Discovery of KDM5A inhibitors: Homology modeling, virtual screening and structure-activity relationship analysis

Article abstract of DOI:10.1016/j.bmcl.2016.03.048

Herein we report the discovery of a series of new KDM5A inhibitors. A three-dimensional (3D) structure model of KDM5A jumonji domain was firstly established based on homology modeling. Molecular docking-based virtual screening was then performed against c

Full text of DOI:10.1016/j.bmcl.2016.03.048

Accepted Manuscript
Discovery of KDM5A Inhibitors: Homology Modeling, Virtual Screening and
Structure-Activity Relationship Analysis
Xiaoai Wu, Zhen Fang, Bo Yang, Lei Zhong, Qiuyuan Yang, Chunhui Zhang,
Shenzhen Huang, Rong Xiang, Takayoshi Suzuki, Lin-Li Li, Sheng-Yong Yang
PII:
S0960-894X(16)30268-2
DOI:
http://dx.doi.org/10.1016/j.bmcl.2016.03.048
BMCL 23691
Reference:
Bioorganic & Medicinal Chemistry Letters
To appear in:
Received Date:
Revised Date:
Accepted Date:
24 October 2015
1 March 2016
14 March 2016
Please cite this article as: Wu, X., Fang, Z., Yang, B., Zhong, L., Yang, Q., Zhang, C., Huang, S., Xiang, R., Suzuki,
T., Li, L-L., Yang, S-Y., Discovery of KDM5A Inhibitors: Homology Modeling, Virtual Screening and Structure-
Activity Relationship Analysis, Bioorganic & Medicinal Chemistry Letters (2016), doi: http://dx.doi.org/10.1016/
j.bmcl.2016.03.048
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Discovery of KDM5A Inhibitors: Homology Modeling, Virtual Screening and
Structure-Activity Relationship Analysis
Xiaoai Wua,b,#, Zhen Fanga,#, Bo Yangc,d,#, Lei Zhonga, Qiuyuan Yanga, Chunhui
Zhanga, Shenzhen Huanga, Rong Xiange, Takayoshi Suzukif, Lin-Li Lic,*, Sheng-Yong
Yanga,*
aState Key Laboratory of Biotherapy and Cancer Center, West China Hospital, and
Collaborative Innovation Center for Biotherapy, Sichuan University, Sichuan 610041,
China.
bDepartment of Nuclear Medicine, West China Hospital, Sichuan University, Sichuan
610041, China.
cKey Laboratory of Drug Targeting and Drug Delivery System of Ministry of
Education, West China School of Pharmacy, Sichuan University, Sichuan 610041,
China.
dDepartment of Pharmacy, Panzhihua Central Hospital, Panzhihua, Sichuan 617067,
China.
eDepartment of Clinical Medicine, School of Medicine, Nankai University, Tianjin,
300071, China.
fGraduate School of Medical Science, Kyoto Prefectural University of
Medicine,Taishogun Nishitakatsukasa-Cho, Kita-ku, Kyoto 403-8334, Japan
# These authors contributed to this work equally
* Corresponding authors. Tel.: +86-28-85164063; fax: +86-28-85164060
E-mail address:yangsy@scu.edu.cn (S.-Y. Yang), or ysylilinli@sina.com (L.-L. Li)

ABSTRACT
Herein we report the discovery of a series of new KDM5A inhibitors. A
three-dimensional (3D) structure model of KDM5A jumonji domain was firstly
established based on homology modeling. Molecular docking-based virtual screening
was then performed against commercial chemical databases. A number of hit
compounds were retrieved. Further structural optimization and structure-activity
relationship (SAR) analysis were carried out to the most active hit compound, 9 (IC50:
2.3 μM), which led to the discovery of several new KDM5A inhibitors. Among them,
compound 15e is the most potent one with an IC50 value of 0.22 μM against KDM5A.
This compound showed good selectivity for KDM5A and considerable ability to
suppress the demethylation of H3K4me3 in intact cells. Compound 15e could be
taken as a good lead compound for further studies.
Keywords
KDM5A, histone lysine demethylation, structure-activity relationship, epigenetics,
small molecule inhibitor

