Identification of inactive conformation-selective interleukin-2-inducible T-cell kinase (ITK) inhibitors based on second-harmonic generation.

Many clinically approved protein kinase inhibitors stabilize an inactive conformation of their kinase target. Such inhibitors are generally highly selective compared to active conformation inhibitors, and consequently, general methods to identify inhibitors that stabilize an inactive conformation are much sought after. Here, we have applied a high-throughput, second-harmonic generation (SHG)-based conformational approach to identify small molecule stabilizers of the inactive conformation of interleukin-2-inducible T-cell kinase (ITK). A single-site cysteine mutant of the ITK kinase domain was created, labeled with an SHG-active dye, and tethered to a supported lipid bilayer membrane. Fourteen tool compounds, including stabilizers of the inactive and active conformations as well as nonbinders, were first examined for their effect on the conformation of the labeled ITK protein in the SHG assay. As a result, inactive conformation inhibitors were clearly distinguished from active conformation inhibitors by the intensity of SHG signal. Utilizing the SHG assay developed with the tool compounds described above, we identified the mechanism of action of 22 highly selective, inactive conformation inhibitors within a group of 105 small molecule inhibitors previously identified in a high-throughput biochemical screen. We describe here the first use of SHG for identifying and classifying inhibitors that stabilize an inactive vs. an active conformation of a protein kinase, without the need to determine costructures by X-ray crystallography. Our results suggest broad applicability to other proteins, particularly with single-site labels reporting on specific protein movements associated with selectivity.

Many clinically approved protein kinase inhibitors stabilize an inactive conformation of their kinase target. Such inhibitors are generally highly selective compared to active conformation inhibitors, and consequently, general methods to identify inhibitors that stabilize an inactive conformation are much sought after. Here, we have applied a high-throughput, second-harmonic generation (SHG)-based conformational approach to identify small molecule stabilizers of the inactive conformation of interleukin-2-inducible T-cell kinase (ITK). A single-site cysteine mutant of the ITK kinase domain was created, labeled with an SHG-active dye, and tethered to a supported lipid bilayer membrane. Fourteen tool compounds, including stabilizers of the inactive and active conformations as well as nonbinders, were first examined for their effect on the conformation of the labeled ITK protein in the SHG assay. As a result, inactive conformation inhibitors were clearly distinguished from active conformation inhibitors by the intensity of SHG signal. Utilizing the SHG assay developed with the tool compounds described above, we identified the mechanism of action of 22 highly selective, inactive conformation inhibitors within a group of 105 small molecule inhibitors previously identified in a high-throughput biochemical screen. We describe here the first use of SHG for identifying and classifying inhibitors that stabilize an inactive vs. an active conformation of a protein kinase, without the need to determine costructures by X-ray crystallography. Our results suggest broad applicability to other proteins, particularly with single-site labels reporting on specific protein movements associated with selectivity.
Protein kinases are the largest enzyme family encoded by the human genome and are comprised of over 500 members [1]. Pharmacological agents directed to them have high potential to treat a number of diseases, spurring numerous lines of research [2]. The overall structure of the catalytic domain of protein kinases is highly conserved [3]. The sequence and the structural topologies of ATP-binding sites of kinases in the active conformation are highly similar, with key residues optimally aligned for catalysis [4]. Accordingly, the conserved ATP-binding pocket creates significant difficulties for designing selective kinase inhibitors [5]. In physiological conditions, most kinases exist in equilibrium between active and inactive conformational states [4]. Inhibitors that stabilize an inactive conformation generally tend to show higher selectivity [6]. Highthroughput screens (HTS) of protein kinases are commonly employed to identify kinase inhibitor candidates. These screens provide hits that inhibit the kinase activity but do not report on the resulting conformational state of the target. Thus, it is important for selectivity considerations to further identify those compounds from the HTS hits that stabilize inactive conformations before moving to crystallography. Second-harmonic generation (SHG) has emerged as a technique for real-time detection and measurement of protein conformational changes in solution and is broadly applicable across many target classes [7][8][9][10][11]. The method is based on tethering proteins labeled with an SHG-active dye to a surface such as a supported lipid bilayer membrane. SHG is highly sensitive to the orientation of the SHG-active molecules, permitting the study of conformational changes in the target protein ( Fig. 1).
In this report, we have developed an SHG assay for selection of inactive conformation inhibitors to Tec family tyrosine kinase, interleukin (IL)-2-inducible T-cell kinase (ITK). ITK is expressed in T cells, mast cells, and NK cells, and it plays an important role in the production of cytokines such as IL-2, IL-4, IL-5, and IL-13 [12,13]. Due to this role of ITK in T cells, inhibition of the kinase could offer the potential for treating T-cell-mediated inflammatory diseases [14][15][16].

