In eukaryotic cells, the peak of protein phosphorylation occurs during mitosis, switching the activities of a significant proportion of proteins and orchestrating a wholesale reorganization of cell shape and internal architecture. Most mitotic protein phosphorylation events are catalysed by a small subset of serine/threonine protein kinases. These include members of the Cdk (cyclin-dependent kinase), Plk (Polo-like kinase), Aurora, Nek (NimA-related kinase) and Bub families, as well as Haspin, Greatwall and Mps1/TTK. There has been steady progress in resolving the structural mechanisms that regulate the catalytic activities of these mitotic kinases. From structural and biochemical perspectives, kinase activation appears not as a binary process (from inactive to active), but as a series of states that exhibit varying degrees of activity. In its lowest activity state, a mitotic kinase may exhibit diverse autoinhibited or inactive conformations. Kinase activation proceeds via phosphorylation and/or association with a binding partner. These remodel the structure into an active conformation that is common to almost all protein kinases. However, all mitotic kinases of known structure have divergent features, many of which are key to understanding their specific regulatory mechanisms. Finally, mitotic kinases are an important class of drug target, and their structural characterization has facilitated the rational design of chemical inhibitors.
The eukaryotic cell cycle is tightly regulated by protein phosphorylation, which peaks during mitosis and is associated with drastic reorganization of cellular structures and protein activities . These mitotic phosphorylation events are co-ordinated temporally and spatially by mitotic kinases, a subset of serine/threonine kinases including members of the Cdk (cyclin-dependent kinase), Plk (Polo-like kinase), Aurora, Nek (NimA-related kinase), Bub and other families . Precise control and co-ordination of these events are crucial for the accurate execution of mitosis. Unsurprisingly, deregulation of mitotic kinases can therefore result in undesirable effects including aneuploidy, resulting in genome instability and leading to cell death or cancer. Considerable effort has therefore been directed towards improving our mechanistic understanding of mitotic kinase regulation.
On the basis of measurements of kinase activity and analysis of the available crystal structures, we take the view that kinase activation is not a simple on–off mechanism, but rather a continuous process in which the kinase may adopt a series of ‘inactive’ and ‘active’ conformations exhibiting a range of activity levels. Transitions between these states are mediated by phosphorylation (either autophosphorylation or phosphorylation by an upstream kinase) and by protein–protein interactions.
In the present review, we first summarize the key features that regulate catalytic activity of kinases in general and then illustrate the structural changes defining the activation states of two exemplar mitotic kinases, Aurora-A and Nek7, which have very different activation pathways. For details of other mitotic kinases, we refer you to our recent comprehensive review of the subject .
Protein kinases have a conserved fold that consists of two lobes: an N-lobe composed of a single β-sheet and one or more α-helices, and a mainly α-helical C-lobe . Crystal structures of protein kinases have revealed the structural basis of their enzymatic function and regulation [3,5–7]. The structures of kinases in their active states are very similar, whereas inactive kinases can adopt a number of different conformations. Three features that can be used to assess whether a kinase structure represents an active or inactive state are the hydrophobic R-spine, the αC-helix and the activation loop (illustrated in Figure 1). In active kinase structures, all three motifs adopt characteristic positions. If any (or all) of these features are displaced, the kinase exhibits lower activity.
Structural motifs of an active kinase
The hydrophobic R-spine traverses the N- and C-lobes of the kinase and is composed of the side chains of four hydrophobic residues: one at the N-terminal end of the β4 strand, one at the C-terminal end of the αC-helix, the phenylalanine residue of the DFG (Asp-Phe-Gly) motif and the histidine residue of the HRD (His-Arg-Asp) motif [8,9]. In active kinase structures, the hydrophobic side chains of these residues are stacked. This serves to bring together the structural features to which they are attached, which are all associated with parts of the catalytic machinery.
Around half-way along the αC-helix is a conserved glutamate which stabilizes a lysine residue on the β3 strand. This lysine residue positions the terminal phosphates of the ATP substrate ready for transfer to the serine, threonine or tyrosine substrate. In the active kinase, the glutamate and lysine residues form a salt bridge that is characteristic of the αC-helix in its ‘in’ conformation. In some inactive kinase structures, such as Cdk2 in the absence of cyclin, the salt-bridge is broken and the αC-helix is rotated and displaced outwards from the N-lobe .
Different kinases vary in both the length and sequence of their activation loops. For most kinases, a specific serine, threonine or tyrosine residue on the activation loop is the primary site of regulatory phosphorylation, either by an upstream kinase [e.g. B-Raf phosphorylation of MEK (mitogen-activated protein kinase/extracellular-signal-regulated kinase kinase)] or in an autophosphorylation reaction (e.g. receptor tyrosine kinases and Aurora-A). Autophosphorylation on the activation loop can be accomplished by either an intramolecular (cis) or intermolecular (trans) mechanism. In active kinases, the phosphorylated serine, threonine or tyrosine residue of the activation loop is co-ordinated by a number of basic residues. During catalysis, the negative charge of the side-chain phosphoryl group stabilizes the transition state of the ATP-substrate phospho-transfer reaction. Upon co-ordination of the phosphorylated residue, the activation loop of a protein kinase becomes ordered. Ordering of the activation loop in the region of the GT (Gly-Thr) motif results in formation of the substrate-binding site, thus promoting catalysis further.
