Spleen tyrosine kinase is a non-receptor tyrosine kinase, overactivation of which is thought to contribute to autoimmune diseases as well as allergy and asthma. Protein kinases have a highly conserved ATP binding site, thus making challenging the design of selective small molecule inhibitors. It has been well documented that some protein kinases can be stabilized in their inactive conformations (Type-II inhibitors). Herein, we describe a protein structure/ligand-based approach to successfully identify ligands that bind to novel conformations of spleen tyrosine kinase. By utilizing kinase protein crystal structures both in the public domain (RCSB) and within Pfizer’s protein crystal database, we report the discovery of the first spleen tyrosine kinase Type-II ligands. Compounds 1 and 3 were found to bind to the DFG-out conformation of spleen tyrosine kinase, while compound 2 binds to a DFG-in, C-Helix-out conformation. In this instance, the C-helix moved significantly to create a large hydrophobic pocket rarely seen in kinase protein crystal structures.
Spleen tyrosine kinase (Syk) is a non-receptor tyrosine kinase targeted in multiple therapeutic programs including autoimmunity, inflammation, and cancer. Activating signals through both B-cell receptors (BCR) and Fc receptors (FcR) rely on Syk to couple membrane proximal events to downstream pathways and gene expression programs. Overactivation of the kinase is thought to contribute to the pathogenesis of autoantibody production from B cells in the case of autoimmune diseases (1), proliferation and resistance to apoptosis in B-cell chronic lymphocytic leukemia (2), and hypersensitivity to allergens in allergy and asthma (3).
Upon antigen binding to the extracellular domains of the BCR complex, members of the intracellular Src kinase family such as Lyn or Fyn become recruited to the intracellular tails of the BCR complex. Here, they phosphorylate the tyrosine of the immunostimulatory tyrosine activating motifs (ITAMs) of the Igα and Igβ chains of the BCR complex. When the ITAM sequences are phosphorylated, they bind the two tandem N-terminal SH2 domains of Syk (4). Syk is in an auto-inhibited conformation when in the cytoplasm. Binding to the phosphorylated ITAMs opens the Syk conformation, increasing the kinase activity (5). The open conformation is further stabilized by phosphorylation on 2 interdomain B (a linker region between the SH2 domains and the C-terminal kinase domain) region tyrosines, Y348 and Y352 by Lyn (6). Both binding through the SH2s to the pITAMs as well as phosphorylation by Lyn have been shown to increase the kinase activity (7).
Activated Syk phosphorylates the adaptor molecule Blnk (SLP-65). Phosphorylated Blnk acts as a scaffold to recruit molecules from distinct pathways to the immunoreceptor complex. PLCγ2 and Btk bind through their SH2 domains to phosphorylated tyrosines on Blnk and together with PI3k activate the Ca++ flux pathway leading to NFAT and NF-κB transcription factor activation. Other phosphorylated tyrosines on Blnk recruit Grb2 and Vav that link through Ras and Rho GTPases, respectively, to ERK/AP-1 map kinase pathways and actin cytoskeletal rearrangements (8).
Protein kinases have a highly conserved ATP binding site, and therefore, designing selective small molecule inhibitors that purely target the ATP site can be challenging. Additionally, reversible ATP site inhibitors (Type-I inhibitors) are competitive with ATP and must compete against high intracellular levels of ATP (5–10 mm) within cells (9), resulting in significant loss in their biochemical efficiency (10). It has been well documented and evidenced that some protein kinases can be stabilized in their inactive conformations by small molecules (Type-II inhibitors). This inactive conformation of the kinase is resonated by the DFG loop movement from its ‘in-state’ to the ‘out-state’ conformation. For example, imatinib (Gleevec) and a number of other known drugs and candidates have been shown in X-ray studies to bind to the inactive conformation of the kinase in a DFG-out conformation (11), where the phenylalanine residue of the DFG motif undergoes a significant movement to partially block the ATP site and the aspartic acid moves to point away from the active site, rendering the kinase inactive, and revealing a large hydrophobic binding pocket not present in the active kinase. The consequences of this binding mode are profound and numerous. By stabilizing an inactive conformation, DFG-out inhibitors can exhibit increased biochemical efficiency (10) and therefore exhibit higher cellular potency relative to Type-I inhibitors (12). Additionally, residues around the newly formed hydrophobic pocket in a DFG-out conformation are less highly conserved relative to the residues that form the ATP site, and as such, the potential exists for designing kinase inhibitors with an increased degree of selectivity (13,14).