Histone lysine demethylation is one of the important histone modifications and
plays a key role in epigenetic regulation.1 Histone lysine demethylation is catalyzed
by the histone lysine demethylases (KDMs). A number of KDMs have been identified
at present.2,3 These KDMs are categorized into two distinct classes based on their
structure and mechanism, namely flavin-dependent histone lysine demethylases LSD1
and LSD2, and the 2-oxoglutarate-dependent, Jumonji domain-containing histone
demethylases.4
Lysine demethylase 5A (KDM5A), also known as RBP2 or JARID1A, is a
typical Jumonji domain-containing demethylase.5 KDM5A can function as a
transcriptional repressor through the demethylation of tri- and dimethylated histone
lysines, such as H3 at lysine 4 (H3K4). A number of studies have shown that
dysregulation of KDM5A is associated with the initiation and development of several
human tumors, including breast cancer, prostate cancer, lung cancer, gastric cancer,
and acute myeloid leukemia (AML).6-8 Recent studies also demonstrated that KDM5A
was involved in drug-resistance of anti-cancer drugs.9-11 KDM5A has thus been
thought as a potential anti-cancer drug target, and discovery of KDM5A inhibitors has
attracted much attention in recent years. Unfortunately, just very limited number of
KDM5A inhibitors have been reported at this moment.3, 12, 13
In this account, we report the discovery of a number of new KDM5A inhibitors
with the aid of a molecule docking based virtual screening and structure-activity
relationship analysis. For the most active compound, we further studied its selectivity
against other KDMs and demethylation activity in intact cells.
Construction of a three-dimensional (3D) structure model of KDM5A with
homology modeling.Since there is no 3D structure of KDM5A reported at present, we
established a 3D structure model of the active site of KDM5A, namely jumonji

domain (JMJD),14,15 with homology modeling in advance; MODELLER16 in the
Discovery Studio (Accelrys, San Diego, CA) (DS) was used for the homology
modeling. The procedure for the model development is briefly described as follows.
Firstly, the amino acid sequence of human KDM5A jumonji domain, which
contains 167 amino acids, was retrieved from UniProt database
(http://www.uniprot.org) (UniProt entry P29375). PSI-BLAST program in DS was
used for the template search. PSI-BLAST searched a number of templates, and four of
them (PDB entries: 3DXT, 2W2I, 2OX0, and 2XML) possessed the minimum
E-values (see Supplementary Table S1), which were used for the subsequent model
construction.
Secondly, “Align Sequence to Templates” protocol in DS was used to align the
target sequence to templates. Then the MODELLER program generated five structure
models (see Supplementary Table S2). The model numbered KDM5A JMJD.M0002
was selected as an initial model for the succeeding optimization because it possesses
the lowest values of PDF Total Energy and PDF Physical Energy.
Thirdly, the generated model was optimized. Here the “standard dynamic
cascade” protocol in DS was used for the model optimization, and the CHARMM
force field17,18 was employed for the dynamic simulation. Various optimization
methods, which differ in the used implicit solvent model and energy minimization
algorithm, led to the generation of a total of 30 models (MS1-MS30, Supplementary
Table S3).
All the generated models were finally evaluated by PROCHECK verification,
Profile 3D evaluation, and Ramachandran Plot analysis. The detailed results are
presented in Supplementary data (Supplementary Figure S1-S3). We finally chose
MS28 as the 3D structure of KDM5A jumonji domain because it has the best

comprehensive performance in all the evaluation tests.
To mimic the active site of KDM5A, Fe(II) was also introduced manually into
the jumonji domain of KDM5A according to the coordination models of similar
KDMs with Fe(II) (For example, KDM4A, PDB entry 2OX0). The finally obtained
structure model is shown in Figure 1.
Figure 1. (A) The KDM5A jumonji domain homology model (MS28). The surface of
2-oxoglutarate (2OG)-binding site is colored in pink, while the cartoon defining the
protein backbone is colored in green. (B) Fe(II)-containing KDM5A jumonji domain
homology model. Fe(II) is colored in orange. Three key residues (Glu591, His589,
His677) in the 2-oxoglutarate (2OG)-binding site are colored in green.
Virtual screening to retrieve new KDM5A inhibitors. Virtual screening based on
molecular docking was carried out using the developed 3D structure of KDM5A
jumonji domain. Chemical libraries used in the virtual screening include Specs (Specs,
Inc. Zoetermeer, The Netherlands), Enamine (Enamine Ltd, Kiev, Ukraine), and
ChemDiv (ChemDiv, Inc. San Diego, CA, USA), which contain about 1.6 million
compounds. All the docking calculations were carried out with the GOLD program.19
Goldscore20 was chosen as the scoring function. From the top ranking compounds,
forty-five compounds were selected for bioactivity validation. Twelve compounds

showed IC50 values less than 100 μM against KDM5A (see Figure 2).
Figure 2. Chemical structures together with bioactivities of retrieved compounds.
Among the twelve compounds, compound 9 is the most active one. We next
performed structure optimization and SAR analysis to this compound.
Structural Optimization and SAR analysis. To improve the potency of
compound 9, structure optimization was carried out. We firstly fixed the
2-(iminomethyl) phenol moiety of compound 9 (left) and changed the C-4 (R2) and
C-5 (R1) positions of pyrimidine (right) with different substituents (see Figure 3). A
total of six compounds were synthesized. The synthetic routes for these compounds
are depicted in Scheme 1. Intermediates 15 and 15a-e were first prepared through
nucleophilic substitution reactions between self-prepared intermediate (14) or
commercially available regents (14a-e) and hydrazine hydrate. Condensation
reactions between these intermediates and 2-hydroxybenzaldehyde gave target
compounds 16a-f.

