Ridaifen G, tamoxifen analog, is a potent anticancer drug working through a combinatorial association with multiple cellular factors

Article abstract of DOI:10.1016/j.bmc.2015.08.001

Ridaifen-G (RID-G), a tamoxifen analog that we previously synthesized, has potent growth inhibitory activity against various cancer cell lines. Tamoxifen is an anticancer drug known to act on an estrogen receptor (ER) and other proteins. However, our prev

Full text of DOI:10.1016/j.bmc.2015.08.001

Bioorganic & Medicinal Chemistry 23 (2015) 6118–6124
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Ridaifen G, tamoxifen analog, is a potent anticancer drug working
through a combinatorial association with multiple cellular factors
Kentaro Ikeda a,y, Shinji Kamisuki a,y, Shoko Uetake b, Akihito Mizusawa b, Nozomi Ota b, Tatsuki Sasaki a,
Senko Tsukuda a, Tomoe Kusayanagi a, Yoichi Takakusagi a, Kengo Morohashi c, Takao Yamori d,à
,
Shingo Dan d, Isamu Shiina b, Fumio Sugawara a,
⇑
a Department of Applied Biological Science, Tokyo University of Science, Chiba, Japan
b Department of Applied Chemistry, Tokyo University of Science, Tokyo, Japan
c Department of Molecular Genetics, The Ohio State University, Columbus, OH, United States
d Division of Molecular Pharmacology, Cancer Chemotherapy Center, Japanese Foundation for Cancer Research, Tokyo, Japan
a r t i c l e i n f o
a b s t r a c t
Article history:
Ridaifen-G (RID-G), a tamoxifen analog that we previously synthesized, has potent growth inhibitory
activity against various cancer cell lines. Tamoxifen is an anticancer drug known to act on an estrogen
receptor (ER) and other proteins. However, our previous studies interestingly suggested that the mech-
anism of action of RID-G was different from that of tamoxifen. In order to investigate the molecular mode
of action of RID-G, we developed a novel chemical genetic approach that combined a phage display screen
with a statistical analysis of drug potency and gene expression profiles in thirty-nine cancer cell lines.
Application of this method to RID-G revealed that three proteins, calmodulin (CaM), heterogeneous
nuclear ribonucleoproteins A2/B1 (hnRNP A2/B1), and zinc finger protein 638 (ZNF638) were the candi-
dates of direct targets of RID-G. Moreover, cell lines susceptible to RID-G show similar expression profiles
of RID-G target genes. These results suggest that RID-G involves CaM, hnRNP A2/B1, and ZNF638 in its
growth inhibitory activity.
Received 17 July 2015
Revised 31 July 2015
Accepted 1 August 2015
Available online 5 August 2015
Keywords:
Tamoxifen
Anticancer drug
Phage display
Target identification
Ó 2015 Published by Elsevier Ltd.
1. Introduction
purification is often used for target identification, we have
employed phage display screen to identify binding proteins of
Classical cytotoxic anticancer drugs, such as taxol, doxorubicin,
and tamoxifen, have been used for cancer chemotherapy. Several
reports suggested that these drugs acted on multiple target pro-
teins to exert cytotoxic effects on cancer cells.1,2 One of them,
tamoxifen, has been widely used in chemotherapy against breast
cancers. The primary target molecule is known to be ER; however,
many studies on the mechanism of action reported that tamoxifen
acted not only on ER, but also on other target proteins such as CaM,
protein kinase C, and proto-oncogene c-Myc.3
Target identification of bioactive compounds can elucidate their
molecular mechanism. Although a variety of methodologies for
identifying target proteins has been reported, comprehensive iden-
tification of multiple target proteins is still difficult.4 Although
chemical proteomics workflow using affinity matrix-based protein
bioactive compounds.5–8 Compared with chemical proteomics
workflow, the phage display screen method has the advantage of
identifying multiple targets, since DNA sequencing of the inserts
in the phage clones detects numerous binding proteins, whereas
liquid chromatography–mass spectrometry analysis of proteins
gives relatively little information. Once cDNA libraries derived
from several cultured cell lines and tissues are constructed, several
types of phage libraries can be used for the screening as libraries of
the binding partner. Arango et al. have recently developed a novel
approach that combines phage display with next generation
sequencing, which allowed the comprehensive identification of
human targets of a bioactive flavonoid.9 Although the phage dis-
play screen method is indeed feasible for comprehensive target
identification, it still has shortcomings. In particular, the phage dis-
play screen method does not provide physiological information on
the interaction between a compound and a target protein. In this
paper, we propose a unique chemical genetic approach in which
phage display screen was combined with a statistical analysis
using drug potency and gene expression profiles to understand
the physiological relevance of the direct association (Fig. 1).
