Specific inhibition of CK2α from an anchor outside the active site

Abstract

The development of selective inhibitors of protein kinases is challenging because of the significant conservation of the ATP binding site. Here, we describe a new mechanism by which the protein kinase CK2α can be selectively inhibited using features outside the ATP site. We have identified a new binding site for small molecules on CK2α adjacent to the ATP site and behind the αD loop, termed the αD pocket. An elaborated fragment anchored in this site has been linked with a low affinity fragment binding in the ATP site, creating a novel and selective inhibitor (CAM4066) that binds CK2α with a K_d of 320 nM and shows significantly improved selectivity compared to other CK2α inhibitors. CAM4066 shows target engagement in several cell lines and similar potency to clinical trial candidate CX4945. Our data demonstrate that targeting a poorly conserved, cryptic pocket allows inhibition of CK2α via a novel mechanism, enabling the development of a new generation of selective CK2α inhibitors.

Fig. 1. (a) The position of the seven molecules of 3,4-dichlorophenethylamine (1, green carbon and chlorine atoms) bound to CK2α (PDB : ; 5CLP), with the fragment bound in the αD pocket highlighted by a red outline. (b) A cross section of the αD pocket with 1 bound. The size of the αD pocket in the partially open CK2α structure, with ethylene glycol bound in the mouth of the pocket (PDB : ; 3WAR), is represented by the red dashed line, whereas the closed pocket (PDB : ; 3FWQ) is represented by the black dashed line. (c) A comparison of the binding mode of 1 (dark green carbons and lighter green chlorines) and 4 (cyan carbons and green chlorines, shown in both binding poses observed in the structure) with hydrogen bonds between the amine of 4 and backbone carbonyls and water molecules indicated with dashed lines, and side chains lining the αD site as sticks. (d) The conformations of the αD loop, Phe121 and Tyr 125 as induced by the binding of fragments shown for complexes with 1 (green, PDB : ; 5CLP) and 3 (magenta, PDB : ; 5CS6), for partially open apo form (pink, PDB : ; 3WAR) and for the closed apo enzyme (blue, PDB : ; 3FWQ). The largest rearrangement of the αD loop is highlighted with the dashed line indicating the shift of Tyr125 between the closed form and protein bound to 3 (shown as sticks).

Fig. 2. (a) Cross section view of the binding mode of CAM4066 (yellow) and CX4945 (dark blue) to the αD pocket and the ATP site of CK2α (PDB : 5CU4). (b) The hydrogen bonding network of CAM4066 with CK2α and three waters, is also shown. (c) A more detailed comparison of CAM4066 (yellow) and CX4945 (dark blue, PDB : ; 3NGA) binding in the ATP site. CAM4066 replicates the salt bridge of CX4945, however CX4945 also interacts with the hinge region which would make it unsuitable for linking to the αD pocket fragments due to the movement of Met163 and the hinge region. (d) A comparison between the binding of 6 and CAM4066 in the ATP site shows a tilt of 10° in order to link to the αD pocket. This indicates that the linker is not optimal. (e) A comparison between the binding of 4 (cyan) and CAM4066 (yellow) in the αD pocket shows that CAM4066 binds slightly higher in the pocket and only displays one binding mode, rather than the two modes displayed by 4. (f) The structure of the linked compound CAM4066. The αD fragment is highlighted in blue and the ATP site fragment is highlighted in purple. (g) The isotherm of CAM4066 binding to CK2α (h) the inhibition of kinase activity by CAM4066, 4 and 6. IC₅₀ is shown for CAM4066 and error bars show the standard error from at least three replicates.

Fig. 3. (a) The results from a kinase selectivity screen where CAM4066 was screened against 52 kinases at 2 μM. Data for kinases that are inhibited by other Ck2α inhibitors (CX4945: red triangles, D11: gray triangles, TF: black triangles, CX5279 blue triangles and 7h white triangles) are highlighted with red bars. The error bars indicate the standard deviation from two replicates of the assay. (b) An alignment of the ATP site and the αD loop of CK2α and the kinases CX4945 inhibits with nM potencies. Alignment is coloured by conservation and the numbering on the top refers to human CK2α. Yellow and brown arrowheads under the alignment indicate residues in the CAM4066 and ATP binding sites, respectively. The green arrowheads highlight the two conserved residues in αD helix. (c) Superpositioning of Clk2 (purple, PDB : ; 3NR9) and CAM4066-bound CK2α (white) with residues in the αD pocket shown as sticks and labelled for both proteins. In the case of Clk2, the αD helix has been removed to allow for the view of the pocket. (d) View of the active sites of Clk2 (purple) and CX4945-bound CK2α (white, PDB : ; 3NGA). (e) Orientation of αD helix in unliganded CK2α and in Clk2 (purple).

Fig. 4. (a) Dose response curve for the inhibition of the growth of HCT116 cells by pro-CAM4066 and CX4945. CX4945 has a GI₅₀ of 4.8 μM and pro-CAM4066 has a GI₅₀ of 8.8 μM (graph for CX4945 shows the mean ± SEM of four independent experiments with each compound concentration in triplicate, and the graph for pro-CAM4066 shows the mean ± SEM of three experiments with each concentration in triplicate). (b) Western blot analysis showing the specific CK2 phosphorylation targets of AKT1 phosphoserine 129 and Cdc37 phosphoserine 13. HCT116 cells were treated with 2 × GI₅₀ of CX4945 (10 μM) or 8 (20 μM) for 24, 48 or 72 hours. For pro-CAM4066 data is shown for two different experiments performed with different batches of inhibitor. (c) and (d) Show results of quantification of the Western blot shown in (b) for AKT1 phosphoserine 129 and Cdc37 phosphoserine 13, respectively. Data are normalised to actin loading control and to DMSO only sample.