WO2018167502A1

WO2018167502A1 - Labelled kinases for drug discovery

Info

Publication number: WO2018167502A1
Application number: PCT/GB2018/050677
Authority: WO
Inventors: Charlotte A. DODSON; James A. H. GILBURT; Liming Ying
Original assignee: Imperial Innovations Limited
Priority date: 2017-03-17
Filing date: 2018-03-15
Publication date: 2018-09-20
Also published as: GB201704298D0

Abstract

A kinase or kinase fragment comprising a first label and a second label, wherein: the kinase or kinase fragment has a first conformation and second conformation; one or both of the first and second label are fluorophores; the first label and second label comprise an interacting pair capable of interacting with each other by static quenching or Dexter quenching or PET (photoinduced electron transfer) or exciplex formation to produce a quenched pair; and wherein the first and second labels are positioned on the kinase or kinase fragment such that when the kinase or kinase fragment is in the first conformation the labels are distal to each other such that static quenching or PET or Dexter quenching or exciplex formation does not occur and when the kinase or kinase fragment is in the second conformation the labels are brought into close proximity such that static quenching or PET or Dexter quenching or exciplex formation of one or both labels occurs.

Description

LABELLED KINASES FOR DRUG DISCOVERY

The present invention provides compositions and methods useful in determining the conformational dynamics of kinase superfamily proteins at the single molecule level. The compositions and methods disclosed herein are considered to be useful in drug discovery and lead compound optimisation, amongst other uses.

Background Kinases mediate the transfer of a phosphate moiety from a high energy molecule (such as ATP) to their substrate molecule. Kinases are classified into broad groups by the substrate they act upon: protein kinases, lipid kinases, carbohydrate kinases. The phosphorylation state of a molecule, whether it be a protein, lipid, or carbohydrate, can affect its activity, reactivity, and its ability to bind other molecules. Therefore, kinases are critical in metabolism, cell signalling, protein regulation, cellular transport, secretory processes, and many other cellular pathways, for example protein kinases are essential for the regulation and signalling of eukaryotic cells and are important drug targets in cancer and inflammatory disease 1-3. There are also kinase family proteins termed pseudokinases that are structurally related but do not have kinase activity, or have negligible kinase activity. See, for example Bailey FP, Byrne DP, McSkimming D, Kannan N, Eyers PA. Biochem J. 2015 Jan 15;465(2):195-211. doi: 10.1042/BJ20141060. The term "kinase" used below is considered to refer to kinase family proteins, for example to encompass pseudokinases, unless the context demands otherwise. Protein kinases all share a common structure / overall fold, while the structure of carbohydrate kinases / lipid kinases / adenylate kinase may not be, and in most cases is not, the same. Many protein kinases themselves are regulated by phosphorylation of a regulatory Ser/Thr/Tyr residue on a region of the kinase known as the activation loop, T- loop, activation loop and P+1 loop, A-loop, or activation segment. Phosphorylation of a kinase on the T-loop is typically considered to be activating. However, phosphorylation can also be inhibitory, for example phosphorylation on a region of the kinase known as the P-loop. In non-protein kinases (for example carbohydrate kinases / lipid kinases / adenylate kinase) there typically isn't anything analogous to the activation loop or ser thr/tyr phosphorylation to activate. There are some exceptional cases, for example the lipid kinase PI3K which appears to be a dual lipid/protein kinase and has an activation loop. Kinases may occupy either an active or inactive conformation. While the active conformation of a kinase is typified by the residues being aligned for catalysis, there may be multiple inactive conformations (eg defined by breaking of conserved spines - termed catalytic or regulatory (Kornev AP, Taylor SS. Biochim Biophys Acta. 2010 Mar;1804(3):440-4. doi: 10.1016/j.bbapap.2009.10.017)), for example by movement of a region of the kinase known as the alphaC helix away from the N-lobe of the kinase (often coupled to breaking of a conserved Lys-Glu salt bridge); by inhibition of the active site by a pseudo-substrate (ref Dalton GD, Dewey WL Neuropeptides 40 (2006) 23-34); by occlusion of the active site by regulatory subunits; by phosphorylation on a region of the kinase termed the glycine-rich loop or P-loop (Welburn, J. P., Tucker, J. A., Johnson, T., Lindert, L, Morgan, M., Willis, A., Noble, M. E., and Endicott, J. A. (2007) J. Biol. Chem. 282, 3173-3181); by the proximity of this loop to the ATP binding site; by the orientation of a conserved DFG motif at the N-terminal end of a region of the kinase known as the activation loop or T-loop; or by the orientation the activation loop itself.

In the context of drug discovery, the orientation of the activation loop and DFG-motifs (which are usually, but not always, coupled) is one of the most important conformational markers. Interconversion between the two major conformations changes the chemical and structural properties of the ATP binding site and the region around it and x-ray crystal structures tend to show inhibitors bound exclusively in one or other conformation. Kinase inhibitors are thus classified into two groups depending on whether they bind an active conformation (type I inhibitors) or an inactive conformation (type li inhibitors). In the context of this classification, the active conformation is typified by the activation loop being oriented to form the protein substrate binding site and the aspartic acid of the conserved DFG-motif at the beginning of this loop pointing into the ATP binding site to coordinate divalent metal cations (usually Mg²⁺ or Mn²⁺) / ATP (termed DFG-in). In the classical inactive conformation both the activation loop and DFG-motif are rotated through 180° and a phenylalanine replaces the aspartic acid in the ATP-binding pocket (DFG-out).

The influence of phosphorylation and interaction of small molecule inhibitors with kinase conformation is often summarized by two models. In the first model, phosphorylation achieves activation by ensuring that the activation loop is in a conformation where the catalytic residues are aligned (Figure 1a) ^-5. In the second, an inactive-conformation kinase bound to a type II inhibitor (an inhibitor whose binding site extends into a specific allosteric pocket adjacent to the ATP-binding site) is in equilibrium with the ligand-free kinase in an active conformation (Figure 1 b). Both models are supported by x-ray crystallography but direct experimental testing of kinase activation loop mobility has proved impossible since activation loop dynamics occur on the same timescale as NMR intermediate exchange ^6-7. Since kinases play a vital role in many fundamental cellular processes and are often associated with disease, selective inhibitors of kinases are an attractive prospect for new therapeutics. It has been considered that the inactive conformation of a kinase is a more attractive target for inhibition with specific inhibitors than the active conformation, since the active conformation is more conserved across kinases, leading to a loss of specificity. More recently it may be argued that the evidence has shown that inactive conformation inhibitors have turned out not to be as specific as people thought they would be (and so invalidated this argument). However, there are also semantic questions around whether an inactive conformation inhibitor must occupy the new allosteric binding site - and how you classify things will affect the answer obtained.

"Type II" kinase inhibitors such as Imatinib contact both the ATP cof actor binding site and an adjacent "allosteric" site available only when the kinase assumes a catalytically inactive conformation where the "Asp-Phe-Gly (DFG)" motif at the N terminus of the activation loop is flipped "out" relative to its conformation in the active state ("in"). This conformation of the kinases is generally accepted to be T-loop inactive, although there are occasional exceptions (eg CD532 bound to Aurora-A which is DFG-in and T-loop inactive; Example 3). In contrast, type I inhibitors including VX-680 (and dasatinib) bind at the ATP site but do not induce a change in the orientation of the DFG motif. These inhibitors bind the kinase in a T-loop active conformation and do not make use of the allosteric pocket. Type II inhibitors are generally more selective than type I inhibitors across the enormous human kinome (518 members), but the reasons for this selectivity advantage are not well understood.

An increased understanding of the mobility of the activation loop and the intrinsic conformation that a kinase occupies can aid in drug screening, lead compound optimisation and choice of chemical strategy to pursue, for example whether you want to aim for a type I or type II inhibitor. It is therefore an object of the present invention to provide the means to directly monitor the conformation and dynamics of kinase molecules in a straightforward manner which can be used on any kinase and that leads to highly accurate data from which various parameters can be observed or calculated. Other prior art methods, such as the FRET based method described by Perdios et al 2017 (Scientific Reports 7: 39841) do not solve this problem. The method of Perdios uses two FRET dyes with a critical distance of 58 A (Perdios, PhD thesis, Imperial College London 2015) and places one dye C-terminal to the kinase domain in a separate tetra-cysteine motif. There is mobility in the location of this dye because the spatial location of the tetra- cysteine motif and of the C-terminus of the kinase can vary in a manner independent of kinase function or activation loop conformation. This means that any distance or FRET change determined between the two fluorophores of Perdios is an average distance reporting on changes in the location of the tetra-cysteine motif and of the activation loop simultaneously. These are uncoupled and this method therefore cannot be used to accurately determine the conformation of the activation loop of kinase. Furthermore, the dyes and labelling positions used by Perdios cannot be transferred to other kinases because the distances between the labelling sites are such that the distance changes upon activation loop movement are not great enough to achieve a useful FRET signal change. This makes the method of Perdios difficult to extrapolate to other kinases beyond the single kinase exemplified. The present invention aims to overcome these problems, and indeed is considered to provide several additional advantages over the methods of the prior art.

Methods for assessing protein dynamics that have been used in other situations include, for example, the method Zhou et al (201 1) Nano Lett 1 1(12), 5482-5488 which uses self- quenching of two rhodamine derivatives in a bacterial actin homologue ParM. The inventors are not aware of any other examples of this method being used to monitor conformational changes in proteins at the single molecule level. The use of such a technique applied to kinases was previously considered to be unlikely to work, since there are many other processes that can quench the fluorescence of TMR, such as blinking, bleaching and energy transfer to the protein backbone. The fluorescence of TMR bound to DNA has previously been shown to be unstable (Di Fiori & Meller (2010) Biophysical Journal 98(10) p2265-2272 http://dx.doi.Org/10.1016/j.bpj.2010.02.008). The method was also previously considered unfeasible as two distinct conformations that do not cause quenching would appear identical, resulting in resolution being very much reduced by this approach. The present inventors appreciated that this could be overcome by careful selection of different labelling positions to eliminate each possibility in turn (see for example Example 7). It was also considered that to use such a method significant screening to identify positions for the fluorophores would need to be carried out. Furthermore, FRET traces were considered to be much more straightforward to interpret than for example the largely untested TMR-dimer quenching trace. Summary of the invention

The invention provides compositions and methods that can be used to determine accurately, on a single molecule level, whether or not a particular kinase is occupying a particular conformation, for example a particular active or inactive conformation, for example a T-loop active or inactive conformation, at any given time. From this, detailed kinetic data can be determined, aiding in drug discovery and lead compound optimisation, for example in the identification of specific inhibitors and of inhibitors promoting a specific conformation, for example a specific conformation of the activation loop. The conformation of the activation loop (T-loop) of protein kinases underlies enzymatic activity and influences the binding of small molecule inhibitors. Using single molecule fluorescence spectroscopy in combination with a kinase comprising at least two strategically placed labels, the inventors have devised a model, applicable to any kinase, in which to study the movement of the T-loop between active and inactive forms of kinase. This approach is considered to enable conformation-specific effects to be integrated into inhibitor discovery across the kinome. As examples, the compositions and methods of the present invention led the inventors to the following conclusions:

• The inhibitor binding pocket is part of an ensemble of structures in very different conformations, and an inhibitor previously considered a type II inhibitor additionally binds a state of the kinase not previously considered

• It is now possible to measure, and thus develop reliable prediction algorithms for, the effect of an inhibitor on the conformation of the activation loop of a kinase in solution

· Potent conformation-independent inhibitors need to bind both active and inactive T-loop kinase conformations and such inhibitor development may need to focus on common structural features

• Conformation-specific inhibitors must optimise discrimination between the dissociation constant (Kd) for each conformation:

The intrinsic position of equilibrium for a kinase influences the inhibitor-bound T- loop conformation; the same compound may thus inhibit different targets in different T-loop conformations • Binding partners and/or phosphorylation states which change the intrinsic position of equilibrium of a target may contribute to unexpected selectivity profiles in cells, or in conformation-specific complex disruption (e.g. Aurora-A/N-Myc by alisertib (MLN8237) and CD523 in small ceil lung cancer (SCLC) and neuroblastoma) · The ability to determine the intrinsic conformational propensity of a kinase has implications for the selection of kinases against which to develop type II inhibitors

The methods of the invention may be incorporated into lead optimisation in the chemistry cycle of drug discovery, and may be used alongside x-ray crystallography and other biophysical techniques. The methods of the invention give quantitative information on the conformation induced and its extent (not binary either / or) and therefore is expected to report on trends across chemical series/scaffolds before optimisation (new mechanism of action relationship between chemistry and activation loop motion). The methods disclosed herein are considered to be particularly useful where the kinase activity is low or absent, for example specific targeting of unphosphorylated kinase, for targeting of pseudokinases and for targeting conformations of a kinase where the activatory ser/thr tyr is hidden from an upstream kinase. The methods disclosed herein are also considered to be useful in type II inhibitor programmes in the absence of x-ray crystallographic information, for example where kinases are difficult or impossible to crystallise, or where the kinase activation loop is not visible in the structure.

Detailed description of the invention

In a first aspect, the invention provides a kinase or kinase fragment comprising a first label and a second label, wherein: the kinase or kinase fragment has (or may have) a first conformation and a second conformation; one or both of the first and second label are fluorophores; the first label and second label comprise an interacting pair capable of interacting with each other by static quenching or Dexter quenching or PET (photoinduced electron transfer) to produce a quenched pair or exiplex formation to produce an exiplex; and wherein the first and second labels are positioned on the kinase or kinase fragment such that when the kinase or kinase fragment is in the first conformation (or first possible conformation) the labels are distal to each other such that static quenching or Dexter quenching or PET or exciplex formation between the first and second labels does not occur; and when the kinase or kinase fragment is in the second conformation (or second possible conformation) the labels are brought into close proximity such that static quenching or Dexter quenching or PET or exciplex formation of one or both labels [or between the first and second label] occurs. The kinase of the invention is capable (or may be capable) of adopting a first conformation and a second conformation. As discussed below, the kinase of the invention may have the amino acid sequence of a native or wild-type kinase, or may be a mutant kinase, including drug-resistant mutants. Typically a first conformation and a second conformation may mean an active T-loop conformation and an inactive T-loop conformation, though in such case it is not necessary for the first conformation to be the active confirmation and the second conformation to be the inactive conformation: equally the first conformation may be the inactive conformation and the second conformation may be the active conformation. The kinase of the present invention, through the associated labels, produces a particular signal if the kinase is in one conformation, and a different signal if the kinase is in another conformation. By assaying single molecules of the kinase of the invention it is possible to determine whether the kinase is in one conformation or another, and also determine when the kinase changes from one conformation to another. As noted above, there are multiple inactive conformations (eg defined by breaking of conserved spines - termed catalytic or regulatory (Kornev AP, Taylor SS. Biochim Biophys Acta. 2010 Mar; 1804(3) :440-4. doi: 10.1016/j.bbapap.2009.10.017)) for example as set out above, and it is considered possible to monitor any of them. However, it is considered that T-loop conformation may be particularly useful to monitor.

The kinase may be any kinase, for example it may be a protein kinase, or a lipid kinase, or a carbohydrate kinases or any other type of kinase. In a preferred embodiment the kinase is a kinase that has a role in disease, for example has a role in the development or progression of cancer. Typically such a kinase may be a protein kinase. Full details of kinases from sequenced genome projects are on www.kinase.com. Comprehensive information on kinases can be found within Uniprot (www.uniprot.org) by searching either for named kinases or for 'protein kinase'. Specific examples include ABL Abelson kinase; Akt protein kinase B or kinase from the transforming oncogene Akt8; ALK anaplastic lymphoma kinase; BTK Bruton tyrosine kinase; CHEK1 (CHK1 ) checkpoint kinase- 1 ; EGFR epidermal growth factor receptor; FAK focal adhesion kinase; Fes Feline sarcoma oncogene kinase; FGFR1 fibroblast growth factor recptor-1 ; FLT3 fetal liver kinase-3; GSK3 glycogen synthase kinase-3p; JH2 Jak homology domain-2; LKB1 serine/threonine-protein kinase STK11 ; MAP2K mitogen activated kinase kinase; MAP3K mitogen activated kinase kinase kinase; MAP4K mitogen activated kinase kinase kinase kinase MAPK mitogen activated kinase; MEK1 mitogen activated kinase kinase- 1 MET mesenchymal epithelial transition factor or hepatocyte growth or scatter factor receptor; mTOR mammalian target of rapamycin; PDGFR platelet- derived growth factor receptor; PDK1 3-phosphoinositide-dependent protein kinase-1 ; PTEN phosphatase and tensin homologue; RAF rapidly accelerated fibrosarcoma; SMAD SMAD is the composite of MAD form drosophila Mothers Against Decapentaplegic and SMA of Caenorhabditis elegans (from gene sma for small body size); STRAD1 STE20- related adapter a; TrkB or NTRK-2 tropomyosin receptor kinase B or neurotrophin receptor kinase-2.

Further examples include: EGFR, LRRK2, SRC, ABL, BTK, MAP kinases, FGFR, MAP pathway kinases, KIT, AKT1 , ERBB2, JAK, FAK2, PLK1 , CHK2, CHK1 , GCK3B, p38a, PKC, AURKA, AURKB, NEK2, NEK6, NEK7, CDK1 , CDK2, CDK5, CDK9, RAF1 , TTK, PAK7, PAK2, LCK, MTOR, VEGF, BRAF, ARAF, CRAF, DAPK3. The kinase may be, for example, a member of a classification group such as a protein kinase, serine/threonine kinase, tyrosine kinase, tyrosine kinase-like kinase, tyrosine-like kinases, receptor tyrosine kinase, non-receptor tyrosine kinase, STE kinase, CMCG kinase, CK1 kinase, CAMK kinase, dual specificity kinase (DYRK), AGC-like kinase, AGC kinase, ATE Kinase, atypical protein kinases, or a pseudokinase, optionally wherein the kinase is Aurora kinase or Aurora-A kinase.

The labelling approach is considered to be applicable for any kinase (including, for the avoidance of doubt, pseudokinases) that has an activation loop. For a kinase without an activation loop or for any kinase which doesn't share the conserved protein kinases fold, where to label would be highly dependent on structure and it may not be possible to identify suitable labelling sites. Accordingly, different labelling sites may be needed (and may not be feasible) for many or most carbohydrate kinases, adenylate kinase, and many lipid kinases. PI3K has an activation loop and a similarish structure, so a similar approach to that for protein kinases (with an activation loop) is considered likely to be appropriate. In principle, the labelled kinase may not work so well (in terms of being useful in revealing changes in structure or activation state) for protein kinases where the activation loop doesn't change. However, this is a property that has not been investigated before, so there is little information on which kinases these will be. It is considered that two such kinases (which may be suitable as negative controls) are PKA and Fes, but this has not yet been confirmed by measurement. It is considered that not all kinases which require phosphorylation for activation will have mobile loops, and the situation is also unclear for kinases which don't require phosphorylation. The labelling approach and methods of the present invention are considered to be useful in identifying whether the activation loop (or other region) or a kinase undergoes a predicted or possible conformational change or not.