⇑
Corresponding author. Tel.: +81 4 7124 1501x3400; fax: +81 4 7123 9767.
E-mail address: sugawara@rs.noda.tus.ac.jp (F. Sugawara).
These authors contributed equally to this work.
Present address: Center for Product Evaluation, Pharmaceuticals and Medical
y
à
Device Agency, Tokyo, Japan.
http://dx.doi.org/10.1016/j.bmc.2015.08.001
0968-0896/Ó 2015 Published by Elsevier Ltd.

K. Ikeda et al. / Bioorg. Med. Chem. 23 (2015) 6118–6124
6119
cell lines. By applying the workflow, we identified CaM, hnRNP
A2/B1, and ZNF638 as plausible target proteins of RID-G.
2. Materials and methods
2.1. T7 phage display screen
T7 phage display screening was performed as described in Sup-
porting information. In brief, an aliquot of the T7 phage library was
added to control or Bio-RID-G-immobilized well, and the mixture
was then incubated for 1 h. After incubation, the wells were
washed with TBST (TBS, 0.1% Tween 20). An elution buffer (TBS,
1% SDS) was then added, and the mixture was incubated for
15 min to recover the remaining phage particles. In order to
amplify the recovered phage particles, each eluate was mixed with
a culture of Escherichia coli BLT5615 (Merck Millipore, Billerica,
MA, USA). The cells were then cultured at 37 °C until cell lysis
was observed. The resulting solution was used for the next round
of biopanning. After six rounds of selection, the DNA sequence of
each phage clone was then analyzed.
2.2. SPR analysis
SPR analysis was performed on a BiacoreÒ 3000 (GE Healthcare,
Buckinghamshire, England). The purified CaM or hnRNP B1 (2–
200 a.a.) was immobilized on a CM5 sensor chip. After the immo-
bilization, appropriate concentrations of compounds were injected
over the flow cells. Binding analyses were carried out with a flow
rate of 20 lL/min at 25 °C. Kinetic parameters were determined
by analyzing the data using the BIAevaluation 4.1 software (GE
Healthcare).
Figure 1. A flowchart showing the identification of proteins for multiple target
drugs.
2.3. Analysis of ERK1/2 phosphorylation level
HeLa cells were cultured in serum-free medium for 48 h and
then incubated with RID-G. The cells were harvested, and the cell
lysates were mixed with an equal volume of 2ꢀ SDS sample buffer.
The sample was then heated at 96 °C for 10 min.
2.4. Expression and purification of recombinant hnRNP B1
Expression and purification of recombinant hnRNP B1 was per-
formed as described in Supporting information. In brief, hnRNP B1
(2–200 a.a.) expression vector was used to transform E. coli Rosetta
2 (DE3) (Novagen). These bacteria were grown in LB medium and
Figure 2. Structures of tamoxifen, RID-G and RID-D.
To demonstrate our strategy, we chose a tamoxifen analog, RID-
G (CAS Registry Number 1020853-04-6) as a multi-target drug
(Fig. 2). We previously synthesized tamoxifen analogs named
ridaifens,10 and one of them, RID-G, exhibited the highest
potency.11,12 The mean value of GI50 (the concentration for 50%
of maximal inhibition) over thirty-nine human cancer cell lines
treated with 1 mM isopropyl thio-b-D-galactoside. After incubation
for 5 h, the cells were harvested and suspended in a buffer. The
cells were disrupted by sonication, and the soluble fraction was
loaded onto a HisTrap HP column (GE Healthcare) using an FPLC
system (AKTA explorer, GE Healthcare). Finally, bound proteins
were eluted using an elution buffer (500 mM NaCl, 8.1 mM Na2-
HPO4, 2.68 mM KCl, 1.47 mM KH2PO4, 0.05% Triton X-100, 5% glyc-
erol, 0.5 M imidazole, pH 7.4).
