Structure of cyclin-dependent kinase 2 (CDK2) in complex with the specific and potent inhibitor CVT-313

The binding of CVT-313 to cyclin-dependent kinase 2 (CDK2) was confirmed using a thermal shift assay and its cellular potency was demonstrated in a CTG-based cell-proliferation assay. The crystal structure of the CDK2–CVT-313 complex was solved to a resolution of 1.74 Å. The structural information obtained from this study is expected to facilitate the design of more potent and selective CDK2 inhibitors for the treatment of cancer.

Abstract

CVT-313 is a potent CDK2 inhibitor that was identified by screening a purine-analogue library and is currently in preclinical studies. Since this molecule has the potential to be developed as a CDK2 inhibitor for cancer therapy, the potency of CVT-313 to bind and stabilize CDK2 was evaluated, together with its ability to inhibit aberrant cell proliferation. CVT-313 increased the melting temperature of CDK2 by 7°C in thermal stabilization studies, thus indicating its protein-stabilizing effect. CVT-313 inhibited the growth of human lung carcinoma cell line A549 in a dose-dependent manner, with an IC₅₀ of 1.2 µM, which is in line with the reported biochemical potency of 0.5 µM. To support the further chemical modification of CVT-313 and to improve its biochemical and cellular potency, a crystal structure was elucidated in order to understand the molecular interaction of CVT-313 and CDK2. The crystal structure of CDK2 bound to CVT-313 was determined to a resolution of 1.74 Å and clearly demonstrated that CVT-313 binds in the ATP-binding pocket, interacting with Leu83, Asp86 and Asp145 directly, and the binding was further stabilized by a water-mediated interaction with Asn132. Based on the crystal structure, further modifications of CVT-313 are proposed to provide additional interactions with CDK2 in the active site, which may significantly increase the biochemical and cellular potency of CVT-313.

1. Introduction  

Cyclin-dependent kinases (CDKs) belong to the family of serine/threonine protein kinases. These are multifunctional enzymes that modify various protein substrates involved in cell-cycle progression, stem-cell self-renewal, epigenetic regulation and DNA repair (Harper & Adams, 2001 ▸; Pines, 1991 ▸; Peyressatre et al., 2015 ▸; Malumbres, 2014 ▸). CDKs catalyze the phosphorylation of their substrates by transferring a phosphate group from adenosine triphosphate (ATP) onto specific amino acids on the target proteins. 20 mammalian CDKs are known to date and each of them performs a specific biological role. CDK1, CDK2, CDK4 and CDK6 act as regulators of different stages of the cell cycle, CDK3, CDK7, CDK8, CDK9 and CDK10 are involved in the regulation of transcription and CDK11 is involved in the regulation of RNA splicing. CDKs themselves are regulated by a set of proteins called cyclins. Each CDK associates with a specific cyclin and is activated by phosphorylation at the active-site threonine by CDK-activating kinase (CAK). Often, the same CDK associates with different cyclin partners in order to accomplish different functions.

Cyclin-dependent kinase 2 (CDK2) is one of the most extensively targeted kinases for drug development because of its involvement in a critical stage of cell division. CDK2 associates with cyclin E and inactivates the retinoblastoma protein by phosphorylation to initiate G1/S transition. CDK2 further associates with cyclin A to facilitate orderly S-phase progression (Sherr, 1994 ▸; Hanahan & Weinberg, 2011 ▸). Aberrant CDK2 activity leads to cell-cycle dysregulation, causing uncontrolled proliferation. The overexpression of CDK2 and associated cyclins is observed in multiple cancer types, making CDK2 an attractive drug-discovery target (Hanahan & Weinberg, 2000 ▸; Gu et al., 1992 ▸; Sherr, 1996 ▸; Knockaert et al., 2002 ▸).