The kinase may be from any organism, for example it may be from a mammal, for example from a human, or it may be from a dog, for example. The kinase may be from a eukaryote or from a prokaryote. Typically the kinase may be a human kinase or an experimental animal kinase, for example a mouse, rat or dog kinase. The kinase may alternatively be from a human parasite - eg Plasmodium falciparum, Schistosoma mansoni, Trypanosoma brucei, Leishmania donovani, Babesia bovis, Toxoplasma gondii, Eimeria tenella, Theileria annulata, Entamoeba histolytica; or from an animal parasite, for example a companion or domesticated animal parasite, for example a parasite of importance in agriculture. The kinase may be homogeneously expressed (i.e. expressed in the same organism from which the kinase originated), or it may be heterologously expressed (i.e. expressed from a different organism from which the kinase originated). The kinase may be an altered form of the wild-type or native kinase, but expressed in the organism from which the kinase originated.

In one embodiment, the kinase is a full length native or wild-type kinase. By native or wild- type kinase we include the meaning of the kinase having the identical amino acid sequence to that of the kinase as found in vivo, for example a native or wild-type human kinase may have the identical amino acid sequence to the kinase as expressed in a particular human cell. In a further embodiment the kinase is not a full length native or wild-type kinase. For example the kinase may be a truncated kinase, for example may be a kinase fragment, or may comprise additional amino acid sequences (or both, for example a tagged fragment). In one embodiment the kinase comprises one or more mutated residues as compared to the native or wild-type amino acid sequence. For example the kinase may have between 1 and 100 mutated residues, for example between 5 and 95 mutated residues, for example between 10 and 90 mutated residues, for example between 20 and 80 mutated residues, for example between 30 and 70 mutated residues, for example between 40 and 60 mutated residues, for example about 50 mutated residues as compared to the native or wild-type amino acid sequence. In one embodiment the kinase has more than 100 mutated residues as compared to the native or wild-type amino acid sequence.

The mutations may be synonymous or non-synonymous. The kinase may be a kinase considered to be a drug-resistant mutant. Such mutants may arise during cancer therapy (as there is a strong selective pressure for mutations to occur), and are considered as targets in their own right.

The kinase may have one or more mutations that mimics phosphorylation, as well known in the field of kinases. The kinase may have one or more mutations that mimic an unphosphorylated state, for example mutation to alanine at a phosphorylation site. Alternatively the kinase is not mutated in a residue on which the kinase is considered to be capable of being phosphorylated, particularly not mutated in a residue whose phosphorylation is considered to be significant in modulation/regulation of the kinase's activity. Thus the kinase may not be a pseudophosphorylated kinase. However, there may be circumstances in which it is instructive or useful to monitor conformation changes in a kinase that is so mutated in a residue on which the kinase is considered to be capable of being phosphorylated. As will be known to those skilled in the art, mutations to D or E may be used as phosphomimics in cell biology, for example. Similarly, chemical biology is beginning to develop ways of chemically phosphorylating non-natural amino acids (eg phosphocysteine) to act as phospho-mimics. Phosphocysteine is a remarkably good mimic of phosphoserine.

In one embodiment, the residues of a particular type, or of a particular amino acid are mutated. For example prior to labelling, site-directed mutagenesis has been performed on the kinase to mutate at least one, for example at least two, for example at least three, for example at least four, for example at least five, for example all of the cysteine residues in undesirable locations, for example all of the native cysteine residues, optionally to mutate the cysteine residues to alanine or serine or any other amino acid, typically a relatively small and/or usually uncharged amino acid or to remove the cysteine residues. In one embodiment, which may be in conjunction with the embodiments discussed above and below, the residues of a particular type, or of an amino acid, that are located in a particular place on the kinase are mutated. For example the residues of a particular type, or of an amino acid that are located in a particular region of the linear sequence are mutated. In a further embodiment, the residues of a particular type, or of an amino acid that are located in a particular region of the three-dimensional structure are mutated. In this embodiment the residues of a particular type, or of an amino acid do not have to be consecutive in the linear amino acid sequence, since the skilled person knows that different regions of the linear sequence come together to form the three-dimensional structure of the kinase. In a particular embodiment, the cysteine residues that are located on the surface of the three-dimensional structure are mutated, for example are mutated to alanine or serine residues. Further, in addition, or alternatively, the kinase may have one or more, preferably two or more particular residues, for example cysteines, introduced into the amino acid sequence such that they appear on the surface of the three-dimensional structure. In a preferred embodiment the native or wild-type cysteines that appear on the surface of the three-dimensional structure are mutated to residues that are not cysteines, and at least one, preferably two or more non-cysteine residues in the native or wild-type kinase sequence are mutated to cysteines, such that the surface of the three-dimensional structure of the kinase has one, preferably two, or more than two cysteines at strategic locations. If a cysteine of the native or wild-type kinase sequence is already in one of said strategic locations, the skilled person will realise that there is no need to change the residue at all. The skilled person will also understand that the above embodiments that refer to, for example, cysteine, may apply equally to any other amino acid, including non- natural amino acids. Mutation of residues to or from cysteine residues (for example) may be of particular relevance with regard to some labelling strategies, in controlling the number of labelling sites that are exposed on the kinase surface/potentially available for labelling.

The skilled person will understand that the aim of introducing the strategically located cysteines as described above is to allow the specific labelling of the kinase, for example with a fluorophore and/or quencher in specific well defined places. Cysteines are considered to provide only one option for labelling the protein. Other residues that may be incorporated include lysine residues. Preferably the residue is a cysteine since typically this residue occurs less frequently in proteins and allows one or two (or more as desired) specific labelling events to occur. However, it is also possible to incorporate non-natural amino acids into specific locations on the kinase. This has the advantage that no prior manipulation of the kinase is required, for example to mutate any undesired cysteines to non-cysteines, and the kinase can be simply mutated to incorporate the non-natural amino acid at the preferred location. As discussed above, it is preferable if the kinase has a defined number, typically two, but potentially could be more, of particular residues which allow labelling of the protein, with for example a fluorophore and/or quencher on the external surface of the protein. Preferred locations are discussed further below.

In one embodiment the types of residues which allow labelling of the protein with, for example, a fluorophore and/or quencher, are the same. This may be the case when for example a single label is to be used, for example tetramethylrhodamine (TMR), and both locations on the protein are labelled with the same label. If the kinase is to comprise two (or more) different labels, it is also possible to achieve this when labelling residues are the same. For example, if equal stoichiometric amounts of each label are used, the expected proportions of kinase labelled with both dye 1 and dye 2 is 50%, dye 1 and dye 1 is 25%, and dye 2 and dye 2 is 25%. These populations can be distinguished, for example via total internal reflection microscopy (TIRF) and these populations can be taken account of statistically. A technique called ALEX (alternating laser excitation; Kapanidis AN1 , Laurence TA, Lee NK, Margeat E, Kong X, Weiss S. Acc Chem Res. 2005 Jul;38(7):523- 33.) may also be used, in which excitation of each dye occurs in turn and co-localisation used to determine molecules which possess one dye of each type. In another embodiment, the residues which allow the labelling of the protein with, for example, a fluorophore and/or quencher are different. This is preferable when at least two different labels are to be used, for example a given fluorophore and a given quencher. In this way the location of the fluorophore and the location of the quencher are known. However, it is considered that the present invention provides an advantage over the prior art since it is not considered to be necessary to know which of the locations has the fluorophore and which of the locations has the quencher, since the kinase produces a signal dependent upon the distance between the two labels. If one binding site can be occluded with a ligand (eg an antibody) while the second is labelled, for example, then there would be preferential labelling, but this may not be possible in all circumstances.

Some preferred reaction chemistries that allow labelling of proteins, including kinases, include dye-maleimide and dye-iodoacetamide which react with thiols (for example on a cysteine residue) to form a covalent linkage; dye-NHS ester which reacts with amines for example lysine residues to form a covalent linkage; dye-azide, for example those using azide-phosphine reactions, azide-alkyne reactions, strained alkene-azide reactions and alkene-tetrazole reactions. In addition, strain-promoted inverse electron demand Diels- Alder cycloaddition reactions between alkene or alkyne (for example norbornenes, bicyclononynes, trans-cyclooctenes and cyclopropenes), or tetrazine groups. In one embodiment, the first and/or second label are attached to the kinase through a thiol reaction, optionally through a thiol on a cysteine residue. Any method which allows the kinase to incorporate at least two labels into specific well defined locations is encompassed by the present invention. This may result in the kinase having an altered amino acid sequence to that of the native or wild-type sequence, or not. In any event, it is preferable that any modification to allow the kinase to be labelled, or the labelling itself, does not affect, or does not substantially affect, the activity or activation profile of the kinase compared to the activity of the native of wild-type kinase, or the non- native and non-wild-type kinase that lacks one or more of the labels. In one embodiment the labelled kinase retains (typically under the same activation circumstances) at least 40% of the catalytic activity of the unlabelled kinase, optionally retains at least 45%, optionally at least 50%, optionally at least 55%, optionally at least 60%, optionally at least 65%, optionally at least 70%, optionally at least 75%, optionally at least 80%, optionally at least 82%, optionally at least 84%, optionally at least 86%, optionally at least 88%, optionally at least 90%, optionally at least 91 %, optionally at least 92%, optionally at least 93%, optionally at least 94%, optionally at least 95%, optionally at least 96%, optionally at least 97%, optionally at least 98%, optionally at least 99%, optionally 100% of the catalytic activity of the unlabelled kinase.

It will also be appreciated that any modification to allow the kinase to be labelled, or the labelling itself, does not affect, or does not substantially affect the conformations that the kinase adopts, or the transition between those conformations. This may be judged by considering activation profile or response to inhibitors, for example. It may also be possible to derive some information by NMR (eg doi: 10.1073/pnas.1318899111 ) or Infra-red spectroscopy (see for example see Cyphers et al (2017) doi: 10.1038/nchembio.2296)

As discussed above, the kinase of the invention may be a mutant kinase. Although kinases are typically considered to occupy a first conformation and second conformation and interconvert between the two, it may be possible that particular kinase mutants or native or wild-type kinases are only capable of occupying a single conformation. Such kinases are also considered potentially to be useful. The skilled person will appreciate that the ability of a kinase to occupy only a single conformation cannot be determined prior to the present invention being carried out. Accordingly, in a further aspect, the invention provides a variant, for example a mutant, for example differing in no more than 1 , 2, 3, 4, 5, 6, 7, 8, 9, 10, 1 1 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 50 or 70 amino acid positions, of a kinase of the invention as described in the first aspect, that can adopt only one conformation. In one embodiment, the variant, for example a mutant differs by more than 70 amino acid positions compared to a kinase of the invention as described in the first aspect, that can adopt only one conformation, for example at least 70 amino acid positions, for example at least 80 amino acid positions, for example at least 90 amino acid positions, for example at least 100 amino acid positions.

The skilled person will appreciate that a mutant which adopts only one conformation would potentially be very valuable in drug screening. At present, screening is performed against multiple conformations at once. Targeting a specific conformation, knowing that the kinase was constrained to this conformation would enable high throughput screening against a single conformation. The methods of the present invention may be useful in identifying such a variant, for example a mutant, protein kinase that can adopt only one conformation. The invention therefore also provides such a variant, a method of identifying such a variant, and methods and uses of such a variant, for example in high throughput screening.

The skilled person will also appreciate that it may be useful to measure the dynamics of the kinase in an un activated (potentially catalytically dead) form. Inhibition of this form may prevent subsequent activation (eg by holding the activation loop in a conformation where the activatory ser/thr/tyr is inaccessible to an upstream kinase). Alternatively, if the kinase activates a second kinase in an allosteric manner (eg B-Raf or EGFR) inhibiting it in a known conformation (one that doesn't activate allosterically) is also considered to be useful, even if the kinase that is being 'inhibited' has no catalytic activity itself. Accordingly, in one embodiment, the labelled kinase retains (typically under the same activation circumstances) less than 40% of the catalytic activity of the unlabelled kinase, optionally less than 30%, optionally less than 20%, optionally less than 10%, optionally less than 5%, optionally less than 1 % of the catalytic activity of the unlabelled kinase. In some instances the kinase under investigation is a mutant relative to the native or wild- type sequence, for example may comprise an amino acid sequence which is the same as a mutant kinase isolated from a cancerous cell, for example a drug resistant mutant kinase. In this case, the mutant kinase may have an elevated or reduced kinase activity compared to the native or wild-type kinase, but this is irrelevant for the present invention. The skilled person will understand that it is preferable that the method used to introduce the two or more labels into the kinase will have no or little further effect on the kinase activity.

Although it is preferred that the presence of the labels does not affect the catalytic activity of the kinase, in some situations it is not considered to be detrimental if the labels do affect the activity of the kinase. For example, it is considered that such a kinase labelled with the appropriate labels (for example a kinase in which the presence of the labels influences the catalytic activity) will still give useful information. The skilled person is capable of taking any information obtained from such a kinase and weighting it compared to other data, knowing the effect that the label(s) has had on the activity of the kinase. For example, such a kinase is still considered to give useful structural information, or information relating to intra- and inter-molecular interaction.

It is also preferred if the method used to introduce the label results in the labels being tightly associated with a given residue. Although methods which employ FLAsH (e.g. Nature Biotechnology 23: 1308-1314 (2005)) and enzymatic ligation (e.g. doi: 10.1073/pnas.091067107) may be used in the present invention, they are not considered to be preferred embodiments since they involve the use of tag sequences or enzyme consensus sequences of at least 12 residues. In addition to these additional sequences potentially affecting kinase conformation, change in conformation and catalytic activity, they also reduce the association of the label with the kinase so that the label is in fact located within a radius of a given location, and not at the actual location itself.

The label may be a tryptophan residue located at a suitable position within the kinase polypeptide chain, but more typically the label is attached to the kinase polypeptide chain rather than being a naturally occurring amino acid forming part of the kinase polypeptide chain.

"3rd Edition of Principles of Fluorescence Spectroscopy"; and "Methods in Molecular Biology Vol 335 - fluorescent energy transfer in nucleic acid probes: designs and protocols - chapter 2", for example, indicate that quenching can occur via FRET (dynamic quenching), or static/contact quenching. Static or contact quenching appears to occur when the molecules are very close to one another, i.e. within non-covalent molecular contact /in molecular contact/in Van der Waals contact or in near Van der Waals contact. This is one form of quenching that is applicable to the present invention. In the present invention the labels are brought very close together in one conformation (quenched) and separated when the kinase is in the other conformation. This conformational change in kinase structure therefore has to be enough to take the fluorophores in and out of being in Van der Walls contact or in near Van der Waals contact.

Accordingly, types of quenching that are considered to be suitable for use in the present invention are static quenching, contact quenching, Dexter quenching and photoinduced electron transfer. This list is not exhaustive and other types of quenching are considered to be suitable for use in the invention, provided they are not FRET quenching. In one embodiment the quenching is non-FRET quenching, i.e. all types of quenching that do not occur by FRET. FRET is a particular type of energy transfer and as discussed here, the skilled person is able to determine whether a particular quenching event occurs through FRET. Preferably the type of non-FRET quenching is a type of quenching that produces an essentially on-off signal (i.e. quenched vs non-quenched).

The kinase of the invention comprises a first label and a second label, wherein one or both of the first and second label are fluorophores.

The kinase of the invention is labelled with two labels (at least) at specific well defined positions. Preferences for the mechanism of attachment of a label to the kinase of the invention are discussed above, for example the label may be attached through a thiol reaction to a specific cysteine in the kinase. The skilled person will understand that there are many methods of labelling a protein, for example a kinase, with a label, for example a fluorophore. All such methods are encompassed by the present invention.

The kinase of the invention produces one signal when in one conformation, and another signal when in another conformation. Accordingly, the two labels have to be capable of interacting with each other to produce either a signal when in close proximity and no signal when distal to one another, or a signal when distal to one another no signal when in close proximity, i.e. quenched.

By the term "label" we include the meaning of a compound or moiety that is capable of reporting information about the entity to which the label is attached. The label may be considered to be a dye. The kinase of the invention comprises two labels, at least one of which is a fluorophore, i.e. a moiety that absorbs radiation predominantly of a particular wavelength and emits radiation predominantly of another wavelength. The second label may be any type of label, but is also preferably a fluorophore or a compound or moiety that is typically considered to be a quencher.

FRET is one mechanism by which a particular fluorescence signal can be quenched, and requires that the emission spectra of the donor (i.e. fluorophore) overlaps with the absorption spectra of the acceptor (quencher). In such a case, in the absence of the quencher/acceptor, the donor will emit a fluorescence signal that can be detected, at a given wavelength. In the presence of the acceptor, energy is transferred from the donor to the acceptor by resonance energy transfer, putting the acceptor in an excited state and returning the donor to the ground state in a non-radiative manner (i.e. no fluorescence is emitted from the donor). In this way, the signal from the donor is said to be quenched. Generally, the acceptor molecule returns to the ground state in a radiative manner, i.e. emits fluorescence. Therefore, unless the quencher/acceptor is capable of returning to ground state without emitting light (for example is not a so-called blackhole quencher, as known to those skilled in the art and discussed in, for example, Johansson Meth Mol Biol 335 Fluorescent Energy Transfer Nucleic Acid Probes: Designs and Protocols (2006)), a signal from the quencher/acceptor is observed. Furthermore, quenching in this way is considered never to be total and so two peaks of emission are observed with FRET, making the data more difficult to understand since a continuous signal is generated of varying intensity depending on the proximity of the two labels, i.e. the signal is not on/off (i.e. a first kinase conformation versus a second kinase conformation) but is rather a continuous gradient.

However, FRET has a number of disadvantages, which include the presence of bleed- through and acceptor photobleaching. The presence of bleed-through (also termed crosstalk and crossover) and cross excitation between spectrally overlapping fluorophores are also important issues that can hamper FRET investigations. This occurs when the fluorescence from one fluorophore is partially recorded in the detection channel for the second fluorophore (ie the wavelengths of the spectra partially overlap). In some cases, the acceptor can be directly excited with light in the wavelength region chosen to excite the donor. Additionally, fluorescence from the donor can leak into the detection channel for the acceptor fluorescence, especially when the emission spectral profiles of the donor and acceptor exhibit significant overlap. Because these two sources of crosstalk arise from the photophysics of organic fluorophores and will most certainly be present for any FRET pair, they must be addressed when FRET is measured. Choosing fluorophores that are well-separated spectrally is an excellent mechanism to reduce crosstalk. However, in most cases the increased spectral separation also reduces the overlap integral, (J(A)), which in practice usually translates to a reduced ability to detect the FRET signal.

