Characterisation and differentiation of kinase binding pockets in PKA and PIM1 by small molecule fragments using protein crystallography

The primary aim of this thesis was the crystallographic analysis of PKA and Pim1 crystals in complex with small fragment-like molecules to identify potential fragment binders in the protein kinase binding pockets. In the first subproject of this thesis, the similarities and differences between both kinases are highlighted. The fragments used for this study were extracted as a part of an in-house library comprising 361 entries. These fragments were analyzed by TSA as a pre-screening method to evaluate their potential binding to both kinases. This pre-screening resulted in 31 putative binding candidates for the PKA, whereas 52 were identified for Pim1. The subsequent crystallographic analysis of these candidates revealed that 15 fragments had bound to PKA and 13 to Pim1. Referred to the number of collected datasets of each protein (31 PKA / 15 Pim1) this led to a hit rate of 48% for PKA and of 87% for Pim1. Due to the efficient pre-screening selection and the difficult crystallisation of Pim1 complexes, only three of the 361 fragments were commonly studied by protein crystallography with both kinases. Fragment F222 was the only one, which bound to PKA and Pim1. The binding mode of this fragment is similar in both proteins. Nevertheless, three striking differences were detected in the ATP-binding pocket, which influences protein-ligand interactions. Pim1 contains Pro123 in the hinge region, which reduces the number of H-bond donors in the ATP-binding pocket by one. PKA contains the hydrophilic side chain of Thr183 in the binding pocket, whereas Pim1 exposes the hydrophobic Ile185 at the same position. Additionally, PKA exhibits a long C-terminal loop, which is located close to the ATP-binding pocket. The side chain of Phe327 penetrates into the binding pocket of PKA and thus reduces the available space of the pocket and narrows the entrance. These differences in the binding pocket also affect the deviating binding modes of F222 in both kinases. As a pre-screening method, TSA had shown that enrichment of crystallographic hits could be achieved with both kinases, as an additional X-ray analysis of randomly picked test candidates outside the TSA hit list revealed a lower hit-rate of 20%. The overall results of the TSA analyses also provide some ideas about required selectivity features of the fragments between PKA and Pim1. In comparison with already published PKA and Pim1 binders, we could detect either known but also new structural scaffolds for the interactions with the protein. In the second subproject of this thesis, complexes of PKA and fragments were analyzed by direct exposure of soaked crystal to X-ray analysis. A full library with 96 entries was tested. Of these 43 were soaked into a PKI containing crystal form and 53 into a PKI free one. There were 30 crystallographic hits found out of 96 collected data sets. This results in a hit-rate of 31%. The obtained hits were split into 15 PKI-containing and 15 PKI-free complex structures. There were 24 fragments detected to bind in the ATP-binding pocket. Since several fragments bound multiple times in one structure, 15 fragments were detected in a remote position leading overall to 39 different binding poses. The binders of the binding pocket were further classified in terms of their interaction motifs. They were sorted into groups such as direct hinge binders (12), water-mediated hinge binder (2), direct DFG-loop binders (16) and water-mediated DFG-loop binders (4). Remote binding fragments were analyzed in terms of the buried surface area with respect to the proximate PKA molecules to obtain some figures-of-merits, whether binding only occurs in the crystallographic packing and would turn out as ‘artefacts’ during an analysis by a method focussing on conditions in the solute phase. Five of these 15 fragments revealed a relatively high percentage of buried surface area with a neighbouring PKA molecule. Due to the high amount of ATP-pocket binders, a topological map of the pocket was drafted, giving some ideas where the detected structural motifs of the fragments assembled in the different sub-pockets. In comparison with another fragment-based study of the Saxty group with PKA/PKB, three fragments of the current study were conspicuous, which showed similarities in structure and binding mode to a nanomolar inhibitor found in their study. These three fragments indicate how a merged ligand could be formed. In the course of this sub-project, a new crystal form of PKA without the bound peptide inhibitor PKI was obtained (s. above). Growing these PKA crystals required a longer time span, but showed better diffraction properties. In relation to PKI-bound structures, mean improvement in resolution of 0.25 Å could be achieved. Additionally, these new crystals allowed the binding of fragments to the peptide binding site, which was successfully documented by the Arg-mimicking fragments J72 and J77. The third sub-project of this thesis is part of a cooperation project with AG Kolb and AG Diederich. Here, the first step to identify fragment molecules that interact with the amino acids Glu121 and Lys67 of the Pim1 kinase was accomplished. Suitable fragments will be combined with another component to achieve additional interactions with Asp128 and Glu171. We used the database tool SCUBIDOO to identify virtual screening hits. The search for suitable fragments resulted in three compounds, which were detected by docking, similarity search and an FCFP4 fingerprint analysis based on the “Chembridge building block library” (4012413, 4012414 and 5175110). The latter fragment was already published in 2012 by Good et al. in complex with Pim1. We could simultaneously confirm the retrieval and subsequently the binding of this fragment by our procedure and the reliability of our strategy. The compound 4012413 was crystallised with Pim1. The electron density resulting from analysis of the dataset revealed that the ligand had presumably bound in the ATP-binding pocket. However, the observed difference electron density was too poorly defined to assign the ligand precisely into the structure. Testing all three fragments by TSA, suggested binding of all three fragments. Further crystallisation of fragment 4012413 and 4012414 is still pending. The forth subproject of this thesis was focused on compounds extracted from the ZINC databank. They were identified as potential Pim1 binders by docking experiments. A sample of six promising ligands was identified and selected showing a promising binding in the docking prediction. An analysis of these ligands by TSA revealed that all compounds, excluding one, had a stabilizing effect on Pim1, which indicates their binding. Three of these ligands were analyzed crystallographically. Two ligands showed a binding to the protein. The compound ZINC08880252 could not be detected as a binder, and it was the only one lacking any stabilizing effect in TSA. Although the two ligands ZINC72154357 and ZINC09314085 showed a binding, their binding modes deviate from the docking prediction. The ligand ZINC09314085 is only partly represented by the difference electron density. An explanation for this incompletely defined binding is still under investigation. Hydrolysis of its central ester group could be a possible reason or a pronounced partial disorder, or residual mobility is in operation. To investigate this, an exchange of the ester by an amide bond has been initiated, and a mass spectrometric analysis is currently performed