Because of the opportunities that Type-II inhibitors may offer, a number of approaches have evolved to identify such ligands. Gray et al. describe a pharmacophore model for the rational design of Type-II inhibitors. Here, they depict a ‘hybrid-design’ approach wherein a hinge binding moiety is identified by its affinity for the kinase and its ability to connect to a Type-II tail in an appropriate manner (15). The tail can then be chosen based on fit to a docking model derived from an appropriate crystal or homology model. This approach was successfully applied in the development of Abl inhibitors (16). Huang et al. more recently applied this approach to the development of Src/Abl dual inhibitors by starting with appropriate Type-I inhibitors. Hybridization to Type-II tails followed by SAR studies led to the identification of potent and orally available DFG-out inhibitors (17).
In addition to the ligand-based approaches described previously, there has been significant progress in the progression of assays to identify DFG-out inhibitors. Klüter et al. developed probes that bind exclusively to the allosteric pocket. A Type-III inhibitor was hybridized with a peptide fragment of β-galactosidase that acted as an enzyme donor (ED) (18). When displaced from the kinase, the ED could then couple with the complementing portion of β-galactosidase (the enzyme acceptor) to generate a chemiluminescent signal. This technology was applied to both p38α and JNK2. Tecle et al. (19) utilized fluorescent probes designed to bind preferentially to the DFG-out conformation of p38α. The probes were used in fluorescence polarization assays with both phosphorylated and non-phosphorylated p38α to identify several DFG-out inhibitors. Munoz et al. (20) also applied a similar approach to p38α.
Simard et al. have described a different approach for an assay to identify DFG-out inhibitors.
Instead of labeling the ligand, they have chosen to label the protein in the region of the activation loop using a cysteine residue introduced via site-directed mutagenesis (21). The cysteine can then be coupled to an environmentally sensitive fluorophore that can reflect changes in its environment as the protein transitions from the DFG-in to DFG-out conformation. This approach has been applied to p38α to explore the SAR of a defined set of inhibitors (22) as well as identify compounds from a high-throughput screen (23).
We are unaware of any Type-II inhibitors reported for Syk. While there is a Syk protein crystal structure with the classic DFG-out inhibitor Gleevec (24), Syk maintains a DFG-in conformation (PDB code 1XBB) (25). Herein, we report on a protein structure/ligand-based approach to identify Type-II inhibitors. We utilized the protein crystal structure inhibitors found to occupy the deep pocket and DFG-out pocket of kinases in the RCSB and Pfizer’s protein crystal structure database to identify Type-II inhibitors of Syk. Screening of these resulted in the identification of two DFG-out inhibitors as well as an inhibitor that binds to a rare C-helix-out conformation.
Identification of Type-II inhibitors
Two approaches were utilized to identify Type-II-like ligands. The first employed ADOPT (26). This in-house protein relational database has a curated version of the RCSB as well as some internal protein crystal structures (27). Within ADOPT is a kinase knowledgebase containing information on 1100 kinase protein crystals representing 148 different kinases. Thirty-nine of these kinases are annotated DFG-out (26,28) covering 97 different ligands. Those ligands that were indicated to occupy a kinase in the DFG-out conformation were selected for screening, of which 14 were availablea.
The second approach involved analysis of our internal protein crystal structures. To identify any compounds that were found to be near ‘deep pocket’/DFG-out space, all kinase protein crystal ligands in the Pfizer protein crystal database were superposed with a structure of P38α in the DFG-out conformation. Six hundred and ninety ligands that reached toward the deep pocket or DFG-out pocket were identified. Compounds that were available for testing from these two analyses (374 total) were screened against the catalytic domain of Syk.
Determination of size of deep pockets
The C-helix-out deep pocket and the DFG-out pocket were defined and measured to gauge the relative sizes of these deep pockets. The size of the pocket is measured by the number of site points (site pts) assigned by Sitemapb to the pocket. Sitemap evenly fills the pocket with site pts at 1 Å intervals. The deep C-helix-out pocket is delineated by a plane defined by the C-alpha of Asp512, C-alpha of Lys402, and C-alpha of Met424; The pocket is defined by the space above this plane.