Materials
Biacore T200 and 4000 optical biosensors and NTA sensor chips were obtained from GE Healthcare (Little Chalfont, Buckinghamshire, UK).
ADP-Glo TM Kinase Assay kit which contains ADP-Glo TM Reagent, kinase detection reagent, ATP, and ADP was purchased from Promega (Madison, WI, USA).
Cloning, expression, and purification of human tyrosine kinase The His 6 -tagged human wild-type ITK kinase domain (WT ITK) was expressed from the pVL1393/His-hITK KD vector. This vector was constructed by cloning the human ITK KD from the human ITK full-length cDNA (OriGene, Rockville, MD, USA) into pVL1393 (Thermo Fisher Scientific, Waltham, MA, USA) using primers that encoded an N-terminal His 6 -tag.
His 6 -tagged human mutant ITK KD (ITK Mutant A) was expressed from the pVL1393/His-hITK KD MutA vector. This vector was constructed by PCRbased site-directed mutagenesis of pVL1393/His-hITK Incident laser light strikes the surface and through total internal reflection creates an evanescent wave. Labeled protein is bound to the surface, and the measured SHG signal intensity depends on the average, net orientation of the dye label relative to the surface normal (z-axis). A conformational change that alters the orientational distribution of the label in space or time results in a change in SHG signal intensity. A decrease in SHG signal intensity is caused by the dye moving more perpendicular to the z-axis, while an increase in SHG signal intensity is caused by the dye moving more parallel to the z-axis. In this way, different conformational states of a protein produce different SHG signals.
KD to incorporate the following changes to the native sequence: one native solvent-exposed amino acid on the aC-helix was substituted with a cysteine, and the native cysteines (C442, C477) in the WT enzyme were mutated to serines. The myc-tagged human YopH was cloned utilizing primers that encoded an N-terminal myc-tag to amplify human YopH from commercially available cDNA (GenScript, Piscataway, NJ, USA). The resulting PCR product was cloned into pVL1393 resulting in expression vector pVL1393/myc-YopH. All plasmids were confirmed by sequence analysis.
Recombinant baculovirus was produced using the BaculoGold Transfection Kit (BD Biosciences, San Jose, CA, USA). For protein production, the cells were coinfected with either pVL1393/His-hITK KD or pVL1393/ His-hITK KD MutA baculovirus, and pVL1393/YopH baculovirus at an appropriate rate to ensure uniform dephosphorylation. Cells were harvested by centrifugation 72 h postinfection and stored at À80°C.
Wild-type ITK and ITK Mutant A were purified from cell lysate on Ni-NTA Superflow (Qiagen, Germantown, MD, USA) and eluted with 50 mM HEPES, pH 8.0, 150 mM NaCl, and 500 mM imidazole. The WT ITK and ITK Mutant A containing fractions were pooled, respectively, and further purified by ion exchange over Q Sepharose Fast Flow (GE Healthcare), followed by gel filtration over a Superdex 200 10/300 GL (GE Healthcare). The final protein buffer contained 50 mM HEPES, pH 8.0, 500 mM NaCl, 10% glycerol, and 1 mM DTT. The protein was concentrated to~2.5 mgÁmL À1 , flash-frozen with liquid nitrogen, and stored at À80°C.
Dephosphorylation of WT ITK and ITK Mutant A by YopH was confirmed by western blotting using the monoclonal phosphotyrosine antibody clone 4G10 from Merck Millipore (Darmstadt, Germany).

