Bioorganic & Medicinal Chemistry Letters 15 (2005) 2851–2856

Homology modeling and docking study of cyclin-dependent

kinase (CDK) 10
Miao Sun, Zesheng Li,* Yuan Zhang, Qingchuan Zheng and Chia-chung Sun
Institute of Theoretical Chemistry, State Key Laboratory of Theoretical and Computational Chemistry, Jilin University,
Changchun 130023, PR China
Received 1 February 2005; revised 19 March 2005; accepted 23 March 2005
Available online 29 April 2005

Abstract—In order to understand the mechanisms of ligand binding and the interaction between the ligand and the cyclin-dependent
kinase 10 (CDK10), a three-dimensional (3D) model of the CDK10 is generated based on the crystal structure of the cyclin-depen-
dent kinase 2 (CDK2) (PDB code 1AQ1) by using INSIGHTII /Homology module. With the aid of the molecular mechanics and
molecular dynamics methods, the last refined model is obtained and is further assessed by PROFILE-3D and PROSTAT , which show
that the refined model is reliable. With this model, a flexible docking study is performed and the results indicate that the Lys39
and Asp94 form hydrogen bonds and have strong nonbonding interaction with adenosine 5 0 -triphosphate (ATP). From the docking
studies, we also suggest that the Leu141, Tyr21, and Val24 in CDK10 are three important determinant residues in binding as they
have strong nonbonding interaction with ATP. The hydrogen bonding interactions also play an important role for the stability of
the complex. Our results may be helpful for further experimental investigations.
Ó 2005 Elsevier Ltd. All rights reserved.

1. Introduction M transition of the cell cycle.16 Three gene isoforms of

human CDK10 were shown in 2000.17 A study by Kas-
Among the estimated 500–1000 human protein kinases,1 ten and Giordano showed that CDK10 interacts with
a family of kinases activated by a family of cyclins, the the N-terminus of the Ets2 transcription factor, which
cyclin-dependent kinases (CDKs), has been extensively contains the highly conserved pointed domain and
studied because of their essential role in the regulation transactivation domain.18 CDK10 was identified that it
of cell proliferation, of neuronal and thymus functions is one of the 19 new genes in nonsmall cell lung cancer.19
and of transcription.2–5 Cell cycle progression is tightly
controlled by the activity of CDKs.6 Ten cyclin-depen- The function of protein is determined essentially by its
dent kinases (CDK1–CDK10) are currently known, of corresponding three-dimensional (3D) structure20 and
which only CDKs1, 2, 3, 4, and 6 intervene directly in one of the major obstacles to further elucidate the
the cell cycle, while CDK7 plays only an indirect role molecular origin of the observed metabolic profiles of
as an activator of these CDKs.7,2 Furthermore, these enzymes is lack of the corresponding 3D struc-
CDK7, CDK8, CDK9 act as regulators of transcrip- tures. The detailed structures of CDK10 and of
tion.8–13 CDK1 controls G2/M transition and CDK2, CDK10–substrate complex remain unknown, and these
CDK3, CDK4, and CDK6 are implicated at G1/S. have hindered the further investigation in CDK10 and
cell cycle. Three isoforms of CDK10 have been princi-
In 1994, a human novel CDC2–related protein kinase pally expressed17 and in the present investigation we
(CDK10) was identified using two independent PCR- construct a 3D model of CDK10 isoform 2 and search
based strategies.14,15 A study by Li et al. using a domi- for the binding site of substrate. The 3D features of
nant negative mutant and an antisense vector led to the model are obtained by a homology modeling proce-
the suggestion that CDK10 was implicated at the G2/ dure based on the X-ray diffraction structure of CDK2
(PDB code: 1AQ1)21 and it is known that the homology
Keywords: Cyclin-dependent kinase; Docking; Molecular dynamics; model is effective for the 3D structure construction of
Molecular modeling. protein.22 The obtained result can be used to explain
* Corresponding author. Tel.: +86 431 8498960; fax: +86 431 substrate specificity and relate enzyme function to its
8498026; e-mail: [email protected] structure.