The DFG-out pocket is delineated by a triangle that is defined by C-alpha of Asp512, C-alpha of Glu420, and C=O of Val491. The pocket is defined by the space directly under this triangle.
Protein crystallization and data collection
Residues 343–635 of human Syk were cloned, expressed, and purified as described (25,29). Crystallization of Syk was achieved using a 6 mg/mL protein stock in 50 mm HEPES pH 7.50, 10% glycerol, 150 mm sodium chloride, 10 mm methionine, and 5 mm DTT. The protein was incubated overnight with 1 mm of the compound and 2% DMSO. Crystals were grown using sitting drop vapor diffusion in a 2-drop MRC plate (Swisssci). 0.5 + 0.5 μL drops were set up with a 80-μL well using a Mosquito (TTP Labtech). For the deep pocket compound 2, the well consisted of 25% (w/v) PEG 1500 or 28% (w/v) PEG MME 2000 and 0.1 m bis-tris pH 6.5. For the DFG-out compound 1, the well contained 25% (w/v) PEG 3350, 0.2 m sodium chloride, and 0.1 m bis-tris pH 6.5. For the final compound 3, the well contained 0.1 m bis-tris pH 6.6 and 30% (w/v) PEG MME 2000. Plates were incubated at 18 °C, and crystals grew within 3–7 days. The crystals were cryoprotected with a quick dip into 80% well solution and 20% ethylene glycol. Diffraction data were collected from flash-frozen crystals at 100 K at beamline 17-ID at the Advanced Photon Source of Argonne National Laboratory. Data were processed using the HKL2000 suite of software (30). Data collection statistics are summarized in Table 1. The structure of Syk complexed with 2 was solved by molecular replacement methods with the CCP4 version of PHASER (31), using apo Syk structure (PDB code: 1XBA) as a search model. After molecular replacement, maximum likelihood-based refinement of the atomic position and temperature factors was performed with REFMAC (32) and autoBUSTER (33), and the atomic model was built with the program COOT (34). The refined structure was then used as a starting model for all other complexes. Crystallographic statistics for the final models are shown in Table 1.
Table 1. Data collection and refinement statistics
aValues in brackets represent statistics for highest-resolution shells.
Cell parameters (Å)
a = 45.9, b = 71.6, c = 84.2
a = 48.0, b = 69.5, c = 83.6
a = 46.8, b = 68.1, c = 82.8
13 968 (700)
16 841 (851)
Average B-factor (Å2)
r.m.s.d. bonds (Å)
r.m.s.d. angle (°)
Syk enzyme assay
A LANCE TR-FRET assay (Perkin Elmer) was used to measure and compare the potency of compounds against Syk kinase domain (residues 343–635). Compounds were prepared in 100% DMSO at 2 mm and serial diluted, and 1 μL was transferred to a 384-well assay plate. Syk kinase domain (0.04 ng/mL), ATP (28 μm), and biotinylated peptide WGJ-2 (800 nm) (Perkin Elmer) in 20 mm HEPES, pH 7.4, 10 mm MgCl2, 1 mm BME, 0.005% Briji35, and 0.05% BSA (19 μL per well) were added to the assay plate and incubated with the compound for 60 min at room temperature. The reaction was stopped with the addition of 90 mm EDTA, 20 mm HEPES, 0.01% BSA, and 0.005% Brij35 (5 μL per well). After 5 min, a mixture of detection reagents Surelight APC-streptavidin (30 μg/mL) and Europium-labeled anti-PT66 (6 nm) in 20 mm HEPES, 0.01% BSA, and 0.005% Brij35 were added and incubated for 15 min (5 μL per well). Plates were read on an Envision instrument (Ex. 320 nm, Em. 665 nm). For each compound concentration, the 665/615 ratio was converted to percent inhibition based on the average ratio of the zero percent effect wells (no compound added) and hundred percent effect wells (no enzyme added) included on each assay plate. Plots of percent inhibition versus compound concentration were fit to a 4-parameter logistic non-linear regression model to calculate compound IC50s.
Our aim was to develop an approach to identifying deep pocket/DFG-out inhibitors that would have a timely impact on the project. We targeted a protein structure/ligand-based approach that we could execute quickly. To this end, we took advantage of the protein kinase crystal structures both within Pfizer and in the RCSB (27). The aim was to identify ligands that potentially bound in a Type-II manner. Two approaches were pursued. The first involved taking advantage of ADOPT (26). This in-house protein relational database has a curated and annotated version of the RCSB, wherein kinase PDBs were annotated as to which conformation the kinase is in (DFG-in or DFG-out) (28). Those ligands that were indicated to occupy a kinase in the DFG-out conformation were selected.