(designated as MG-MID) of RID-G was 0.85
lM, while MG-MID of
tamoxifen was 7.41
l
M.13 Moreover, GI50 profiles across thirty-
nine cell lines (designated as JFCR39 by the Japanese Foundation
for Cancer Research) suggested that the mode of action of RID-G
was ER-independent and different from that of tamoxifen.13 Since
RID-G, like tamoxifen, was expected to affect multiple target pro-
teins to exert cytotoxicity against cancer cells, we applied our
strategy on a comprehensive identification of RID-G-binding pro-
teins by using GI50/gene expression profiles of JFCR39 cell lines
(Fig. 1). Phage display screen with RID-G-immobilized matrix iden-
tified RID-G-binding peptides, and their interactions were vali-
dated by binding analyses of each single clone phage. The
candidate genes were further analyzed to investigate the relation-
ship of GI50 values of RID-G with gene expression profiles in the
3. Results
3.1. Selection of RID-G-binding proteins using a T7 phage
display screen
RID-G and biotinylated RID-G (Bio-RID-G) were synthesized
using the Mukaiyama reductive coupling reaction,14,15 and the Shi-
ina esterification (Fig. S1, Supporting information).16–18 Since our
purpose was to identify RID-G-binding proteins in a comprehen-
sive manner, we used cDNA libraries derived from multiple cell
lines such as Jurkat cells (human T-cell leukemia), TE8 cells

6120
K. Ikeda et al. / Bioorg. Med. Chem. 23 (2015) 6118–6124
Table 1
Peptides displayed on phage clones obtained after the screening
No.
Protein name (region)
Identity (%)
4
WW domain-binding protein 4 (117–250 a.a.)
97
7
8
9
10
11
13
15
Eukaryotic translation initiation factor 4 gamma 1(1079–1171 a.a.)
Cysteine and glycine-rich protein 1 (135–193 a.a.)
Calmodulin (1–75 a.a.)
93.5
98.3
98.7
99.4
90.8
96.5
99.1
Zinc finger protein 638 (1313–1479 a.a.)
Ankyrin repeat domain-containing protein 11 (1203–1394 a.a.)
Plasminogen activator inhibitor 1 RNA-binding protein (115–286 a.a.)
Heterogeneous nuclear ribonucleoproteins A2/B1 (14–130 a.a.)
(human esophageal squamous cell carcinoma), and HUVECs, and
human lung tumors. Each T7 phage library was incubated with
Bio-RID-G immobilized on wells, and affinity selection was per-
formed. After several rounds of selection, binding phage clones
were amplified, and each DNA sequence was analyzed (Table S1,
Supporting information). The translated amino acid sequences of
sixteen clones were summarized in Table 1. Out of sixteen clones,
eight clones (No. 4, 7, 8, 9, 10, 11, 13, and 15) had high homology
with human proteins and were further analyzed.
3.2. Binding analyses of single clone phage and gene expression
profiles in cancer cell lines
To confirm specific binding of RID-G to each peptide, we per-
formed a binding assay using single clone of T7 phage particles dis-
playing each peptide. Each phage clone was applied to
nonimmobilized or Bio-RID-G-immobilized wells, and the recovery
rate was calculated (Fig. 3A). Phage clones Nos. 4 and 7 showed
lower values for ratios of recovery rates between Bio-RID-G (L)
A
B
Figure 3. Binding analyses of single clone phage and gene expression profiles in JFCR39 cells. (A) The affinity binding of each single clone to RID-G or RID-D. Each single clone
was incubated into the control well, Bio-RID-G- or Bio-RID-D-immobilized well; the phage titer in each eluate was compared to the input phage titer and shown as recovery
rate. EV is an empty phage vector. Mean values are shown the standard error (n = 3). L/N is the ratio between the recovery rates of Bio-RID-G and the control. (B)
Correlations between RID-G sensitivity and expression level of gene encoding each RID-G-binding candidate protein. Expression level of five genes, CSRP1 (cysteine and
glycine-rich protein 1), CALM1 (calmodulin), ZNF638 (zinc finger protein 638), ANKRD11 (ankyrin repeat domain-containing protein 11), SERBP1 (plasminogen activator
inhibitor 1 RNA-binding protein) and HNRNPA2/B1 (heterogeneous nuclear ribonucleoproteins A2/B1) were plotted for two groups (HS group: GI50 < 1.0
lM and LS group:
GI50 > 1.5
l
M). Student’s t test was used to examine significance of differential expression (*P < 0.05; **P < 0.01).

K. Ikeda et al. / Bioorg. Med. Chem. 23 (2015) 6118–6124
6121
and control (N) (L/N) than of phage clones with no insert (EV).