0960-894X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.03.088
2852 M. Sun et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2851–2856

This article describes the modeling of the human

CDK10 and CDK10 complexes with adenosine 5 0 -tri-
phosphate (ATP). The investigation is made to gain fur-
ther insight into the structural basis for complex of
CDK10 with ATP. In order to design novel and higher
affinity ligands, an understanding of the interaction be-
tween ATP and CDK10 at the molecular level would
be valuable.

2. Methods

All simulations are performed on the SGI O3800 work-

stations using INSIGHTII software package developed by
Biosym Technologies. The sequence of CDK10 is ob-
tained from the databank in the National Biomedical
Research Foundation (NBRF) of Georgetown Univer-
sity Medical Center (http://pir.georgetown.edu). Mole-
cular minimization (MM) and molecular dynamics
(MD) calculations are carried out with the aid of the
force field CVFF.

2.1. Homology of CDK10

Homology modeling is usually the method of choice

when a clear relationship of homology between the se-
quence of target protein and at least one known struc-
ture is found. This approach would give reasonable
results based on the assumption that the tertiary struc- Figure 1. The alignment of 1AQ1 and CDK10.
tures of two proteins will be similar if their sequences
are related.23 The homology protein is searched by FAS-
TA program. The sequence identity between CDK10 and 2.2. Identification of the binding site of CDK10
reference protein CDK2 (PDB code 1AQ1) is 43.667%.
The alignment of CDK10 (target) and CDK2 (template) The Binding-Site module is a suite of programs in IN-
is shown in Figure 1. Sixteen residues from 298 to SIGHTII for identifying and characterizing protein bind-
314 at the C-terminus are removed from the CDK10 ing sites and functional residues from protein. In this
model, since there is no good template for the fragment study, the ACTIVESITE-SEARCH program was used to
and these residues are far away from the catalysis search the protein binding site by locating cavities in
domain of CDK10. For modeling 3D structure of CDK10 structure, which can be used to guide the pro-
CDK10, the program MODELER 24 was used. MODELER tein–ligand docking experiment. Through comparing
is an implementation of an automated approach to the conserved residues in CDKs and combining the
comparative modeling by satisfaction of spatial searched results, we can predict the binding site of
restraints.25–27 CDK10.

The initial structure of modeling is revised by means of 2.3. Docking the substrate into the binding site
refining loops and rotamers, checking bonds and adding
hydrogen atoms, and then molecular mechanics (MM) Molecular docking can fit molecules together in a favor-
and molecular dynamics (MD) simulations were used able configuration to form a complex system. The struc-
to optimize the initial modeling structure. For energy tural information from the theoretically modeled
minimization, the 20,000 iterations of steepest descent complex may help us to clarify the catalytic mechanism
(SD) calculation were performed and then the conju- of enzyme. The 3D structure of the ATP was built with
gated gradient (CG) calculation was carried out until the BUILDER program and the geometry was optimized.
0.005 kcal mol
1 Å
1 of convergences on the gradient. For the reason of taking the interacting mode of CDK10
After the above simulations, the homology model is ob- with ATP, the advanced docking program Affinity29 was
tained by carrying out the molecular dynamics (MD) used to perform the automated molecular docking. The
calculation with the use of Discover3 software package28 best binding structures of the ligand to the receptor
in INSIGHTII . Explicit solvent model TIP3P water is also based on the energy of the ligand/receptor complex were
used. The homology model is solvent with a 10 Å water automatically found and in this procedure a combina-
cap from the center of mass of CDK10. Finally, a con- tion of Monte Carlo type and stimulated annealing pro-
jugate gradient energy minimization of full protein is cedure to dock a guest molecule with a host was
performed until the root mean-square (RMS) gradient employed. A key feature is that the ÔbulkÕ of the recep-
energy is lower than 0.001 kcal mol
1 Å
1. In this step, tor, defined as atoms not in the binding site specified,
the quality of the initial model is improved. held rigid during the docking process, while the binding
 M. Sun et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2851–2856 2853

site atoms and ligand atoms were moveable. The poten-

tial of complex was assigned according to the CVFF
force field. Nonbonding interactions were used for cell
multipole approach. When docking the ATP, Mg2+ is
considered in the complex. Finally, the docked com-
plexes of CDK10 with ligands were selected according
to the criteria of interacting energy combined with geo-
metrical matching quality. These complexes of CDK10
from docking study were used as the starting conforma-
tion for further energetic minimization and geometrical
optimization before the final models were achieved.