ADP-Glo TM inhibitory assay
Enzymatic activity of WT ITK and ITK Mutant A was measured using the ADP-Glo TM Kinase Assay [30] (Promega). Activity measurements were taken following the manufacturer's protocols. The ADP production was measured in assay buffer (25 mM HEPES, pH 7.5, 25 mM KCl, 10 mM MgCl 2 , 1 mM DTT, 0.005% Tween 20, 0.1% BSA) in a 384-well plate format. The reactions were carried out at room temperature in a total volume of 5 lL for 60 min with 70 lM ATP and at excess substrate (SLP76 [18]) concentrations. ADP-Glo TM reagent was added at room temperature for 40 min to stop the kinase reaction and deplete the unconsumed ATP, leaving only ADP and a very low background of ATP. Furthermore, kinase detection reagent was added to convert ADP to ATP and introduce luciferase and luciferin to detect ATP. The plate was read by a Paradigm (Molecular Devices, LLC, Sunnyvale, CA, USA) in luminescent mode.

SPR assay
Wild-type ITK was immobilized onto an NTA sensor chip using the capture coupling method that results in the capture in a nonrandom orientation by the His 6 -tag prior to amine coupling to produce functional and stable surfaces [31]. The surface of the sensor chip was activated with a mixture of 0.1 molÁL À1 N-hydroxysuccinimide and 0.4 molÁL À1 N-ethyl-N'-(dimethylaminopropyl) carbodiimide for 5 min at 10°C. WT ITK (10 lgÁmL À1 ) in running buffer (25 mM HEPES, pH 7.5, 25 mM KCl, 10 mM MgCl 2 , 1 mM DTT, 0.005% Tween 20) was injected for 10 min. Typical immobilization levels ranged from 2900 to 4900 resonance units.
The compounds were diluted directly into the running buffer containing 5% DMSO. Serially diluted compounds were injected in a separate cycle for 60-s association followed by 60-s dissociation at a flow rate of 30 lLÁmin À1 . Competition experiments were carried out in the presence of ADP at a 1000 lM concentration diluted in the running buffer.
All data analysis was performed using Biacore T200 or Biacore 4000 evaluation software. Sensorgram data were solvent-corrected and double-referenced. For kinetic analysis, data were fit to 1 : 1 interaction model. For steady-state analysis, responses at equilibrium were plotted against the compound concentration and fit to a 1 : 1 Langmuir binding isotherm.

SHG assay protein labeling
Interleukin-2-inducible T-cell kinase Mutant A was labeled with SHG2-maleimide (thiol-reactive dye) in 50 mM sodium phosphate, pH 6.5, 150 mM NaCl, and 10% glycerol. One hundred micromolar of protein was incubated with 500 lM of dye at room temperature for 90 min. The reactions were then centrifuged at 4°C for 20 min before the reaction was stopped by removing excess dye by buffer exchange with a Zeba TM column (Thermo Fisher) into storage buffer (50 mM HEPES, pH 8.0, 500 mM NaCl, 10% glycerol, 1 mM DTT). The degree of labeling was measured by UV/Vis spectroscopy and was measured to be 1.3 (dyes/protein). The resulting conjugate, ITK Mutant A-SHG2, was flash-frozen in liquid nitrogen and stored at À80°C until use.

SHG assay for tool compounds using ITK Mutant A-SHG2
Five microliter of ITK Mutant A labeled with SHG2maleimide (ITK Mutant A-SHG2) was preincubated at room temperature with 20 lM compound for 60 min in assay buffer (25 mM HEPES, pH 7.5, 25 mM KCl, 10 mM MgCl 2 , 1 mM DTT, 0.005% Tween 20). 10 lL of the protein/compound mixture was added to 10 lL of assay buffer on the Ni/NTA bilayer surface. Protein was allowed to tether to the surface through the Ni/NTA:His 6 -tag interaction at room temperature for 90 min, at which point the SHG signal was measured (Fig. 3A-D).
In a competition assay, after SHG readout #1 (Fig. 3D), unbound protein and excess compound were washed out, and then, SHG baseline signal was measured (Fig. 3F). The washed well was then injected with 20 lL of 10 lM test compound, and the SHG signal was measured again (Fig. 3H). The time between measurement of SHG intensity #2 (Fig. 3F) and measurement of SHG intensity #3 (Fig. 3H) was 5 min.
To calculate the percent change in SHG intensity (% D SHG ), the SHG intensity of readout #2 (Int #2 ) was subtracted from the SHG intensity of readout #3 (Int #3 ) and then divided by the initial second-harmonic intensity (Int #2 ) according to the following equation:

SHG assay for binning HTS hits
The compounds to be analyzed were mixed with 500 nM ITK Mutant A-SHG2 and allowed to incubate at room temperature for 60 min. The protein/ compound mixtures were added to the Ni/NTA bilayer surface with a 1 : 1 dilution for a final protein concentration of 250 nM in a 384-well plate format. Each compound was run at 2.5, 10, and 40 lM for lower potency compounds (IC 50 values ≥ 1 lM) or at 0.31, 1.3, and 5.0 lM for higher potency compounds (IC 50 values of < 1 lM). Protein/compound mixtures were allowed to attach to the bilayer surface for 90 min at room temperature, and the SHG signal was then measured (Fig. 3D). All experiments were performed in triplicate. SHG signal is reported in photon counts per second (cps).

Characterization of protein and tool compounds
His 6 -tagged human ITK KD (WT ITK) was utilized for both the biochemical and surface plasmon resonance (SPR) assays. WT ITK with a cysteine substituted in the aC-helix (ITK Mutant A) was used for the SHG assay. The engineered Cys residue was inserted in order to conjugate the dye to a site-specific location that reports on the structural change between the active and inactive conformations of the protein. The amino acid substitutions in ITK Mutant A (C442S/C477S, and Cys substitution on aC-helix) were shown to have no impact on enzymatic activity (Fig. 4). To allow the protein to adopt both an active and an inactive conformation, all proteins were coexpressed with the tyrosine phosphatase, YopH, in baculovirus to fully dephosphorylate them. Dephosphorylation was validated by western blot using a monoclonal phosphotyrosine antibody ( Fig. 5). We prepared 14 tool compounds with known biochemical ITK inhibitory activity and binding affinity summarized in Fig. 2 and Table 1, respectively.

SHG assay development
Interleukin-2-inducible T-cell kinase Mutant A-SHG2 was preincubated with each of the 14 tool compounds to demonstrate that we could distinguish inactive conformation inhibitors, active conformation inhibitors, and nonbinders by SHG signal differences. A schematic of the assay is shown in Fig. 3A-D. The results of this panel showed large SHG signal compared to apo protein for compounds F, H, I, and J (Fig. 6), all of which are known active conformation inhibitors. Compound G is also an active conformation inhibitor but showed an intermediate SHG signal, which could be due to its lower solubility. All other compounds in the panel showed similar SHG signal intensity to the apo form of the protein. Compounds A-E are known inactive conformation inhibitors, and compounds K-N are known nonbinders. These two classes of compounds cannot be distinguished from one another in the current assay setup. However, they can be readily distinguished from the active conformation inhibitors.

SHG assay validation
To show that ITK Mutant A-SHG2 maintained its activity on the surface, compound injection after immobilization of the protein was performed. A schematic of the assay is shown in Fig. 3A-H. ITK Mutant A-SHG2 was preincubated with compound E, a low-affinity inactive conformation inhibitor (K D of 1.1 lM for WT ITK). The protein/compound mixture was injected onto the Ni/NTA bilayer surface and allowed to tether to the surface. Unbound protein/compound was washed out, and then, a high-affinity inactive conformation inhibitor (compound A, K D of 1.3 nM for WT ITK), a high-affinity active conformation inhibitor (compound J, K D of 3.5 nM for WT ITK) or buffer, was injected onto the complex on the surface.
Upon compound A injection, the SHG signal decreased by 24%, suggesting that more conjugate molecules shifted to the inactive conformation (Fig. 7). When compound J was injected, there was a dramatic increase of 49% in SHG signal, suggesting a switch from the inactive to the active conformation. The observed SHG signal change correlates with the preincubation results in Fig. 6, which shows relatively low SHG signal for inactive conformation inhibitors and high SHG signal for active conformation inhibitors. The signal changes upon competitive binding with the high-affinity inhibitor also demonstrate that ITK Mutant A-SHG2, when stabilized by preincubation with compound E, remains functional on the surface. Injection of buffer showed a decrease in the SHG intensity of À1.7% indicating no conformational change.