The second approach involved analysis of our internal protein crystal structures. In this case, the search was not restricted to validated DFG-out inhibitors, but extended to any compounds that were found to be near ‘deep pocket’/DFG-out space. As the DFG-out/DFG-in conformations are at some level a continuum, we wanted to identify all compounds with a potential to reach deeply into the pocket. All of our kinase structures were superposed and the ligands extracted. The ligands were then analyzed in the context of P38α in a DFG-out conformation. Those that extended into the DFG-out pocket were selected as potential DFG-out inhibitors. Compounds that ‘broke’ through the surface to the back were considered ‘deep pocket’ inhibitors and were selected as well. Compounds that were available from the Pfizer compound library were screened against Syk.
Screening resulted in the identification of three compounds (1, 2, and 3, Figure 1). Protein–ligand crystals structures were obtained for each. Compound 3 (Figure 2C) places the phenylamino-pyridopyrimidinone at the hinge, making a classic two-point interaction with Ala451 and the protein adopting the DFG-out conformation. The trifluoromethylbenzamide occupies the DFG-out pocket with the amide N-H making an interaction with Glu420 common to DFG-out inhibitors (14). While the amide carbonyl does not make a formal H-bond with the N-H of Asp512, it is likely weakly interacting with a distance of 3.5 Å.
Compound 1 (Figure 2A) places the quinoline at the hinge resulting in a one-point interaction with Ala451 and the protein adopting the DFG-out conformation. The terminal aryl group occupies the DFG-out pocket. The urea interacts with both the C-helix Glu420 and Asp512 in a typical DFG-out motif. The overall protein structures for compounds 1 and 3 are very similar, each with the DFG-out and the C-helix in, the C-helix Glu420 interacting with the conserved Lys402.
The dimethoxyquinoline 2 (Figure 2B) resides at the hinge, making a one-point interaction with the N-H of Ala451 as seen for compound 1. While compounds 1 and 2 are very similar with the pyrido group extending under the conserved Lys402 for each, the protein of compound 2 does not adopt the DFG-out conformation. The cyclopropyl group of compound 1 directs the aryl ring toward the DFG-out pocket. In the case of compound 2, the pyridinone directs the terminal benzyl group to the hydrophobic pocket resulting from the significant movement of the C-helix. The pyridinone carbonyl makes an internal H-bond to rigidify the structure in the appropriate conformation, as well as target Lys402. The amide makes a hydrogen bond with the N-H of Asp512.
The protein structure/ligand-based approach described resulted in the rapid identification of two kinase conformations not previously reported for Syk, DFG-out and C-helix out. Note that this approach did not require the development of complex assays or protein constructs. We identified that DFG-out conformations of Syk are more remarkable considering that the DFG-out inhibitor Gleevec binds to Syk in a DFG-in conformation (PDB code 1XBB) (25). A recent analysis of kinases and their propensity to adopt DFG-out conformations also suggested that Syk is unlikely to do so (35).
The C-helix-out conformation for compound 2 is especially striking. The significant movement of the C-helix results in the formation of a large hydrophobic pocket occupied by the benzyl group off the pyridinone. Evaluation of the occupation of this space by kinase inhibitors from the RCSB was performed by using the kinome.mdb database available in MOE 2010.10c. Analysis suggests that relatively few inhibitors penetrate this deeply. As can be seen in Figure 3, only three ligands from the RCSB occupy this space (B, C, D, spring green). In all three cases, the kinase is C-MET. This contrasts sharply with 112 PDB’s representing 16 kinases that occupy the DFG-out pocket. Thus, this rare deep C-helix-out pocket may offer new opportunities for the design of selective Syk inhibitors.
While only the previously described ligands enter into the deep C-helix-out pocket, there are other ligands that extend significantly beyond the gatekeeper. A more expansive list of compounds was identified by broadening the space beyond the gatekeeper occupied by the ligands (Figure 5). Note that even in this case, there are only eight deep C-helix-out ligands covering four kinases (C-MET, Aurora-A, PLK, and HCK). Lapatinib (A) (36) (red in Figure 4) is included as a reference, as it is often cited as the classic inhibitor that reaches past the gatekeeper (14,15,36). While it does induce a C-helix-out conformation, the ligand falls well short of the deep C-helix-out pocket. In this case, the size of the pocket is quite small.
