Virtually all recurrent molecular alterations in acute myeloid leukemia (AML) functionally converge to cause signal transduction pathway dysregulation that drives cellular proliferation and survival. The mammalian target of rapamycin complex 1 (mTORC1) is a rapamycin-sensitive signaling node defined by the interaction between mTOR and raptor. Constitutive mTORC1 activity is nearly universal in AML. However, pharmacologic inhibition with rapamycin or second-generation mTOR kinase inhibitors has shown limited anti-leukemic activity in both preclinical models as well as patients, suggesting that addiction to this oncogene is not a recurrent event in AML. Here we report that sustained mTORC1 activity is nonetheless essential for the cytotoxicity induced by pharmacologic activation of AMP-activated protein kinase (AMPK) in AML.

Our studies employed a novel AMPK activator called GSK621. Using CRISPR and shRNA-mediated silencing of the AMPKa1 catalytic subunit, we showed that AMPK activity was necessary for the anti-leukemic response induced by this agent. GSK621-induced AMPK activation precipitated autophagy, as demonstrated by western blotting, immunofluorescence, flow cytometry and electron microscopy. Blocking autophagy via shRNA-mediated knockdown of ATG5 and ATG7 protected AML cells from cytotoxicity resulting from treatment with GSK621, suggesting that autophagy promotes cell death in the context of active AMPK. GSK621 cytotoxicity was consistently observed across twenty different AML cell lines, primary AML patient samples and AML xenografts in vivo. GSK621-induced AMPK activation also impaired the self-renewal capacity of MLL-ENL- and FLT3-ITD-induced murine leukemias as measured by serial methylcellulose replating assays.

Strikingly, GSK621 did not induce cytotoxicity in normal CD34+ hematopoietic progenitor cells. We hypothesized that the differential sensitivity to GSK621 could be due to the difference in amplitude of mTORC1 activation in AML and normal CD34+ cells. In contrast to most reported cellular models in which AMPK inhibits mTORC1 both directly (through raptor phosphorylation) and indirectly (through TSC2 phosphorylation), sustained mTORC1 activity was seen following GSK621-induced AMPK activation in AML. Inhibition of mTORC1 either pharmacologically (using rapamycin) or genetically (using shRNAs targeting raptor and mTOR) abrogated AMPK-induced cytotoxicity in AML cells, including primary AML patient samples. This protective effect was mediated by mTORC1-dependent modulation of the ATF4/CHOP stress response pathway. The ultimate functional consequence was that, rather than diminishing GSK621-induced cytotoxicity, persistent mTORC1 activity was in fact synthetically lethal with AMPK activity in AML cells. This synthetic lethality could be recapitulated in normal CD34+ progenitors by constitutive activation of mTORC1 using a lentivirally-transduced myrAKT construct. It could also be enhanced in AML cells by mTORC1 overactivation induced by CRISPR-mediated deletion of TSC2.

Taken together, these data show that the magnitude of mTORC1 activity determines the degree of cytotoxicity triggered by AMPK activation. This finding may have important implications for AMPK and mTORC1 signaling pathways in cancer biology more broadly. Context-dependent permissiveness towards mTORC1 activation may amplify the response to cytotoxic stress, such as that resulting from AMPK activation by GSK621. Our results therefore support AMPK activation as a promising therapeutic strategy in AML and other mTORC1-active malignancies which warrants further investigations in clinical trials.


Brusq:GSK: Employment. Nicodeme:GSK: Employment.

Author notes


Asterisk with author names denotes non-ASH members.