Figure 3. Schematic showing the focus of structural modifications.
Scheme 1. Synthetic routes for compounds 16a-f. Reagents and conditions: (a)
Morpholine, K2CO3, EtOH, rt, 6h; (b) N2H4·H2O, DIEA, EtOH, 90°C, sealed-tube,
over night; (c) N2H4·H2O, EtOH, rt-60°C, over night; (d) 2-hydroxybenzaldehyde,
MeOH, 40°C, overnight.
Bioactivities of the synthesized compounds are displayed in Table 1; the
bioactivities used in the SAR analysis were the inhibition rates at the fixed
concentration of 10 μM of compounds. From Table 1, we can see that compound 16f,
which contains a carboxyl group at the R2 position and a hydrogen atom at the R1
position, is the most active one. Here one may notice that compound 16b has an
inhibition rate of -13%. This type of minus inhibition rate in a single concentration
does not mean that this compound is an activator. It could be due to the experimental
error since all compounds with minus inhibition rates did not show activating effect
on the KDM5A enzyme at western blot assays (data not shown).

Table 1. Bioactivities of compounds 9 and 16a-f
Compound R1
R2
morpholyl
morpholyl
H
Inhibition rate@10μMa
58 ± 6.4
F
H
F
9
16a
16b
16c
16d
16e
16f
42 ± 8.7
-13 ± 3.5
19 ± 1.3
Cl
H
H
H
H
H
32 ± 3.1
CH3
44 ± 3.4
COOH
88 ± 1.3
aAll assays were conducted in duplicate.
We then fixed R1 as hydrogen and R2 as a carboxyl group, and varied the
substitution group of 2-amino. Two new compounds were prepared (see Scheme 2).
These compounds were easily prepared through a condensation between 15e and
aldehydes. Table 2 lists the bioactivities of the selected compounds. Obviously,
compound 15e, which bears a hydrazine group in 2-position of pyrimidine, displayed
more potent activity compared with 16f.
Scheme 2: The synthetic route for compounds 17a and 17b. Reagents and conditions:

(a) Benzaldehyde (or furan-2-carbaldehyde) , MeOH, 40°C, over night.
Table 2. Bioactivities of compounds 16f, 15e, 17a-b.
Compound
16f
R
Inhibition rate@10μMa
88 ± 1.3
99 ± 0.1
NH2
15e
(IC50 = 0.22 ± 0.01 μM)
17a
17b
0 ± 0.4
3 ± 1.3
aAll assays were conducted in duplicate.
We finally fixed the left part as the optimal amino group and again changed the
C-4 (R2) and C-5 (R1) positions of pyrimidine (right) with different substituents. Eight
compounds were synthesized. As illustrated in Scheme 3, compounds 19, 22 and
23a-c were synthesized through nucleophilic substitution reactions between
commercially available materials and several desired nucleophilic reagents.
Intermediate 3-(2-chloropyrimidin-4-yl) benzoic acid (compound 20) was prepared
based on a palladium-catalyzed coupling between 2,4-dichloro-pyrimidine and
3-boronobenzoic acid (Buchwald conditions).21 Target compounds 25a-h were
synthesized following the procedure for compounds 15a-e which was presented above.
Bioactivities of these compounds are presented in Table 3. From Table 3, we can see
that the newly synthesized compounds did not surpass 15e in terms of the bioactivity.

