraged as therapeutic targets.
We had previously used moderate throughput techniques to define somatic DNA copy number aberrations in neuroblastoma, and showed that hemizygous deletions at chromosome 1p36 and 11q23 were associated with an aggressive neuroblastoma phenotype in small pilot series. We then performed a definitive study in 915 neuroblastoma cases from the Children’s Oncology Group (COG) and demonstrated that deletions at 1p36 and 11q23 were each independently associated with risk for relapse, and the 11q deletions were an independent predictor of patient survival in multivariable analyses (Attiyeh et al., N Engl J Med 2005). These genomic factors are now used internationally within the COG to stratify therapy in ongoing prospective clinical trials (Ambros et al., Br J Cancer 2009; Maris et al., Lancet 2007).
During the past five years we have implemented microarray technologies to more thoroughly define the neuroblastoma genome. We first utilized gene expression profiling to define the molecular signature of neuroblastoma and to precisely define subsets based on specific gene expression profiles (Wang et al., Cancer Res 2006). These data were key to further understanding of Wnt/beta-catenin pathway activation in a subset of tumors (Liu et al., Oncogene 2008), and the myc pathway activation and/or absence of expression from genes that normally drive neuronal differentiation are highly associated with adverse patient outcome (Fredlund et al., Proc Natl Acad Sci U S A 2008).
We used BAC and then SNP microarrays to further define the genomic DNA signatures of primary neuroblastomas and cell line models, including developing several computational advances to precisely and accurate detect alterations on these arrays (Attiyeh et al., Genome Res 2009; Diskin et al., Genome Res 2006; Mosse et al., Genes Chromosomes Cancer 2007; Mosse et al., Genes Chromosomes Cancer 2005). These experiments allowed us to accurately define copy number alterations in neuroblastoma, integrate these with expression signatures (Wang et al., Cancer Res 2006), and define genomic subsets correlated with phenotypic subsets. Taken together, these data provided a clear map of aberrations associated with phenotype, with ongoing work focused on developing a SNP-based assay to assist in our therapy allocation algorithm.
We have used these genomic datasets to map critical oncogenes and tumor suppressors. We have discovered that the prognostically relevant 1p36 deletions target multiple genes. First, we used an integrated genomics and functional screening strategy to identify miR-34a (1p36) and miR-34c (11q23) as tumor suppressors in neuroblastoma (Cole et al., Mol Cancer Res 2008). There was a perfect correlation of absent miR-34a expression in cell lines with a 1p36 aberration and phenotypic effect after mimetic add-back. BCL2 and MYCN were identified as miR-34a targets and likely mediators of the tumor suppressor phenotypic effect. In collaborative research projects, we also helped define CHD5 (Fujita et al., J Natl Cancer Inst 2008) and KIF1Bbeta (Schlisio et al., Genes Dev 2008) as neuroblastoma suppressor genes. Ongoing work is seeking to leverage our findings, and we have developed an orthotopic liver model of neuroblastoma to test the feasibility and efficacy of miR-34a mimetic replacement via lipid nanoparticle delivery vehicles.
Currently, we our continuing our efforts using genome scale technologies to discover the critical events necessary to develop high-risk neuroblastoma in order to develop better therapies. We were recently awarded an NCI TARGET grant (www.target.cancer.gov; Maris, PI) that is modeled after the Cancer Genome Atlas Project, but with a more focused goal of not only discovering somatic mutations, but those that are relevant in terms of identifying potential therapeutic targets. In the past year we have completed SNP and exon array-based array assessment of tumor DNA and RNA, respectively, for over 400 neuroblastomas, and have also generated genome-scale assessment of DNA methylation and miRNA expression in about 100 samples. This allowed us to identify 117 candidate genes for complete resequencing in 188 cases, data just generated in the last few weeks. Taken together, we have completed a thorough genomic characterization of high-risk neuroblastoma, and this coming year will focus on leveraging these data. One early and important “hit” was the discovery that gain-of-function mutations in the ALK tyrosine kinase domain are also somatically acquire, that occurred co-incident with the discovery of germline mutations (Mosse et al., Nature 2008). We have subsequently proven that these mutations are oncogenic drivers, and have shown that a dual c-MET/ALK inhibitor was potently cytotoxic against neuroblastoma models in vitro and in vivo in a mutation-specific manner (presented at ASCO this year, manuscript submitted). A phase 1 trial based on these data has recently opened, representing a remarkably facile translation of a basic science discovery to the clinic in a little over a year.