There are many reports of the development of CDK2 inhibitors, mainly by targeting the ATP-binding site in the protein (Li et al., 2015 ▸). Based on their binding modes, kinase inhibitors are classified as type I and type II inhibitors. Type I inhibitors competitively target the ATP-binding site in the active conformation of the kinase, whereas type II inhibitors bind to the ATP-binding pocket in its inactive conformation. The active and inactive conformations of the kinase are dependent on the position of the DFG loop (Vijayan et al., 2015 ▸). Several type I inhibitors of CDKs have successfully passed the preclinical development stage and are currently in various phases of clinical trials. Flavopiridol (Tolero Pharma), roscovitine (Cyclacel) and dinaciclib (Merck & Co.) are a few such examples (Li et al., 2015 ▸; Malumbres et al., 2008 ▸). The Novartis Institute of Biomedical Research has developed a potent type II inhibitor of CDK2 named K03861 (Alexander et al., 2015 ▸). Although these molecules primarily target CDK2, some of them also inhibit other related kinases such as CDK1, CDK5, CDK9 and/or CDK4/6 (Law et al., 2015 ▸; De Azevedo et al., 1996 ▸, 1997 ▸; Carlson et al., 1996 ▸; Liu et al., 2012 ▸; Goodyear & Sharma, 2007 ▸; Flynn et al., 2015 ▸; Martin et al., 2013 ▸). The search for more specific CDK2 inhibitors resulted in the identification of CVT-313 {2,2′-[(6-{[(4-methoxyphenyl)methyl]­amino}-9-(1-methylethyl)-9H-purin-2-yl)imino]bis(ethanol); Fig. 1 ▸}, a purine-based analogue with moderate potency against CDK2 (IC₅₀ ≃ 0.5 µM) and an approximately eightfold selectivity over CDK1 and a >400-fold selectivity over CDK4. However, CVT-313 is equipotent against CDK5 (Jorda et al., 2018 ▸). The mode of binding of CVT-313 (as a type I or type II binder) has not previously been explored.

Figure 1.

Chemical structure of CVT-313.

CVT-313 has shown promising results both as a single agent and in combination with the CDK4 inhibitor indolocarbazole in cell-based assays (Jalili et al., 2012 ▸). Jalili and coworkers reported a synergistic effect of CVT-313 with indolocarbazole in a panel of melanoma cell lines. Both CVT-313 and indolocarbazole demonstrated poor potency against the melanoma cell line LOXIMVI, with <20% inhibition at 5 and 0.5 µM, respectively. However, when a combination of CVT-313 (2 µM) and indolocarbazole (0.25 µM) was used, cell-growth inhibition was about 80%, demonstrating a significant synergistic effect. Similar results were also observed against a panel of other melanoma cell lines (Jalili et al., 2012 ▸).

In the present study, the binding of CVT-313 to CDK2 was confirmed using a thermal shift assay and its cellular potency was demonstrated in a CTG-based cell-proliferation assay. Since a crystal structure of the CDK2–CVT-313 complex will be very useful in identifying the interactions and the binding mode of the ligand, the crystal structure of the CDK2–CVT-313 complex was solved to a resolution of 1.74 Å. This is the first report of a crystal structure of CDK2 with bound CVT-313. The structural information obtained from this study is expected to facilitate the design of more potent and selective CDK2 inhibitors for the treatment of cancer.

2. Materials and methods  

2.1. Chemicals, bacterial strains and vectors  

Escherichia coli strain BL21 (DE3) was obtained from Invitrogen, India; Ni–NTA beads and the Superdex 200 10/300 column were procured from Qiagen, USA and GE Healthcare Biosciences, India, respectively. The pET-21a-His-CDK2 construct was custom-synthesized by GenScript, USA. CVT-313 was obtained from Sigma–Aldrich, India.