3. Results and discussion

3.1. Homology modeling of CDK10

Figure 3. The evaluation of the CDK10 final structure of by PROFILE-
By means of FASTA program, the sequence of CDK10
3D program.
was compared to all the known proteins in PDB, and
the results showed that 1AQ1 had the best sequence
identity (43.667%) with CDK10, so 1AQ1 was used to
model the 3D structure of the CDK10. All the side and 1AQ1, and the main results are listed in Table 1.
chains of CDK10 were set by AUTO _ROTAMER program From Table 1, we can see that the U–W values of
which uses the library proposed by Ponder and Rich- 87.5% residues in CDK10 fall within the favored regions
ards.30 In order to remove the steric contacts and get a of the Ramachandran plot, and 85.6% of 1AQ1 locate in
stable conformation, the energy minimizations of the favored regions. From Table 1, we can also see that
20,000 iterations and dynamics simulations of 150 ps eight bond angles deviate from the formal values in the
were performed. The variation of potential energy with 297 residues for CDK10, and for 1AQ1, two bond
time during the 150 ps of molecular dynamics simula- angles have greater values than the reference values in
tions is plotted in Figure 2. From Figure 2, we can see the 298 residues. The root mean-square deviation
that the potential energy falls rapidly in the first 20 ps, (RMSD) of the residue C-a atoms between CDK10
and then it decreases with very low deviation between and 1AQ1 is 0.61 Å, which is in the reasonable range.
two steps, and the dynamics simulations tend to equili- By the comparison of the 3D structure of CDK10 with
brium at 150 ps. Thus, we chose the conformation at that of 1AQ1 and the assessments mentioned above,
150 ps as the final 3D structure for the further study. we think that our final 3D structure of CDK10 is
reliable.
The final structure was further checked through PRO-
FILE-3D and PROSTAT program. The checked results of The final structure of CDK10 is presented in Figure 4,
PROFILE-3D are presented in Figure 3. From Figure 3, which shows that the final structure includes 10 a-helices
we know that all residues are scored positive, which and eight b-sheets. It is well known that for the structure
means that these residues are reasonable. The PROSTAT of 1AQ1, nine a-helices, and eight b-sheets are involved.
program was used to check the dihedral angles, bond From Figure 4, we can see that the structure of CDK10
lengths, and bond angles in the structure of CDK10 is typical for the kinase family (i.e., bilobal) and com-
prises an N-terminal domain of b-sheets and a larger
C-terminal domain mainly constituted by a-helices.
The N-lobe of CDK10 also contains the PISSLRE helix,
which correspond to the PSTALRE helix in CDK2.
This helix is another feature found in all CDKs and is
important for interaction with cyclins.31 During the
MD simulation, we found that the most variable

Table 1. The results of dihedral angles, bonds, and angles of 1AQ1

and CDK10 checked by PROSTAT program
Target protein 1AQ1 CDK10
%U–W Angles in core 87.5% 85.6%
Ramachandran region
Number bond distances with 0 0
significant deviations
Number bond angles with 2 8
significant deviations
Number residues examined 298 297
Figure 2. The potential energy as a function of simulations time (ps).
2854 M. Sun et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2851–2856

Figure 5. The binding site of CDK10.

