Identification of a First in Class GSK3-Alpha Selective Inhibitor As a New Differentiation Therapy for AML

Isoform-selective enzyme targeting poses a potential clinical challenge when there is significant homology in the protein activity domains. Glycogen synthase kinase 3-alpha (GSK3α) has been identified previously as a therapeutic target to promote differentiation in acute myeloid leukemia (AML) (Banerji et al, JCI, 2012). The GSK3 family consists of two highly homologous, but functionally non-redundant, isoforms (α and β). These proteins are serine-threonine kinases implicated in a diversity of cellular processes. Concurrent inhibition of both isoforms is known to induce β-catenin stabilization and can therefore increase self-renewal of leukemia initiating cells. Isoform-selective GSK3 suppression does not increase β-catenin levels (Doble et al, Dev Cell, 2007) and thus may offer a new therapeutic approach. GSK3α selective inhibitors, with cell-based activity, have not been reported. Our collaborative study aimed to identify first-in-class GSK3α selective inhibitors and evaluate their activity in AML, a disease in great need for new therapeutic approaches.

A series of potent and selective inhibitors for GSK3α/β were discovered through a screening campaign supported by the NIH's Molecular Libraries Probe Centers Network (MLPCN) program. While these compounds are equipotent for GSK3α/β, they demonstrate remarkable selectivity versus the larger kinome (> 100x selective vs 311 kinases). Analysis of the high resolution x-ray crystal structure obtained with BRD9421, one of the lead compounds, and GSK3β revealed a tridentate hydrogen bond binding mode within the ATP catalytic domain. One of these interactions, is to aspartic 133 in GSK3β. This same residue in GSK3α is a glutamic acid and represents one of the two amino acid differences within the ATP binding domain between these two enzymes. GSK3α and β share 98% homology in their catalytic domains, but the exploitation of this hydrogen bond interaction through rational drug design offered an excellent opportunity to produce selective inhibitors for GSK3α and GSK3β. BRD0705 was thus identified as an 8-fold GSK3α isoform-selective inhibitor over GSK3β (IC50 GSK3α: 0.045 vs. GSK3β: 0.350 μM).

We first validated BRD0705 isoform-selectivity in a panel of AML cell lines and primary patients (n=3 cell lines and n= 2 patient samples). Treatment with BRD0705 resulted in a dose- and time-dependent decrease in tyr279 GSK3α phosphorylation without any effect on tyr216 GSK3β phosphorylation. No increase in β-catenin protein levels by western blotting and no nuclear localization by immunofluorescence were observed. GSK3α inhibition with BRD0705 led to morphological and surface marker changes (CD11b, CD11c, CD14, CD117) consistent with AML differentiation in a panel of AML cell lines (n=4) and primary AML samples (n=5). BRD0705 also induced a G2/M arrest and impaired methylcellulose colony formation in AML cell lines and primary patient samples. BRD0705 possesses favorable pharmacokinetic properties and was well tolerated without any observed adverse events in C57BL/6 mice with repeated dosing. In vivo efficacy studies in AML models are ongoing.

In conclusion, we identified a first-in-class GSK3α isoform-selective inhibitor and conducted preclinical studies validating BRD0705 as a promising new differentiation therapeutic candidate in AML. Small-molecule inhibition of GSK3α induced myeloid differentiation, cell cycle arrest and impaired colony formation without β-catenin stabilization in AML.