2.2. Expression and purification of recombinant CDK2  

CDK2 cDNA corresponding to amino acids 1–298 (the full length) was synthesized and subcloned into pET-21a vector. Transformants of the E. coli BL21 (DE3) strain containing the pET-21a-CDK2 construct were induced at an OD₆₀₀ of ∼0.6 using 0.2 mM isopropyl β-d-1-thiogalactopyranoside at 18°C. The cells were harvested after 16 h of induction. The cell pellet was washed with lysis buffer (10 mM Tris–HCl pH 7.5, 25 mM NaCl, 1 mM EDTA, 1 mM PMSF) and was treated with 100 µg ml⁻¹ lysozyme for 30 min. The cells were lysed by sonication and the cell debris was clarified by centrifugation at 12 000 rev min⁻¹ for 40 min at 4°C. The supernatant was bound to Ni–NTA beads pre-equilibrated with 10 mM Tris–HCl pH 7.5, 25 mM NaCl, 1 mM EDTA and eluted with 300 mM imidazole. Ni–NTA fractions containing the CDK2 protein were pooled and further purified by size-exclusion chromatography using a Superdex 200 column. The purified protein was stored in a buffer consisting of 10 mM HEPES pH 7.5, 25 mM NaCl, 1 mM EDTA, 1 mM DTT.

2.3. Thermal shift assay  

The melting temperatures of the CDK2 protein alone and in the presence of the inhibitor CVT-313 were determined using an ABI Prism 7500 instrument (Applied Biosystems, USA) using SYPRO Orange dye as an indicator. 5 µM CDK2 protein in 10 mM HEPES pH 7.5, 25 mM NaCl, 1 mM EDTA, 1 mM DTT was incubated with either 10 or 100 µM CVT-313 for 1 h, followed by the addition of SYPRO Orange dye. Protein Thermal Shift version 1.1 (Applied Biosystems) was used to determine the melting temperatures.

2.4. Cell-viability assay  

A549 (ATCC CCL185) cells grown in complete Ham’s F-12 medium were plated onto 96-well plates (2000 cells per well) and incubated for 24 h. CVT-313 was added to the cells at increasing concentrations (0.0015–10 µM) in triplicate to generate a dose–response curve. The final concentration of DMSO in the assay was 0.1%. After 72 h of incubation, the assay was terminated using 50 µl CellTiter-Glo A reagent (Promega) and the number of viable cells was quantified. Luminescence readings were taken using a Victor-3 multimode reader (Perkin Elmer, USA). The data were analyzed using GraphPad Prism version 4.

2.5. Crystallization of CDK2 with CVT-313  

Crystals of CDK2 were grown using the hanging-drop vapor-diffusion method. Briefly, the protein was concentrated to 18 mg ml⁻¹ and mixed with reservoir buffer consisting of 0.1 M Tris pH 8.0, 0.1 M NaCl, 20% PEG monomethyl ether 550 in a 1:1(v:v) ratio, and drops were set up in a 24-well plate. The plates were incubated at 18°C and crystals appeared after four days of incubation. The apo crystals were soaked in reservoir buffer containing 0.5 mM CVT-313 for three days to bind the compound.

2.6. X-ray data collection and structure determination  

Crystals of CDK2 were flash-cooled at 100 K using 20% glycerol in reservoir buffer as a cryoprotectant. The diffraction data sets were collected on the MX2 beamline at the Australian Synchrotron to a maximum resolution of 1.74 Å. Data collection, reduction and scaling were performed using Blu-Ice (McPhillips et al., 2002 ▸), XDS and the CCP4 suite version 7.1 (Kabsch, 2010 ▸; Winn et al., 2011 ▸). The structure was solved by molecular replacement using PDB entry 3qxp (CDK2 in complex with the inhibitor RC-3-89; Schönbrunn et al., 2013 ▸) as a search model. Alternate cycles of restrained refinement and manual rebuilding were performed with REFMAC version 5.2.00013 (Murshudov et al., 2011 ▸) and Coot version 6 (Emsley et al., 2010 ▸), respectively. 5% of the reflections were randomly excluded from refinement to monitor the free residual factor (R _free). After completing the protein model, the presence of electron density corresponding to the bound ligand was identified and fitted to the ligand model to complete the structure refinement (Table 1 ▸).