Acceptor photobleaching can also be a problem in FRET experiments. When this occurs, high fluorescence is observed for the donor, and zero fluorescence for the acceptor. Since this signal is often similar to that of the low FRET conformation, it can be difficult to distinguish the two. For either low FRET or acceptor photobleaching, the onset of the two signals will be anticorrelated, making them even harder to distinguish. When choosing a FRET pair to use, one of the principal issues is that the donor and acceptor fluorophores might exhibit significantly different brightness levels when imaged together. Although in theory this discrepancy should not be a problem, in practice because most instruments can measure only a limited dynamic range, dual fluorophore imaging may result in one channel that is saturated (for the brighter fluorophore) while the other channel is dominated by systematic noise (for the dimmer fluorophore). A typical single molecule setup will capture donor and acceptor signals on two sides of the same detector (eg create an image where the left side is donor fluorescence and the right hand side is acceptor fluorescence) meaning that camera settings cannot be optimised for each fluorophore individually. Since the present invention does not require FRET, but uses instead static quenching, contact quenching, Dexter quenching and photoinduced electron transfer, this is not an issue. The present invention does not require the detection of two different brightness levels such that one may bleach the other. The present invention detects instead an on/off signal. The signal is either there, and the kinase is in one conformation, or the signal is absent and the kinase is in the other conformation.

One other disadvantage of the use of FRET is that FRET also requires two different dyes (one donor, one acceptor). This means that in order for a FRET measurement to be used to monitor conformational change in kinases, one dye of each type must be attached to the kinase. Although the use of two different dyes is encompassed by the present invention, and labelling strategies are discussed above, in one embodiment it is preferred that the two labels are the same label, in which case labelling the kinase is straightforward. However, by using FRET two different labels must be used. In general, the labelling chemistry for each dye is the same, meaning that each dye attaches randomly to a labelling site. This means that in a labelling reaction with two different dyes, there is a statistical probability that any individual kinase will be labelled with either two donor fluorophores or two acceptor fluorophores. While the ratio of donor and acceptor in the labelling reaction can be altered to change these statistics (in particular to decrease the amount of donor to reduce the population of double-donor labelled molecules which create a high false 'low- FRET' population), this will not remove the presence of kinases with two identical labels. The expected populations of double-donor: donor-acceptor: double-acceptor labelled molecules can be calculated from the ratio of donor: acceptor using probabilities (1 :1 donor: acceptor results in a labelling ratio of 1 :2:1 ; 1 :4 donor: acceptor results in a labelling ratio of 1 :8:16). The statistical nature of using two different dyes in a labelling reaction means that the generation of double-labelled kinase is inefficient and wasteful. It also means that a considerable fraction of single molecules measured will need to be discarded. This is wasteful and can be time-consuming.

The efficiency of FRET depends on the distance between the two molecules (r) and decays with its inverse sixth power (r^{1 6}). For every pair of donor / acceptor molecules, there is a distance defined as the critical distance (Ro) which is the distance at which resonance energy transfer is 50% efficient. FRET is due to long-range dipolar interactions between excited state donor and acceptor.

The rate of energy transfer, k-r(r) is

where τ_D is donor lifetime in absence of acceptor, r is the centre-to-centre distance between donor and acceptor, R₀ is the critical distance (Forster distance).

The expected FRET efficiency, E, is

where r is the centre-to-centre distance between donor and acceptor and R₀ the critical distance (Forster distance). When calculating expected FRET efficiencies from protein structures, the length of the linker between the protein and the dye must also be taken into consideration. For two dyes on short linkers, this distance is a total of -10 A.

The distances over which existing single molecule FRET pairs operate is considered to make it unsuitable for use in the present invention. To work on the scale of the very small conformational changes in the kinase between the active and inactive T-loop conformations, in a domain as small as the kinase domain, a FRET pair must have a small enough critical distance to give a measureable change in signal (i.e. from signal to unequivocally different signal) between distances such as 5 A and 32 A. For example, a standard single molecule FRET pair with a critical distance of 56 A would give expected FRET efficiencies of 100% at 5 A and 97% at 32 A, clearly too small to observe. A putative FRET pair with a critical distance of 30 A would give expected FRET efficiencies of 100% and 40% which would be observable. Two FRET dyes closer than 0.5 x R₀ will give 100% FRET at all times. Two dyes further apart than 2 x R₀ will give 0 % FRET at all times. It is balancing the R₀ of the FRET pair with the distance between the dyes that enables FRET to be used to measure conformational change.

Whilst there is nothing to stop FRET occurring over a very short distance and providing useful information, the problem is finding a suitable FRET pair. It would need 1. Small dyes (so that the location of the dye reports accurately on the location of the protein residue to which it is attached); 2. Dyes with high molecular brightness (molecular brightness = extinction coeff x quantum yield) so that we can see them at the single molecule level; 3. FRET pairs with a critical distance < 40A. 4. Commercially available dyes with short linkers (otherwise the critical distance is effectively increased). Preferably also: 5. Excited with visible light (using a UV laser increases the safety screening required). 6. Emitting >450 nm and < 850nm (to ensure high efficiency detection by camera).

However, there are other mechanisms by which the fluorescence of a fluorophore can be quenched, termed static or contact quenching, photoinduced electron transfer (PET), exciplex formation and Dexter quenching, as discussed above. These processes cause a reduction in fluorescence intensity of a fluorophore in which dissipation of the fluorophore's electronic energy occurs as heat. In static or contact quenching, the fluorophore and quencher come into non-covalent molecular contact at the Van der Waals radii, allowing interaction between the electron clouds. This interaction enables the return of an excited state electron in the fluorophore to ground state in a non-radiative manner. Static quenching is due to short-range interactions between the fluorophore and quencher and decreases exponentially with the distance between the two. In practice, quenching can form an essentially "on/off system. In photoinduced electron transfer (PET), upon photoexcitation and molecular contact between an electron donor (D) and an electron acceptor (A), an excited state complex is formed between the electron donor and acceptor, and electron transfer occurs (to form D⁺A-). Electron transfer does not occur in the ground state because this reaction is not energetically favourable. The fluorescent species may be either the electron donor or acceptor, but in both cases the result of the electron transfer is that fluorescence emission from the fluorophore is quenched. Emission from the excited state complex (exciplex) may or may not occur, but this will be at a different wavelength from that of the original fluorophore and so can easily be distinguished. As D⁺A- decays to ground state, the extra electron on the acceptor is returned to the electron donor in a non-radiative manner down an energy gradient. Dexter quenching is a quantum mechanical effect with no classical analogy. Dexter quenching also involves electron transfer from donor to acceptor in an excited state, and is paired with electron transfer in the ground state from acceptor to donor which may either be stepwise or concerted with the initial transfer. For Dexter quenching,

where k(r) is the rate of quenching, r is the centre to centre distance between donor and acceptor, L is the sum of the Van de Waals radii of donor and acceptor and J is the spectral overlap:

where ίϋ(λ) is the wavelength-dependent fluorescence of the donor,

the wavelength- dependent absorption spectrum of the acceptor and λ the wavelength.

Exciplex formation is the formation of a fluorescent excited state complex between two molecules. The excited state complex may have a different absorption and fluorescence emission profiles to the non-complexed molecules. In particular, the fluorescence of the exciplex is often red-shifted (moved to higher wavelengths) compared with the emission of the non-complexed molecules.

The present invention makes use of any of these forms of quenching, and in one embodiment, static or contact quenching.

The skilled person will be well aware of the particular fluorophores and/or quenchers that are capable of use in the present invention.

Because of the necessity for Van de Waals' contact, the quenching required by the present invention occurs over shorter distances than FRET. In practice, the distance changes that can be reported on will depend on the length and flexibility of the dye linkers in the context of the labelled protein. This is readily calculated by the skilled person. For example, the estimated distance between S (from thiol) and label/dye for both iodoacetemide and maleimide linkers is ~5 A. Doubling this (for two labels/dyes) gives 10 A, adding 2 A for dye interaction distance gets to an assay sensitivity at 12 A separation which is in line (+/- error due eg from distance to effective centre of dye) with the published experimental estimate of -15 A (obtained using TMR). Since Van der Waals' contact is needed whatever dye pair is used, the estimate really depends mainly on the length of the linkers, not the identity of the dye. So if (for some reason) a linker with 10 carbons was used (rather than one with 2 carbons for example), an extra 22 A (16 x 1.4) would have to be added, making the distance for assay sensitivity 34 A (12 + 22). Things would be further complicated particularly with a longer linker because the linker will almost certainly not be fully extended at all times - so increasing length typically isn't an easy/desirable way to tune assay sensitivity.

The rate of energy transfer, k_E(r), for static quenching, is

where r is the centre-to-centre distance between fluorophore and quencher and r_c is the distance of closest approach at molecular contact. A is expected to have a value near 10¹³ s^-1 for interactions between orbitals and values of β are typically near 1 A^-1. This expression ignores the effect of diffusion on quenching.

The Van der Waals radii would be calculated using 'standard chemistry' that will calculate the size of the electron cloud for any given molecule, as well known to those skilled in the art.

Accordingly, the first label and second label comprise an interacting pair capable of interacting with each other by static quenching to produce a quenched pair. As discussed above, the skilled person will understand which labels are suitable for use in the present invention. Such labels and methods of determining which labels would be suitable for use in static quenching are discussed in 3^rd Edition of Principles of Fluorescence Spectroscopy. for example.

An example of labels that are considered to be suitable for use are any of the dyes in table I of Marras et al, Nucleic Acids Research, 2002, Vol. 30 No. 21 e122. TMR-TMR self-quenching is considered to be particularly useful because it doesn't require labelling with more than one type of dye.

Fluorescein, BODIPY, Cyanine and rhodamines self-quench which is considered to be beneficial in the present invention. However fluorescein, for example, may not be suitable for the present invention because it is considered not to be bright enough for single- molecule measurements and has poor photostability. Table 8.1 on page 279 of 3^rd Edition of Principles of Fluorescence Spectroscopy lists some fluorophores and quenchers but it is considered that most of them are not useful, for example because the fluorophore properties are unsuitable (for example not bright enough) or the quencher unsuitable to immobilise on the protein or too general (eg oxygen, iodide, nitric oxide, sulphur dioxide).

Methods in Molecular Biology Vol 335 - fluorescent energy transfer in nucleic acid probes: designs and protocols - chapter 2 teaches:

Static quenching is the same as contact quenching and lists dyes and dark quenchers (Table 1 - dyes for oligos; Table 2 - dark quenchers). These dyes and quenchers are considered generally to be useful in the present invention. Marras et al 2002 discloses one method (involving hybridisation of labelled oligonucleotides) by which the skilled person could determine whether a particular pair of fluorophores or fluorophore/quencher pair is capable of static quenching. Such a method is well within the skill of the skilled person. The method for screening F/Q pairs uses blunt end hybridised olignucleotides labelled with interacting labels leading to static quenching whilst a staggered end oligonucleotide duplex labelled with interacting labels acts through FRET. Another method for determining whether static quenching or FRET is occurring may involve measuring fluorescence lifetime. Static quenching is considered to produce no change in the lifetime. The skilled person will be able to distinguish between a signal generated by FRET and a signal generated by other forms of quenching (i.e. be able to determine whether he is or is not working within the scope of the invention) since the signal obtained by FRET decreases with increasing distance (centred around the critical distance) between a dye pair, while that from quenching has essentially a step change from fluorescence to quenched fluorescence over a very short distance of non-covalent molecular contact. Additionally, in FRET there is an anticorrelated fluorescence signal between two reporter dyes (ie fluorescence of one dye increases when the fluorescence of the other dye decreases) which can be detected eg by the technique of ALEX. This is not the case in other forms of fluorescence quenching. In the case of static quenching, a ground state complex is formed between fluorophore and quencher which tends to result in a change in the absorption profile of the dye. A FRET interaction leaves the absorption profiles of the dyes unchanged. In the case of PET, an interaction can often be predicted by reference to the electrochemical potentials of the donor and acceptor. FRET interactions are unrelated to the electrochemical potentials of the dye pair.

Dependent upon the exact combination of labels, the labels chosen typically may be capable of quenching fluorescence when the labels are separated by (centre-to-centre) less than 20 A, 19 A, 18 A, 17 A, 16 A, or less than 15 A, optionally less than 14 A, optionally less than 13 A, optionally less than 12 A, optionally less than 11 A, optionally less than 10 A, optionally less than 9 A, optionally less than 8 A, optionally less than 7 A, optionally less than 6 A, optionally less than 5 A, optionally less than 4 A, optionally less than 3 A, optionally less than 2 A, optionally less than 1 A, optionally less than 0.5 A, optionally less than 0.25 A; and do not quench fluorescence when separated by at least 20 A, optionally at least 22 A, optionally at least 24 A, optionally at least 26 A, optionally at least 28 A, optionally at least 30 A, optionally at least 35 A, optionally at least 40 A. For a single molecule, fluorescence is essentially either on or quenched (/^'e there is no such thing as partial quenching in practice). 'Partial quenching' is more typical of FRET (not encompassed by the present invention).

By separated by we include the meaning of the centre-to-centre distance between the two labels. The organic dyes useful in the present invention may typically be planar molecules. The Van de Waals' contact radius for a planar molecule is likely to be closer to an ovoid with the plane at its centre than a sphere.

In one embodiment the kinase comprises two labels and both labels are fluorophores, for example both labels are TMR, or alternatively both labels may be different fluorophores. Embodiments in which both of the labels are fluorophores are considered to be advantageous since in the unquenched state, a brighter signal is obtained since there are two fluorophores in the excited state, rather than one, leading to reduced noise in the detection system. Where two fluorophores are used, the presence of two photobleaching steps at the end of the experimental measurement acts as in internal diagnostic for a doubly labelled molecule. Since when the fluorophores are in contact the signal is almost totally quenched, there is a large difference in the signal emitted between the two conformations of the kinase. The absence of total quenching allows contact quenching to be distinguished from other photophysical effects.

This is particularly true in embodiments in which the two fluorophores are substantially the same fluorophore, for example where both fluorophores are TMR. This means that the light emitting from the fluorophores is of the same, or substantially the same wavelength, resulting in a more intense signal than if the fluorophores were different fluorophores with different emission spectra. Accordingly in a preferred embodiment, the first and second labels are both fluorophores which are substantially the same fluorophore. In another embodiment the first and second fluorophore are the same fluorophore, by which we mean are the same type of fluorophore and not that the first and second fluorophore are physically the same molecule In a preferred embodiment the first and second label are both TMR.

The use of such a technique applied to kinases was previously considered to be unlikely to work, since there are many other processes that can quench the fluorescence of TMR, such as blinking, bleaching and energy transfer to the protein backbone. The fluorescence of TMR bound to DNA has previously been shown to be unstable (Di Fiori & Meller (2010) Biophysical Journal 98(10) p2265-2272 http://dx.doi.Org/10.1016/j.bpj.2010.02.008). The method was also considered to not resolve the difference between two conformations that do not cause quenching appearing identical. The present inventors appreciated that this could be overcome by careful selection of different labelling positions to eliminate each possibility in turn (Example 7). It was also considered that to use such a method significant screening to identify positions for the fluorophores would need to be carried out. Furthermore, FRET traces are considered to be much more straightforward to interpret than for example a TMR-dimer quenching trace which was, prior to the present invention, a largely untested method.

Despite all of the above perceived disadvantages with the current approach, the inventors have successfully used the present invention to determine essential information about the interconversion between conformations of a kinase. However, embodiments in which the first and second fluorophore are different fluorophores are also encompassed by this invention. Accordingly, in one embodiment the first label and the second label are fluorophores and are different fluorophores. Suitable pairs may be determined from, for example, Table 1 of Marras, supra. It will be apparent to the skilled person that whilst one label is a fluorophore, the other label may be a quencher, or a label that is known as a quencher. Suitable quenchers are those that, as discussed above, quench the fluorescence of the fluorophore when in very close proximity, for example when they are approximately less than 15 A apart (centre-to- centre), but which do not quench the fluorescence of the fluorophore when the quencher and fluorophore are more than approximately 20 A apart (centre-to-centre), for example. Given the very narrow range of distances over which static quenching or PET or Dexter quenching or exciplex formation can be used, it will be apparent to the skilled person that consideration has to be given to the placing of the labels on the kinase. The location of the labels should be such that when the kinase is in one conformation the labels are separated such that quenching does not occur, but that when the kinase is in the other conformation quenching does occur. This defines the region of non-covalent molecular contact or region of Van der Waals contact between the two dyes which can be predicted by someone skilled in the art. Accordingly, in one embodiment the location of the labels on the kinase are such that when the kinase is in one conformation the labels are not in molecular contact, and when the kinase is the other conformation the labels are in molecular contact. The skilled person will readily be able to determine the optimum places to locate the labels, for example based on three-dimensional structures, or computer modelling. However, a limited amount of trial and error is to be expected by the skilled person and does not represent an undue burden. Accordingly, several pairs of locations may be assayed to identify the optimum locations. Factors to be considered will be apparent to those skilled in the art and may include one or more of achieving an acceptable change between signal / no signal; acceptable signal/noise ratio; absence of artefacts from potentially native quenchers on protein; labels do not interfere with binding of ligands; labels do not change intensity upon binding of ligands. As an example, quenched signal with TMR is low, but above background: it is not necessary for the quenched signal to be the same as background. Techniques that can be used to assess if a given pair of labels are working as intended include obtaining expected result upon addition of ligands; comparison with previously characterised standard (if available); expected result with control experiments (single labelling position); absence of unexplained dye blinking (fast fluctuations to background fluorescence).

In a preferred embodiment, the first or second label is located on the T-loop (activation loop). The T-loop/activation loop stretches between the conserved DFG and APE motifs of protein kinases and some other kinases such as PI3K. While the amino acid sequences of some kinases do not contain the two motifs in full (eg human Nek2 has a DLG motif and human Aurora- A a PPE motif), homology alignments - either by sequence, structure or function - enable the correct region to be defined. The conserved D (which functionally coordinates divalent metal ions (usually Mg2+) and the phosphate groups of ATP) is often absent in pseudokinases (proteins sharing the protein kinase fold, but with low or zero catalytic activity). However this does not prevent definition of the activation loop region. Existing alignments can also be used in identifying the activation loop/T-loop, for example http://kinase.com/human/kinome/phylogeny.html, an example of a comm u n ity-accepted peer reviewed reference alignment

In one embodiment, one label is located on the T-loop and the other label is located at a position that gives the optimal difference between fluorescence signal when the kinase is in one conformation and when the kinase in the other conformation.

In a preferred embodiment, only one label is located on the activation loop of the kinase.

In one embodiment, the first label is a fluorophore and is located on the activation loop/T- loop. In another embodiment the second label is a quencher and is located on the activation loop/T-loop.

The kinase of the invention may comprise a first or second label located on the N-lobe, the Glycine-rich loop / P-loop / phosphate anchor of the kinase. All of these regions are defined by structure/sequence homology and will be well known to those skilled in the art. See, for example Pearce, Komander,& Alessi (2010) Nature Reviews Molecular Cell Biology 11 , 9-22 doi:10.1038/nrm2822; Johnson, Noble & Owen (1996) Cell, Vol. 85, 149- 158; Taylor & Radzio-Andzelm (1994) Structure 2:345-355 (note that the conclusions of this paper have been shown to be more generalizable than the authors imagined at the time); Fabbro, D., Cowan-Jacob, S. W. and Moebitz, H. (2015), Br J Pharmacol, 172: 2675-2700. doi:10.1 11 1/bph.13096. The first or second label may also be on the C-lobe other than the activation loop (eg the M373C/S283C construct described). The other label may be located on the T-loop, for example.