can see that the binding pocket has b-strand-loop-a-

helix supersecondary structure and an activation loop.
The b-strand-loop-a-helix supersecondary structure
consists of the residues of Ile16-Gly17-Glu18, Gly22-
Ile23-Val24, Lys39, Val41, Leu86, in b-strand, Gly19,
Thr20, Tyr21, Val70, Tyr90, Cys91, Ala151 in loop,
and Leu155 in a-helix. Other CDKs and protein kinases
have similar overall structures and almost superimpos-
Figure 4. The final structure of CDK10. The helices are colored red able catalytic clefts that include the glycine-rich ATP
and sheets are colored orange. binding domains and the catalytic base, Lys33.32
Through the alignment we know that the Lys33 in
CDK2 is corresponding to Lys39 in CDK10. Except
domain of CDK10 is the T-loop, which shows that this Gly22, Ile23, and Tyr90, other residues, which com-
domain of CDKs is the most variable part of these pro- posed the binding site in CDK10 are conserved with
teins, and it might be a suitable region in which to look CDK2.
for drug selectivity.
3.3. Explicit characterization of the complexes
3.2. Identification of the binding site of CDK10
The validated substrate-free CDK10 was used to charac-
The CDK10 is folded into the typical bilobal structure, terize complexes of this enzyme with typical substrates.
with the smaller N-terminal domain consisting predomi- The significant compound ATP was selected for study.
nantly of b-sheet structure and the larger C-terminal do- The phosphorylation and ATP-hydrolysis reactions play
main consisting primarily of a-helices. There are no important roles in signal transduction and regulation of
significant differences in the domain orientations be- many proteins, especially protein kinases. The molecular
tween the inhibitor–enzyme complex and the ATP– structure of ATP consists of a purine base, a ribose ring,
CDK10 complex. The sequences and structures of and three phosphate groups. The 3D structure of ATP
CDK10 and CDK2 should be conserved because their was built with the BUILDER program and the geometry
main biological functions are similar. Thus, it is pre- was further optimized by using DISCOVER _3 program.
dicted that the ATP binds in a manner similar in both
structures. Those protein kinases that bind ATP have Hydrogen bonds play an important role for structure
the classic mononucleotide-binding fold, which contains and function of biological molecules, especially for the
a b-strand-loop-a-helix supersecondary structure, which enzyme catalysis. The hydrogen bonds presented in the
is a characteristic feature of the phosphate-binding complexes are listed in Table 2 and Figure 6. We can
site.32,33 In order to calculate the interactions between see from Table 2 and Figure 6 that there are eight hydro-
the active site of CDK10 and the substrates, the ATP- gen bonds between ATP and CDK10. The purine base
binding pocket was defined as a subset that contains res- of ATP has a hydrogen bond with the enzyme between
idues in which any atoms are within 6.0 Å from ATP. the NH2 group at the position of N7 and side-chain carb-
The binding-site was searched by Binding-Site module, oxyl of Cys91. The ribose hydroxyl group of ATP forms
which can be used to guide the protein–ligand docking two hydrogen bonds with the side-chain carboxyl of
experiment. The residues comprising ATP pocket of Asp94. The c-phosphate groups form three hydrogen
CDK10 are displayed in Figure 5. From Figure 5, we bonds with the residues Thr20, His132, and Asn139 of
 M. Sun et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2851–2856 2855

Table 2. Hydrogen bonds between ATP and CDK10 Table 3. The total interaction energy (Etotal), van der Waals energy
CDK10 ATP Hydrogen bond (Evdw) and electrostatic energy (Eele)
Residues Evdw (kcal/mol) Eele (kcal/mol) Etotal (kcal/mol)
Residues Atom Atom Length (Å) Angle (°)
Cys91 O Purine base NH 1.95 154.25 Asp94 2.199
37.633
35.434
Lys39 1.007
14.473
13.466
Thr20 O c-Phosphate O
2.14 165.29
Leu141
5.176 0.385
4.791
His132 OH c-Phosphate O
2.36 154.47
Tyr21
3.398
1.268
4.666
Asn139 NH c-Phosphate O
1.88 168.67
Lys39 H b-Phosphate O 1.68 166.30 Val24
3.496
0.757
4.253
Thr20
2.619
0.845
3.464
Tyr21 H b-Phosphate O
1.78 167.18
Tyr90
2.667
0.432
3.099
Asp94 O Ribose OH 1.62 157.23
Ile16
2.702
0.166
2.868
Asp94 O Ribose OH 1.62 177.03
Lys136
1.324
1.428
2.752
Cys91
1.297
1.434
2.731
Ser138
2.931 0.268
2.663
Glu92
1.108
1.435
2.543
His132
2.332
0.166
2.498
Gly19
2.447 0.106
2.341
Ala37
2.08
0.257
2.337
Met88
2.184
0.127
2.311
Ala151
1.796
0.513
2.309
Asp152
4.953 2.942
2.011
Glu18
1.79 0.135
1.655
Val70
1.69 0.059
1.631
Val161
1.414
0.052
1.466
Asn139 0.516
1.858
1.342
Sum
43.682
58.949
102.631