These results indicated that peptides displayed by phage clones
Nos. 4 and 7 had low affinity with RID-G; thus, we eliminated these
clones from the RID-G-binding candidate proteins. We also used
ridaifen D (RID-D) (CAS Registry Number 939819-26-8) (Fig. 3A)
as a negative control, whose cytotoxicity was lower than that of
Ca2+. These data indicated that Ca2+ was essential for the binding
between RID-G and CaM.
We further analyzed the affinity of RID-G to CaM by surface
plasmon resonance (SPR) method (Fig. 4B). Various concentrations
of RID-G were injected on the surface of the CM5 sensor chip
where CaM proteins had been immobilized. In the presence of
Ca2+, RID-G bound to CaM in a dose-dependent manner, and the
RID-G; MG-MID of RID-D was 14.8
lM, whereas that of RID-G
was 0.85
l
M.13 Recovery rates of clones Nos. 8, 9, 10, 13 and 15
KD value obtained by global fitting was 4 lM. The binding was
with Bio-RID-G were significantly higher than those with Bio-
RID-D, whereas the difference in the affinity of clone No. 11 for
Bio-RID-G and for Bio-RID-D was not statistically significant
(P > 0.05). Thus, the peptide displayed by phage clone No. 11
(ankyrin repeat domain-containing protein 11) was unlikely to
be involved in RID-G cytotoxicity. Therefore, we focused on five
clones, No. 8 (cysteine and glycine-rich protein 1), No. 9 (CaM),
No. 10 (ZNF638), No. 13 (plasminogen activator inhibitor 1 RNA-
binding protein), and No. 15 (hnRNP A2/B1) for further analysis.
Next, the correlation between RID-G cytotoxicity and gene
expression was determined using the expression level of each
gene, which may encode the candidate of RID-G-binding proteins.
The cytotoxicity of RID-G in JFCR39 cell lines had been previously
evaluated,13 and the expression level of each gene in the JFCR39
cell lines was determined by DNA microarray (unpublished data).
Thirty-nine cell lines in the JFCR39 collection were divided into
two groups: high sensitivity group (designated as ‘HS’) and low
sensitivity group (designated as ‘LS’) for the treatment of RID-G.
impaired by the addition of EGTA, which was consistent with the
single-clone binding assay (Fig. 4A). We also compared the affinity
of RID-G with that of tamoxifen (Fig. 4B). Interestingly, tamoxifen
showed low affinity binding to CaM in our SPR experiment, sug-
gesting that the affinity of RID-G was higher than that of
tamoxifen.
Considering the high affinity of RID-G to Ca2+–CaM and the
known CaM antagonist property of tamoxifen, we hypothesized
that RID-G could function as a CaM antagonist in cells. Several
CaM antagonists are known to activate ERK, a critical mediator of
Ca2+–CaM signaling, and to suppress cancer cell proliferation.20
To clarify whether the binding of RID-G with Ca2+–CaM could have
an antiproliferative activity against cancer cells, we examined
phosphorylation levels of ERK1/2 in HeLa cells. HeLa cells showed
moderate sensitivity to RID-G with an IC50 value of 2.4 lM (Fig. S2,
Supporting information). HeLa cells were treated with RID-G, and
the phosphorylation level of ERK1/2 was assessed. As a result,
RID-G induced phosphorylation of ERK1/2, and the phosphoryla-
tion was sustained up to 2 h, which was consistent with the prop-
erty of known CaM antagonists (Fig. 4C).20 The increase in ERK1/2
phosphorylation was also observed in the presence of FBS (Fig. 4D).
These data suggested that RID-G bound to Ca2+–CaM to inhibit its
signaling in cancer cells.
The cells with GI50 < 1.0
lM were classified as HS, whereas the cells
with GI50 > 1.5 M as LS. The expression levels of candidates CSRP1,
l
CALM1, ZNF638, SERBP1, and HNRNPA2/B1 and of the nonspecific
binding protein ANKRD11 were plotted in each group (Fig. 3B). As
a result, three genes (CALM1, HNRNPA2B1, and ZNF638) were differ-
entially expressed between the HS and LS groups (P < 0.05),
whereas CSRP1, SERBP1, and ANKRD11 were not (P > 0.05). CALM1
gene expression was high in the LS group, whereas the other two
genes were high in the HS group. The expression of ANKRD11
was not significantly different between the two groups, which
was consistent with the binding analysis of single clone phage
(No. 11) for Bio-RID-G and Bio-RID-D (Fig. 3A). Five irrelevant
genes (e.g., TUBB) were also analyzed by the same method and
found to be not significantly different between the two groups
with different sensitivities to the RID-G treatment (data not
shown). These data suggested that the three genes encoding the
proteins CaM, hnRNPA2/B1, and ZNF638 were involved in RID-G-
mediated inhibition of cancer cell proliferation. Considering the
single-clone binding assay and the correlation analysis, we hypoth-
esized that RID-G could bind to the three proteins and modulate
their functions, thus exerting cytotoxicity. Therefore, we further
analyzed these proteins.
