c-Src Binds to the Cancer Drug Ruxolitinib with an Active Conformation

The cancer drug Ruxolitinib is a potent janus kinase inhibitor approved for the treatment of the myeloproliferative neoplasms. In addition, Ruxolitinib has weak inhibitory activity against a panel of other kinases, including Src kinase. There is no structural information of Ruxolitinib binding to any kinase. In this paper, we determined the crystal structure of c-Src kinase domain in complex of Ruxolitinib at a resolution of 2.26 Å. C-Src kinase domain adopts the DFG-in active conformation upon Ruxolitinib binding, indicating Ruxolitinib is a type I inhibitor for c-Src. Ruxolitinib forms two hydrogen bonds with Met341, a water-mediated hydrogen bond with Thr338, and a number of van der Waals contacts with c-Src. Ruxolitinib was then docked into the ligand-binding pocket of a previously solved JAK1 structure. From the docking result, Ruxolitinib also binds JAK1 as a type I inhibitor, with more interactions and a higher shape complementarity with the ligand-binding pocket of JAK1 compared to that of c-Src. Since Ruxolitinib is a relatively small inhibitor and there is sizeable cavity between Ruxolitinib and c-Src ligand-binding pocket, we propose to modify Ruxolitinib to develop more potent inhibitors to c-Src.


Introduction
Protein kinases catalyze the transfer of a phosphoryl group from adenosine triphosphate (ATP) to serine, threonine or tyrosine residues of its substrate proteins [1]. Such posttranslational modifications serve as a mechanism to modulate enzymatic activity or molecular interactions of substrate proteins in response to endogenous and exogenous signals [1]. Phosphorylation plays a critical role in signal transduction and regulates numerous cellular processes including cell adhesion, invasion, proliferation, survival and angiogenesis [2]. Over-expression or mutations of protein kinases can lead to a variety of human diseases such as cancer and autoimmunity. Protein kinases are therapeutic targets for the treatment of human diseases [3]. A prototypical example, Imatinib, targets BCR-Abl, a constitutively active form of the Abl kinase that leads to chronic myeloid leukemia (CML), and is very successful in the treatment of this disease [4].
Because of a high degree of sequence conservation within the kinase domain, it is not surprising that most kinase inhibitors tend to have limited target specificity. Off-target effects can be beneficial in some cases, but can lead to side effects in other cases. Every kinase inhibitor has its unique and highly unpredictable target spectrum [5]. Understanding the mechanism behind the target specificity is an important goal that would enhance the use of existing kinase inhibitors and benefit the process of inhibitor development. For example, the structural information of Imatinib binding kinase was not only studied in complex with its intended target kinase Abl [6], but also studied in complex with other kinases, including c-Src, Lck, p38 [7][8][9]. These studies greatly help us understand the basis of kinase inhibition, selectivity, and potential off-target effects. In addition, these studies provide a structural scaffold for the development of new kinase inhibitors of different kinases.
Protein kinase inhibitors are typically divided into three subtypes: type I, type II and type III inhibitors. Type I inhibitors occupy the pocket primarily filled by ATP, and a catalytically important Asp-Phe-Gly (DFG) motif is kept in active conformation (referred to as DFG-in conformation). A typical example of a type I kinase inhibitor is the second-generation BCR-Abl inhibitor, Dasatinib. Type II inhibitors, such as Imatinib, occupy the ATPbinding pocket and an additional region, and the DFG motif is rotated by ,180u with respect to the active conformation (referred to as DFG-out conformation) [10][11][12]. Type III inhibitors bind regulatory domains outside the ATP-binding pocket, and modulate the kinase activity in an allosteric manner. Because the amino acids outside the ATP-binding pocket are less conserved relative to those in the pocket, it has been proposed that it might be easier to achieve kinase selectivity with type II or type III inhibitors.
A single residue within the ATP-site of protein kinases, termed the gatekeeper, plays an important role in forming the specificity pocket, and controls sensitivity to a variety of small molecule inhibitors. The gatekeeper residue varies among different protein kinases. Some kinases have a small residue (e.g. Thr, Ala, or Gly) at this position, and are readily targeted by structurally diverse classes of inhibitors. Other kinases possess a larger residue (e.g. Phe) at this position, and are more resistant [13]. Mutation of the gatekeeper residue is a common mechanism of resistance to kinase inhibitors. For example, substitution of BCR-Abl gatekeeper Thr-315 to Ile has led to resistance to Imatinib [14][15][16].  Ruxolitinib is a potent janus kinase inhibitor for the treatment of the myeloproliferative neoplasms (MPNs) [17]. It has potent inhibitory activity against JAK1 (IC 50 = 3.3 nM) and JAK2 (IC 50 = 2.8 nM), moderate activity against TYK2 (IC 50 = 19 nM) and weak activity against JAK3 (IC 50 = 428 nM) and a panel of other kinases, including Src kinase [18,19]. JAK2 mutations and activation play a fundamental role in the etiology of human MPNs. For example, about half of patients with MPNs carry a gain-of-function mutation in the JAK2 gene (JAK2 V617F) [17,18]. Ruxolitinib inhibits the dysregulated JAK2 signaling pathway, and showed significant benefits in clinical trials. In 2011, Ruxolitinib was approved for the treatment of intermediate or high-risk myelofibrosis [20].
There is no structural information of Ruxolitinib binding to any kinase. In this paper, we investigate the structural basis for Ruxolitinib binding to a kinase, using c-Src kinase domain as a prototypical system. By determining the structure of a Ruxolitinib complex at 2.26 Å resolution, we showed that the c-Src kinase domain adopts the DFG-in active conformation. Ruxolitinib forms two hydrogen bonds with Met341, a water-mediated hydrogen bond with Thr338, and a number of van der Waals contacts with c-Src. We then docked Ruxolitinib into the ligand-binding pocket of a previously solved JAK1 structure. From our docking result, Ruxolitinib also binds JAK1 as a type I inhibitor, with more interactions and higher shape complementarity with the ligandbinding pocket of JAK1 than that of c-Src. Since there is sizeable cavity between Ruxolitinib and the c-Src ligand-binding pocket, it is highly possible to modify Ruxolitinib to develop more potent inhibitors to c-Src.

