Specificity rendering 'hot-spots' for aurora kinase inhibitor design: the role of non-covalent interactions and conformational transitions

Abstract

The present study examines the conformational transitions occurring among the major structural motifs of Aurora kinase (AK) concomitant with the DFG-flip and deciphers the role of non-covalent interactions in rendering specificity. Multiple sequence alignment, docking and structural analysis of a repertoire of 56 crystal structures of AK from Protein Data Bank (PDB) has been carried out. The crystal structures were systematically categorized based on the conformational disposition of the DFG-loop [in (DI) 42, out (DO) 5 and out-up (DOU) 9], G-loop [extended (GE) 53 and folded (GF) 3] and αC-helix [in (CI) 42 and out (CO) 14]. The overlapping subsets on categorization show the inter-dependency among structural motifs. Therefore, the four distinct possibilities a) 2W1C (DI, CI, GE) b) 3E5A (DI, CI, GF) c) 3DJ6 (DI, CO, GF) d) 3UNZ (DOU, CO, GF) along with their co-crystals and apo-forms were subjected to molecular dynamics simulations of 40 ns each to evaluate the variations of individual residues and their impact on forming interactions. The non-covalent interactions formed by the 157 AK co-crystals with different regions of the binding site were initially studied with the docked complexes and structure interaction fingerprints. The frequency of the most prominent interactions was gauged in the AK inhibitors from PDB and the four representative conformations during 40 ns. Based on this study, seven major non-covalent interactions and their complementary sites in AK capable of rendering specificity have been prioritized for the design of different classes of inhibitors.

Figure 1. MSA of the kinase domains of AK and other sequentially similar kinases obtained from kinbase.

A pairwise sequence alignment of the AK sequence (AURKA_HUMAN) against the annotated kinome present in Kinbase v1.1 was done using blast-p (Table S3 in File S2). The kinase domains of the sequentially similar sequences were retrieved from Kinbase v1.1 and a MSA was constructed with ClustalW 2.1. Jalview 2.8.0 has been used to view the alignment and the colour scheme is as per ClustalX. The red dots represent kinases which have been crystallized with different AK inhibitors (Table 2). The labelled residues indicate the possible sites for target specific inhibitor binding in AK based on conservation.

Figure 2. Conformational variations in the major structural motifs of AK concomitant with DFG-flip.

In the figure, the two conformations of AK, active DFG-in (PDB: 3E5A, white) and the inactive DFG-out (up) conformation (PDB: 3UNZ, yellow) are superimposed. The major structural motifs of DFG-in are multi-coloured while that of DFG-out (up) are depicted in yellow. The figure displays the key components of the AK kinase active site: G-loop, αC-helix, gatekeeper (GK) hinge, DFG-loop, A-loop; and their critical residues: salt bridge formers Lys162, Glu181 (αC-helix) and Phe275 (DFG-loop). The arrows depict the differences in the two conformations resulting due to the DFG- and A-loop flip, αC-helix rotation and G-loop folding. The G-loop is in the folded conformation (G_F), αC-helix is in the 'in' conformation (C_I) and A-loop in 'in' conformation (A_I) in the DFG-in structure shown in figure while in the DFG-out (up) structure, the G-loop is in the extended form (G_E), αC-helix is in the 'out' conformation (C_O) and A-loop in 'out (up)' conformation (A_OU). The co-crystals have been removed for clarity.

Figure 3. The three major DFG-loop conformations observed in AK.

The figure displays the synergy between the salt-bridge and cation-π interactions in different DFG-conformations of AK. The interacting partners are the conserved Lys162 (β3), Glu181 (αC-helix), Phe275 (DFG-loop) and Arg255 (HRD motif, the conserved triad found in the catalytic loop of most kinases).

Figure 4. Major interaction sites in AK.

The graphs depict the frequency of the most prominent interactions with different regions of the binding site formed by a) different classes of AK inhibitors from PDB and b) by the inhibitors bound to the four representative conformations during the 40 ns MD simulation. The four classes of inhibitors depicted in fig. 4a are the type I inhibitors which bind to the conserved ATP site in the DFG-in conformation, the type I¹/₂ which explore an additional back-pocket (BP) formed by the GK in addition to the ATP site in the DFG-in conformation, type III which bind to the allosteric pocket (HPII) in the DFG-out conformation and the type II which explore both the ATP and allosteric pockets in the DFG-out conformation. The details of the simulated systems and inhibitors in fig. 4b have been given in Table 1. The legend 4b describes the system, PDB id of the starting structure, its bound inhibitor and conformation of the major structural motifs. The x-axis represents interactions formed by different pharmacophore and their complementary sites in the binding pockets as given in Table 3.

Figure 5. Non-covalent interactions based specificity rendering hot-spots for the design of Aurora kinase inhibitors.

The AK binding site has been partitioned into six sub-pockets namely back-pocket (BP), adenine-pocket (AP), sugar-pocket (RP), phosphate-pocket (PP), solvent-pocket (SP) and hydrophobic allosteric-pocket (HPII). All possible pharmacophore features found in different classes of AK inhibitors have been mapped onto these six sub-pockets. The pharmacophore features constitute the H-bond donor (HD), H-bond acceptor (HA), aromatic moiety (Ar), linker (L) and HPII binder (hydrophobic). The Venn diagram shows intersections of the six sub-pockets and hot-spots. The colour of the ring represents the pharmacophore endeared by different sub-pockets. The cartoon representation of the binding-site sub-pockets shows the key interacting residues occupying each sub-pocket. The seven hot-spots highlight the possible non-covalent interactions formed by the key interacting residues of each sub-pocket with different pharmacophore features to achieve specificity.

Figure 6. Comparative analysis of the specificity hot-spots explored by AK inhibitor in other kinases.

Three kinases a) PKA b) ABL and c) CDK identified as potent co-targets for AK inhibitors have been modelled in the DFG-out (up) conformation. The six AK sub-pockets and its pharmacophore features have been overlapped on the binding-sites of three kinases. The inset shows the influence of the key-residues of different kinases on the binding of AK inhibitors as well as the likely sub-sets of AK specificity rendering sites and non-covalent interactions compatible to other kinases.