Optimization of protein concentration for a more high-throughput SHG assay
To develop a high-throughput SHG screen, we attempted to reduce the labeled protein concentration in the assay to conserve material. During the initial phase of assay development, we incubated 5 lM ITK Mutant A-SHG2 with 20 lM compound. This reaction was then diluted 1 : 1 with assay buffer, and the final concentrations of protein and compound were 2.5 and 10 lM, respectively. Holding the compound concentration constant, we tested final protein concentrations of 500 and 250 nM. Both protein concentrations gave similar results with SHG signals of active conformation inhibitors above three times the standard deviation (+3r) of the signal level (5300 cps) of the protein bound to compound E (Fig. 8). Therefore, we performed the binning experiments at 250 nM final protein concentration (160 ng per well), and a +3r difference of the signal bound to compound E, the lowest affinity inactive form binder among the tool compounds, was used as a criterion for separation between active conformation inhibitors and inactive conformation inhibitors/nonbinders.

Conformational binning of hits from a HTS with the SHG assay
Next, we used the optimized SHG assay to distinguish between active and inactive conformation inhibitors within the set of 105 hits previously identified in a high-throughput biochemical screen; the screen only identified whether they are inhibitors but did not provide any information about their conformational mechanism of action (MOA). Inhibitory activity of these compounds was lower than 10 lM, and most of these binders showed a typical 1 : 1 interaction and stoichiometry by SPR. The compounds were classified into low potency (IC 50 values ≥ 1 lM) and high potency (IC 50 values < 1 lM) molecules. Each compound was evaluated at 2.5, 10, and 40 lM for low potency or at 0.31, 1.3, and Derivative of D 4.6 9 10 À7 1.1 9 10 À6 F Active Staurosporine 1.3 9 10 À8 1.2 9 10 À8 G Active GSK Comp 13 3.5 9 10 À9 3.6 9 10 À9 H Active Genentech Comp 19 1.3 9 10 À9 2.1 9 10 À9 I Active Sanofi US 1.8 9 10 À9 1.5 9 10 À9 J Active Genentech GNE-9822 1.1 9 10 À9 3.5 9 10 À9 K Nonbinder Novartis Imatinib > 3.0 9 10 À5 > 3.0 9 10 À5 L Nonbinder GSK GW-2580 > 3.0 9 10 À5 > 3.0 9 10 À5 M Nonbinder Plexxikon PLX-4720 > 3.0 9 10 À5 > 3.0 9 10 À5 N Nonbinder AstraZeneca Gefitinib > 3.0 9 10 À5 > 3.0 9 10 À5 Inhibitory activities and affinities of each compound against His 6 -tagged human ITK KD (WT ITK) were measured by ADP-Glo TM Kinase Assay and SPR assay, respectively. K D of compound D was not measured due to covalent inhibitor. 5.0 lM for high potency compounds. The results of the screen showed a clear difference between active and inactive conformation inhibitors (Fig. 9). Compounds were classified as active conformation inhibitors when the resulting SHG signal was above the +3r signal (7000 cps) of the control compound (compound E) at the highest inhibitor concentration. In contrast, compounds below the +3r of compound E were classified as inactive conformation inhibitors, low-affinity active conformation inhibitors, or nonbinders. Of 105 compounds, 22 were determined to be inactive conformation inhibitors, weak active conformation inhibitors, or nonbinders. However, based on the HTS assay results and SPR follow-up (data not shown), all 105 compounds were shown to bind ITK, with the affinity of the weakest binders at~10 lM. Therefore, all 22 compounds were neither weak active conformation inhibitors nor nonbinders, suggesting that these compounds with low SHG signal were inactive conformation inhibitors. As examples, SHG signals of J-31 (IC 50 of 5.5 lM for WT ITK) and J-95 (IC 50 of 3.3 lM for WT ITK) are shown in Fig. 10. SHG signals at all concentrations of these compounds are comparable in intensity to compounds A and E, indicating that they too are inactive conformation inhibitors.

Kinase selectivity of ITK inhibitors
Having determined by SHG that J-31 and J-95 are inactive conformation inhibitors, we next evaluated their selectivity in a broad kinase panel using biochemical assays. We screened against 54 kinases in the  low solubility. J-31 only inhibited one kinase (CLK1) at > 65% and J-95 only inhibited two kinases (HCK and LCK) at > 65% (Table 2), indicating good specificity of these compounds.

Determination of binding site
To help identify potential binding sites on ITK and provide valuable information for further optimization, we used the Biacore assay to examine competition between the inhibitors and ADP. J-31 and J-95 bound to ITK in the absence of ADP (K D of 3.4 and 1.0 lM, respectively) and did not bind in the presence of ADP (Fig. 11). Therefore, both compounds would probably be ATP-site binders that stabilize the inactive conformation of ITK.