In one embodiment, the first and second label are located on residues of the kinase that correspond to 373, 224, 367 and/or 283 of the human Aurora-A kinase. The skilled person will appreciate that kinases share homology and will be able to locate the corresponding position of for example residue 373, 224, 367 and/or 283 of the human Aurora-A kinase an another kinase. Accordingly in one embodiment the first and second labels are located on residues of the kinase that correspond to residues 373, 224, 367 and/or 283 of human Aurora-A kinase. In a further embodiment the first and second label are located on residues corresponding to M373, K224, N367 and/or S283 of the human Aurora-A kinase or the equivalent residue or position in a different kinase. In one embodiment the first and second label are located on residue M373C, K224C, N367C and/or S283C of the human Aurora-A kinase or the equivalent residue or position of a different kinase.

For example, the skilled person is able to identify the corresponding position of for example residues 373, 224, 367 and/or 283 of the human Aurora A kinase on other kinases, for example by sequence alignment or alignment of the 3-D structures of the kinases. In some embodiments these residues are mutated to cysteines, for example where they are not already cysteines so that the kinase may be labelled according to the methods described herein.

Further guidance on the selection of the specific labelling site and/or labels is now provided.

In order to predict whether a conformational change will result in a change in fluorescence signal, initial modelling should be carried out using the pdb files of a crystal structure of the kinase in each conformation, or using a homology model of the kinase (if a crystal structure is not available). The aim of the modelling is to predict at least two labelling sites - one on the activation loop (or region of the kinase expected to move between the different conformations) and one on a second region of the kinase (which is not expected to move between the different conformations). Ideally these labelling sites will be on the surface of the protein to enable labelling reactions to occur. These labelling sites should also be chosen so as to minimise the effect of the labels on the activity and binding of the kinase. Therefore, the labelling sites should avoid regions of the kinase known (or predicted) to be important in kinase activity or the binding of ligands. Such regions include protein-protein interaction sites, the active site, the substrate binding site (particularly the C-terminal end of the activation loop if possible) and the DFG-motif itself. In the case of a kinase about which little is known, predictions can be made about the likelihood of a region to be involved in kinase activity or ligand binding from sequence conservation with other kinases or co-variation of mutation with known ligands.

Before selecting the distance change to be selected, the dyes to be used need to be selected. This is because it is necessary to know the linker length between the fluorophore and the point at which it will be coupled to the protein. The length of this linker can vary with the commercial preparation of the dye (ie from different manufacturers) and can be estimated by summing the bond lengths of each bond in the chemical structure. Once the total length of the linker is known for each dye, this should be summed together to give the total overall linker length. For the quenched conformation, the two chosen labelling sites on the protein should be closer than the total linker length (approximately 0.5 - 0.9x total linker length). This is firstly to enable the dye molecules (or dye/quencher pair) to interact with each other (there must be molecular contact for quenching to occur), and secondly because the dyes (or dye / quencher pair) must interact not only when the two linkers are both stretched to their furthest extremes, but at points within their cones of rotation other than this (when their linking residues are still close in space). When measuring distances on the structure of the protein, the distances should be measured between the a or β carbons of the amino acids in question. This is because the sidechains shown in the protein structure are unlikely to be fixed in space (as they are in the crystal structure), but to be dynamic and moving in space in solution. Measuring distances between atoms further down the amino acid side chain will therefore introduce errors in the measured the distances. Measuring between Ca or C(i atoms (particularly Cβ atoms) is particularly appropriate for dye attachment using thiol-maleimide chemistry because the sulphur of the cysteine residue is in the same position as the Cβ atom on the wild-type residue. This means that the distances measured will be as accurate as possible.

For the fluorescent (unquenched) conformation, the two chosen labelling sites on the protein should be further apart than the total linker length (ideally »1 .5x). This is to ensure that the dyes (or dye/quencher pair) cannot interact in the second conformation.

If at all possible, the dye labelling sites should also be chosen to be away from any wild- type tryptophan residues on the protein surface. This is because the amino acid tryptophan can quench many fluorescent dyes, and could therefore interfere with the assay. If it is not possible to be distant from a native tryptophan, careful control measurements should be made with single-labelled protein to ensure that tryptophan quenching does not occur, or the assay modified to exploit this quenching (eg by using a dye-tryptophan quenching pair such as MR121 - tryptophan).

Careful selection of the dye labelling sites in the manner laid out above will reduce the need for extensive experimental screening.

As discussed above, it is important that when the kinase changes conformation, it takes the labels in and out of non-covalent molecular contact. Accordingly, in one embodiment, the relative positions of the first and second label are such that when the kinase is in a first conformation the labels are not in non-covalent molecular contact. In another embodiment the relative positions of the first and second label are such that when the kinase is in a second conformation the labels are within non-covalent molecular contact. Preferably the relative positions of the first and second label are such that when the kinase is in a first conformation the labels are not in non-covalent molecular contact and when the kinase is in a second conformation the labels are within non-covalent molecular contact.

As discussed above, static quenching and PET and Dexter quenching and exciplex formation occur over a smaller distance than FRET. Accordingly in one embodiment the first and second label are positioned on the kinase such that when the kinase is in a first conformation the centres of the labels are separated by at least 17 A, optionally at least 18 A , optionally at least 19 A , optionally at least 20 A, optionally at least 22 A, optionally at least 24 A, optionally at least 26 A, optionally at least 28 A, optionally at least 30 A, optionally at least 35 A, optionally at least 40 A. In a further embodiment, the first and second label are positioned on the kinase such that when the kinase is in a second conformation the labels are separated by 15 A or less than 15 A, optionally less than 14 A, optionally less than 13 A, optionally less than 12 A, optionally less than 11 A, optionally less than 10 A, optionally less than 9 A, optionally less than 8 A, optionally less than 7 A, optionally less than 6 A, optionally less than 5 A, optionally less than 4 A, optionally less than 3 A, optionally less than 2 A, optionally less than 1 A, optionally less than 0.5 A, optionally less than 0.25 A.

In a preferred embodiment, the first and second label are positioned on the kinase such that when the kinase is in a first conformation the centres of the labels are separated by at least 17 A, optionally at least 18 A , optionally at least 19 A , optionally at least 20 A, optionally at least 22 A, optionally at least 24 A, optionally at least 26 A, optionally at least 28 A, optionally at least 30 A, optionally at least 35 A, optionally at least 40 A, and when the kinase is in a second conformation the labels are separated by 15 A or less than 15 A, optionally less than 14 A, optionally less than 13 A, optionally less than 12 A, optionally less than 11 A, optionally less than 10 A, optionally less than 9 A, optionally less than 8 A, optionally less than 7 A, optionally less than 6 A, optionally less than 5 A, optionally less than 4 A, optionally less than 3 A, optionally less than 2 A, optionally less than 1 A, optionally less than 0.5 A, optionally less than 0.25 A.

Typically, the centre-to-centre distance in the second conformation may be less than 7 A, 5 A, 2 A, 1 A, <1 A. As an example, in the linkers used by the commercial dye manufacturer Attotech (www. atto-tec.com). the distance from the dye to the N of the reactive group (iodoacetamide or maleimide) is approx. 6.5 A (estimated using C-C bond length of 1.4 A, C-N bond length of 1 .8 A). From the N of the reactive group to the S of the cysteine residue on the protein is -5.3 A for iodoacetamide and -4.9 A for maleimide (estimated using bond lengths above and C-S bond length of 2.1 A). Thus the distance from dye to C of cysteine is -13 A for each labelling site. With two labelling sites, the Cp of the two labelled cysteine residues would be expected to be -20 A apart when dye quenching occurs. This distance will change if the linker length of the dye changes or if a different labelling chemistry is used.

The skilled person will appreciate that once the two labels are separated by sufficient distance, no quenching is observed. For example, if a particular pair of labels results in quenching when the centre-to-centre distance is 7 A or less, then for those particular labels to be suitable for use in the present invention, the labels have to be positioned such that when the kinase is in one conformation the labels are separated by no more than 7 A, and when the kinase is in a second conformation, the labels are separated by more than 7 A, resulting in a fluorescence signal being produced. This applies to all separation distances mentioned herein. For example, in one embodiment the centre-to-centre distance of the dyes when the kinase is in one conformation is 5 A or less, and when the kinase is in the second conformation the centre-to-centre distance of the dyes is more than 5 A. In the former conformation there is quenching, and in the latter conformation there is essentially not quenching.

The skilled person will appreciate that it is the combination of the label, the linker, and the position of the attachment site of the label, for example the position of the amino acid residue to which the label and/or linker is attached that determines the spatial arrangement of the two labels and whether or not the labels interact appropriately (i.e. quench) when the kinase is in one conformation, and do not quench when the kinase is in a second conformation. The skilled person is well aware of these parameters and is readily able to determine the required label, labelling site and linker required. A particular pair of labels, for example two fluorophores or a fluorophore/quencher pair is considered to be suitable for use in the present invention if the fluorescence signal obtained when the kinase is in the first conformation and the fluorescence signal obtained when the kinase is in the second conformation differs by sufficient amount that it is possible to determine beyond noise that the two values are different, which may be, for example, at least 20%, 30%, 40%, 50%, 60% or 70%, optionally differs by at least 75%, optionally at least 80%, optionally at least 82%, optionally at least 84%, optionally at least 86%, optionally at least 88%, optionally at least 90%, optionally at least 91 %, optionally at least 92%, optionally at least 93%, optionally at least 94%, optionally at least 95%, optionally at least 96%, optionally at least 97%, optionally at least 98%, optionally at least 99%, optionally 100%. The skilled person will easily be able to determine these values. One way to test that the dye pair fulfils the criteria above is by molecular modelling of the 3-D structure of the protein and the expected volume explored by the dye molecules taking into account the different degrees of freedom of each bond in the dye / linker region. An experimental way to test a dye pair is to make use of a ligand or phosphorylation state of the kinase expected to promote one or other conformation of the activation loop. If addition of the ligand increases the population of the expected state, this is evidence that the dye pair is working as expected. For example, in the case of Aurora-A, TPX2 is expected (from x-ray crystallography and enzyme kinetic measurements) to promote the T-loop active conformation and CD532 expected (from x-ray crystallography) to promote the T-loop inactive conformation (Example 3). Also in the case of Aurora-A, phosphorylation on Thr288 is expected (from x-ray crystallography and from enzyme kinetic measurements) to promote the T-loop active conformation and dephosphorylation on this residue to promote the T-loop inactive conformation (Example 7). An alternative way to test the efficacy of a dye pair is to carry out analogous experiments by NMR on kinase without fluorescent dyes (eg www.pnas.org/cq i/doi/10.1073/pnas.13188991 1 1 ) and determine whether two different techniques provide similar results.

The signal from the labels may be determined by any known means in the art. In one embodiment the signal is determined by TIRF. The detection method will depend on the dye pair used. For a conventional quenching pair (static, Dexter quenching, exciplex formation or PET), the method needs to detect a change in fluorescence intensity. For a dye pair where homo-FRET occurs (ie FRET occurs between two identical dyes because there is substantial overlap between the emission and excitation spectra, enabling resonance energy transfer from the donor to a second dye molecule) there will be no change in fluorescence intensity after energy transfer (one dye will be emitting). Thus, for homo-FRET, a different detection method is necessary (fluorescence anisotropy). It is considered that there is no reason in principle why a homo-FRET pair cannot be used in the present invention, although the detected signal is expected to be weak.

In one embodiment the kinase may comprise three or more labels. As an example, a third label could be used in conjunction with a corresponding label on an inhibitor or activator to assay binding. As a further example, an alternative would be to have two different fluorophores (of different colours) which can both be quenched by the same quencher (on the activation loop). This way, there would be an anti-correlated signal between the two fluorophores which would act as an internal control. Thus, for example, two fluorophores (eg one on K224C, one on M373C) and one quencher (on S283C) would be a very good system.

In one embodiment, the kinase of the invention is one which has a low kinase activity; and/or is difficult to crystallise or cannot be crystal ized; and/or wherein the activation loop is not visible in the structure.

The kinase of the application is considered to have a role in the screening of drugs, for example screening for novel kinase inhibitors, and in lead compound optimisation, particular where several different kinases of the invention as used simultaneously.

Accordingly in one embodiment the invention provides a kinase of the invention which is immobilised on a solid support, for example as part of an array or library.

In a second aspect the invention provides a panel/array/library of kinases as defined above. In one embodiment the panel/array/library may comprise at least 5 different kinases, optionally at least different 10 kinases, optionally at least 15 different kinases, optionally at least different 20 kinases, optionally at least different 30 kinases, optionally at least 40 different kinases, optionally at least 50 different kinases, optionally at least 75 different kinases, optionally at least 100 different kinases, optionally at least 150 different kinases, optionally at least 200 different kinases, optionally at least 250 different kinases, optionally at least 300 different kinases, optionally at least 350 different kinases, optionally at least 400 different kinases, optionally at least 500 different kinases. In a further embodiment the kinases in the panel/array/library are immobilised on a solid support, for example a functionalised glass coverslip (that can be imaged through), but there are other suitable options.

In a further embodiment, the kinases of the panel/array/library are from the same family.

In a further embodiment the panel, array or library may comprise multiple aliquots of the same kinase. Such a panel, array or library is considered to be useful in the screening of agents for example potential inhibitors, for action against a particular kinase. Preferably the library comprises both multiple kinases as described above, and multiple aliquots of each kinase. Such a panel is considered to be useful both in the screening of agents for their effects on the conformation of a particular kinase, and simultaneously screening the global effect of that agent across a wide range of kinases.

Kinases may be grouped by, for example phylogenetics (ie sequence conservation), by occurrence in the same disease (eg the cancer panel, the inflammatory disease panel), by signalling pathway (eg the B-RAF / ERK2 pathway), by likelihood of responding to similar chemistries (eg kinases which tend to give off-target effects together with the main target kinase), by synthetic lethality (ie kinases which are lethal to the cell if both are knocked out), by disease resistant mutants (eg kinases and commonly or predictably occurring resistance mutations so that both can be screened together).

The assay may be used in vivo,. The kinase may be labelled ex-vivo and then introduced into the cell by for example micro-injection or electroporation. Alternatively, the kinase may be expressed using non-native amino acids and the cells washed with dye (which can cross the cell membrane). Under these conditions, labelling would occur in the cell itself, and free dye be removed by multiple washing of the cells. Suitable controls may be performed by carrying out the assay in vitro in the presence of an equivalent cell lysate, for example.

The skilled person will understand that the features of any embodiment discussed above in relation to the kinase of the invention, the types of label and position of the label may be incorporated with features of any other embodiment. For example, the kinase may be a human kinase, mutated to introduce specific cysteine residues, the first label and the second label may both be fluorophores, and one label may be located on the T-loop and one label may be located on the N-lobe.

A third aspect of the invention provides a method for determining the intrinsic conformational propensity of a kinase wherein the method comprises determining the conformation of a kinase as defined above, for example in relation to the first and second aspects of the invention. In a preferred embodiment the method comprises determining the conformation of single kinase molecules. Methods of determining the conformation of a single kinase molecule are discussed in the Examples, and include the attachment of the kinase to a solid support, for example a functionalised glass coverslip, followed by microscopy, for example TIRF. Other methods may include light sheet microscopy and confocal microscopy. In a preferred embodiment the conformation of many single kinase molecules is determined, either sequentially or simultaneously, preferably simultaneously.

Since the conformation of a kinase protein may change from active to inactive and vice versa, in one embodiment the conformation of a single kinase molecule, or multiple single kinase molecules is determined over time. The length of time may be a few minutes, for example up to 1 minute or more, for example up to 2 minutes or more, for example up to

3 minutes or more, for example up to 4 minutes or more, for example up to 5 minutes or more, for example up to 10 minutes or more, for example up to 15 minutes or more, for example up to 20 minutes or more, for example up to 30 minutes or more. Potentially one could monitor for hours if the interconversion is slow, for example for up to 1 hour or more, for example up to 2 hours or more, for example up to 3 hours or more, for example up to

4 hours or more, for example up to 6 hours or more, for example up to 12 hours or more, for example up to 18 hours or more, for example up to 24 hours or more. In one embodiment it is considered that the longer the period of time the kinase is monitored for,, the fewer molecules one needs to measure. There are approximately 32,000 frames in each histogram that the inventors have provided - these (for example) could come from a small number of molecules for a long time or a larger number of molecules for a shorter time.

It is considered that approximately 32,000 frames is a suitable number of frames to obtain reliable data. However, the skilled person will appreciate that the number of frames obtained may be greater than this, or may be less than this, depending upon the circumstances required. For example, if the interconversion is very slow, a greater number of frames may be required to enable a sufficient number of conversion events is monitored. For example, a suitable number of frames may be at least 1 ,000, for example at least 2,000, for example at least 5,000 frames, for example at least 10,000 frames, for example at least 20,000 frames, for at least 30,000 frames, for example at least 32,000 frames, for example at least 40,000 frames, for example at least 50,000 frames, for example at least 60,000 frames, for example at least 70,000 frames for example at least 80,000 frames, for example at least 90,000 frames, for example at least 100,000 frames or more.

Accordingly therefore the invention provides a method for determining the timescale of conformation change of a kinase, wherein the method comprises measuring the conformation of a kinase according to the invention. The skilled person will appreciate that the frame rate may also be changed, for example the frame rate may be 80ms long or for example may be 150ms long. The skilled person will also appreciate that the data can be collected when the duty cycle of the experiment is less than 100%, i.e. there are gaps in time which are much shorter than the rate of conformational change for example between frames when no data is collected. The skilled person will understand that the method of determining the conformation of a kinase molecule is dependent upon the labels used. For example, the kinase comprises at least one fluorophore, so the method of determining the conformation of a kinase molecule involves the use of techniques that can measure the amount of fluorescence signal displayed by a particular kinase molecule. The meaning of this signal varies dependent on the relative positions of the two labels.

For instance, in one embodiment, when the kinase is in an active conformation, the two labels are separated such that static quenching does not occur, and a fluorescence signal is obtained; when the kinase is in an inactive conformation the two labels are brought together in close proximity, i.e. into non-covalent molecular contact such that the signal is quenched and no fluorescence is emitted. In this instance, detection of a fluorescence signal indicates that that particular kinase molecule is in an active conformation. Kinase molecules that are in an inactive conformation are largely undetectable (they typically have a quenched low but non-zero fluorescence which means that their presence can be detected) since the signal is quenched. In a different embodiment, when the kinase is in an active conformation, the two labels are brought together in close proximity, i.e. into non- covalent molecular contact such that the signal is quenched and no fluorescence is emitted; when the kinase is in an inactive conformation the two labels are separated such that for example static quenching does not occur, and a fluorescence signal is obtained.

The data obtained from the methods of the invention allow the skilled person to determine or calculate the dissociation constant (Kd), conformation-specific Kd, equilibrium constant (Keq) and/or change in Gibbs free energy. Accordingly in a further aspect the invention provides a method to calculate the dissociation constant (Kd), Kd, conformation-specific Kd, equilibrium constant (Keq) and/or change in Gibbs free energy of a kinase, for example using the method of determining the intrinsic conformational propensity of a kinase according to an earlier aspect of the invention.

