Exploiting an Asp-Glu "Switch" in Glycogen Synthase Kinase 3 to Design Paralog Selective Inhibitors for Acute Myeloid Leukemia Treatment

Glycogen synthase kinase 3 (GSK3), a key regulatory kinase in the WNT pathway, remains a therapeutic target of interest in many diseases. While dual GSK3A/B inhibitors have entered clinical trials, none have successfully translated to clinical application. Mechanism-based toxicities, driven in part by the inhibition of both GSK3 paralogs and subsequent β-catenin stabilization, are a concern in the translation of this target class to the clinic. Knockdown of GSK3A or GSK3B individually does not increase β-catenin levels and offers a conceptual resolution to targeting GSK3: paralog-selective inhibition of GSK3A or GSK3B (Doble et al., Dev Cell, 2007). However, no such chemical tools exist. The design of selective ATP competitive inhibitors poses a drug discovery challenge due to the high homology (95% identity, 100% similarity) in their ATP binding pockets.

Taking advantage of an Asp133 ® Glu196 "switch" in their hinge binding motif, we present a rational design strategy towards the discovery of a chemically matched and paralog selective set of GSK3 inhibitors (GSK3A inhibitor BRD0705, GSK3B inhibitor BRD3731 and GSK3AB dual inhibitor BRD0320). Paralog kinase selectivity was validated using multiple biochemical and in cellulo approaches.

These selective inhibitors provided insights into the therapeutic targeting of GSK3 in acute myeloid leukemia (AML) where GSK3A has been identified as a therapeutic target using genetic approaches (Banerji et al., JCI, 2012). We first validated that our GSK3A selective compound (BRD0705) significantly inhibits kinase function and does not stabilize β-catenin in multiple AML cell lines (n=5) and primary patient samples (n=5), mitigating potential neoplastic concerns, particularly in an AML context. In contrast, GSK3B inhibition showed a context dependent effect on β-catenin stabilization. Indeed, BRD3731 induced β-catenin stabilization and nuclear translocation in a subset of the AML cell lines, supporting the safety concerns in GSK3B targeting in AML (Guezguez et al., Cancer Cell, 2016).

Selective inhibition of GSK3A by BRD0705 induced myeloid differentiation and impaired colony formation in AML cell lines (n=6) and primary AML samples (n=5), while no effect was observed on human CD34 hematopoietic cells. Moreover, BRD0705 paralog selectivity was confirmed using GSK3A CRISPR KO isogenic clones in U937 AML cell line to validate absence of phenotypic effects when GSK3A is no longer expressed. Furthermore, the transcriptional effects of BRD0705, evaluated using an RNAseq approach, were consistent with differentiation induction and impaired stemness. Importantly, we observed no increase in β-catenin related signatures. In support of this gene expression data, a limited dilution experiment performed in a syngeneic mouse model driven by the oncogenic translocation MLL-AF9 revealed a 3.79 fold decrease in leukemia initiating cells (LIC) after BRD0705 pre-treatment and re-injection into secondary recipient mice. No significant difference in LIC numbers was observed after BRD3731 or BRD0320 pre-treatment.

BRD0705 possesses favorable pharmacokinetic properties and displays efficacy with prolonged survival in several mouse models of AML (HL-60 and MV4-11 orthotopic xenograft models and an MLL-AF9 syngeneic mouse model). Interestingly, in the MV4-11 model, we extended testing to BRD3731 (GSK3B) and BRD0320 (dual GSK3A and GSK3B). Neither induced a significant effect on mice survival, confirming the specific role of GSK3A in some AML subsets.

These studies validate that paralog selective inhibition of GSK3A using BRD0705 is feasible and offers a promising therapeutic approach in AML.