3.4. Binding analyses of RID-G to hnRNPA2/B1
HnRNP A2/B1 is a member of the hnRNP family known as a
splicing factor and has been reported to be highly expressed in
many cancers.21–24 Knockdown of this protein induces cell death
by apoptosis in various cancer cells, but not in normal cells.25
Therefore, RID-G–hnRNP A2/B1 interaction can be involved in
RID-G-mediated cytotoxicity against cancer cells. To confirm the
interaction, SPR analysis using a recombinant hnRNP A2/B1 pep-
tide was performed. His-tagged N-terminal region of hnRNP A2/
B1 (hnRNP A2/B1 [2–200 a.a.]) corresponding to the region dis-
played on the RID-G-binding phage particles was purified and used
for SPR analysis. As a result, the response was increased in a dose-
dependent manner after adding RID-G, and the KD value was 3
(Fig. 5). Since this KD value was comparable to that of CaM (4
and to the GI50 value of RID-G (0.85 M), the binding of RID-G to
lM
l
M)
l
hnRNP A2/B1 might partially inhibit cancer cell growth.
3.3. Effect of RID-G on CaM
4. Discussion
Tamoxifen has been also known to be a CaM antagonist.3 Con-
sidering the structural similarity between RID-G and tamoxifen,
CaM is the most reliable candidate among the target proteins.
CaM is a major calcium-binding protein that is responsible for
the regulation of a wide range of cellular signaling. Upon the bind-
ing of Ca2+ to the EF-hand motif of CaM, its conformation change
occurs, allowing the interaction with several Ca2+–CaM target pro-
teins.19 Since the phage clone No. 9 contains a calcium binding
region of CaM, we investigated whether the binding of RID-G with
phage clone No. 9 was Ca2+-dependent by performing single-clone
binding assay in the presence of EGTA, a Ca2+ chelator as described
previously (Fig. 4A).20 The addition of EGTA remarkably decreased
the binding of Bio-RID-G to the CaM peptide displayed on phage
particles, and the binding was recovered by adding an excess of
In this paper, we have identified three proteins, CaM, hnRNPA2/
B1, and ZNF638 as candidate target proteins of RID-G, a tamoxifen
analog, by T7 phage display screen. The validation of target pro-
teins is generally a time-consuming process. In this study, candi-
dates obtained from phage display screen were subjected to two
analyses: (1) single-clone binding assay using RID-D as a negative
control compound without biological activity, and (2) a statistical
analysis of the gene expression profiles combined with RID-G cyto-
toxicity. The chemical genetic approach could be useful for the
identification of multiple target proteins of small molecules, espe-
cially multi-target drugs such as tamoxifen.
We demonstrated that RID-G had higher affinity for CaM than
tamoxifen and acted as a CaM antagonist in cells, like tamoxifen

6122
K. Ikeda et al. / Bioorg. Med. Chem. 23 (2015) 6118–6124
B
A
C
2.5 µM RID-G
p-ERK1/2
-
+
-
+
-
+
-
+
total ERK1/2
D
2.5 µM RID-G
-
+
-
+
-
+
-
+
-
+
-
+
p-ERK1/2
total ERK1/2
Figure 4. Effect of RID-G on calmodulin (CaM). (A) Effect of Ca2+ and EGTA on the binding of RID-G to single phage clones displaying CaM (1–75 a.a.). The phage particles were
incubated with Bio-RID-G in the presence or absence of 2 mM EGTA with or without excess amount of Ca2+ (10 mM), and the recovery rates were calculated. Mean values are
shown the standard error (n = 3). (B) SPR analysis of RID-G or tamoxifen binding to CaM. CaM was immobilized on a flowpath of CM5 sensorchip, and a solution of RID-G or
tamoxifen at indicated concentrations was injected to protein-immobilized flowpath in the presence of 5 mM EGTA or 5 mM Ca2+. The binding responses (in RU) were
recorded as a function of time (in sec), and the results were analyzed by using BIA evaluation 4.1. (C) and (D) Effect of RID-G on ERK 1/2 phosphorylation. HeLa cells were
starved in medium without FBS for 48 h and then treated with 2.5 lM RID-G in the absence (C) or presence (D) of 10% FBS for the indicated time points. An anti-phospho ERK
1/2 antibody was used to detect ERK 1/2 phosphorylation (p-ERK 1/2). The protein levels of ERK 1/2 were used as internal controls (total ERK 1/2).