The methods of the invention are considered to be useful in aspects of drug discovery and lead compound optimisation, for example in the discovery or optimisation of specific kinase inhibitors. The invention therefore provides a method for determining the effect of an inhibitor or activator or combination of inhibitor and activator on kinase conformation, wherein the method comprises any of the methods disclosed herein, for example the method of determining the intrinsic conformational propensity of a kinase according to an earlier aspect of the invention, or the method to calculate the dissociation constant (K_d), conformation-specific Kd, equilibrium constant (K_eq) and/or change in Gibbs free energy of a kinase according to the method of an earlier aspect of the invention. The methods of the invention are considered to be particularly useful where the active of the kinase is low, or for kinases that are difficult to crystallise or which cannot be crystalized, or wherein the activation loop is not visible in the structure.

It will be appreciated that the methods disclosed herein are not restricted solely to the investigation of the effects of inhibitors or activators, the effects of any agent on kinase conformation can be determined. The invention therefore provides method of determining the conformation of a kinase that is induced by an agent. An agent includes the meaning of, for example, small molecule compounds which may be inhibitors or activators, and also includes for example biological agents, for instance an antibody, nanobody, peptide and/or a synthetic mimic of these agents. Any of these, and other agents, are considered to potentially be an allosteric activator / inhibitor (via stabilisation of one conformation / conformation-specific binding).

Since the cellular (or extracellular) environment is complex, there may be agents or factors that are present in vivo that affect the conformation of the kinase, or equilibrium of the active and inactive conformation of the kinase, or the affinity of the kinase for certain agents, for example inhibitors. Some of these cellular agents or factors are known, for example the TPX2 protein that activates Aurora A kinase, and can be used in conjunction with the methods disclosed herein. However, there are likely to be some factors that are present in the cellular or extracellular environment and which affect the kinase in some way, but which are unknown. It one embodiment therefore the methods disclosed herein are carried out in the presence of a cell lysate, or other fluid which comprises the molecules from the same environment that the kinase was originally taken from, or is medically relevant, for example. For example, Aurora A kinase is found to be overexpressed in some breast tumours. It is considered to be appropriate therefore to carry out the methods of the invention (including those discussed both above and below), for example determining the intrinsic conformational propensity of a kinase; or calculating the dissociation constant (K_d), conformation-specific Kd, equilibrium constant (Keq) and/or change in Gibbs free energy of a kinase: or determining the effect of an inhibitor or activator or combination of inhibitor and activator on kinase conformation; or for example methods involving the use of a panel, library or array); in the presence of a tumour cell lysate, preferably a breast tumour cell lysate. Alternatively it may not be considered to be important that the cell lysate or other fluid is relevant to the kinase under investigation, for example the methods disclosed herein may be carried out in any environment, for example in the presence of an imaging buffer, as described in the examples, or a generic cell lysate. This may give more accurate data on how the kinase actually behaves in vivo, for example taking into account potentially unknown activators. A background control could, for example, be a single-labelled protein (ie protein with only one labelling site). Such a control would expect to display a continuous high fluorescence intensity (the same as for a similar control measurement in imaging buffer).

The inventors of the present invention have provided data (see the Examples) which indicate that in order to develop new drugs, for example kinase inhibitors, it is useful if the effects of inhibitors and activators are assessed in combination. It is important to determine the relative affinities for an activator and inhibitor to either conformation of the kinase to determine the required affinity of an inhibitor to enable the equilibrium of active to inactive conformation to be shifted to the inactive conformation.

It is our understanding that the affinity of a given inhibitor or activator for a given active or inactive conformation is at present unknown. The present invention will allow this to be determined, and also allow the effects of combinations of inhibitors and activators to be determined. For example if an activator that is found in vivo has a strong affinity for the active conformation of the kinase, then an inhibitor may be required that has an even stronger affinity for the inactive conformation. This will also depend on the intrinsic conformational preference of the kinase because this preference will also bias the underlying population towards / away from the active conformation.

The methods of the invention will also allow the skilled person to predict the effect of an inhibitor or activator on kinase conformation. This depends on what is known about the relative KdS of the inhibitor/activator for active and inactive conformation of the kinase. If these can be predicted (eg from in silico studies), then equation 18 of Example 5, for example, can be used to predict the effect of the inhibitor on overall conformation. However, the main aim of the invention is measuring the effect on conformation directly. The methods disclosed herein are also useful in the design of a new kinase inhibitor, or strategic modification of a known kinase inhibitor, since the effects of the inhibitor on kinase conformation and equilibrium between conformations can now be determined. Accordingly, the invention also provides a method for designing a kinase inhibitor or the strategic modification of a known kinase inhibitor. For example in one embodiment, the method enables the discrimination of the drug or inhibitor between Kd for each conformation to be measured. The methods of the invention may also involve the use of the array/library or panel of different kinases as described above. For example, kinase inhibitors may be considered to be useful if they specifically inhibit the kinase of interest, but no others, or inhibit a small group of kinases, or inhibit multiple kinases within the same pathway. Using such an array, library or panel, which gives a clear active/inactive conformation readout allows the rapid detection of those potential inhibitors that have selectivity for a given kinase, and those that have a more global effect on the conformation of kinases.

The present invention allows the determination of the intrinsic propensity of a kinase to occupy the active conformation. As discussed above, the kinase may be a native or wild- type kinase, but also may be a mutant. In some instances it may be considered to be useful to generate a kinase that has an increased propensity for a particular conformation, compared to for example the native or wild-type kinase. In other instances a kinase mutation may arise in naturally (eg a drug-resistant mutant). The present invention allows the assessment of any mutation of the kinase on the equilibrium of active to inactive conformation. Accordingly, the present invention provides a method for generating a mutant kinase that has an intrinsic propensity to occupy the active conformation, wherein the method comprises determining the intrinsic propensity for a kinase to occupy a particular conformation according to the methods of the invention, introducing one or more mutations into said kinase and determining the effects of the mutation on the intrinsic propensity of the kinase to occupy the active conformation, and selecting the mutant kinase or mutant kinases that have a higher intrinsic propensity to occupy the active conformation.

Similarly, the invention provides a method for generating a mutant kinase that has an intrinsic propensity to occupy the inactive conformation, wherein the method comprises determining the intrinsic propensity for a kinase to occupy a particular conformation according to the methods of the invention, introducing one or more mutations into said kinase and determining the effects of the mutation on the intrinsic propensity of the kinase to occupy the inactive conformation, and selecting the mutant kinase or mutant kinases that have a higher intrinsic propensity to occupy the inactive conformation. A kinase occupying an active conformation would be expected to have increased activity, so would be of benefit in a biotechnologicai process. A screen which compared affinities between 'active' and 'inactive' kinases could be used to screen for specific-conformation inhibitors. This kind of screen would be very well suited to high-throughput applications. It will be evident to the skilled person that one or more of the agents disclosed herein, for example the kinase of the invention, reagents required to make the kinase of the invention (for example the labels), or reagents required to perform the methods of the invention, may be provided as part of a kit of parts. For example, the invention provides a kit comprising at least one kinase according to the invention. The kit may comprise 2 or more kinases according to the invention. In addition, or alternatively, the kit may further comprise one or more small molecules, for example inhibitors, activators or other compounds. The kit may also comprise imaging agents or for example cell lysates or other assay buffers

Kit components may include physiological binding partners, for example polypeptides; kinase substrates; mutant kinases (eg drug-resistant mutants arising physiologically or otherwise). As an example a kinase may be pre-immobilised to a surface (for example of a flow cell). Anti-fading reagents (for example trolox, protocatechuic acid, protocatechuate 3,4- dioxygenase) may also or in addition be included.

Compounds that affect the activity of kinases are often used in medicine. For example kinase inhibitors are often used in the treatment of cancer. It will be appreciated therefore that the methods of the invention can be used to identify or select a suitable kinase inhibitor for use in medicine. Accordingly, the invention provides a kinase inhibitor for use in treating a subject in need thereof, wherein the kinase inhibitor has an affinity for the said kinase that is higher than the affinity of known activators, wherein the affinity of the inhibitor and affinity of known activator(s) is determined by the methods of the invention. In one embodiment, the subject in need thereof is suffering from a disease or condition, for example has Down's syndrome or irritable bowel syndrome. In another embodiment the subject in need thereof may be pregnant. As discussed above, the affinity of a particular inhibitor can also be determined in the presence of, for example, a cell lysate. In one embodiment therefore the invention provides a kinase inhibitor for use in treating a subject in need thereof, for example wherein the subject has cancer, wherein the kinase inhibitor shifts the equilibrium of the active and inactive conformation towards the inactive conformation in the presence of a cell lysate, wherein the equilibrium of the active and inactive conformation is determined by the methods of the invention.

The invention also provides a method of treating a subject in need thereof, wherein the method comprises administering a kinase inhibitor, wherein the kinase inhibitor shifts the equilibrium of the active and inactive conformation towards the inactive conformation in the presence of a known activator or a cell lysate, wherein the equilibrium of the active and inactive conformation is determined by the methods of the invention.

The invention also provides the use of a kinase inhibitor in the manufacture of a medicament for use in medicine, for example for use in treating cancer, wherein the kinase inhibitor shifts the equilibrium of the active and inactive conformation towards the inactive conformation in the presence of a known activator or a cell lysate, wherein the equilibrium of the active and inactive conformation is determined by the methods of the invention. For kinases with low or no activity in vitro, the present method has advantages (ie it is possible). In general, it also has advantages over enzymatic activity for example in the instances where the therapeutic aim is to dislodge a binding partner. Eg MLN8237 and CD532 aim to increase the population of T-loop inactive kinase in order to disrupt an Aurora-A / N-Myc interaction. Disruption of the interaction leads to proteolytic degradation of N-Myc (a driver of oncogenesis in small cell lung cancer and neuroblastoma).

The invention also provides a method for identifying a kinase inhibitor suitable for use in treating a disease involving a kinase which it is desirous to inhibit, wherein the method comprises:

a) determining the affinity of the said kinase for an activator; and

b) selecting an inhibitor with a higher affinity for the kinase than the affinity of the activator for the kinase. The methods of the invention may be useful in partitioning a known overall affinity into the affinity for active / inactive conformations. Figure 11 (4), for example, teaches selecting an inhibitor which moves the position of equilibrium to an inactive T-loop conformation rather than just selecting one with a higher affinity than that of the activator.

Methods and uses of the kinase of the invention are considered to have advantages over the prior art, not least because only very small amounts of kinase protein are required to determine the active versus inactive conformation. For example, in one embodiment just 5pmol protein is used per slide in order to determine the active or inactive conformation of single kinase molecules.

References 1. Cohen, P., Protein kinases - the major drug targets of the twenty-first century?

Nature

Reviews Drug Discovery 2002, 1 (4), 309-315.

2. Barnes, P. J., New anti-inflammatory targets for chronic obstructive pulmonary disease. Nat

Rev Drug Discov 2013, 12 (7), 543-59.

3. Wu, P.; Nielsen, T. E.; Clausen, M. H., FDA-approved small-molecule kinase inhibitors. Trends

Pharmacol Sci 2015, 36 (7), 422-39.

4. Meng, Y.; Roux, B., Locking the active conformation of c-Src kinase through the phosphorylation of the activation loop. J Mot Biol 2014, 426 (2), 423-35.

5. Johnson, L. N.; Noble, M. E. M.; Owen, D. J., Active and inactive protein kinases: Structural basis for regulation. Cell 1996, 85 (2), 149-158.

6. Vogtherr, M.; Saxena, K.; Hoelder, S.; Grimme, S.; Betz, M.; Schieborr, U.; Pescatore, B.; Robin, M.; Delarbre, L; Langer, T.; Wendt, K. U.; Schwalbe, H., NMR characterization of kinase p38 dynamics in free and ligand-bound forms. Angewandte Chemie-lntemational Edition 2006, 45 (6),

993-997.

7. Morando, M. A.; Saladino, G.; D'Amelio, N.; Pucheta-Martinez, E.; Lovera, S.; Lelli, M.; Lopez- Mendez, B.; Marenchino, M.; Campos-Olivas, R.; Gervasio, F. L., Conformational Selection and Induced Fit Mechanisms in the Binding of an Anticancer Drug to the c-Src Kinase. Sci Rep 2016, 6,

24439. 8. Bayliss, R.; Sardon, T.; Vernos, !.; Conti, E., Structural basis of Aurora-A activation by TPX2 at the mitotic spindle. Molecular Cell 2003, 12 (4), 851 -862.

9. Dodson, C. A.; Bayliss, R., Activation of Aurora-A Kinase by Protein Partner Binding and

Phosphorylation Are Independent and Synergistic. Journal of Biological Chemistry 2012, 287 (2),

1 150-1 157.

10. Eyers, P. A.; Mailer, J. L, Regulation of Xenopus Aurora A activation by TPX2. J Biol Chem

2004, 279 (10), 9008-15.

1 1 . Anderson, K.; Yang, J.; Koretke, K.; Nurse, K.; Calamari, A. ; Kirkpatrick, R. B.; Patrick, D.; Silva, D.; Tummino, P. J.; Copeland, R. A.; Lai, Z., Binding of TPX2 to Aurora A alters substrate and inhibitor interactions. Biochemistry 2007, 46 (36), 10287-95.

12. Bavetsias, V.; Faisal, A.; Crumpler, S.; Brown, N.; Kosmopoulou, M.; Joshi, A.; Atrash, B.; Perez-Fuertes, Y.; Schmitt, J. A.; Boxall, K. J.; Burke, R.; Sun, C; Avery, S. ;

Bush, K.; Henley, A.; Raynaud, F. I.; Workman, P.; Bayliss, R.; Linardopoulos, S.; Blagg, J., Aurora isoform selectivity: design and synthesis of imidazo[4,5-b]pyridine derivatives as highly selective inhibitors of Aurora-A kinase

in cells. Journal of medicinal chemistry 2013, 56 (22), 9122-35.

13. Dodson, C. A.; Kosmopoulou, M.; Richards, M. W.; Atrash, B.; Bavetsias, V.; Blagg, J.; Bayliss, R., Crystal structure of an Aurora-A mutant that mimics Aurora-B bound to MLN8054: insights into selectivity and drug design. Biochemical Journal 2010, 427, 19- 28.

14. Geuns-Meyer, S.; Cee, V. J.; Deak, H. L; Du, B.; Hodous, B. L; Nguyen, H. N.; Olivieri, P. R.; Schenkel, L. B.; Vaida, K. R.; Andrews, P.; Bak, A.; Be, X.; Beitran, P. J.;

Bush, T. L; Chaves, M. K. ; Chung, G.; Dai, Y.; Eden, P.; Hanestad, K.; Huang, L.; Lin, M. H.; Tang, J.; Ziegler, B.; Radinsky, R.; Kendall, R.; Patel, V. F.; Payton, M., Discovery of N-(4-(3-(2-aminopyrimidin-4-yl)pyridin-2- yloxy)phenyl)-4-(4-methylthiophen-2-yl)p hthalazin-1 -amine (AMG 900), a highly selective, orally bioavailable inhibitor of aurora kinases with activity against multidrug-resistant cancer cell lines. Journal of medicinal chemistry 2015, 58 (13), 5189-207.

15. Gustafson, W. C; Meyerowitz, J. G.; Nekritz, E. A.; Chen, J.; Benes, C; Charron, E.; Simonds, E. F.; Seeger, R.; Matthay, K. K.; Hertz, N. T.; Eilers, M.; Shokat, K. M.; Weiss, W. A., Drugging MYCN through an allosteric transition in Aurora kinase A. Cancer ce// 2014, 26 (3), 414-27. 16. Kollareddy, M.; Zheleva, D.; Dzubak, P.; Brahmkshatriya, P. S.; Lepsik, M.;

Hajduch, M., Aurora kinase inhibitors: progress towards the clinic. Investigational new drugs 2012, 30 (6), 241 1- 32.

17. Lawrence, H. R.; Martin, M. P.; Luo, Y.; Pireddu, R.; Yang, H.; Gevariya, H.; Ozcan, S.; Zhu, J. Y.; Kendig, R.; Rodriguez, M.; Elias, R.; Cheng, J. Q.; Sebti, S. M.; Schonbrunn, E.; Lawrence, N. J., Development of o-Chlorophenyl Substituted Pyrimidines as Exceptionally Potent Aurora Kinase Inhibitors. Journal of medicinal chemistry 2012, 55 (17), 7392-416.

18. Manfredi, M. G.; Ecsedy, J. A.; Meetze, K. A.; Balani, S. K.; Burenkova, O.; Chen, W.; Galvin, K. M.; Hoar, K. M.; Huck, J. J.; LeRoy, P. J.; Ray, E. T.; Sells, T. B.; Stringer, B.; Stroud, S. G.; Vos, T. J.; Weatherhead, G. S.; Wysong, D. R.; Zhang, M. K.; Bolen, J. B.; Claiborne, C. F., Antitumor activity of MLN8054, an orally active small-molecule inhibitor of Aurora A kinase. Proceedings of the National Academy of Sciences of the United States of America 2007, 704 (10), 4106-4111.

19. Richards, M. W.; Burgess, S. G.; Poon, E.; Carstensen, A.; Eilers, M.; Chesler, L.; Bayliss, R., Structural basis of N-Myc binding by Aurora-A and its destabilization by kinase inhibitors. Proc Natl Acad Sci U S A 2016.

20. Brockmann, M.; Poon, E.; Berry, T.; Carstensen, A.; Deubzer, H. E.; Rycak, L.; Jamin, Y.; Thway, K.; Robinson, S. P.; Roels, F.; Witt, O.; Fischer, M.; Chesler, L.; Eilers,

M., Small molecule inhibitors of aurora-a induce proteasomal degradation of N-myc in childhood neuroblastoma. Cancer cell 2013, 24 (1 ), 75-89.

21. DuBois, S. G.; Marachelian, A.; Fox, E.; Kudgus, R. A.; Reid, J. M.; Groshen, S.; Malvar, J.; Bagatell, R.; Wagner, L.; Maris, J. M.; Hawkins, R.; Courtier, J.; Lai, H.; Goodarzian, F.; Shimada, H.; Czarnecki, S.; Tsao-Wei, D.; Matthay, K. K.; Mosse, Y. P., Phase I Study of the Aurora A Kinase Inhibitor Alisertib in Combination With Irinotecan and Temozolomide for Patients With Relapsed or Refractory Neuroblastoma: A NANT (New Approaches to Neuroblastoma Therapy) Trial. J Clin Oncol

2016, 34 (12), 1368-75.

22. Burgess, S. G.; Bayliss, R., The structure of C290A:C393A Aurora A provides structural insights into kinase regulation. Acta Crystallographica Section F-Structural Biology Communications

2015, 71, 315-319.