Discussion
Second-harmonic generation is a nonlinear optical technique in which two photons of equal energy are combined by a nonlinear material or molecule to generate one photon with twice the energy [32,33].
Recently, the SHG technique has been shown to be a powerful method for studying biomolecular structure and conformational changes. In this study, we developed an SHG-based assay to identify compounds that stabilized the ITK aC-helix Glu-out inactive conformation without the need to determine costructures by X-ray crystallography. From among a set of 105 HTS hits, we identified 22 inactive conformation inhibitors by excluding compounds with SHG signal comparable to the active conformation inhibitor control. It is important for precise compound binning that the SHG-active dye is conjugated at a suitable site on the target protein and that the activity of an immobilized target protein is maintained. Multiple approaches  100  100  SYK  100  61  TAOK1  100  91  TEC  100  86  TIE2  100  100  TRKC  100  97  TSSK1B 100 100 can be used to label proteins. Generally WT proteins can be labeled randomly at lysines through amine-reactive chemistry. In parallel, mutant proteins with engineered cysteines can be labeled through thiol-reactive chemistry to introduce dyes to targeted residues. Although most published assays of SHG-detected conformational changes rely on random labeling [9], we adopted a site-specific labeling strategy to distinguish inactive conformation inhibitors. To select the site of the cysteine mutant for introducing the SHG dye, we focused on movements of the aC-helix. This helix exhibits a large movement between the inactive and active conformations in the reported ITK X-ray structures (Protein Data Bank ID code: 4L7S [21], 4MF1 [22], 4PQN [25] for active conformations and 3MJ2 [17], 4M14 [19], 4HCU [20] for inactive conformations). A single cysteine mutant was introduced into the kinase aC-helix resulting in an SHG assay that reports on the conformational state of the labeled aC-helix upon binding active or inactive conformation inhibitors. The ITK Mutant A-SHG2 conjugate was found to discriminate well between the inactive and active conformation inhibitors with a set of 14 tool compounds, proving that site-specific labeling of an engineered cysteine in the aC-helix is effective to distinguish active and inactive conformations in this study.
To ascertain whether ITK Mutant A-SHG2 maintained its activity on the surface during the assay, we ran an order of addition experiment. The SHG assay resolved the displacement of a low-affinity inactive conformation inhibitor (compound E, K D of 1.1 lM for WT ITK) with a high-affinity inactive conformation inhibitor (compound A, K D of 1.3 nM for WT ITK) and a high-affinity active conformation inhibitor (compound J, K D of 3.5 nM for WT ITK; Fig. 7). Therefore, ITK can be stably tethered to the surface and produce functional responses, as expected.
Next, we used the SHG assay to select inactive conformation inhibitors from a set of 105 HTS hits. All compounds were previously validated to bind ITK by Biacore, but could not be binned into conformational class without requiring X-ray cocrystal structure determination. NMR and X-ray crystallography have been generally used to detect conformational shifts, but these methods have low to mid-throughput, require large amounts of sample, and are often challenging. In contrast, the SHG assay is very efficient in terms of protein consumption and throughput. For instance, 500 lg of protein would be required for screening about 100 compounds at three concentrations in a triplicate. Moreover, there is no target protein size limitation with SHG. SHG can be a powerful tool, providing information about ligand-induced conformational changes. In the assay format described here we did not attempt to quantitatively measure the affinity of compounds as we used an SPR assay instead. Rather, we focused on selecting inactive form binders from hits qualified and validated by SPR and inhibitory assays. Of 105 compounds, 22 compounds were determined by SHG to be inactive conformation inhibitors, providing crucial structural information about the compound MOA and supplementing the typical screening data consisting of potency and affinity routinely measured by orthogonal methods. J-31 and J-95 were selected as examples and were evaluated in a broad kinase panel. Both compounds demonstrated high selectivity for ITK.
The identification of a subset of compounds that stabilize the inactive conformation of ITK among a larger pool of hits using an SHG assay creates an opportunity for rapidly developing novel highly selective kinase inhibitors against this kinase and, in principle, any kinase target in a high-throughput manner. We believe that applying similar SHG-based assays opens the way to discover high-selectivity inhibitors across many other members of the kinome and with other protein classes as well.