The NBL-TARGET consortium group is now planning for a major resequencing effort of 300 genomes from 100 high-risk neuroblastoma cases (germline, tumor DNA and tumor RNA for each case). In a unique collaboration with funding from the NCI and NHGRI, we have engaged the three major genome-sequencing centers to complete this project in 12-18 months. This unprecedented experiment is powered to discover all of the major driver mutations in this disease. Embedded in this project is functional validation of top hits with the goal of developing new therapeutic strategies based on these data.
Our laboratory is highly focused on impacting patient care, and thus has developed a significant amount of effort towards drug target identification and preclinical evaluation of pharmaceuticals both in vitro and in vivo. We are the neuroblastoma laboratory for the Pediatric Preclinical Testing Program (www.pptp.stjude.org), an NCI contract to prioritize drugs for development in pediatric cancer. This xenograft-based screening system has produced studied dozens of compounds, and has produced several hits in the past year, including AURKA, CENPE and an oncolytic virus (SVV), each of which are in or moving towards a clinical trial in children with refractory solid tumors. This program has been highly successful in fast tracking certain pharmacologics, but equally important, de-prioritizing other drugs that showed no evidence for preclinical activity (15+ publications, not listed here). The broader context of the PPTP is to determine, in an unbiased fashion, if anti-cancer drug activity in xenograft models is predictive of activity in the clinic. Importantly, the xenografts are all very highly characterized at a molecular level, allowing us to use xenograft genomics to help define molecular correlates of anti-tumor activity (Neale et al., Clin Cancer Res 2008). We are just now completing Phase 1 and 2 studies of many of the drugs identified in the PPTP, and our own work on the AURKA inhibitor (we are leading the trial and performing the correlative biology) should be particularly instructive (Maris, Cancer Cell 2009).
In collaborations with GlaxoSmithKline and Merck pharmaceuticals, we have matched our laboratory observations to their cancer therapeutics pipeline. This has lead to the discovery of CENPE being a major oncogenic driver of transgenic murine neuroblastomas, and functional evidence for this protein being relevant to human disease as well (manuscript submitted). A phase 1 trial is planned as soon as the adult recommended phase 2 dose is established in an ongoing trial. We have a new program exploring metabolic targets in neuroblastoma in collaboration with other AFCRI investigators, and these experiments will significantly ramp up this coming year. Preliminary data strongly implicate fatty acid synthase as druggable oncogenic driver of high-risk neuroblastoma. Finally, as mentioned above, we are working closely with Merck/Rosetta to develop a miR-34 mimetic as a therapeutic using lipid nanoparticle-based delivery of small RNAs.
Ultimately, we seek to exploit our basic and translational research discoveries in the clinic. To assure that the infrastructure is present to accomplish this goal, I developed a clinical trials team at CHOP focused on the disease, and we currently receive referral from across the country. I employ two advanced practice nurses, and clinical research associate and a social worker to manage our population of large and complex patients with refractory neuroblastoma, and have gathered the philanthropic support to establish an endowed travel fund to support under privileged children accessing our experimental therapies. In addition, for the past four years I have served as Chair of the Children's Oncology Group (COG) Neuroblastoma Disease Committee, and thus have responsibility for all Phase 1 – 3 neuroblastoma clinical trials in our 238 institution consortium. I am also a lead investigator and Scientific Review Committee member for the New Approaches to Neuroblastoma Therapy Phase 1 consortium.
We have played a leadership role in the development of targeted radiotherapy for neuroblastoma using 131I-metaiodobenzylguanidine. We completed the pivotal Phase 2 study (N=167) to establish the very high anti-tumor activity of this agent in highly refractory disease (Matthay et al., J Clin Oncol 2007), performed Phase 1 dose escalation trials of the agent with high dose chemotherapy (Matthay et al., J Clin Oncol 2006), and as a rapid double infusion (Matthay et al., J Clin Oncol 2009), and now are poised to open our first study using the agent in newly diagnosed patients. We have also played a pivotal role in the development of other novel therapeutics (Fox et al., Clin Cancer Res 2008; Fox et al., Clin Cancer Res 2006; Wagner et al., J Clin Oncol 2009), and I am the Chair or Vice-Chair of multiple open studies in the COG and NANT. We have developed a system to become much more facile in drug development for children with cancer as highlighted by the 18 month time-line from discovering the first ALK mutation in a hereditary neuroblastoma patient to a Phase 1/2 clinical trial in our national Phase 1 consortium.