23. Rowan, F. C; Richards, M.; Bibby, R. A.; Thompson, A.; Bayliss, R.; Blagg, J., Insights into

Aurora-A Kinase Activation Using Unnatural Amino Acids Incorporated by Chemical Modification. Acs Chemical Biology 2013, 8 (10), 2184-2191.

24. Zhou, R. B.; Kunzelmann, S.; Webb, M. R.; Ha, T., Detecting Intramolecular Conformational

Dynamics of Single Molecules in Short Distance Range with Subnanometer Sensitivity. Nano Letters

2011 , 11 (12), 5482-5488.

25. Ha, T.; Tinnefeld, P., Photophysics of Fluorescent Probes for Single-Molecule Biophysics and Super-Resolution Imaging. In Annual Review of Physical Chemistry, Vol 63, Johnson, M. A.; Martinez, T. J., Eds. Annual Reviews: Palo Alto, 2012; Vol. 63, pp 595- 617.

26. Schuster, J.; Brabandt, J.; von Borczyskowski, C, Discrimination of photoblinking and photobleaching on the single molecule level. J Lumin 2007, 127 (1), 224-229.

27. Mutch, S. A.; Fujimoto, B. S.; Kuyper, C. L; Kuo, J. S.; Bajjalieh, S. M.; Chiu, D. T., Deconvolving single-molecule intensity distributions for quantitative microscopy measurements. Biophys J 2007, 92 (8), 2926-43.

28. Agafonov, R. V.; Wilson, C; Otten, R.; Buosi, V.; Kern, D., Energetic dissection of Gleevec's selectivity toward human tyrosine kinases. Nature Structural & Molecular Biology 2014, 21 (10), 848-

853.

29. Wilson, C; Agafonov, R. V.; Hoemberger, M.; Kutter, S.; Zorba, A.; Halpin, J.; Buosi, V.; Otten, R.; Waterman, D.; Theobald, D. L; Kern, D., Using ancient protein kinases to unravel a modern

cancer drug's mechanism. Science 2015, 347 (6224), 882-886.

30. Dar, A. C; Lopez, M. S.; Shokat, K. M., Small molecule recognition of c-Src via the Imatinib- binding conformation. Chem Biol 2008, 15 (10), 1015-22.

31. Seeliger, M. A.; Ranjitkar, P.; Kasap, C; Shan, Y.; Shaw, D. E.; Shah, N. P.; Kuriyan, J.; Maly, D. J., Equally potent inhibition of c-Src and Abl by compounds that recognize inactive kinase conformations. Cancer Res 2009, 69 (6), 2384-92.

32. Aleksandrov, A.; Simonson, T., Molecular dynamics simulations show that conformational selection governs the binding preferences of imatinib for several tyrosine kinases. J Biol Chem 2010,

285 (18), 13807-15.

33. Lin, Y. L.; Meng, Y.; Jiang, W.; Roux, B., Explaining why Gleevec is a specific and potent inhibitor of Abl kinase. Proc Natl Acad Sci U S A 2013, 110 (5), 1664-9.

34. Lovera, S.; Sutto, L.; Boubeva, R.; Scapozza, L.; Dolker, N.; Gervasio, F. L., The different flexibility of c-Src and c-Abl kinases regulates the accessibility of a druggable inactive conformation. J Am Chem Soc 2012, 134 (5), 2496-9.

35. Xiao, Y.; Lee, T.; Latham, M. P.; Warner, L. R.; Tanimoto, A.; Pardi, A.; Ahn, N. G., Phosphorylation releases constraints to domain motion in ERK2. Proc Natl Acad Sci U S A 2014 111 (7), 2506-1 1.

36. Lamichhane, R.; Solem, A.; Black, W.; Rueda, D., Single-molecule FRET of protein- nucleic acid and protein-protein complexes: surface passivation and immobilization. Methods 2010, 52 (2), 192-

200.

37. Heron, N. M.; Anderson, M. ; Blowers, D. P.; Breed, J.; Eden, J. M.; Green, S.; Hill, G. B.; Johnson, T.; Jung, F. H.; McMiken, H. H.; Mortlock, A. A.; Pannifer, A. D.; Pauptit, R. A.; Pink, J.; Roberts, N. J.; Rowsell, S., SAR and inhibitor complex structure determination of a novel class of potent and specific Aurora kinase inhibitors. Bioorg Med Chem Lett 2006, 16 (5), 1320-3.

Figure legends

Figure 1 - a) Current model of kinase activation: phosphorylation on the activation loop locks the kinase in an active T-loop conformation; b) Current equilibrium model of type II inhibitor binding: active apo kinase is in equilibrium with inhibited kinase in an inactive T- loop conformation; c) Proposed equilibrium model showing TMR-labelling sites (stars) and expected fluorescence signal; d) Conformational change in and labelling sites in Aurora- A. i) Active T-loop conformation, ii) Inactive T-loop conformation. Activation loop (DFG- PPE) shown in cartoon representation and dark grey. Labelling sites used in Examples shown as filled spheres. ADP shown in stick representation in (i). Distances between labelling sites in different T-loop conformations indicated by labelled arrows, e) Kinase activity assay for unlabeled pseudo-wildtype Aurora-A (open squares) and TMR labelled K224C/S283C (filled triangles). Activity shown as [ADP] produced over course of 1 hr reaction. The difference in activity between pseudo-wildtype and labelled protein cannot be accounted for by incomplete protein labelling (labelling efficiency in Example 6). f) Example trace of double-labelled K224C/S283C Aurora-A single molecule showing background-subtracted fluorescence intensity over time.

Figure 2 - Fluorescence intensity distribution for phosphorylated TMR-labelled K224C/S283C Aurora-A. a) Intensity histogram of unliganded Aurora-A; b) Dwell time histogram of quenched, inactive T-loop, conformation; c-k) Fluorescence intensity distributions with c) 1 mM ATP; d) 3 mM kemptide; e) 1 mM AMP-PNP; f) 1 mM AMP- PNP and 3 mM kemptide; g) 5 μΜ TPX2; h) 5 μΜ TPX2, 1 mM AMPPNP and 3 mM Kemptide; i) 10 μΜ MLN8054; j) 10 μΜ CD532; k) 5 μΜ TPX2 + 10 μΜ MLN8054; I) Summary of conformational preferences of Aurora-A under different conditions. Error bars show propagated fitting error.

Figure 3 - Schemes of ligand residency time

Figure 4 - Occupancy of Aurora-A conformations under different conditions.

ND Not determined.

a Error of loop occupancy is propagated from the fitting error of histograms and is≤ 2. b Keq = [Inactive T-loop]/[Active T-loop]. Error on Keq is propagated from the fitting error of histograms and is <0.1.

c ΔG inactive-active = -RTIn(Keq). Error on ΔG inactive-active is propagated from fitting error of histograms and is < 0.1 kcal moM .

n Km measured in this study (Figure 8). Error on value is ± 70 μΜ.

Figure 5 - Fitted parameters from intensity histograms

aThe difference of -500 a.u. between the modes of S283C and K224C is likely to be due to the differing molecular environments of each dye molecule within the structure of the protein.

"Error is fitting error (width) or error propagated from this (calculated mode).

Figure 6 - Naming conventions for different conformations of Aurora-A.

a) Active T-loop (active kinase; left; PDB 10L5) and inactive T-loop (inactive kinase; right;

PDB 2WTV) conformations, b) DFG-in (active kinase bound to ADP; white; PDB 10L5) and DFG-out (inactive kinase bound to compound 13 37; black, PDB 2C6E) conformations, c) Active T-loop (active kinase bound to ADP; white; PDB 10L5) and inactive T-loop conformations (inactive kinase bound to MLN8054; black; PDB 2WTV), and the non- conventional conformation of the DFG motif of aurora-A kinase bound to MLN8054 (black, inset) in which the aspartic acid residue is orientated in the same direction as the phenylalanine residue (DFG-up), compared to the DFG-in conformation of aurora-A binding TPX2 (white), d) The inactive T-loop (aurora- A kinase bound to CD532 inhibitor; black; PDB 4J8M) and active T-loop (active kinase bound to ADP; white; PDB 10L5) conformations, and the DFG-in (black; inset) conformation of the DFG motif of aurora-A bound to CD532, which is aligned in the same conformation as that of active aurora-A kinase (white).

Figure 7 - PEG labelling reaction. M - marker (sizes indicated on left); '-' - Aurora-A K224C/S283C control; '+' - Aurora-A K224C/S283C reacted with 10x molar excess of PEG maleimide. A maximum of two AurA-PEG bands are observed, indicating that only two thiol groups are available for reaction with maleimide.

Figure 8 - Peptide Km for double labelled Aurora-A. Activity shown as [ADP] produced over the course of a 1 hr reaction.

Figure 9 - Flow cell and tethering strategy, a) Cartoon representation of the conjugate chain of linkers anchoring Aurora-A to the internal surface of a coverslip, and the angle of the incident laser light, b) Construction of the flow cell used to image the fluorescent molecules. In this image, the glass slide is at the rear and the coverslip at the front.

Figure 10 - Fluorescence intensity histograms for single labelled protein a) S283C; b) K224C; c) S283C in the presence of 10 μΜ MLN8054 d) S283C in the presence of 10 μΜ CD532.

Figure 11 - Potential outcomes of combining activators and inhibitors. Opacity of image indicates relative concentration of each species. 1) A protein kinase is in equilibrium between inactive and active T-loop conformations; 2) Adding an activator moves the equilibrium to the right and increases the population of the active T-loop conformation; 3)- 5) Adding an inhibitor will have different effects depending on the relative affinities of the inhibitor and activator for the different T-loop conformations of the kinase; 3) K_d. inhibitor, inactive

^≈ Kd, inhibitor, active. The inhibitor has a similar affinity for active and inactive conformations of the kinase. It does not move the position of equilibrium and forms a triple complex in active and inactive conformations. The overall population of active and inactive T-loop conformations is as for kinase with activator alone; 4) K_d, inhibitor, inactive < K_d, inhibitor, active; K_d, inhibitor, inactive ~ Kd, activator, active The inhibitor binds more strongly to the inactive T-loop conformation. It binds the inactive T-loop conformation about as strongly as the activator binds the active T-loop conformation. The inhibitor increases the population of the inactive T-loop conformation. Some of these kinase molecules will be bound to inhibitor alone, others will be bound to inhibitor and activator in a triple complex (depending on the relative affinities of the activator and inhibitor for the inactive T-loop conformation). There is still a population of kinase in the active T-loop conformation. The majority of this is bound to activator, some is also bound to inhibitor in a triple complex. There may be a small population of free activator in solution depending on its overall affinity for the kinase; 5) K_d,

overall. The inhibitor binds very strongly to the inactive T-loop conformation and barely binds the active conformation at all. It binds the inactive T-loop conformation more strongly than the activator binds the active conformation. It thus moves the position of equilibrium to the left and increases the population of the inactive T-loop conformation. The activator has a weak affinity for the inactive T-loop conformation and so only a small proportion of activator binds the kinase. The weaker the affinity of the activator for the inactive conformation, the less triple complex is formed in the inactive T-loop conformation and the greater the population of free activator in solution. Depending on the relative affinities of the inhibitor for the inactive T-loop conformation and the activator for the active T-loop conformation, a small population of activator-bound kinase may still be present.

Figure 12 - Effect of small x-axis translations on quality of data fit and reported peak areas. a) Data fits of translated idealized data. We modelled an idealized low intensity data peak using parameters based on our experimental results for CD532 (filled circles). Data was translated along the x-axis by amounts indicated in figure legend and fitted to equation (19). b) Area under each fitted curve in (a), relative to the original curve. Negative x-axis values illustrate effect of curve truncation.

Figure 13 - a) An 'inverse' data set using (S283C/M373C) where we expect a fluorescence signal for the inactive conformation and a quenched signal for the active conformation. There is 26% inactive T-loop conformation and the equilibrium constant (K_eq) is 0.3 (compared with 23% and 0.3 for K224C/S283C). b) The original construct (K224C/S283C), but coexpressed with lambda phosphatase to produce unphosphorylated protein (lambda phosphatase is removed during the protein purification process). There is 51 % inactive T- loop conformation and Keq = 1.1.

The K224C/S283C construct gives a quenched fluorescence when the two dye molecules are close together (T-loop inactive conformation) and a high fluorescence signal when they are far apart (T-loop active conformation). We interpret this to mean that they are two conformations of the kinase (T-loop active and inactive), but formally our experiments do not differentiate between potential multiple conformations when the dyes are apart from one another (eg T-loop active conformation and an intermediate conformation where the T-loop is neither in the active nor inactive conformations). In order to answer this question, we designed a construct where one dye was placed on the C-lobe of the kinase (M373C) and one on the activation loop, such that when the kinase was in the T-loop active conformation we would expect the fluorescence signal to be quenched. If only two conformations of the T-loop were adopted (active and inactive), then this construct would give exactly opposite results to the K224C/S283C construct. If additional intermediate conformations were adopted, the quenched population in 373C/S283C would be smaller than the high fluorescence population in K224C/S283C, and the difference between them equal to the population of the intermediate conformation. Our experiment measured a quenched population in M373C/S283C equal in size to the high fluorescence population of K224C/S283C, indicating that only two conformations are present (T-loop active and T-loop inactive).

In order to test the effect of phosphorylation on the conformation of the kinase T-loop, we carried out our experiments on labelled K224C/S283C kinase which had been co- expressed with lambda phosphatase. The phosphatase is removed in the protein purification process, but effectively prevents autophosphory!ation of Aurora-A while still being expressed in E.coli (ref Dodson, Yeoh et al, 2013, Sci Signal). In the absence of ATP, unphosphorylated Aurora-A is stable. Our experiments showed that the population of the T-loop inactive conformation is greater in the unphosphorylated kinase, consistent with phosphorylation increasing the activity of the kinase.

Figure 14 - Kinases adopt 2 major conformations

a) For over 20 years we have known that protein kinases adopt two major conformations. Despite billions of dollars of drug discovery in a class of molecules pivotal to cell function, almost nothing is known about the process of interconversion between different conformations of protein kinases because prior to the present invention this was inaccessible to experimental methods, b) The present inventors have developed an assay that can monitor kinase conformation using fluorescence. Dyes, or labels, are covalently attached to kinase molecules. In one embodiment one dye/label is on the activation loop, and one dye/label is on the N-lobe. The assay is suitable for TIRF microscopy and i) directly measures the relative populations of each conformation; and ii) directly measures the timescale of interconversion. The measurements give fluorescence intensity of single kinase molecules over a defined time period. The measurements are considered to end when the dyes photobleach, which when using two flurophores as the labels is a two-step process. Prior to photobleaching the dyes fluctuate between high and quenched fluorescence (active and inactive conformations). This provides the first measurements observing individual molecules of any kinase interconverting between two conformations, and the first evidence that the DFG-out conformation is transiently accessed by uninhibited kinase.

Figure 15 - Table detailing advantages of the present invention over prior art methods. Figure 16 - Fluorescence v distance between dyes for a set of model compounds.

Examples Example 1 - labelling of the kinase

As discussed above, the influence of phosphorylation and interaction of small molecule inhibitors with kinase conformation can be summarized by two models (Figure 1 a and 1 b).

In order to test these models directly, we used single molecule fluorescence spectroscopy to monitor the conformation of the kinase activation loop. We implemented our assay in Aurora- A kinase, a mitotic kinase whose catalytic activity can be increased in the presence of its protein binding partner TPX2 8-1 1 and inhibited with numerous drug-like small molecules 11-18. Interest in Aurora-A and its conformational plasticity has recently increased due to the discovery that inhibiting Aurora-A with MLN8054 or CD532, both inactive-conformation inhibitors, disrupts the Aurora-A-N- Myc interaction, leading to degradation of /V-Myc and offering an alternative therapeutic strategy for /V-Myc-driven tumors that is currently under clinical evaluation with MLN8237 (alisertib) ^{15, 1 9-21}. These factors make Aurora-A an ideal clinically-relevant system in which to study the heterogeneity and dynamics of T- loop conformation which underlie the observed structural and catalytic properties of protein kinases. However, it is considered that the inventive model devised by the inventors is applicable to all kinases and using this method detailed information about the conformational dynamics of any kinase can be obtained.

To aid clarity in this document, we use the terminology 'active T-loop' and 'inactive T-loop' to refer to the two orientations of the activation loop (Figure 1 c&d and Figure 6a). We reserve the terms DFG-in and DFG-out to discuss the detailed orientation of the DFG motif (Figure 6b). We labelled a pseudo-wildtype construct of Aurora-A (C290A/C393A) 22-23 _wjt TMR on cysteine residues introduced into the activation loop (S283C) and the N-lobe of the kinase (K224C; Figure 1 d). The fluorescence of TMR is quenched when two molecules are closer than ~15 A and this phenomenon has previously been used to probe small distance changes at the single molecule level 24 Control reactions confirmed that only two sites were available for coupling (Figure 7).

TMR-labelled phosphorylated Aurora-A is catalytically active (Figure 1 e and Figure 8) with an activity similar to unlabeled pseudo-wildtype protein.

-Example 2 - The majority of phosphorylated Aurora-A occupies an active T-loop conformation We immobilized labelled protein on a cover slip using its N-terminal His-tag (Figure 9) and imaged fluorescence intensity using total internal reflection fluorescence (TIRF) microscopy. Single molecule fluorescence traces originating from double-labelled Aurora-

A molecules were identified by the presence of two-step photobleaching at their end ²⁵. Before photobleaching, double-labelled molecules exhibited a single high fluorescence intensity and transiently entered a low intensity, quenched, state (example trace in Figure 1f). The quenched state was never observed after the first photobleaching event, nor in control measurements on single-labelled protein (K224C and S283C), and thus was not due to photophysical dye blinking 26. Photobleaching occurring directly from the quenched state resulted in an increase in fluorescence to that expected for a single dye molecule. We therefore assigned changes in our fluorescence signal to changes in the position of the Aurora-A T-loop: high fluorescence indicating an active T-loop, quenched fluorescence an inactive T-loop.

Our measured fluorescence intensity histogram shows two peaks indicating that phosphorylated Aurora-A occupies both active and inactive T-loop conformations (Figure 2a). Consistent with the distributions measured for single labelled controls (Figure 10 a&b) and observed elsewhere in the literature ²⁷, our data fit to the sum of two log normal distributions.

The relative areas of the peaks indicate that in solution the majority of phosphorylated Aurora-A adopts an active T-loop conformation, with a minority proportion (23% ± 2%) in an inactive T-loop conformation (Figure 4). This is consistent with the active T-loop conformation observed in x-ray structures of uninhibited kinase and with the high catalytic activity of phosphorylated Aurora-A 9. Contrary to the model of a locked phosphorylated T-loop, our data indicate that the activation loop of phosphorylated Aurora-A exists in a dynamic structural equilibrium with an equilibrium constant Keq of

0.3 + 0.1.