Ostensibly, one of the key drivers for targeting the DFG-out pocket is that it is a well-defined hydrophobic pocket of reasonable size. In the case of p38α (J, Figure 5), the DFG-out pocket is 65 site pts (Methods). The DFG-out pocket for 3 was found to be 49 site pts. The DFG-out pocket for compound 1 is similarly sized at 59 site pts. The C-helix-out deep pocket of compound 2 compares favorably with a size of 58 site pts. The C-MET ligands B and C are 67 and 58 site pts, respectively. Thus, these deep penetrating C-helix out inhibitors create a rare hydrophobic pocket comparable in size to classic DFG-out pockets.
The deep C-helix-out pocket inhibitors (B, C, D, and 2) differ significantly from the recently described Type-1 1/2 inhibitor (14), which also extends beyond the gatekeeper. Type-1 1/2 pharmacophore included kinases that were in the DFG-in conformation, whereas the C-helix could be ‘in’ or ‘out’. Key to the Type-1 1/2 pharmacophore are interactions with the acid of the conserved C-helix glutamate (Glu420 for Syk) and the N-H of the DFG aspartate. None of the deep C-helix-out pocket inhibitors (2, B, C, and D) interact with the C-helix glutamate. The less deeply penetrating I [described by Zucotto et al. (14)] and H do interact with the C-helix glutamate. It is noteworthy that in these two instances, the C-helix is pulled forward resulting in movement of the ligand away from the deep C-helix-out pocket. As part of a review on Type-II inhibitors, Backes et al. (36) covered C-helix-out inhibitors including A and E. E occupies a deep pocket with a size of 54 site pts with no interaction with the C-helix glutamate.
Kinase selectivity data were obtained from Invitrogend. Compound 3 demonstrated >50% inhibition at 1 μm for eight of 36 kinases tested. Compounds 1 and 2 demonstrated >50% inhibition for four of 36 and five of 35 kinases, respectively. While compound 3 did inhibit a higher percentage of the kinases, it should be noted that it is significantly more potent. The deep C-helix-out inhibitor 2 did not demonstrate improved selectivity compared with the DFG-out inhibitor 1. It was hypothesized that targeting such a rarely seen pocket would endow the ligand with improved selectivity. However, the flexibility about the amide and pyridinone ring of 2 likely allows it to bind to the DFG-out conformation of kinases as well as the C-helix-out conformation reported here. Docking studies indicate that compound 2 can dock to docking models derived from the DFG-out inhibitor 1 (not published). Moreover perusal of the RCSB finds two similar analogs inducing DFG-out conformations. The ligands of 3CE3 (37) and 3B8Q (38) have Tanimoto coefficients (using FCF6 fingerprints) of 0.81 and 0.78 to compound 2, respectively. Appropriately constrained analogs would better take advantage of the deep C-helix-out pocket and may result in improved kinome selectivity.
Herein, we describe a protein structure/ligand-based design approach to successfully identify ligands that bind to novel conformations of Syk. By utilizing kinase protein crystal structures both in the public domain (RCSB) and within Pfizer’s protein crystal database, three Type-II ligands were found. Compounds 1 and 3 bind to the DFG-out conformation of Syk, while compound 2 induces the C-helix to move to the ‘out’ conformation. In this instance, the C-helix moved significantly to create a large hydrophobic pocket rarely seen in kinase protein crystal structures. It was anticipated that binding to this rare conformation may instill the ligand with greater kinome selectivity. While compound 2 was not found to be more kinome selective than the similar DFG-out ligand 1, it may be that the flexibility of the ligand allows it to bind to DFG-out conformations as well as C-helix out, thus expanding the potential number of kinase targets. As conformationally constrained analogs that selectively probe the deep C-helix out pocket are developed, it is anticipated the kinome selectivity profile will improve.
While we cannot disclose all of the ligands used in this study we have deposited in the supplementary material those ligands that were derived from the RCSB.
SiteMap Schrodinger, LLC, New York, NY, 2011.
Molecular Operating Environment (MOE), 2010.10; Chemical Computing Group Inc., 1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2010.