25
(Fig. 4). However, a previous study showed that ER-binding activity
of RID-G was lower than that of tamoxifen.13 These differences in
20
Conc. (µM)
the affinities could explain a distinct mechanism of action between
RID-G and tamoxifen.13 Recently, it was reported that tamoxifen
4
15
disrupted CaM–Fas interaction,26 and several CaM antagonists
3
and tamoxifen induced apoptosis through a Fas-related mecha-
2
10
nism in cholangiocarcinoma and other cancer cell lines.27–29 Our
1
result suggested that higher expression of CaM decreased sensitiv-
ity of cancer cells to RID-G (Fig. 3B). Therefore, CaM is the most
reliable target among the selected proteins because related to
RID-G-mediated cytotoxicity.
We have also identified other two candidate proteins, hnRNP
A2/B1 and ZNF638. HnRNP A2/B1 is known to be highly expressed
in many cancers and regulate splicing of oncogenes. Our SPR anal-
ysis demonstrated that RID-G had a similar affinity for hnRNP A2/
B1 as for CaM (Fig. 5). Additionally, gene expression profiles in cell
lines affected by RID-G suggested that the cancer cell lines express-
ing hnRNP A2/B1 at relatively higher levels were likely to be sensi-
5
0
0.5
0
-5
-50
0
50
100
150
200
Time (sec)
Figure 5. SPR analysis of the binding of RID-G to hnRNPA2/B1. HnRNPA2/B1 was
immobilized on a flowpath of CM5 sensorchip, and a solution of RID-G at indicated
concentrations was injected to the protein-immobilized flowpath. The binding
responses (in RU) were recorded as a function of time (in sec), and the results were
analyzed by using BIA evaluation 4.1.

K. Ikeda et al. / Bioorg. Med. Chem. 23 (2015) 6118–6124
6123
100
90
80
70
60
50
40
30
20
10
0
0
50
100
150
200
250
ZNF638
Figure 6. Correlation map of the expression levels of CaM, hnRNP A2/B1, and ZNF638, and the GI50 of RID-G. Horizontal and longitudinal axes indicate ZNF638 and CALM1
expression levels, respectively. The colors of each circle, white and black, indicate HNRNPA2B1 expression level. White or black circle represents a lower or higher expression
level, comparing to the intermediate value of HNRNPA2B1 expression levels in JFCR39 cell lines. The red area indicates a lower CALM1expression level and a higher ZNF638
expression level using as reference the intermediate value of each protein expression level in JFCR39 cell line. The circle size indicates the GI50 values of RID-G (the smaller
one shows a weaker effect and the bigger one shows a stronger effect). The arrow indicates the DMS114 cell line, showing the lowest GI50 among the JFCR39 cell lines.
tive to RID-G treatment (Fig. 3B). It can be speculated that RID-G
might be able to impair splicing activity of hnRNP A2/B1 through
direct binding and block aberrant proliferation of cancer cells
dependent on hnRNP A2/B1 overexpression. Golan-Gerstl et al.
revealed that hnRNP A2/B1 regulated alternative splicing of tumor
suppressors and oncogenes in glioblastoma.30 Therefore, hnRNPA2/
B1 is a potential new target for glioblastoma and other cancers
therapy. Although the effects of RID-G on hnRNPA2/B1 cellular
function still remain to be investigated, RID-G could be a potential
new anticancer drug targeting hnRNP A2/B1.
us to predict RID-G effectiveness in other cancer cells by measuring
the expression levels of the three genes.
Our approach could be applied for multi-target anticancer
drugs, aiming at tailor-made medicine in the future. Once the pro-
teins associating to multi-target drugs are identified by our phage
display screen strategy, and the profiles of growth inhibition activ-
ities of the drugs against a panel of cell lines are determined, it is
possible to make a correlation map using gene expression levels
of the target proteins obtained. Eventually, this map would be use-
ful to predict drug effectiveness for clinical usage.