It has been suggested that the free energy penalty of interconverting between active and inactive conformations of Src and Abl underlies the selectivity of type II inhibitors such as Imatinib, although this has been challenged 28-31 Consequently there have been efforts to calculate the free energy difference between these conformations using a range of computational techniques. None of these studies explicitly distinguishes between the orientation of the DFG-motif and the overall orientation of the T-loop (which are usually assumed to be coupled), however examination of the protocols suggests that they report on the interconversion between DFG-in active T-loop and DFG-out inactive T-loop conformations 32-34 Our experimental measurements report directly on the interconversion between active and inactive T-loop conformations (without measuring the orientation of the DFG- motif), and the free energy difference we obtain for phosphorylated

Aurora-A from our measured Keq is 0.7 ± 0.1 kcal mol^-1 (Δ G inactive-active) . This is very similar to the value of 0.8 kcal mol^{- 1} measured very recently for the two unassigned conformations of phosphorylated Erk2 35 and considerably less than the calculated values for Src ³² , ³⁴.

We measured the microscopic rate constant for adopting an active T-loop conformation, kactive, by constructing a histogram of the duration of each quenching event (Figure 2b). The data fitted well to a single exponential and yielded a value of 2.3 ± 0.1 s-1. We were unable to measure kinactive directly as the observation time of individual molecules was limited by photobleaching, but we calculated this to be 0.7 ± 0.2 s^{- 1} from Keq.

•Example 3 - Influence of kinase ligands

To determine how kinase substrates influence the conformation of the activation loop, we measured fluorescence intensity distributions of Aurora-A in the presence of saturating levels of ATP, kemptide (7-residue peptide substrate), and AMP-PNP (Figure 2c-f). Neither kemptide nor AMP-PNP changed the position of equilibrium from that of apo kinase while, surprisingly, ATP alone and AMP-PNP- kemptide both slightly increased the population of the inactive T-loop conformation (Figure 4). We next measured the intensity distribution of Aurora-A in the presence of saturating levels of TPX2 (Figure 2g). Occupancy of the active T-loop conformation was increased (Figure 4), consistent with the increased catalytic activity of the enzyme. In order to build a picture of the enzyme poised for maximal activity, we measured the distribution of the Aurora-A- TPX2 complex bound to AMP-PNP and kemptide (Figure 2h). This adopted a predominantly active T-loop conformation, similar to that of the Aurora-A-TPX2 complex alone.

MLN8054 and CD532 are both nanomolar inhibitors of Aurora-A, and x-ray structures show that each binds in an inactive T-loop conformation (Figure 6c and 6d) ^{1 3, 1 5, 1 8}. Although both are referred to as type II inhibitors, neither extends into the allosteric hydrophobic pocket, and neither is captured binding the kinase with a canonical DFG-out motif. Aurora-

A bound to MLN8054 adopts an unusual DFG conformation previously termed DFG-up ^{1 3}, and Aurora-A bound to CD532 is DFG-in ⁵.

In order to determine the effect of these inhibitors on the activation loop of Aurora-A in solution, we repeated our assay in the presence of each (Figure 2i&j). Both inhibitors resulted in a large increase in the population of the inactive T-loop conformation, consistent with crystal structures (Figure 4). In order to control for possible quenching of TMR by high concentrations of inhibitor, we carried out additional experiments with single labelled protein (Figure 4 and 10) and, as a final verification, counted the number of fluorescent molecules observed within each control (35 ± 6 kinase alone, 34 ± 5 with MLN8054; n=12).

TPX2 and MLN8054 increase the population of opposing T-loop conformations. In order to determine how these binding partners together influence Aurora-A, we measured the conformation of the activation loop in the presence of both. Kinetic studies have shown that the presence of TPX2 increases the Ki of MLN8054 more than 4-fold ^{1 3} and similar changes have been observed for VX680 and GSK623906A ^{1 1}. In the presence of MLN8054 and TPX2 together, Aurora-A adopts a predominantly active T-loop conformation, similar to that for TPX2 alone (Figure 2k and Figure 4).

In order to determine whether this represents a mix of binary Aurora-A-MLN8054 and Aurora-A-TPX2 complexes or a population of an Aurora-A-TPX2-MLN8054 triple complex, we calculated the expected experimental result for the mix of binary complexes based on the published affinities of the two ligands ^{9, 1 3} (Example 5). Our calculation indicated that a mix of binary complexes would result in a measured inactive T-loop population of 43%, inconsistent with the experimental result (15%, Figure 4). We therefore concluded that our experimental samples contained a triple complex of Aurora-A-TPX2- MLN8054. We are unable us to quantify the extent of triple complex formation, but to account for the experimental results it must be the majority species and must occupy a majority active T-loop conformation.

All our measurements indicate that phosphorylated Aurora-A is not locked into a single conformation, and that it spontaneously interconverts between active and inactive T-loop conformations in solution. This includes Aurora-A in the presence of saturating quantities of each of the small molecule inhibitors MLN8054 and CD532 and Aurora-A with TPX2. In order to explain our results, we hypothesized that either the ligand residence time is short (and rearrangement of the activation loop occurred in the unliganded kinase; eg Scheme 1 of Figure 3) or that Aurora-A bound to saturating quantities of ligand can interconvert between T-loop conformations.

We investigated ligand residence time by using Keq of the kinase alone and the known ligand Kd to predict the observed Keq in the presence of ligand. Figure 3A, an equivalent scheme for TPX2, and Figures 3B and 3C all predicted values of Keq which were 2-4 orders of magnitude different from those measured (Example 5). We therefore concluded that ligand-bound Aurora-A Interconverts between T-loop conformations and modelled our data using Figure 3D, a cyclical equilibrium in which ligands bind both active and inactive T-loop conformations of Aurora-A, and T- loop interconversion occurs in both free and liganded kinase molecules. This surprising conclusion is supported by x-ray crystallography: three PDB structures (2WTV, 3H10 and 2X81) show Aurora bound to MLN8054 or a similar compound in an inactive T-loop conformation, while in a fourth low resolution structure (2WTW) the crystal packing is incompatible with an inactive T-loop and the MLN8054-bound kinase adopts an active T-loop conformation ¹³. All four Aurora-MLN8054 structures superpose within the N and C lobes of the kinase, varying only in the orientation of the activation loop and in the exact angle between the two lobes. Crucially, the position of MLN8054 relative to the N lobe is identical across all structures, implying that no change in the binding mode of the inhibitor is required to interconvert between active and inactive T-loop conformations. We propose that 2WTW and 2WTV represent two extreme cases of the Aurora-A- MLN8054 conformational ensemble. Interconversion between these two conformations is brought about by movement of the activation loop without the inhibitor leaving the binding site. In order to partition the measured Ki values for MLN8054 into separate dissociation constants for active and inactive T-loop conformations we derived expressions for the different binding reactions and equilibria (see Example 5). Under our experimental conditions, Keq nh (the equilibrium constant for Aurora-A bound to MLN8054) is numerically equal to the measured Keq in the presence of inhibitor and the conformation- specific dissociation constants are thus 1.0 nM (Kd, active) and 0.4 nM (Kdjnactive). In the presence of TPX2, the conformation-specific dissociation constants of MLN8054 are equal and remain unchanged (within the limits of experimental accuracy) at 3.3 nM. These results show that the reported change in Ki of MLN8054 on the addition of TPX2 13 reflects a true change in the Kd of MLN8054, and cannot be accounted for by changes in the proportion of active / inactive T-loop conformers alone.

In addition to providing absolute values of conformation-specific dissociation constants, equation (18) (Example 5) provides an insight into the drivers of inactive / active T-loop conformations. The position of equilibrium between these two conformations for any binding partner depends on the ratio Kd,active I Kdjnactive and the intrinsic position of equilibrium for the kinase Keq, free (constant for each protein). In other words, the proportion of molecules in an inactive T-loop conformation is driven by the degree to which an inhibitor can discriminate between inactive and active T-loop conformations, not solely by the net overall inhibitor-kinase affinity.

Example 4 - Conclusions

Our reported measurements derive from in vitro experiments. While we anticipate that the exact values we have determined may change in the cellular environment, we see no reason why the principle of conformational equilibria and the models of inhibitor binding that we have established will not translate into cell-based contexts. Our results thus have a number of consequences for drug discovery.

Firstly, we have shown that in addition to binding the inactive T-loop conformation as observed in x- ray structures, at least two Aurora-A inhibitors bind a conformation of the kinase (the active T-loop conformation) not previously considered. It is now possible to measure (and thus develop validated prediction algorithms for) the effect of an inhibitor on kinase conformational equilibria in solution. Secondly, potent conformation-independent inhibitors will need to bind both active and inactive T-loop conformations of the kinase and structure-based drug design may need to focus on common structural features of these. Thirdly, design of inactive T-loop inhibitors should aim to maximize differential binding to the inactive T-loop conformation. This is particularly important for inhibitors such as CD532 which aim to induce a specific conformation of the kinase in order to disrupt a physiological interaction (eg Aurora-A with N -Myc). While such inhibitors retain affinity for the active T- loop conformation (ie achieve less than 100% population of the inactive conformation), we expect that a small proportion of the complex (eg Aurora-A-CD532-/V-Myc) in the active T-loop conformation will always be present, even at saturating concentrations of inhibitor. Fourthly, for those kinases where the intrinsic position of equilibrium lies far below 1 (Keg, free 1 ), an inactive T-loop inhibitor must achieve greater discrimination of Kd values between active and inactive T-loop conformations than an inhibitor for a kinase where Keg, free≥ 1. We therefore predict that some inhibitors may appear to be inactive T-loop inhibitors when bound to one target and active T-loop inhibitors when bound to another. When comparing the effect of an inhibitor on two kinases for which it has equal overall affinity, differential effects on kinase conformation may contribute to cellular phenotypes. We predict that kinase phosphorylation state will also affect the value of Keg, free. A fifth consequence of this work is the discovery that modification of the target protein (eg by binding a physiological protein partner such as TPX2) can change the binding affinity of an inhibitor beyond that which would be predicted solely from changing the position of conformational equilibrium. Distinguishing between allosteric partners such as TPX2 and purely scaffolding partners may potentially be possible from x-ray structures, from enzyme activity assays or even by inference from physiological function (eg catalytic activation versus substrate recruitment). This outcome means that early decisions over the best form of the enzyme to target are still important.

Sixthly, success in allosteric disruption using an ATP-competitive small molecule to displace a conformation-specific physiological ligand will depend on the conformational discrimination of the protein-protein interaction to be disrupted (Figure 11 ). By carefully matching inhibitors and interactions targeted for disruption, it may potentially be possible to achieve specificity between different binding partners of the same kinase. In practice, the limits of this approach will need to be found experimentally. Example 5 - Reaction scheme modelling

Calculation of the expected population of inactive T-loop conformation in the presence of two ligands simultaneously

In order to determine whether the binding of activator and inhibitor ligands is mutually exclusive, we can simplify the system under consideration to the following scheme:

Expressions for the dissociation constants of TPX2 and MLN8054 are as follows:

where [AurA] is the concentration of free Aurora-A, [TPX2] the concentration of free TPX2, [MLN] the concentration of free MLN8054, and [AurA-TPX2] and [AurA-MLN] are the concentrations of Aurora-A bound to TPX2 and MLN8054 respectively. Rearranging (1) for [AurA] and substituting into (2) gives the following expression:

Since our experiments were carried out under conditions where both ligands are present in vast excess, [TPX2] and [MLN] can be approximated by the total concentration of each Iigand. This enables a numerical value for the ratio of ligated Aurora-A molecules to be calculated as 0.015. Weighting this value by the measured percentages of inactive T-loop conformation for each Iigand individually gives the expected population of the inactive T- loop conformation in a binding model where each Aurora-A molecule binds only one Iigand at a time. A simple model for inhibitor binding (Scheme 1 , main text)

We can write down expressions for the dissociation constant of inhibitor from the inactive T-loop conformation and for the overall equilibrium constant as follows:

where K_eq is the equilibrium constant for the complete system, K_eq* is the equilibrium constant is the dissociation constant for inhibitor from the

inactive T-loop conformation.

The measured dissociation constant,

can be expressed as

and rearranged to give an expression for Kd, inn, inactive.

Using experimental values of K_eq* = 0.3 (Table I, apo conditions),

nM and [MLN8054] = 10 μΜ we obtain calculated values of nM and K_eq = 30,000

- vastly different to the experimental value K_eq = 0.7.

A simple model for TPX2 binding

From a scheme equivalent to scheme 1 in the main text, where TPX2 binds the active T- loop conformation and active T-loop remains in equilibrium with inactive T-loop, we can derive expressions as follows:

Using experimental values of and [TPX2] = 5 μΜ, gives

calculated values of nM and K_eq - 0.0004 - again vastly different to the

experimental value K_eq = 0.2.

Induced fit (Scheme 2) and conformational selection (Scheme 3) following conformational interconversion

Expressions for K_d and for K_eq can be written down / derived in a similar manner to equations (4), (5) and (6). Experimental values can then be compared with those calculated. Binding reactions in full thermodynamic cycle (Scheme 4, main text)

The conformational equilibrium constants can be defined as follows:

where [Inactive] and [Active] are the concentrations of unbound kinase in the indicated T- loop conformations, and [Inactive-inh] and [Active-inh] are the concentrations of inhibitor- bound kinase in inactive and active T-loop conformations respectively.

Likewise, the dissociations constants for inhibitor dissociation from inactive and active T- loop conformations are:

The measured dissociation constant ,K_d,is

Substituting into (13) and (14) from (9), (10) and (12) gives

which can be rearranged and substituted once more to express in terms of

experimental parameters:

Equation (1 1) can also be rearranged and substituted into from (9), (10) and (15) to give

Rearranging equations (15) and (18) gives expressions for enabling

both quantities to be calculated. Similar relationships hold when the enzyme is bound to TPX2 throughout. Example 6 - Materials and methods

Buffers

Labelling buffer: 50 mM TRIS-HCI pH 7.5, 200 mM NaCI, 5 mM MgC . 10% glycerol Kinase buffer: 50 mM TRIS-HCI pH7.5, 200 mM NaCI, 5 mM MgC , 10% glycerol, 1 mM DTT Imaging buffer: 0.3 mg/mL BSA, 50 mM TRIS-HCI pH7.5, 200 mM NaCI, 5 mM MgCb, 10% glycerol, 5 mM protocatechuic acid, 0.1 μΜ protocatechuate 3,4-dioxygenase, 1 % DMSO and 5 mM Trolox

Inhibitors

MLN8054 was purchased from Selleck Chemicals. CD532 was prepared according to the literature method ¹⁵.

Protein expression

S283C, K224C and S283C/K224C point mutations were generated in His-tagged C290A/C393A Aurora-A kinase domain (residues 122-403) using QuikChange. Like wildtype Aurora-A, C290A/C393A autophosphorylates in E.coli, and is purified pre- phosphorylated on Thr288 ^2Z. An unstructured linker of 24 residues separates the His-tag from the N-terminus of the kinase domain and ensures free rotation of the immobilized kinase in solution. The linker sequence is as follows: MHHHHHHSSGLVPRGSGMKETAAAKFEENLYFQGA. All proteins were expressed in E.coli and purified as previously described ⁹.

TMR labelling of surface cysteine residues of Aurora-A

A 200 pL sample of Aurora-A was buffer exchanged into labelling buffer using a 5 mL desalting column. 10 mM of 5'-tetramethylrhodamine iodoacetemide (TMRIA) in DMSO was added to the eluent protein solution at a molar ratio of 1 :15 protein TMRIA for single mutant Aurora-A samples (K224C, S283C) and a ratio of 1 :20 protein:TMRIA for double mutant Aurora-A (S283C/K224C) and incubated at 4°C on rollers in the dark overnight. The reaction was quenched with 1 M DTT at a ratio of at least 10: 1 DTT MRI A, the sample concentrated and unreacted TMRIA removed using a 10 mL desalting column equilibrated with labelling buffer supplemented with 1 mM DTT. Labelling efficiency was determined by the ratio of total protein to total dye (total protein and dye quantified by absorbance at 280 nm and 514 nm respectively) assuming random labelling. Samples were frozen at -80°C for future use.

The labelling efficiency of K224C/S283C was 120% labelling overall. Assuming 60% labelling at each site, this results in 36% double labelled, 48% single labelled and 16% unlabelled protein.

Kinase activity assay Aurora-A kinase activity was determined using the ADP-Glo™ Kinase Assay kit (Promega), following the manufacturer's instructions. Briefly, 25 μΙ kinase reactions were carried out in 96 -we 11 plates for 1 hour at room temperature in Kinase buffer. The reaction was stopped and the remaining ATP depleted by addition of 25 μΙ ADP-Glo reagent for 40 minutes. Kinase detection reagent (50 μΙ) was then added to convert ADP to ATP and allow the luciferin/luciferase reaction to take place. Plates were incubated at room temperature for 1 hour, and luminescence was detected using a BioTek luminescence plate reader. The amount of ADP produced was determined using a standard curve, run in each plate alongside the other assays. Peptide Km experiments were carried out using a twofold serial dilution from 2 mM Kemptide in buffer.

Functionalization and PEGylation of glass slides and coverslips

Aurora-A molecules were bound to the surface of functionalized coverslips and glass slides for fluorescence microscopy. Two 1.5 mm holes were drilled in the microscope slide approximately 5 mm apart. The coverslips and slides were super-cleaned with separate washes of Alconox, ethanol and MiliQ water solutions. The coverslips and slides were then functionalized with an amino-group through incubating with 1 in100 dilution of 3- aminopropyltheithoxysilane (sold as Vectabond) in methanol / 5% acetic acid for 20 minutes. Glass surfaces were PEGylated by incubating with 28.5 mM PEG-SVA:biotin- PEG-SVA solution (7:1 ratio) in 10 mM NaHCC solution for 2 hours, followed by washing with MiliQ water and drying with N2 gas.

Construction of flow cell Functionalized slides and coverslips were used to construct flow cells as follows. 1.5 mm diameter tubing (Agilent Technologies, UK) was glued into each 1.5 mm hole with epoxy resin glue and cut flush on the functionalized side with a razor. Two self-adhesive 10 x 10 mm, 25 μΙ_ Gene Frames (Therm of isher Scientific, UK) were stacked around the inlet and outlet holes on the functionalized side of the slide and a functionalized coverslip glued on top using epoxy resin (functionalized side towards the slide) (Figure 9b). Solutions were injected along the inlet tubing at a low (<0.25 mUmin) flow rate. Immobilization of Aurora-A

Aurora-A was tethered to the glass coverslip inside the flow cell using a biotinylated PEG- neutravidin-biotinylated anti-His-His-tagged protein strategy as follows ^{24 36}: i) Flow cells were first incubated in NeutrAvadin (0.1 mg/mL) in 50 mM TRIS-HCI pH7.5, 50 mM NaCI for one hour to bind to the covalently bonded PEG-biotin conjugate. Excess NeutrAvidin was eluted with a -1.5 mL wash of 50 mM TRIS-HCI pH 7.5, 50 mM NaCI solution, ii) A biotinylated anti-His antibody (Qiagen, UK) at 1 in 1000 dilution in 30 Mg/mL BSA, 50 mM NaCI, 50 mM TRIS-HCI pH 7.5 was incubated for 10 minutes to conjugate to the NeutrAvadin. Excess antibody was eluted with a 1.5 mL wash of 50 mM TRIS-HCI pH 7.5, 50 mM NaCI solution, iii) In low light conditions, a solution of labelled Aurora-A (<10 nM) in labelling buffer supplemented with 0.3 mg/mL BSA was injected in the flow cell and equilibrated for 10 min, before being washed out with ~1.5 mL imaging buffer. For measurements with inhibitors, activators or substrates, 0.5 mL of solution containing the target binding molecule in imaging buffer (made up to a final concentration of 1 % DMSO as necessary) was injected and allowed to incubate in darkness for 10 minutes. CD532 and MLN8054 were dissolved in DMSO to a concentration of 1 mM and added to the final solution at a ratio of 1 :100.

