In recent years, targeting cancer stem cells (CSCs) has become the focus of intense interest in diverse malignancies, including those of hematopoietic origin, and particularly AML. CSCs are capable of continuous regeneration and, as a consequence, may serve as an unlimited reservoir to restore the bulk population of cells following ablation (e.g., by chemotherapy). Significantly, AML stem cells appear to be relatively resistant to conventional cytotoxic agents, such as ara-C or daunorubicin, which very effectively kill bulk populations of AML blasts, raising the possibility that failure to eradicate AML stem cells (AML SCs) may be responsible for or contribute to treatment failure in this disease. On the other hand, evidence has emerged that certain targeted agents, particularly inhibitors of the NF-κB pathway, may be particularly effective in eliminating AML SCs.1 One such agent, the sesquiterpene lactone parthenolide, has been shown to induce AML SC death in association with NF-κB inactivation and induction of oxidative injury.2 Such findings have prompted efforts to identify other compounds with similar characteristics.
In a recent study, Hassane, et al. describe a novel genetic approach specifically designed to achieve this aim. Using the multi-institutional Gene Expression Omnibus (GEO) as a platform, they hypothesized that compounds capable of eradicating AML SCs would exhibit a gene profile array similar to that of parthenolide. In silico screening of this public database by two separate search procedures yielded two compounds, celastrol and 4-hydroxy-2-nonenal (HNE), whose signatures mimicked that of parthenolide. Interestingly, both of these terpenoid compounds shared with parthenolide the ability to ablate the bulk, progenitor, and SC populations of leukemic cells, disrupt NF-κB signaling, and induce oxidative stress. The authors concluded that mining of large, public, gene array databases may provide an extremely valuable resource for drug discovery by helping to classify both new and old drugs according to the genetic perturbations that they produce. More specifically, in the case of AML, such computational tools may help in identifying the relatively small subset of agents likely to be effective against the AML SC, rather than the bulk population of blasts.
In Brief
While the concept of employing sophisticated computational methods in conjunction with large genetic databases to discover new agents is clearly an exciting one, the ultimate success of this strategy in the context of AML will depend upon multiple factors. First, while it seems intuitive that agents capable of ablating both AML SC and bulk populations should prove superior to agents that only ablate bulk populations, this hypothesis has not yet been formally tested. In this regard, the entry of the parthenolide analog LC-1 into the clinical arena, including studies involving AML, should begin to address this question. The possibility also exists that targeting specific genetic signatures may primarily yield agents mimicking the pharmacodynamic properties of the index compound. Nevertheless, the novel drug discovery approach described here clearly has tremendous potential, and its validation over the years to come is awaited with great interest.
References
Competing Interests
Dr. Grant indicated no relevant conflicts of interest.