There are few reports on cellular functions of ZNF638; Mueller
and co-workers recently reported that ZNF638 played a role as a
transcriptional coregulator of adipocyte differentiation via induc-
5. Conclusions
tion of peroxisome proliferator-activated receptor c, and ZNF638
In summary, we have identified three proteins, CaM, hnRNPA2/
B1, and ZNF638 as candidate target proteins of RID-G, a tamoxifen
analog, by a chemical genetic approach. In this approach, we com-
bine phage display screen with a statistical analysis using gene
expression profiles of JFCR39 cell lines. Our results suggest that
RID-G binds to the three proteins and modulates their functions
to exert cytotoxicity against cancer cells.
also interacts with splicing regulators to influence alternative
splicing.31,32 To examine the interaction between RID-G and
ZNF638, we attempted to express a recombinant ZNF638 peptide
in E. coli. However, because ZNF638 peptide might be toxic for
E. coli, we failed to obtain the recombinant peptide. Although fur-
ther studies to determine the binding are needed in the future, the
statistical analysis of gene expression profiles and cytotoxicity sug-
gested that cancer cell lines expressing high levels of ZNF638 were
clearly sensitive to RID-G treatment (Fig. 3B).
Acknowledgements
Considering all our data collectively, we hypothesized that RID-
G could bind to the three proteins and modulate their activity to
elicit its cytotoxicity against cancer cells. Expression profiles of
the three proteins in the JFCR39 cell lines were used to make a cor-
relation map, as shown in Figure 6. The red area indicates the cell
lines with a low CALM1 expression level and a high ZNF638 expres-
sion level, and black circle represent cell lines with a high
HNRNPA2B1 expression level. Our correlation analysis of gene
expression with RID-G cytotoxicity suggested that cells expressing
CALM1 at a low level and HNRNPA2B1 and ZNF638 at a high level
were expected to be relatively sensitive to RID-G treatment
(Fig. 3B). Indeed, cell lines shown as black circles in the red area
had relatively low GI50 values and are designated as larger circles.
The DMS114 cell line (lung cancer, indicated by an arrow), satisfied
all the three conditions. Interestingly, this cell line was the most
sensitive to RID-G treatment among the JFCR39 cell lines with
This work was supported by a Health and Labour Sciences
Research Grants from the Ministry of Health, Labour and Welfare,
Japan to I.S. (11103425).
Supplementary data
Supplementary data associated (amino acid sequences and fre-
quency displayed on phage clones after the screening, synthetic
scheme of RID-G and Bio-RID-G, cytotoxicity of RID-G against HeLa
cells, supplementary materials and methods, 1H and 13C NMR spec-
tra for all novel compounds) with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.bmc.2015.08.001.
References and notes
1. Aoki, S.; Morohashi, K.; Sunoki, T.; Kuramochi, K.; Kobayashi, S.; Sugawara, F.
Bioconjugate Chem. 2007, 18, 1981.
the lowest GI50 value (0.19 lM). This correlation map may enable

6124
K. Ikeda et al. / Bioorg. Med. Chem. 23 (2015) 6118–6124
2. Kusayanagi, T.; Tsukuda, S.; Shimura, S.; Manita, D.; Iwakiri, K.; Kamisuki, S.;
Takakusagi, Y.; Takeuchi, T.; Kuramochi, K.; Nakazaki, A.; Sakaguchi, K.;
Kobayashi, S.; Sugawara, F. Bioorg. Med. Chem. 2012, 20, 6248.
3. Mandlekar, S.; Kong, A. N. Apoptosis 2001, 6, 469.
4. Tashiro, E.; Imoto, M. Bioorg. Med. Chem. 2012, 20, 1910.
5. Tsukuda, S.; Kusayanagi, T.; Umeda, E.; Watanabe, C.; Tosaki, Y.; Kamisuki, S.;
Takeuchi, T.; Takakusagi, Y.; Shiina, I.; Sugawara, F. Bioorg. Med. Chem. 2013, 21,
311.
16. Shiina, I.; Ibuka, R.; Kubota, M. Chem. Lett. 2002, 286.
17. Shiina, I.; Kubota, M.; Oshiumi, H.; Hashizume, M. J. Org. Chem. 2004, 69, 1822.
18. Shiina, I.; Fukui, H.; Sasaki, A. Nat. Protoc. 2007, 2, 2312.
19. Zhang, M.; Tanaka, T.; Ikura, M. Nat. Struct. Biol. 1995, 2, 758.
20. Shim, J. S.; Lee, J.; Park, H. J.; Park, S. J.; Kwon, H. J. Chem. Biol. 2004, 11, 1455.
21. Zhou, J.; Allred, D. C.; Avis, I.; Martínez, A.; Vos, M. D.; Smith, L.; Treston, A. M.;
Mulshine, J. L. Breast Cancer Res. Treat. 2001, 66, 217.
