High-throughput sequencing efforts have identified numerous recurrently mutated genes in human acute myeloid leukemia (AML) yet 5-year overall survival has not changed for 30 years. While a proportion of these mutations are gain of function mutations, the majority are assumed to be loss of function (hypomorphic) mutations that are not readily druggable. One approach to overcome this problem is to target ?synthetic lethal? partners of a given mutation, genes that are compatible with cell survival when inhibited alone but when combined with a loss of function mutation, the synthesis of the two inactivations results in cell death. We recently described a synthetic lethal interaction between the IDH1 mutation and Bcl-2 in primary AML cells which is now in clinical trial using ABT-199 as monotherapy or in combination with chemotherapy for AML. To date, synthetic lethal partners have been identified through large-scale functional screens predominantly in cell lines using shRNA, siRNA or other gene knockdown approaches which are laborious and expensive and do not necessarily phenocopy the in vivo human microenvironment of leukemia evolution. Here, we propose a novel approach to use high-throughput pan-cancer data to identify synthetic lethal partners of recurrent epigenetic mutations in AML (IDH1, DNMT3A, WT1 and cohesin). We propose to identify candidate synthetic lethal partners for mutations in AML using Boolean implication mining and experimental validation with primary AML samples. First, we will analyze genomic and gene expression data from the The Cancer Genome Atlas across numerous human cancers and identify rules based on mutual exclusion. We hypothesize that, across multiple cancers, synthetic lethal partners of a mutation will be amplified more frequently or deleted less frequently, with concordant changes in expression, in primary tumor samples harboring the mutation, and that these changes can be detected by Boolean implication mining. We will focus on four recurrent mutations in AML associated with epigenetic regulation in AML (i) IDH1, (DNA hyper-methylation), (ii) DNMT3a (DNA hypo-methylation) (iii) cohesin (chromatin modifier) (iv) WT-1 (DNA hyper-methylation). We will compare our computational hits with genome-wide high complexity shRNA screens using inducible cell-lines that have mutation-specific leukemogenic phenotypes, allowing us to refine and improve the positive predictive value of our method and determine the most important input variables for accurately predicting synthetic lethal relationships from primary tissue. Importantly, our method has identified a druggable enzyme, ACACA (acetyl CoA carboxylase) which we predicted and have now validated to be synthetic lethal with the IDH1 R132 mutation using both gene targeting and small molecules (TOFA) in primary AML cells and solid tumors.
In the last 10 years we have learnt which mutations and how many are present inside the DNA of acute myeloid leukemia cells but we are yet to see any improvements in the clinical outcome of this disease. We have developed a new method to rapidly match the right therapy for the right mutation using large data sets from many other cancers beyond leukemia. So far our method has found a number of novel targets for treatment in IDH1- mutated leukemia and so we propose to extend this for many other mutations and test our results in pre-clinical models of human leukemia.
| Tvorogov, Denis; Thomas, Daniel; Liau, Nicholas P D et al. (2018) Accumulation of JAK activation loop phosphorylation is linked to type I JAK inhibitor withdrawal syndrome in myelofibrosis. Sci Adv 4:eaat3834 |
| Sinha, Subarna; Thomas, Daniel; Chan, Steven et al. (2017) Systematic discovery of mutation-specific synthetic lethals by mining pan-cancer human primary tumor data. Nat Commun 8:15580 |
| Thomas, Daniel; Majeti, Ravindra (2017) Biology and relevance of human acute myeloid leukemia stem cells. Blood 129:1577-1585 |
| Thomas, Daniel; Majeti, Ravindra (2017) Optimizing Next-Generation AML Therapy: Activity of Mutant IDH2 Inhibitor AG-221 in Preclinical Models. Cancer Discov 7:459-461 |