residues of Leu141, Tyr21, and Val24 in CDK10 are

three important determinants in binding as they have
strong interactions with the ligand. As shown in Tables
2 and 3, these results can serve as a guide to selection of
candidate sites for further experimental studies of site-
directed mutagenesis.
Figure 6. The hydrogen bonding interaction of complex CDK10–ATP.

4. Conclusions
CDK10, and the b-phosphate group forms two hydro-
gen bonds with Tyr21 and Lys39. In this work, we have constructed a three-dimensional
model of the CDK10 by INSIGHTII /Homology module.
To determine the key residues that comprise the binding After energy minimization and molecular dynamics
pocket of the model, the interaction energies of the sub- simulations, this refined model structure is obtained.
strates with each individual amino acid in the enzyme The last refined model is further assessed by PROFILE-
were also calculated. Significant binding-site residues 3D and PROSTAT , and the results show that this model
in the models were identified by the total interaction en- is reliable. The stable structure is further used to per-
ergy between the substrates and each amino acid resi- form the docking of ATP. Through the docking studies,
dues in the enzyme. This identification, compared with the model structures of the ligand–receptor complex are
a definition based on the distance from the substrate, obtained. The docking results indicate that conserved
can clearly show the relative significance for every resi- amino acid residues in CDK10 play an important role
due. The total interaction energies (lower than in maintaining a functional conformation and are di-

1.0 kcal/mol) between the ATP and each individual rectly involved in binding to donor and acceptor sub-
amino acid of CDK10 are listed in Table 3. As seen from strates. The interactions of the CDK10 and ATP
Table 3, Asp94, Lys39, Leu141, Tyr21, Val24, Tyr20, proposed in this study are useful to understand the po-
and Tyr90 have strong interaction energies with ATP. tential mechanisms of the CDK10 and ATP. As is well
The residue Asp94 has the strongest interaction with known, hydrogen bonds play an important role for
ATP and the interaction energy is
35.434 kcal/mol. structure and function of biological molecules, especially
The residue Lys39 is another key residue, which has for the enzyme catalysis. The residues of Lys39 and
strong interaction with ATP, which forms a hydrogen Asp94 are important for strong hydrogen bonding inter-
bond with CDK10 and the interaction energy with action with ATP and play a major role in catalysis of
CDK is
13.466 kcal/mol. Both of the residues Lys39 CDK10. Furthermore, these residues, as well as the
and Asp94 have strong electrostatic interaction with others in Table 2 and in Table 3, are suggested as candi-
ATP. They play important roles in the combination dates for further experimental studies of structure–func-
between the ATP and CDK10. On the other hand, the tion relationships.
2856 M. Sun et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2851–2856

