In this issue of Blood, Hassane and colleagues describe the discovery of new therapeutic agents effective against leukemic initiating cells through an in silico approach, demonstrating the power of collective knowledge deposited in publicly available gene-expression databases.
Putative leukemic stem cells (LSCs) are thought to play an important role in the pathogenesis of acute myelogenous leukemia (AML), and in disease initiation and relapse.1 This suggests that it may be important to identify therapeutic strategies that specifically target LSCs while maintaining normal hematopoietic stem cells (HSCs). Previously, Guzman et al reported the activity of a natural compound, parthenolide (PTL), as a potent in vitro agent that selectively affects leukemic initiating cells but not normal progenitors.2
In their most recent study, Hassane and colleagues use the gene-expression signature derived from the treatment of primary AML with PTL to search for similar signatures in publicly available gene-expression profiles and thereby identify candidate compounds that may have a similar ability to eradicate LSCs. Using this strategy, they identify 2 new agents, celastrol and 4-hydroxy-nonenal (HNE), that can effectively target cell death of bulk AML cells, as well as functional subsets of AML progenitors and cells detected in the surrogate in vivo assay for human LSCs termed severe combined immunodeficiency leukemia-initiating cells (SL-ICs). These agents do not affect hematopoietic progenitors harvested from healthy individuals. The identified chemicals are a diverse set of compounds that mimic the PTL gene-expression signature and share mechanistic targets with PTL, specifically the capacity to inhibit the NF-κB survival signal and the induction of oxidative stress, also suggesting that these pathways could be commonly targeted in the design of additional anti-LSC agents.
Most interestingly, these compounds have been identified using a gene-expression signature derived from primitive AML CD34+ cells to search through a broad spectrum of expression data obtained from several treatments performed on different nonhematopoietic cell types. This raises 2 important points. First, there may be no need to look at specific cancer stem cell (CSC) expression profiles to identify agents that specifically target CSCs if the initial search criteria are based on CSC drug-gene response, and second, different types of cancer might share common tumor-specific mechanisms of growth and survival that can be exploited to design new, effective anticancer therapies.
Of course, other important issues remain: although these compounds effectively eradicate AML at the bulk, progenitor, and SL-IC level in vitro, there is no evidence of an effect in vivo, which is a likely clinical approach to avoid complex ex vivo culture treatment. In addition, these data need to be interpreted with caution. Celastrol and HNE clearly inhibit the leukemic-initiating cell function, but the effectiveness of these compounds on established leukemia versus leukemia initiation detected in surrogate in vivo SL-IC models still needs to be determined. Moreover, the relevance of the SL-IC assay in the detection of cells that have clinical significance for AML disease progression in patients still needs to be clarified, especially because there may be varying subclasses of SL-ICs that are detectable in different murine recipients and delivery methods in human-mouse xenograft models. Finally, the pharmacological use of the agents identified in the present study is difficult, due to the high concentration necessary to achieve SL-IC–specific eradication; analogs with better pharmacological properties or combinations that may lower the effective dose per agent need to be explored.
Altogether, Hassane and colleagues make an important step forward in the discovery of new compounds that may effectively target putative LSCs while sparing normal HSCs. Perhaps more importantly, they provide a powerful experimental strategy for discovering new drugs and targets by using publicly available data. These findings underscore the importance of open-access data and the need for annotation standards and consistent microarray normalization to enhance the utility of the underestimated knowledge contained in databases like Gene Expression Omnibus for LSCs, but also for other solid tumors that seemingly contain CSCs.3
Conflict-of-interest disclosure: The authors declare no competing financial interests. ■