TIRF imaging and single molecule data analysis

Samples were illuminated in TIRF using a -2.0 mW 514 nm laser. Filters were ZET514 (excitation; Chroma), ZT514rdc (dichroic; Chroma) and both HQ525 longpass (emission; Chroma) and 595RDF60 (emission; Omega). Fluorescence was captured by a CoolView EM 1000 camera using an 80 ms/frame capture speed and 2x2 pixel binning with 500 frames per video.

The captured tiff video of the fluorescent molecules was processed with custom written I DL code. In summary, the complete tiff stack was used to generate a time-averaged image. Background intensity values were subtracted from this. High-intensity fluorescent molecules were identified by scanning for pixels above an intensity threshold and recording the peak location if the surrounding pixels were below one standard deviation. In the video stack, the intensity of the 8x8 pixel area surrounding each peak was enhanced by multiplying by the weighted values of a 3D Gaussian curve centered on the high intensity pixel, and the sum intensity of the curve in each frame was collected as intensity vs time.

The fluorescence intensity over time of each molecule was visualized using custom Matlab code. The fluorescence intensity trace was smoothed with a 3-point moving average, and double-labelled fluorescent molecules were identified manually as traces ending with a clear two-step photobleaching event (Figure 1f). Double-labelled fluorescence intensity traces were isolated, and at least 100 molecules (average of 121 frames per molecule) combined to create each intensity histogram (each bar within the histogram indicates the total number of frames for which a particular fluorescence intensity was observed). Histograms were normalized to a total area of 1 for easy visual comparison.

We discovered that some particularly low-intensity frames of quenched events were omitted from our histograms due to the background subtraction (negative intensities). Such frames were infrequent stochastic variations in intensity rather than complete events, but omission of these values led to poor definition of the left hand side of the low intensity histogram peak, partly because the function must pass through the origin. This propagated into poor fits of the double log normal function which could be overcome by a small x-axis translation to make all values positive. We modelled the effect of such a translation on idealized data sets and determined that unnecessarily translating well-defined data introduced an eventual error on reported curve areas of -2%, equivalent to the fitting error on our experimental measurements. We additionally determined that failure to include these low intensity values by only fitting positive background-subtracted intensities was likely to introduce errors of -3% (Figure 12). We thus applied an x-axis translation to all our data by adding 300 a.u. to all fluorescent intensities. This value was chosen as the smallest number required to make all intensity bins positive.

Data fitting Fluorescence intensity distributions were fit to log normal (19) or the sum of two log normal (20) distributions in Prism using the equations below:

The mode (peak) of each distribution is e^ and the area under each curve is Aa /2n. Example 7 - Alternative labelling site

In our assay, a quenched fluorescence signal indicates that the two dyes are in close proximity, while a high fluorescence signal indicates that the two dyes are spatially separated. Our previous results on labelled K224C/S283C clearly show that there are at least two populations of Aurora-A kinase in dynamic equilibrium, but it is not formally possible to be certain whether the high fluorescence ensemble measured is a single ensemble of Aurora-A in the T-loop active conformation or whether it constitutes two (or more) separate ensembles, one in the T-loop active conformation and one (or more) in a conformation intermediate between T-loop active and inactive conformations. In order to distinguish between these two possibilities, we designed a second construct, M373C/S283C, in which one dye is placed on the C-lobe of the kinase (position 373) and one on the activation loop as before (position 283). This construct is expected to give a high fluorescence signal in the T-loop inactive conformation and a quenched fluorescence signal in the T-loop active conformation. If only two conformations of Aurora-A are in dynamic equilibrium, we would expect M373C/S283C to give opposite populations of quenched and unquenched protein to K224C/S283C, corresponding to identical T-loop active and inactive conformations. If an intermediate between T-loop active and inactive conformations of Aurora-A is populated, we would expect the quenched population of M373C/S283C to be smaller than the high fluorescence population of K224C/S283C and vice versa.

Our single molecule histogram of M373C/S283C (figure 13a) is an exact inverse of that for K224C/S283C (figure 2a), indicating 74% of the kinase is in the T-loop active conformation - in excellent agreement with the 77% measured for K224C/S283C (figure 4). This means that there are exactly two conformational ensembles of the Aurora-A activation loop: T- loop active and T-loop inactive. Example 8- unphosphorylated Aurora-A

In order to determine the effect of phosphorylation at Thr288 on the conformational equilibrium of the Aurora-A activation loop, we coexpressed K224C/S283C with the serine / threonine phosphatase from bacteriophage lambda (lambda phosphatase). This dephosphorylates Aurora-A inside E.coli and is removed from the sample during purification (there is no tag on lambda phosphatase), enabling a uniform sample of unphosphorylated Aurora-A to be produced (ref Dodson, Yeoh et al 2013 Sci Signal.). In the absence of phosphorylation, the proportion of Aurora-A in the T-loop active conformation decreases, resulting in an equilibrium constant of 1.1 and only 48% of the kinase being in T-loop active conformation (figure 13b). This is consistent with the lower catalytic activity of the unphosphorylated kinase.

Example 9 - Fluorescence v distance between dyes for a set of model compounds Fluorescence was measured for 1 uM solutions of a series of end-labelled TMR-labelled polyproline helices of different lengths. Samples were excited at 521 nm and emission measured in the range 530-800 nm at 20 °C. Excitation slits were set to 5 nm and emission slits to 2 nm. Fluorescence reported is the area under each curve in the stated range. Distance modelling used the PDB-reported structure of residues 160-169 of entry 10WL which form a polyproline II helix. All residues in this helix were mutated to proline using Pymol software (all geometries easily accommodated) and a glycine residue was added to each end (also using Pymol) to model dye attachment points. TMR was modelled as an ellipsoid with radii 6.8 A, 4.5 A, 2.2 A, a width of 1.8 A and a linker length based on the distance from the Calpha carbon of the terminal residues. The distance reported on the x- axis in the graph (Figure 16) is the distance between the mean position of dyes as calculated for each helix by the FPS software (v 1.1 ; http://www.mpc.hhu.de/en/software.html; Kalinin, S., Peulen, T., Sindbert, S., Rothwell, P. J., Berger, S., Restle, T., Goody, R. S., Gohlke, H., and Seidel C. A. M. A toolkit and benchmark study for FRET-restrained high-precision structural modeling. Nat. Methods 9 (2012); doi: 10.1038/NMETH.2222). The data is fit to a sigmoid curve with a hill slope of 1 and the fitted midpoint of curve determined to be 16.8 A (Figure 16).

Claims

1 ) A kinase or kinase fragment comprising a first label and a second label, wherein: the kinase or kinase fragment has a first conformation and second conformation; one or both of the first and second label are fluorophores; the first label and second label comprise an interacting pair capable of interacting with each other by static quenching or Dexter quenching or PET (photoinduced electron transfer) or exciplex formation to produce a quenched pair; and wherein the first and second labels are positioned on the kinase or kinase fragment such that when the kinase or kinase fragment is in the first conformation the labels are distal to each other such that static quenching or PET or Dexter quenching or exciplex formation does not occur and when the kinase or kinase fragment is in the second conformation the labels are brought into close proximity such that static quenching or PET or Dexter quenching or exciplex formation of one or both labels occurs.

2) The kinase of claim 1 wherein the first or second label is located on the activation loop.

3) The kinase of any of claims 1 or 2 wherein only one label is located on the activation loopof the kinase.

4) The kinase of any of claims 1-3 wherein the first or second label is located on the N-lobe, the Glycine-rich loop, the P-loop or the C-lobe of the kinase, optionally wherein the first and second label are located on residues of the kinase that correspond to residues 373, 224, 367 and/or 283 of human Aurora-A kinase, optionally wherein the first and second label are located on residues corresponding to M373, K224, N367 and/or S283, optionally wherein the first and second label are located on residues corresponding to M373C, K224C, N367C and/or S283C. 5) The kinase of any of claims 1-4 wherein one label is located on the activation loop and the other label is located at a position that gives the optimal difference between fluorescence signal when the kinase is in one conformation and when the kinase in the other conformation.

6) The kinase of any of claims 1-5 wherein the first label and second label are fluorophores.

7) The kinase of any of claims 1 -6 wherein the first label and second label have substantially the same fluorescence properties.

8) The kinase of any of claims 1-7 wherein the first label and the second label are fluorophores and are the same fluorophore. 9) The kinase of any of claims 1-7 wherein the first label and the second label are fluorophores and are different fluorophores.

10) The kinase of any of claims 1-5 wherein the first label is a fluorophore and the second label is a quencher.

1 1) The kinase of any of claims 1-10 wherein the first label is a fluorophore and is located on the activation loop.

12) The kinase of any of claims 1 -5 and 10 wherein the second label is a quencher and is located on the activation loop.

13) The kinase of any of claims 1-12 wherein the relative positions of the first and second label are such that when the kinase is in a first conformation the labels are not in Van der Waals contact or in near Van der Waals contact.

14) The kinase of any of claims 1-13 wherein the relative positions of the first and second labels are such that when the kinase is in a second conformation the labels are in Van der Waals contact or in near Van der Waals contact. 15) The kinase of any of claims 1-14 wherein the first and second label are positioned on the kinase such that when the kinase is in a first conformation the labels are separated by (centre-to centre) at least 17 A, optionally at least 18 A , optionally at least 19 A , optionally at least 20 A, optionally at least 22 A, optionally at least 24 A, optionally at least 26 A, optionally at least 28 A, optionally at least 30 A, optionally at least 35 A, optionally at least 40 A. 16) The kinase of any of claims 1-15 wherein the first and second label are positioned on the kinase such that when the kinase is in a second conformation the labels are separated by (centre to centre) 15 A or less than 15 A, optionally less than 14 A, optionally less than 13 A, optionally less than 12 A, optionally less than 11 A, optionally less than 10 A, optionally less than 9 A, optionally less than 8 A, optionally less than 7 A, optionally less than 6 A, optionally less than 5 A, optionally less than 4 A, optionally less than 3 A, optionally less than 2 A, optionally less than 1 A, optionally less than 0.5 A, optionally less than 0.25 A.

17) The kinase according to any of claims 1-16 wherein the fluorescence signal obtained when the kinase is in the first conformation and the fluorescence signal obtained when the kinase is in the second conformation differs by at least 20%, 30%, 40%, 50%, 60% or 70%, optionally differs by at least 75%, optionally at least 80%, optionally at least 82%, optionally at least 84%, optionally at least 86%, optionally at least 88%, optionally at least 90%, optionally at least 91%, optionally at least 92%, optionally at least 93%, optionally at least 94%, optionally at least 95%, optionally at least 96%, optionally at least 97%, optionally at least 98%, optionally at least 99%, optionally 100%. 1 8) The kinase of any of claims 1-17 wherein one of the labels comprises:

TMR, MR121 , BODPIY, Atto655, Rhodamine 6G, Oregon green 488,

Fluorescein; and/or

the other label comprises:

TMR, tryptophan, BODIPY, Oregon green 488 or fluorescein; ionally wherein the first label and second label are (respectively)

TMR - TMR;

MR121 - tryptophan ;

BODPIY-BODIPY;

Atto655 - tryptophan;

Rhodamine 6G - tryptophan;

Oregon green 488 - Oregon green 488;

Fluorescein - fluorescein; or wherein the second label and first label are (respectively):

TMR - TMR;

MR121 - tryptophan ;

BODPIY-BODIPY;

Atto655 - tryptophan;

Rhodamine 6G - tryptophan;

Oregon green 488 - Oregon green 488;

Fluorescein - fluorescein. 19) The kinase of any of claims 1 -18 wherein the kinase is a protein kinase, serine/threonine kinase, tyrosine kinase, tyrosine kinase-like kinase, tyrosine-like kinases, receptor tyrosine kinase, non-receptor tyrosine kinase, STE kinase, CMCG kinase, CK1 kinase, CAMK kinase, dual specificity kinase (DYRK), AGC-like kinase, AGC kinase, ATE Kinase, atypical protein kinases, or a pseudokinase, optionally wherein the kinase is Aurora kinase or Aurora-A kinase.

20) The kinase of any of claims 1-19 wherein the first and/or second label are attached to the kinase through a thiol reaction, optionally through a thiol on a cysteine residue. 21) The kinase of any of claims 1 -20 wherein prior to labelling, site-directed mutagenesis has been performed on the kinase to mutate at least one, optionally at least two, optionally at least three, optionally at least four, optionally at least five, optionally all of the cysteine residues in undesirable locations, optionally all of the native cysteine residues, optionally to mutate the cysteine residues to alanine or serine or any other amino acid, or to remove the cysteine residues.

22) The kinase according to any of claims 1 -21 wherein the labelled kinase is catalytically active, optionally wherein the kinase retains at least 40% of the catalytic activity of the unlabelled kinase, optionally retains at least 45%, optionally at least 50%, optionally at least 55%, optionally at least 60%, optionally at least 65%, optionally at least 70%, optionally retains at least 75%, optionally at least 80%, optionally at least 82%, optionally at least 84%, optionally at least 86%, optionally at least 88%, optionally at least 90%, optionally at least 91 %, optionally at least 92%, optionally at least 93%, optionally at least 94%, optionally at least 95%, optionally at least 96%, optionally at least 97%, optionally at least 98%, optionally at least 99%, optionally 100% of the catalytic activity of the unlabelled kinase. 23) The kinase according to any of claim 1-22 wherein the kinase comprises a third label, optionally comprises more than three labels.

24) The kinase according to any of claims 1-23 wherein the fluorescence signal obtained does not decay.

25) The kinase according to any of claims 1-24 wherein the kinase is immobilised on a solid support. 26) The kinase according to any of claims 1 -25 wherein the interaction between the labels is detected using microscopy, optionally total internal reflection fluorescence microscopy (TIRF), or single molecule spectroscopy.

27) A method for determining the intrinsic conformational propensity of a kinase, or for determining the dynamic properties of a kinase, or for determining the timescale of conformation change of a kinase, wherein the method comprises measuring the conformation of a kinase according to any of claims 1-26.

28) The method according to claim 27 wherein the method comprises determining the fluorescence of single kinase molecules.

29) The method according to any of claims 27 or 28 wherein the fluorescence intensity over time of single kinase molecules is determined. 30) The method according to any of claims 27-29 wherein the fluorescence is determined by microscopy, optionally TIRF, or single molecule spectroscopy.

31 ) The method according to any of claims 27-30 wherein the Ka (association constant), Kd (dissociation constant = 1/Ka), conformation-specific Kd, and/or change in Gibbs free energy is calculated.

32) A method to calculate the Ka, Kd, conformation-specific Kd, or change in Gibbs free energy of a kinase wherein the method comprises the method according to any of claims 27-31. 33) A method for determining the effect of a binding partner, inhibitor or activator or any combination of binding partner, inhibitor and activator on a kinase, wherein the method comprises the method according to any of claims 27-32. 34) A method for predicting the effect of an inhibitor or activator or combination of inhibitor and activator on kinase conformation wherein the method comprises the method according to any of claims 27-33.

35) A method for designing a kinase inhibitor wherein the method comprises the method according to any of claims 27-34.

36) A method for improving/designing conformation specific binding partner, inhibitors or activators, optionally to optimise discrimination between Kd for each conformation, wherein the method comprises the method according to any one of claims 2-35.

37) A method of determining the effect of an agent on a kinase, wherein the method comprises the method according to any one of claims 27-36.

38) A method of determining the conformation of a kinase that is induced by an agent wherein the method comprises the method according to any one of claims 27-37.

39) A method of identifying a kinase inhibitor or activator specific to a particular said kinase or group of kinases, wherein the method comprises screening a library/panel of different kinases labelled according to claims 1-26.

40) A method for generating a mutant kinase that has an intrinsic propensity to occupy the active conformation, wherein the method comprises determining the intrinsic propensity for a kinase according to any of claims 1 -26 to occupy a particular conformation, introducing one or more mutations into said kinase and determining the effects of the mutation on the intrinsic propensity of the kinase to occupy the active conformation, and selecting the mutant kinase or mutant kinases that have a higher intrinsic propensity to occupy the active conformation.

41) A method for generating a mutant kinase that has an intrinsic propensity to occupy the inactive conformation, wherein the method comprises determining the intrinsic propensity for a kinase according to any of claims 1 -26 to occupy a particular conformation, introducing one or more mutations into said kinase and determining the effects of the mutation on the intrinsic propensity of the kinase to occupy the inactive conformation, and selecting the mutant kinase or mutant kinases that have a higher intrinsic propensity to occupy the inactive conformation. 42) A panel/array/library of kinases as defined in any of claims 1 -26.

43) The panel/array/library according to claim 42 wherein the kinases are immobilised on a solid support. 44) The panel/array/library according to any of claims 42 or 43 wherein the panel/array/library comprises at least 5 kinases, optionally at least 10 kinases, optionally at least 15 kinases, optionally at least 20 kinases, optionally at least 30 kinases, optionally at least 40 kinases, optionally at least 50 kinases, optionally at least 75 kinases, optionally at least 100 kinases, optionally at least 150 kinases, optionally at least 200 kinases, optionally at least 250 kinases, optionally at least 300 kinases, optionally at least 350 kinases, optionally at least 400 kinases, optionally at least 500 kinases.

45) The panel/array/library according to any of claims 42-44 wherein the kinases are from the same family.

46) A kit comprising at least one kinase according to any of claims 1-26.

47) The kit according to claim 62 comprising 2 or more kinases according to any of claims 1-26.

48) The kit according to any of claims 46 and 47 further comprising one or more small molecules or other agents.

49) The kit according to any of claims 46-48 wherein the kit further comprises one or more physiological binding partner (optionally a polypeptide) and/or one or more mutant kinase.

50) A kinase inhibitor for use in treating a subject in need thereof, wherein the kinase inhibitor has an affinity for the said kinase or one or other conformation of said kinase that is higher than the affinity of known activators for the kinase or one or other conformation of the kinase, wherein the affinity of the inhibitor and affinity of known activator(s) is determined by the method according to any of claims 27-32. 51) A method for identifying a kinase inhibitor suitable for use in treating a disease or condition involving a kinase which it is desirous to inhibit, wherein the method comprises: a) determining the affinity of the said kinase for an activator; and

b) selecting an inhibitor with a higher affinity for the kinase than the affinity of the activator for the kinase,

optionally wherein the affinity of the said kinase is determined according to the method of any of claims 27-32.

52) A method according to any of the above claims wherein the propensity of the kinase to adopt an inactive or active conformation is determined in the presence of a cell lysate.