22. Yan-Sanders, Y.; Hammons, G. J.; Lyn-Cook, B. D. Cancer Lett. 2002, 183, 215.
23. Lee, C. L.; Hsiao, H. H.; Lin, C. W.; Wu, S. P.; Huang, S. Y.; Wu, C. Y.; Wang, A. H.;
Khoo, K. H. Proteomics 2003, 3, 2472.
6. Takakusagi, Y.; Takakusagi, K.; Sugawara, F.; Sakaguchi, K. Expert Opin. Drug
Discovery 2010, 5, 361.
7. Morohashi, K.; Sahara, H.; Watashi, K.; Iwabata, K.; Sunoki, T.; Kuramochi, K.;
Takakusagi, K.; Miyashita, H.; Sato, N.; Tanabe, A.; Shimotohno, K.; Kobayashi,
S.; Sakaguchi, K.; Sugawara, F. PLoS One 2011, 6, e18285.
8. Shimura, S.; Ishima, M.; Nakajima, S.; Fujii, T.; Himeno, N.; Ikeda, K.; Izaguirre-
Carbonell, J.; Murata, H.; Takeuchi, T.; Kamisuki, S.; Suzuki, T.; Kuramochi, K.;
Watashi, K.; Kobayashi, S.; Sugawara, F. J. Am. Chem. Soc. 2013, 135, 18949.
9. Arango, D.; Morohashi, K.; Yilmaz, A.; Kuramochi, K.; Parihar, A.; Brahimaj, B.;
Grotewold, E.; Doseff, A. I. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, E2153.
10. Shiina, I.; Sano, Y.; Nakata, K.; Kikuchi, T.; Sasaki, A.; Ikekita, M.; Hasome, Y.
Bioorg. Med. Chem. Lett. 2007, 17, 2421.
24. Lee, C. H.; Lum, J. H.; Cheung, B. P.; Wong, M. S.; Butt, Y. K.; Tam, M. F.; Chan, W.
Y.; Chow, C.; Hui, P. K.; Kwok, F. S.; Lo, S. C.; Fan, D. M. Proteomics 2005, 5, 1160.
25. Patry, C.; Bouchard, L.; Labrecque, P.; Gendron, D.; Lemieux, B.; Toutant, J.;
Lapointe, E.; Wellinger, R.; Chabot, B. Cancer Res. 2003, 63, 7679.
26. Fernandez, T. F.; Samal, A. B.; Bedwell, G. J.; Chen, Y.; Saad, J. S. J. Biol. Chem.
2013, 288, 21898.
27. Ahn, E. Y.; Pan, G.; Oh, J. H.; Tytler, E. M.; McDonald, J. M. Am. J. Pathol. 2003,
163, 2053.
28. Pawar, P.; Ma, L.; Byon, C. H.; Liu, H.; Ahn, E. Y.; Jhala, N.; Arnoletti, J. P.;
McDonald, J. M.; Chen, Y. Clin. Cancer Res. 2009, 15, 1288.
29. Vickers, S. M.; Jhala, N. C.; Ahn, E. Y.; McDonald, J. M.; Pan, G.; Bland, K. I. Ann.
Surg. 2002, 235, 872.
11. Anlifeire, A.; Hatori, M.; Morita, A.; Shiina, I.; Nakata, K.; Tosaki, Y.; Wang, Y.-
W.; Ikekita, M.; Li, G. Chin. J. Cell Biol. 2011, 33, 635.
12. Shiina, I. Jpn. Kokai Tokkyo Koho (24 April 2008), JP2008094836.
13. Guo, W. Z.; Wang, Y.; Umeda, E.; Shiina, I.; Dan, S.; Yamori, T. Biol. Pharm. Bull.
2013, 36, 1008.
30. Golan-Gerstl, R.; Cohen, M.; Shilo, A.; Suh, S. S.; Bakàcs, A.; Coppola, L.; Karni, R.
Cancer Res. 2011, 71, 4464.
31. Du, C.; Ma, X.; Meruvu, S.; Hugendubler, L.; Mueller, E. J. Lipid Res. 2014, 55,
1886.
14. Yu, D.; Forman, B. J. Org. Chem. 2003, 68, 9489.
15. Mukaiyama, T. Angew. Chem., Int. Ed. Engl. 1977, 16, 817.
32. Meruvu, S.; Hugendubler, L.; Mueller, E. J. Biol. Chem. 2011, 286, 26516.