Acute Myeloid Leukemia (AML) affects both adults and children and is molecularly heterogeneous. Based on high prevalence of mutations in a number of genes including DNMT3A, IDH1, and IDH, targeted therapies are under development for treatment of AML with these mutations. However, these targeted therapies developed based on mutations in older adults can not be used in children and young adults because such mutations, although highly prevalent in older adults are rare or not observed in younger patients. This brings into question the utility of “trickle down therapeutics”, where therapies discovered/developed in older adults can simply be used in younger patients. This stems from a lack of understanding of the biology of AML in children and younger adults and comprehensive characterization of the disease. Nevertheless, AML is the leading cause of childhood leukemia mortality and the most common form of childhood cancer. Therefore, there is an urgent need to systematically characterize pediatric AML across age groups so that this heterogeneous disease can be molecularly stratified for the development of personalized treatment regimes.
Dr. Bolouri (Human Biology Division) and Dr. Meshinchi (Clinical Research Division), together with their collaborators from several institutions in the US, Canada, and Europe took on this task of comprehensively characterizing the mutational, transcriptional and epigenetic landscapes of samples from more than 1000 children with AML. The results of this Children’s Oncology Group-National Cancer Institute (COG-NCI) funded project, known as the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) AML initiative was recently published in Nature Medicine, and explains how AML is molecularly distinct in infants, children, adolescents and young adults from that previously published in older adults.
The age of TARGET AML participants ranges from 8 days to 39 years. Participants were stratified into the following age groups: Infants (<3 years old), children (3-14 years old), and adolescents/ young adults (AYA; 15-39 years old). The analysis of somatic DNA sequence alterations by whole geome and targeted capture sequencing revealed that younger AML patients harbored a very different mutational landscape compared to adults, despite a low mutation rate in both pediatric and adult AML. Specifically, genes such as NRAS, KRAS, CBL, GATA2, SETD2, PTPN11 and WT1 were mutated more frequently in younger patients, whereas DNMT3A, IDH1, RUNX1, TP53, and NPM1 were mutated more often in older patients. Besides a pattern of increasing or decreasing mutation rates across age groups evident in certain genes, this pediatric-adult difference also includes genes that have distinct mutation sites. These genes include WT1, CBL, GATA2, MYC and FLT3, and have several frequently recurrent alterations in pediatric AML that are distinct from those identified in adult AML. Interestingly, WT1 mutations tend to be of clonal origin in pediatric patients. Many somatic structural DNA changes, particularly hybrid genes formed from two previously independent genes, also known as gene fusions, were primarily observed in pediatric AML. Similarly, aberrations affecting epigenetic regulators and their frequency also differ between children and adults. The authors collected DNA methylation and mRNA expression data in a cohort of pediatric and adult AML patients and identified dozens of genes with promoter hypermethylation resulting in transcriptional silencing. Intriguingly, two DNA methylation signatures were able to significantly predict poorer event-free survival in both the pediatric and adult participants.
The authors also examined expression patterns of microRNA (miRNA) in pediatric AML, and found four distinct groups that correlated with specific genomic alterations such as NPM1 mutations. Differential expression of miRNAs in pediatric and adult AML, including miR-330, previously shown to have oncogenic potential in AML were observed.
This study involved a large number of subjects and definitely demonstrated that many genes characteristically mutated in AML are altered at widely variable frequencies across age groups in pediatric AML. Additionally, the authors also reveal that DNA methylation and miRNA expression profiles complement DNA alterations, and can be used to stratify pediatric AML patients. Currently, patients are stratified for treatment by considering the effect of each somatic alteration in isolation. Given the results of this study, such clinical practices would be inadequate. Dr. Bolouri stated “We have identified sub-groups of patients with very poor prognosis, who have distinct, well-defined genomic and epigenomic profiles. We are trying to find out what makes these particular cases of AML so aggressive.” Notably, Dr. Bolouri also revealed that they have some interesting preliminary results suggesting that different subsets of AML patients may have distinct metabolic and immune profiles, with direct therapeutic implications.
This TARGET AML data set will serve as a foundation for the development of personalized treatment regimes and pediatric-specific classification schemas, forming a truly valuable resource. Despite already having a large cohort of patients, Dr. Bolouri alluded to the need for more patient samples “Because each AML sub-type is rare, we need more patient samples. The Meshinchi lab has already started characterizing more patient samples, and by the end of 2018, we expect to have genome-wide mRNA, microRNA, and DNA-methylation data for at least 2000 additional patients.”
With the substantial size of these datasets, one would wonder about the challenges faced by computational biologists. We asked Dr. Bolouri, who coordinated the data analysis of this massive study, to share some of the challenges. “The importance of rigorous data curation, tracking, and version control. With a large multi-year project, often the data gets re-processed or re-analyzed using different pipelines and tools. Keeping track of all the different versions of the results and exactly how they were generated is really critical” said Dr. Bolouri. He also added that “There are often many different methods, tools, or pipelines for any given analysis task. I found it very useful to compare the results from multiple approaches. You get a sense of which findings are robust, and which depends on specific insights or assumptions.”
Bolouri H, Farrar JE, Triche Jr T, Ries RE, Lim EL, Alonzo TA, Ma Y, Moore R, Mungall AJ, Marra MA, Zhang J, Ma X, Liu Y, Liu Y, Guidry Auvil JM, Davidsen TM, Gesuwan P, Hermida LC, Salhia B, Capone S, Ramsingh G, Zwaan CM, Noort S, Piccolo SR, Kolb EA, Gamis AS, Smith MA, Gerhard DS and Meshinchi S. 2018. The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions. Nature Medicine. 24:103-112. doi:10.1038/nm.4439
Funding was provided by the Children’s Oncology Group, National Institutes of Health, St. Baldrick’s Foundation, Alex’s Lemonade Stand, TARGET Pediatric AML, the Arkansas Biosciences Institute and the Jane Anne Nohl Hematology Research Fund.