Scheme 3. Synthetic routes for compounds 25a-h. Reagents and conditions: (a)
Morpholine, K2CO3, EtOH, 0°C, 4 h; (b) 3-boronobenzoic acid, NaHCO3, Pd(PPh3)4,
H2O, (CH2OMe)2, 7 h, 80°C, overnight; (c) 4-aminobenzonitrile, K2CO3, EtOH,
0°C-rt, 4 h; (d) substituted-phenol, DIEA, acetone, 60°C, overnight; (e) N2H4·H2O,
EtOH, 60°C, overnight.
Table 3. Bioactivities of compounds 15, 15a-e and 25a-h.
Compound
15
R1
H
F
R2
morpholyl
H
Inhibition rate@10μMa
-45 ± 9.90
10 ± 4.70
15a
Cl
H
H
H
H
15b
-18 ± 8.30
8 ± 0.77
H
15c
CH3
15d
-6 ± 0.22
COOH
15e
99 ± 0.12

(IC50 = 0.22 ± 0.01 μM)
-35 ± 4.30
H
F
CF3
25a
25b
morpholyl
0 ± 6.02
H
H
H
H
25c
25d
25e
25f
-72 ± 0.27
-50 ± 18.03
-121 ± 14.83
26 ± 2.40
NO2
25g
25h
14 ± 0.50
-42 ± 1.41
-CH=CH2-CH2=CH-
aAll assays were conducted in duplicate.
Selectivity of compound 15e. The selectivity of compound 15e for KDM5A
against other KDMs was evaluated. In this test, KDM3A, KDM6B, KDM4E,
KDM3C, and KDM5B were included. The IC50 values of compound 15e against the
selected KDMs are shown in Table 4. Compound 15e displayed moderate inhibitory
activity against KDM3C and KDM4E with IC50 values of 1.9 ± 0.32 μM and 13 ±
1.50 μM, respectively, and very weak activity against KDM3A (IC50 = 43 ± 2.50 μM).
Compound 15e displayed almost no activity against KDM6B and KDM5B (IC50 >
100 μM). These results indicate that compound 15e has a good selectivity for KDM5a
against other selected KDMs.
Table 4. Inhibitory activities of compound 15e against selected KDMs.
IC50 (μM)a
Compound
KDM3A KDM6B KDM4E
KDM3C KDM5B
KDM5A

43 ± 2.50
> 100
13 ± 1.50 1.9 ± 0.32 > 100
0.22 ± 0.01
15e
a All assays were conducted in duplicate.
Predicted interaction model between compound 15e and KDM5A. Molecular
docking was adopted to predict the binding model of the most active compound 15e in
the active pocket of KDM5A. Again, GOLD was used for the docking study. The
predicted interaction model between compound 15e and KDM5A is depicted in
Figure 4. From Figure 4, we can see that there is a strong electrostatic interaction
between the carboxyl group of 15e and Fe(II). One of the carboxyl oxygen atoms also
forms a hydrogen bond with His589. In addition, a pyrimidine nitrogen atom (1-N)
forms a hydrogen bond with Lys607.
Figure 4. Predicted binding model of compound 15e in the active pocket of KDM5A.
Compound 15e is colored in pink, Fe(II) is in orange and residues of KDM5A are in
green. Hydrogen bonds are shown in blue dashed line.
Demethylation activity of compound 15e in intact cells. Western blot assays
were used to detect the demethylation activity of compound 15e in intact cells. In

these assays, breast cancer cell line ZR-75-1 was chosen because ZR-75-1 expresses a
high level of KDM5A.22 The results (see Figure 5) showed that the expression level of
H3K4me3 increases with the increase of the concentration of compound 15e,
implying that the demethylation activity of KDM5A was suppressed dose
dependently.
Figure 5. Western blot analysis showing the ability of 15e to inhibit KDM5A in
ZR-75-1 cells
In summary, virtual screening together with bioactivity validation led to the
retrieval of a number of hit compounds against KDM5A. The most active hit
compound, compound 9, was chosen for structural optimization to improve its
potency. A series of compounds containing pyrimidin-2-amine were synthesized. The
SAR analysis led to the discovery of a potent KDM5A inhibitor, namely compound
15e. This compound showed a good inhibitory activity and selectivity for KDM5A,
and considerable ability to inhibit the demethylation of H3K4me3 in intact cancer
cells. Collectively, compound 15e could be a good lead compound, and deserves
further studies.