Protein Expression and Purification
Chicken c-Src (residues 251-533) was prepared as previously described [21]. In brief, the protein was expressed with a 66 Histagged in Escherichia coli BL21DE3 cells, in the presence of the YopH phosphatase. The protein was first purified by Ni-NTA beads. The 66 His-tag was then removed by TEV cleavage. After cleavage, anion exchange chromatography and size-exclusion chromatography were utilized to further purify the protein. Protein in 50 mM Tris(pH 8.0), 100 mM NaCl, 5% Glycerol, 1 mM DTT was concentrated to 10 mg/ml and flash frozen for storage at 280uC.

Crystallization
The c-Src/ruxolitinib crystals were obtained at 18uC using the hanging drop vapor diffusion method.

Data Collection and Structure Determination
Data was collected at the Shanghai Synchrotron Radiation Facility (SSRF), beamline BL17U, and the Advanced Light Source (Lawrence Berkeley National Laboratory) beamlines 5.0.2 and 8.2.1. Data was reduced using HKL2000 [22]. Structure determination was carried out as described previously [23,24]. Initial phase determination was performed by molecular replacement with Phaser from the CCP4 package [25], using chain A of a previously solved c-Src structure(PDB code 2SRC) [26] as the search model. The structure was refined using Phenix.refine and Coot from the Phenix package [27]. The statistics of the crystallographic analysis are presented in Table 1. Graphical representations of structure were prepared using PyMol (DeLano Scientific, San Francisco, CA). The coordinates and structural factors have been deposited in the Protein Data Bank with accession code 4U5J.

Molecular Docking and structural analysis
Molecular docking was carried out using Molsoft ICM-Pro [28]. Ruxolitinib was docked into the ligand-binding pocket of JAK1 (pdb code: 3EYG) [29]. Volume of the ligand-binding pocket was calculated using POCASA [30]. Diagrams of protein-ligand interactions was generated using LIGPLOT [31].