Table 1. X-ray data-collection and structure-refinement statistics for the CDK2–CVT-313 complex (PDB entry 6inl).

Values in parentheses are for the outermost shell.

Data collection
Space group	P212121
a, b, c (Å)	53.645, 71.994, 72.075
Resolution range (Å)	53.030–1.740 (1.77–1.74)
Total No. of reflections	194317 (10438)
Unique reflections	29380 (1548)
Completeness (%)	99.8 (96.2)
R merge	0.035 (0.140)
〈I/σ(I)〉	21.5 (6.5)
Multiplicity	6.6 (6.7)
Refinement
Resolution (Å)	50.94–1.750
No. of reflections	27342
Completeness (%)	99.8
R work/R free	0.198/0.250
R.m.s. deviations
Bond lengths (Å)	0.019
Bond angles (°)	1.989
Ramachandran plot
Residues in most favored region (%)	95.86
Residues in allowed region (%)	4.14
Average B factors (Å2)
Macromolecule atoms	29.697
Ligand atoms	28.634
Solvent atoms	38.831

3. Results  

3.1. Thermal stabilization of CDK2 by CTV-313  

In the thermal denaturation assay, the melting temperature (T _m) of CDK2 was determined to be 45°C. Addition of CVT-313 significantly enhanced the thermal stability of the protein. The T _m shifts observed were 6.8 and 10.2°C at 10 and 100 µM CVT-313, respectively (Fig. 2 ▸, Table 2 ▸). The data suggest strong binding of CVT-313 to CDK2 to form a stable complex.

Figure 2.

Thermal melting curves of CDK2 alone and in the presence of 10 or 100 µM CVT-313.

Table 2. Stabilization effect of CVT-313 on CDK2.

Ligand	ΔT m ± SD (°C)
CVT-313, 10 µM	6.8 ± 0.12
CV-T313, 100 µM	10.2 ± 0.19

3.2. Anti-proliferation assay  

The effect of CVT-313 at different concentrations on the viability of A549 cells was determined using a CellTiter-Glo A (CTG) cell-viability assay. As shown in Fig. 3 ▸, following treatment with CVT-313 the viability of A549 cells decreased in a dose-dependent manner, with an IC₅₀ of 1.28 µM.

Figure 3.

Determination of the IC₅₀ value of CVT-313 for the inhibition of proliferation of A549 cells. The concentration range of CVT-313 used was 0.0015–10 µM.

3.3. CVT-313-bound structure of CDK2  

The crystal structure of the CDK2–CVT-313 complex was determined to a resolution of 1.74 Å, and CDK2 was found to be monomeric in nature. It has a typical bilobal structure, with the N-terminal domain (Met1–Asp86) predominantly being comprised of a seven-stranded parallel β-sheet (β1, β2, β3, β4, β5, β6 and β7) sandwiched by a C-terminal domain consisting primarily of ten α-helices, with one at the top (α1) and eight at the bottom (α2, α3, α4, α5, α6, α7, α8 and α9), and another helix, α10, located in the C-terminal tail of the protein (Fig. 4 ▸). The two domains are connected by a peptide strand (residues Glu81–Leu83) which acts as a hinge linker ensuring their flexible rotation. The C-terminal domain has two important segments: the phosphorylation and activation segments. The phosphorylation segment contains the catalytic residue (Glu51) required for the phosphorylation of Thr160 in the T-loop for its activation. The activation segment spans the region between the conserved DFG motif (Asp145-Phe146-Gly147) and the APE motif (Ala170-Pro171-Glu172). The unique PSTAIRE motif (Pro45–Glu51) in CDK2 that has a key role in its interaction with the cyclin subunit is found in the N-terminal α1 helix. Density for Pro45 and Ser46 and Leu147–Glu162 was missing in the CDK2–CVT-313 crystal structure, but did not affect the determination of the overall structure (De Azevedo et al., 1996 ▸; Ikuta et al., 2001 ▸). The final refined structural model of the CDK2–CVT-313 complex contains amino-acid residues 1–298. The structural models were generated using PyMOL. The coordinates of the CDK2–CVT313 crystal structure have been deposited in the Protein Data Bank (PDB entry 6inl).