Acknowledgments
This work was supported by the National Natural Science Foundation of China
(81325021, 81473140, 81573349), the Major State Basic Research Development
Program of China (2013CB967204) and PCSIRT (NO: IRT13031).
Supporting Information: Supplementary data associated with this article can be
found in the online version.
REFERENCES
1. Mair, B.; Kubicek, S.; Nijman, S. M.; Trends Pharmacol Sci. 2014, 35, 136.
2. Suzuki, T.; Miyata, N. J. Med. Chem. 2011,54, 8236.
3. Pachaiyappan, B.; Woster, P. M. Bioorg. Med. Chem. 2014, 24, 21.
4. Suzuki, T.; Ozasa, H.; Itoh, Y.; Zhan, P.; Sawada, H.; Mino, K.; Walport, L.;
Ohkubo, R.; Kawamura, A.; Yonezawa, M.; Tsukada, Y.; Tumber, A.; Nakagawa,
H.; Hasegawa, M.; Sasaki, R.; Mizukami, T.; Schofield, C. J.; Miyata, N. J. Med.
Chem. 2013, 56, 7222.
5. Blair, L. P.; Cao, J.; Zou, M. R.; Sayegh, J.; Yan, Q. Cancers (Basel) 2011, 3,
1383.
6. Beshiri, M. L.; Holmes, K. B.; Richter, W. F.; Hess, S.; Islam, A. B.; Yan, Q.;
Plante, L.; Litovchick, L.; Gevry, N.; Lopez-Bigas, N.; Kaelin, W. G., Jr.;
Benevolenskaya, E. V. Proc. Natl. Acad. Sci U S A. 2012, 109, 18499.
7. Chaturvedi, C. P.; Somasundaram, B.; Singh, K.; Carpenedo, R. L.; Stanford, W.
L.; Dilworth, F. J.; Brand, M. Proc. Natl. Acad. Sci U S A. 2012,109, 18845.
8. Ge, Z.; Li, W.; Wang, N.; Liu, C.; Zhu, Q.; Bjorkholm, M.; Gruber, A.; Xu, D.
FASEB J. 2010,24, 579.
9. Dannenberg, J. H.; Berns, A. Cell 2010,141, 18.

10. Sharma, S. V.; Lee, D. Y.; Li, B.; Quinlan, M. P.; Takahashi, F.; Maheswaran, S.;
McDermott, U.; Azizian, N.; Zou, L.; Fischbach, M. A.; Wong, K. K.; Brandstetter,
K.; Wittner, B.; Ramaswamy, S.; Classon, M.; Settleman, J. Cell 2010,141, 69.
11. Hou, J.; Wu, J.; Dombkowski, A.; Zhang, K.; Holowatyj, A.; Boerner, J. L.; Yang,
Z. Q. Am. J. Transl. Res. 2012,4, 247.
12. Chan, S. W.; Hong, W. J. Biol. Chem. 2001,276, 28402.
13. van Zutven, L. J.; Onen, E.; Velthuizen, S. C.; van Drunen, E.; von Bergh, A. R.;
van den Heuvel-Eibrink, M. M.; Veronese, A.; Mecucci, C.; Negrini, M.; de Greef,
G. E.; Beverloo, H. B. Genes Chromosomes Cancer 2006,45, 437.
14. Spannhoff, A.; Hauser, A. T.; Heinke, R.; Sippl, W.; Jung, M. Chem.Med.Chem.
2009,4, 1568.
15. Colas, C.; Pajor, A. M.; Schlessinger, A. Biochemistry. 2015, 54, 4900.
16. Eswar, N.; Eramian, D.; Webb, B.; Shen, M. Y.; Sali, A. Methods. Mol. Biol. 2008,
426, 145.
17. Dwyer, P. J.; Vander Valk, R. J.; Caltaldo, V.; Demianicz, D.; Kelty, S. P. J. Phys.
Chem. A. 2014, 118, 11583.
18. Patel, S.; Brooks, C. L., 3rd. J. Comput. Chem. 2004,25, 1.
19. Bissantz, C.; Folkers, G.; Rognan, D. J. Med. Chem. 2000, 43, 4759.
20. Yamagishi, M. E.; Martins, N. F.; Neshich, G.; Cai, W.; Shao, X.; Beautrait, A.;
Maigret, B. J. Mo.l Model. 2006, 12, 965.
21. William, A. D.; Lee, A. C. H.; Goh, K. C.; Blanchard, S.; Poulsen, A.; Teo, E. L.;
Nagaraj, H.; Lee, C. P.; Wang, H.; Williams, M.; Sun, E. T.; Hu, C.; Jayaraman, R.;
Pasha, M. K.; Ethirajulu, K.; Wood, J. M.; Dymock, B. W. J. Med. Chem. 2012,
55, 169.
22. Hou, J.; Wu, J.; Dombkowski, A.; Zhang, K.; Holowatyj, A., Boerner, J. L.; Yang,

Z.-Q. Am. J. Transl. Res. 2012, 4, 247.

Graphical abstract
Discovery of KDM5A Inhibitors: Homology Modeling, Virtual Screening and
Structure-Activity Relationship Analysis
Xiaoai Wu1,2,#, Zhen Fang1,#, Bo Yang3,4,#, Lei Zhong1, Qiuyuan Yang1, Chunhui
Zhang1, Shenzhen Huang1, Rong Xiang5, Takayoshi Suzuki6, Lin-Li Li3,*,
Sheng-Yong Yang1,*