Ruxolitinib binding to the ATP-binding pocket of c-Src
To see how cancer drug Ruxolitinib interacts with a kinase, we determined the X-ray structure of c-Src in complex with Ruxolitinib to a resolution of 2.26 Å . The statistics of data collection and model refinement are listed in Table 1. Each asymmetric unit contains two molecules of c-Src. There is welldefined electron density for Ruxolitinib in both c-Src molecules, which appears to be bound with full occupancy ( Figure S1). The structure of the c-Src is composed of a bi-lobed architecture (Nlobe and C-lobe) that is typical for protein tyrosine kinases. The ATP binding site of c-Src kinase is located at the folding cleft between the N-and C-terminal lobes, surrounded by the Hinge region, P-loop, Helix aC, and activation loop ( Figure 1A). Part of the activation loop (amino acids 412-423) exhibits no electron density, indicating this region is flexible. Therefore this region is not included in the pdb.
Ruxolitinib binds in the ATP-binding cavity of c-Src, with unambiguous electron density observed for the inhibitor ( Figure  S1). It is orientated such that the pyrrolopyrimidine rings point toward the hinge region, the cyclopentane ring points toward the C-lobe, while the propanenitrile group points toward the P-loop ( Figure 1A). The pyrrolopyrimidine rings form two hydrogen bonds with the main chain atoms of Met341 ( Figure 1B), as well as a water-mediated interaction with Src gatekeeper residue Thr338 ( Figure 1C). Besides the hydrogen bonds, Ruxolitinib also forms a number of van der Waals contacts with c-Src ( Figure S2).
DFG motif locates at the N-terminal end of the activation loop, and its conformation plays a pivotal role in kinase activity [32]. DFG-in and DFG-out conformations represent two extreme states of a continuum of possibilities [32]. Intermediate states are observed in some cases [26]. Three distinct conformations of the kinase domains of Src family kinases: a DFG-in active conformation, a DFG-intermediate inactive conformation, and a DFG-out inactive conformation [32]. In the DFG-in conformation (Figure 2A and 2B, pdbs 3LCK and 3DQW), the Aspartate side chain faces into the ATP binding site, and the phenylalanine side chain faces into the protein. In the DFG-out conformation ( Figure 2E, pdb 2OIQ), the backbone of the DFG motif is flipped, the aspartate sdie chain faces away from the ATP-binding site, while the phenylalanine side chain occupies the ATP-binding site. For example, the kinase domain of c-Src adopts DFG-out conformation upon binding to Imatinib [33]. In the DFG-intermediate conformation ( Figure 2D, pdb 2SRC), the DFG-motif adopts a conformation between the above two conformations. We compared the DFG-motif conformation in Src/Ruxolitinib complex with these three conformations. Apparently, upon Ruxolitinib binding, c-Src kinase domain is kept in DFG-in conformation ( Figure 2C).
The kinase domain of c-Src adopts different conformations upon binding of different inhibitors. In the c-Src/Imatinib complex, Imatinib occupies not only the ATP-binding site, but also extends beyond the DFG motif and occupies an additional region. The DFG motif is in the DFG-out conformation ( Figure 3A). In the c-Src/Ruxolitinib complex, Ruxolitinib only occupies the ATP-binding site, and c-Src adopts a DFG-in conformation. The P-loop moves closer to the C-lobe, leading to a tighter ATP-binding site. This could be due to an induced-fit mechanism ( Figure 3B). Based on these observations, Imatinib is a type II inhibitor for c-Src while Ruxolitinib is a type I inhibitor for c-Src.
Most kinase inhibitors are ATP-competitive and belong to type I inhibitors. The ATP-binding pocket is conserved among members of the kinase family. In addition to the ATP-binding pocket, type II inhibitor binds a less-conserved additional pocket. Type II inhibitor acts by inducing a conformation change such that the kinase is no longer able to function. It is proposed that type II inhibitor could achieve better selectivity. Nonetheless, biological efficacy, especially clinical efficacy, will be the ultimate arbiter.