Figure 4.

Ribbon representation of the monomeric structure of CDK2 (green) with the bound ligand CVT-313 (blue). This figure was created using PyMOL.

In the CDK2–CVT-313 crystal structure, F _o − F _c electron density was clearly observed for CVT-313 after an initial round of rigid-body refinement. The ligand is bound in the ATP-binding site in the cleft between the N-terminal and C-terminal lobes. CVT-313 replaces ATP by occupying the site where the adenine ring of the ATP would bind and interacts directly with Leu83, Asp86 and Asp145 through hydrogen bonds. The core purine moiety of CVT-313 makes two hydrogen bonds through its N3 and N6 atoms to the backbone NH and CO groups of Leu83 (N–N distance of 3.2 Å; O–C distance of 2.8 Å) at the hinge region. One of the hydroxy groups of CVT-313 forms a hydrogen bond to the side-chain carbonyl O atom of Asp145 (O–O distance of 3.2 Å) in the DFG motif. The other hydroxy group forms a hydrogen bond to the side-chain carbonyl group of Asp86 (O–O distance of 2.6 Å). A water-mediated interaction was observed between the O atom of CVT-313 and the side-chain carbonyl O atom of Asn132 (Fig. 5 ▸). The isopropylpurine group of CVT-313 forms an edge-to-face aromatic interaction with the side chain of Phe80. Five water molecules are involved in the network to generate solvent contacts in the solvent-exposed region.

Figure 5.

(a) The 2F _o − F _c electron-density map contoured at the 1σ level for the bound CVT-313 is shown as a blue mesh and the F _o − F _c electron-density map contoured at the 2σ level is shown as a red mesh. (b) Binding-site structure of CVT-313 bound to CDK2. The interacting residues within a 4.0 Å distance of CVT-313 are labeled. Hydrogen bonds are represented as dotted lines.

3.4. Comparison of apo and CVT-313-bound structures of CDK2  

The structure of the CDK2–CVT-313 complex was compared with that of apo CDK2 (PDB entry 1pw2; Wu et al., 2003 ▸) and was found to be similar, with a root-mean-square deviation of 0.221 Å for 227 C^α atoms. The CVT-313–CDK2 complex structure shows the DFG-in conformation, as expected for a type I kinase inhibitor (Fig. 6 ▸ a). For comparison, the DFG-out conformation of the type II inhibitor K03861 (PDB entry 5a14) bound to CDK2 is shown in Fig. 6 ▸(b). In the DFG-out conformation, the phenylalanine residue moves out of the hydrophobic pocket and occupies the ATP-binding pocket (Alexander et al., 2015 ▸).

Figure 6.

(a) DFG-in conformation of the CDK2–CVT-313 complex with Phe146 in the hydrophobic pocket. (b) DFG-out conformation of the CDK2–K03861 complex with Phe146 moved away from the hydrophobic pocket.

4. Discussion  

Cyclin-dependent kinases are one of the most sought-after classes of drug-discovery targets for various types of cancers and for inflammatory and neurodegenerative diseases. Three drug molecules targeting CDK4/6 have been approved for the treatment of recurrent breast cancer (palbociclib, ribociclib and abemaciclib) and the pan-CDK inhibitor Flavopiridol has been approved for lymphocytic leukemia. CDK2 is a known target for the development of cancer drugs because of its involvement in cell-cycle regulation and proliferation, and there are many CDK2 inhibitors that are already in advanced stages of clinical trials. Since no CDK2-specific inhibitors have been clinically approved to date, continuous efforts are being made to develop anti-CDK2 inhibitors from purine-based and pyrimidine-based scaffolds. CVT-313 is a promising candidate that was identified from a purine-based scaffold with significant potential for development as a drug molecule.