The process of kinase activation is therefore one of physically arranging these three structural elements in the correct places for catalysis. In the course of this arrangement, both the substrate-binding sites and the active site are assembled.
Aurora-A: activation by phosphorylation and protein partner binding
Aurora-A has several mitotic functions, including promoting G2/M transition, centrosome maturation and spindle assembly, as well as non-mitotic functions such as in cilia [11,12]. Aurora-A can be activated by both phosphorylation and the binding of protein partners [13–15]. The best studied of these partners at the molecular level is the N-terminal region of TPX2 (residues 1–43), although activation by a number of other proteins has been shown in cells . Activation of Aurora-A can be considered to be a two-step process in which activation loop phosphorylation and protein partner binding work together to achieve maximum catalytic activity.
Contrary to expectation, unphosphorylated Aurora-A in the absence of binding partners (the lowest activity state of Aurora-A) possesses basal catalytic activity . Crystal structures of this state (e.g. PDB codes 4DEE or 1OL6) reveal that two of the three structural motifs of activity are already adopted: the αC-helix is ‘in’ and the hydrophobic R-spine is assembled. The only feature characteristic of an inactive kinase structure is the activation loop, which is in a partially disordered state that is incompatible with substrate binding. This unexpected adoption of a partially active structure by an unphosphorylated kinase helps to explain why unphosphorylated Aurora-A possesses some catalytic activity: catalysis presumably occurs when the activation loop transiently adopts the correct conformation for substrate binding. Aurora-A is unusual in having a glutamine residue as the αC-helix component of the R-spine, and we speculate that the hydrogen bond formed between this side chain and the β-sheet of the N-lobe stabilizes the assembly of the R-spine in the absence of activation loop phosphorylation.
Aurora-A achieves a medium-activity state either by autophosphorylation on Thr288 or by binding of the partner protein TPX2 (targeting protein for Xklp2). The TPX2-bound unphosphorylated kinase is approximately 10-fold more active than the unphosphorylated kinase alone, whereas activity of the phosphorylated unbound kinase is increased approximately 160-fold. The crystal structure of phosphorylated Aurora-A (PDB code 1OL7) shows a partially ordered activation loop, expected to aid substrate binding, although in this structure the phosphothreonine residue is solvent-exposed and not co-ordinated by any basic residues. Fully active Aurora-A is both phosphorylated and bound to TPX2, and has a fully ordered activation loop (PDB code 1OL5). This increases its activity by a factor of almost 500 over that of the unphosphorylated kinase alone . Depending on its activation state, Aurora-A exhibits one of four very clearly defined levels of catalytic activity ranging over two and a half orders of magnitude (Figure 2).
Aurora-A activation by protein partner binding and autophosphorylation
Aurora-A is localized to the centrosome at the G2/M transition as well as to the mitotic spindle during M-phase. Immunofluorescence microscopy images show that centrosomal Aurora-A is strongly phosphorylated, whereas Aurora-A bound to TPX2 to spindle microtubules is weakly phosphorylated [16,17]. It would appear that, at the centrosome, Aurora-A is primarily activated by phosphorylation, whereas on the mitotic spindle the action of PP6 (protein phosphatase 6) (the cellular phosphatase for Aurora-A ) keeps Aurora-A in a mainly dephosphorylated state and ensures that activation is primarily via TPX2 binding.
Activation of Nek7: release of autoinhibition by Nek9
Nek7 functions in mitotic spindle assembly, although its precise function is unclear and there are no well-defined substrates [19–22]. Activity of Nek7 is increased by the binding of the C-terminal non-catalytic domain of the related kinase Nek9, which may also act as a physiological upstream kinase . To date, Nek7 has been crystallized only in its unphosphorylated state (PDB codes 2WQM and 2WQN) . The conformation of the unphosphorylated Nek7 displays the three classic features of an inactive kinase state: the hydrophobic R-spine is disassembled, the αC-helix is in the ‘out’ position (and hence the Lys63–Glu82 salt bridge is broken) and the activation loop is disordered (Figure 3). Tyr97 is involved in all three of these inactive features. It is the β4 residue that forms part of the hydrophobic R-spine, and its side chain is rotated out of the position that is found in active kinase structures. Indeed, the side chain of Tyr97 points downwards into the active site, forms a hydrogen bond to the backbone amide of Leu180 of the atypical DLG (Asp-Leu-Gly) motif (typically DFG in most kinases), potentially locking it in a conformation incompatible with the typical backbone conformation of an active kinase (DFG-in). The side chain of Tyr97 is wedged between the gatekeeper Leu111 and Leu86 of the αC-helix, which prevents the αC-helix from rotating into the ‘in’ position.