Comparison of binding pockets in Src and JAK
Ruxolitinib is a Janus kinase inhibitor, with reported IC 50 values of 2.8 nM for JAK2, 3.3 nM for JAK1, 19 nM for TYK2, and 428 nM for JAK3 [18]. Lck, a Src family kinase, is inhibited by Ruxolitinib with a reported IC 50 value of 3.6 uM [19]. Using kinase assays, we found that Ruxolitinib inhibited c-Src with an IC 50 of 2.92 uM (Table S1). Therefore, Src kinases are inhibited by Ruxolitinib to a much lesser extent than JAK kinases. The sequence identity between Src and JAK family kinases is ,34% in the kinase domain ( Figure 4A & 4B). Besides the fact that JAK kinases have several extra stretches of amino acids ( Figure 4A, highlighted in green box), We identified several other key differences between Src kinase family and JAK kinase family. Thr338, the gatekeeper residue in c-Src, is replaced with Met in the JAK family; Met 341 of c-Src, the residue forming two hydrogen bonds with Ruxolitinib, is substituted with Leucine in the JAK family ( Figure 4A).
We then superimposed our structure with a JAK1 structure (pdb: 3EYG) [29]. Both kinase domains share a similar overall structure with a RMSD of 2.92 Å . The P-loop of JAK1 is closer to the C-lobe, leading to a tighter pocket, compared to that of C-Src ( Figure 5A). We then used POCASA [30] to analyze the ligandbinding pockets of c-Src (our structure) and JAK1 (pdb code: 3EYG). JAK1 has a smaller pocket with a volume of 311 Å 3 , while c-Src has a bigger pocket with a predicted volume of 450 Å 3 ( Figure 5B & 5C).   To better understand how Ruxolitinib differentiates these kinase families, we docked Ruxolitinib into a previously solved JAK1 structure (pdb: 3EYG) [29]. The docking result shows that Ruxolitinib only occupies the ATP-binding site, and the DFG motif is kept in DFG-in active conformation ( Figure 6A), suggesting that Ruxolitinib is also a type I inhibitor for JAK1. This is consistent with previous finding using competition binding assays [5]. Ruxolitinib is orientated such that the pyrrolopyrimidine rings point toward the hinge region. Instead of forming two hydrogen bonds with Met341 in c-Src, the pyrrolopyrimidine rings form two hydrogen bonds with the main chain atoms of Glu957 and Leu959 ( Figure 6B). A major difference between Ruxolitinib binding to c-Src and JAK1 lies on the orientation of the cyclopentane ring. It points toward to the C-lobe in the c-Src/ Ruxolitinib complex, while points toward N-lobe when binding to JAK1 ( Figure 6B). Another notable difference is that the propanenitrile group of Ruxolitinib makes contacts with JAK1, while it does not interact with c-Src ( Figure S3).
There is a high shape complementarity between Ruxolitinib and the JAK1 ATP-binding pocket ( Figure 6C), leading to more contacts with the residues surrounding the pocket than that of c-Src ( Figure S3). In the c-Src/Ruxolitinib complex, there is a sizeable cavity adjacent to Ruxolitinib in the ligand-binding pocket, especially between Ruxolitinib and the Helix aC of c-Src ( Figure 6D). More interactions and higher shape complementarity could be the reason why Ruxolitinib is a more potent inhibitor for JAK kinases than for Src kinases.
Ruxolitiib is a relatively small inhibitor with a molecular weight of 306.37 g/mol. Adding some extra groups to Ruxolitinib could lead to the discovery of new compounds which are more potent inhibitors for c-Src. For example, adding some chemical groups to fill the cavity between Ruxolitinib and Helix aC of c-Src might greatly improve new compound's affinity with Src kinases. However, because of the increase in ligand volume, the new compound would likely to be too big to fill the JAK ligand-binding pocket, therefore could be a poor inhibitor for JAK kinases.

Conclusions
In this paper, we present a crystal structure of Ruxolitinib bound to the kinase domain of c-Src. Although Ruxolitinib is not a potent inhibitor of c-Src, the crystal structure shows that the c-Src kinase domain is kept in the DFG-in active conformation that is characteristic of type I inhibitor. In addition, we docked Ruxolitinib into the ligand-binding pocket of a previously solved JAK1 structure. From our docking result, Ruxolitinib also binds JAK1 as a type I inhibitor. Compared with its binding to c-Src, Ruxolitinib has more interactions and a higher shape comple-mentarity with JAK1 ligand-binding pocket, which could be the main reason why Ruxolitinib is a much more potent inhibitor against JAK kinases than Src kinases. In our structural analysis, we noticed that there is sizeable cavity between Ruxolitinib and c-Src ligand-binding pocket, suggesting that modifying Ruxolitinib could lead to the development of potent c-Src inhibitors.