In the current study, significant thermal stabilization of CDK2 (ΔT _m ≃ 7°C) by CVT-313 has been demonstrated. This is in line with the reported biochemical potency of CVT-313 to inhibit CDK2 activity (IC₅₀ = 0.5 µM). The biochemical potency was translated into cellular activity, as observed in its growth-inhibitory potency towards A549 lung carcinoma cells (IC₅₀ of 1.2 µM). A similar antiproliferative activity has been reported against other tumor cell lines, including human colon adenocarcinoma and human pancreatic carcinoma (Brooks et al., 1997 ▸), showing the potential use of CVT-313 against different types of cancers. Further chemical modification of CVT-313 is necessary to achieve better potency to make it a drug-development candidate. For such an approach, binding and structural information on CVT-313 in complex with CDK2 would be very useful. The structure of CVT-313 bound to CDK2 determined in the present study shows that CVT-313 binds at the hinge region, like other known hinge binders. Although CVT-313 stabilizes the protein with multiple interactions at the hinge region and the DFG motif, the relatively lower biochemical potency could be explained by comparing the structure of CDK2 bound to CVT-313 with those of CDK2 bound to dinaciclib (PDB entry 4kd1; Martin et al., 2013 ▸) and to roscovitine (PDB entry 2a4l; De Azevedo et al., 1997 ▸). In CVT-313-bound CDK2, the activation loop remains highly flexible owing to a lack of interactions in this region. Electron density for the activation loop is not seen in the crystal structure of the CDK2–CVT-313 complex owing to its high flexibility. On the other hand, the activation loop becomes rigidified upon the binding of both dinaciclib and roscovitine owing to an elaborate network of hydrogen bonds and van der Waals interactions. The rigidity of the activation loop is seen in both the dinaciclib-bound and the roscovitine-bound CDK2 structures (Fig. 7 ▸). The flexibility of the activation loop is one of the main reasons for the observed lack of potency in the case of CVT-313 (Martin et al., 2013 ▸).

Figure 7.

Superimposition of the crystal structures of the CDK2–CVT-313 (green), CDK2–dinaciclib (pink) and CDK2–roscovitine (yellow) complexes, showing the differences in the conformation of the activation loop. The discontinuous activation loop observed in CVT-313-bound CDK2 (green) and the highly ordered continuous activation loops in dinaciclib-bound CDK2 (pink) and roscovitine-bound CDK2 (yellow) are indicated by arrows.

From the detailed structural analysis of the complex, further modifications to improve the potency and drug properties of CVT-313 can be proposed. Near the gatekeeper amino acid, Phe80, space is available to expand the ligand size. Substitution of the isopropyl ring with a favorable halogen (F, Cl, Br or I) may allow the ligand to form halogen–π interactions with the side chain of Phe80. This substitution may help the ligand to displace the position of the C-helix by translation and rotation into a conformation that is incompatible with cyclin binding, as reported in the literature (Martin et al., 2017 ▸). A second ligand-modification strategy can be applied at the methyl group-binding site, which the crystal structure shows to be a polar pocket with many bound water molecules. Therefore, replacing the nonpolar methyl group at the methoxy­benzylamino position with a polar group such as an OH, CO, NH₃ or COOH group could be useful. This substitution is expected to establish either a direct interaction or a water-mediated interaction with the nearby hydrophilic amino acids Glu8 and Lys20.

In conclusion, our data show that CVT-313 is a CDK2 inhibitor with moderate cellular potency. X-ray crystallo­graphic studies and thermal stability data show that CVT-313 forms a stable complex with CDK2. The interactions identified from the crystal structure have provided significant input for the modification of the structure of CVT-313 and optimization of its properties to develop it as a more potent and selective CDK2 inhibitor for the treatment of a wide range of cancers.

Supplementary Material

PDB reference: CDK2–CVT-313 complex, 6inl