Acknowledgements 13. Tassan, J. P.; Jaquenoud, M.; Leopold, P., et al. Proc.
Natl. Acad. Sci. U.S.A. 1995, 92, 8871–8875.
This work is supported by the National Science Founda- 14. Grana, X.; Claudio, P. P.; De Luca, A., et al. Oncogene
tion of China (20333050, 20073014), Doctor Founda- 1994, 9, 2097–2103.
15. Brambilla, R.; Draetta, G. Oncogene 1994, 9, 3037–
tion by the Ministry of Education, Foundation for
3041.
University Key Teacher by the Ministry of Education, 16. Li, S.; MacLachlan, T. K.; De Luca, A., et al. Cancer Res.
Key subject of Science and Technology by the Ministry 1995, 55, 3992–3995.
of Education of China, and Key subject of Science and 17. Sergere, J. C.; Thuret, J. Y.; Le Roux, G., et al. Biochem.
Technology by Jilin Province. Biophys. Res. Commun. 2000, 279, 738–739.
18. Kasten, M.; Giordano, A. Oncogene 2001, 20, 1832–
1838.
References and notes 19. Singhal, S.; Amin, K. M.; Kruklitis, R., et al. Cancer Biol.
Ther. 2003, 2, 291–298.
1. Mitch, K.; Jessie, E.; Vincent, M., et al. Genome Biol. 20. Lai, L. H. In Protein Structural Prediction and Molecular
2002, 9, 1–12. Design; Beijing University Press, 1993; 467, pp 1–15.
2. Morgan, D. O. Annu. Rev. Cell Dev. Biol. 1997, 13, 261– 21. Lawrie, A. M.; Noble, M. E.; Tunnah, P.; Brown, J. A.,
291. et al. Nat. Struct. Biol. 1997, 4, 796.
3. Progress in Cell Cycle Research: Editions ‘‘Life in Pro- 22. Luthy, R.; Bowie, J. U.; Eisenberg, D. Nature 1992, 356,
gress’’; Meijer, L., Jezequel, A., Roberge, M., Eds.; 83–85.
Station Biologique deRoscoff, 2003; Vol. 5, pp 235–248. 23. Kroemer, R. T.; Doughty, S. W.; Robinson, A. J., et al.
4. Vogt, P. K.; Reed, S. I. Current Topics in Microbiology and Protein Eng. 1996, 9, 493–498.
Immunology; Springer, 1998, 169pp. 24. Sali, A.; Blundell, T. L. J. Mol. Biol. 1993, 234, 779–
5. Pavletich, N. P. J. Mol. Biol. 1999, 287, 821–828. 815.
6. Norbury, C.; Nurse, P. Annu. Rev. Biochem. 1992, 61, 25. Sali, A.; Overington, J. P. Protein Sci. 1994, 3, 1582–1596.
441–470. 26. Sali, A.; Potterton, L.; Yuan, F., et al. Proteins 1995, 23,
7. Harper, J. W.; Adams, P. D. Chem. Rev. 2001, 101, 2511– 318–326.
2526. 27. Sali, A. Curr. Opin. Biotechnol. 1995, 6, 437–451.
8. Hengartner, C. J.; Myer, V. E.; Liao, S. M., et al. Mol. 28. Discover 3 User Guide, San Diego: MSI, USA, 1999.
Cell. 1998, 2, 43–53. 29. Affinity User Guide, San Diego: MSI, USA, 1999.
9. Serizawa, H.; Makela, T. P.; Conaway, J. W., et al. 30. Ponder, J. W.; Richards, F. M. J. Mol. Biol. 1987, 193,
Nature 1995, 374, 280–282. 775–791.
10. Garriga, J.; Peng, J.; Parreno, M., et al. Oncogene 1998, 31. Sharma, P.; Steinbach, P. J.; Sharma, M.; Amin, N. D.;
17, 3093–3102. Barchi, J. J.; Pant, H. C. JBC 1999, 274, 9600–9606.
11. Akoulitchev, S.; Chuikov, S.; Reinberg, D. Nature 2000, 32. Schulz, G. E.; Schirmer, R. H. Nature 1974, 250, 142–144.
407, 102–106. 33. Schulz, G. E.; Schiltz, E.; Tomasselli, A. G.; Frank, R.;
12. Wei, P.; Garber, M. E.; Fang, S. M., et al. Cell 1998, 92, Brune, M.; Wittinghofer, A.; Schirmer, R. H. Eur. J.
451–462. Biochem. 1986, 161, 127–132.