Activation of Nek7 by Nek9 binding and phosphorylation
The structure of inactive Nek7 suggests that Tyr97 has an autoinhibitory function, and this is confirmed by the Y97A mutant of Nek7 which is approximately 5-fold more active than the wild-type protein . Autoinhibition is necessary for maintaining low activity levels of Nek7 (and the highly related Nek6, also downstream of Nek9) during interphase. Autoinhibition is reversed during mitosis, when Nek9 binds Nek7. Addition of the Nek9 non-catalytic CTD (C-terminal domain) to wild-type Nek7 results in kinase activity comparable with that of the Y97A mutant alone, suggesting that this autoinhibitory mechanism is relieved. Thus we postulate that addition of the Nek9 CTD causes a conformational change in Nek7 either through an allosteric mechanism or through the induction of (auto)phosphorylation, flipping the aromatic ring of Tyr97 into the upwards position. This would allow the hydrophobic R-spine residues to align, the two leucine residues (Leu111 and Leu86) to pack together to form a hydrophobic core and the αC-helix to rotate into the ‘in’ position, thus priming the enzyme for catalysis (Figure 3).
To achieve full Nek7 activation, Ser195 on the activation loop must be phosphorylated, which presumably results in formation of the active conformation of the activation loop . This can occur through autophosphorylation or transphosphorylation by the kinase domain of Nek9. We are currently characterizing the activation pathway of Nek7 in order to dissect out the roles of Nek9 binding and of phosphorylation.
Exploiting kinase conformation in drug discovery
Knowledge of kinase-specific mechanisms of activation can be exploited in drug discovery. In the case of Aurora-A, the flexibility of the activation loop shown in the unphosphorylated kinase has been exploited in the design of kinase inhibitors that bind an inactive conformation in which the activation loop is flipped from the position found in the active conformation to a position close to the ATP-binding pocket. Structures with these inhibitors have demonstrated that even the phosphorylated activation loop of Aurora-A can be induced to adopt a non-native conformation similar to that adopted by Src in the presence of Glivec . In the classic form of this conformation found in Src (DFG-out), the aspartate and phenylalanine residues of the DFG motif each rotate through approximately 180°, effectively swapping places [6,25]. In the case of Aurora-A bound to the inhibitor MLN8054, the gross positioning of the activation loop is as expected. However, the local positioning of the activation loop in the vicinity of the ATP-binding pocket is unique, and takes advantage of a number of polar residues in the activation loop which form hydrogen bonds with both the inhibitor and the αC-helix. In this structure (PDB code 2WTV), both the aspartate and phenylalanine residues of the DFG motif point up towards the top of the N-lobe and we have thus termed this conformation DFG-up .
The Tyr-down inactive conformation observed in unphosphorylated Nek7 is exploited in inhibitors of the related kinase Nek2 based on aminopyridine and aminopyrazine scaffolds [24,27]. Structures of Nek2 in apo form, bound to ATP analogues and in the presence of two other series of chemical inhibitors show Tyr70 (equivalent to Tyr97 in Nek7) adopting the same position found in active kinase structures [28–30]. In contrast, the aminopyridine/aminopyrazine compounds induce Tyr70 in Nek2 to jut down into the ATP-binding pocket in a manner similar to Tyr97 in Nek7. The benzoic acid moiety of these compounds lies at the centre of a network of hydrogen bonds that includes the phenolic hydroxy group of Tyr70, an interaction that probably stabilizes the Tyr-down conformation. Approximately 10% of human kinases have a tyrosine residue at the equivalent position to that of Tyr70 in Nek2, and we suggest that selective inhibitors that target these kinases could be developed based on the induction of a Tyr-down conformation.
Protein kinases share a common fold and a common set of regulatory features (the R-spine, the αC-helix and the activation loop) that must be assembled in order for catalysis to occur. However, the biochemical mechanisms by which activation is implemented vary. In the present review, we have illustrated how structural biology and biochemical methods have enabled us to determine the activation mechanisms of Aurora-A and Nek7. Although each recruits the help of a protein binding partner, the role of this partner appears to be different in each kinase. Furthermore, we have recently found that Aurora-A and Nek7 undergo autophosphorylation through distinct and different mechanisms (C.A. Dodson, T. Haq, S. Yeoh and R. Bayliss, unpublished work). A detailed understanding of the mechanisms of activation of each kinase enables us to understand the action of the kinase in the normal and diseased cell. The differences between these mechanisms can also be exploited in drug design, inspired by the ways in which existing inhibitors interact with the different activation states of the kinase.
Exploring Kinomes: Pseudokinases and Beyond: A Biochemical Society Focused Meeting held at Robinson College, Cambridge, U.K., 24–26 March 2013. Organized by Dario Alessi (Dundee, U.K.), Patrick Eyers (Liverpool, U.K.) and James Murphy (Walter and Eliza Hall Institute of Medical Research, Australia). Edited by Patrick Eyers and James Murphy.
R.B. is a Royal Society Research Fellow and thanks Cancer Research UK for support [programme grant number C24461/A12772]. A.M.F. acknowledges support from the Wellcome Trust [programme grant number 082828] and the Association for International Cancer Research.