Bisphenol A derivatives act as novel coactivator-binding inhibitors for estrogen receptor β

Bisphenol A and its derivatives are recognized as endocrine disruptors based on their complex effects on estrogen receptor (ER) signaling. While the effects of bisphenol derivatives on ERα have been thoroughly evaluated, how these chemicals affect ERβ signaling is less well understood. Herein, we sought to identify novel ERβ ligands using a radioligand competitive binding assay to screen a chemical library of bisphenol derivatives. Many of the compounds identified showed intriguing dual activities as both ERα agonists and ERβ antagonists. Docking simulations of these compounds and ERβ suggested that they bound not only to the canonical binding site of ERβ but also to the coactivator binding site located on the surface of the receptor, suggesting that they act as coactivator-binding inhibitors (CBIs). Receptor–ligand binding experiments using WT and mutated ERβ support the presence of a second ligand-interaction position at the coactivator-binding site in ERβ, and direct binding experiments of ERβ and a coactivator peptide confirmed that these compounds act as CBIs. Our study is the first to propose that bisphenol derivatives act as CBIs, presenting critical insight for the future development of ER signaling–based drugs and their potential to function as endocrine disruptors.

Bisphenol A and its derivatives are recognized as endocrine disruptors based on their complex effects on estrogen receptor (ER) signaling. While the effects of bisphenol derivatives on ERα have been thoroughly evaluated, how these chemicals affect ERβ signaling is less well understood. Herein, we sought to identify novel ERβ ligands using a radioligand competitive binding assay to screen a chemical library of bisphenol derivatives. Many of the compounds identified showed intriguing dual activities as both ERα agonists and ERβ antagonists. Docking simulations of these compounds and ERβ suggested that they bound not only to the canonical binding site of ERβ but also to the coactivator binding site located on the surface of the receptor, suggesting that they act as coactivator-binding inhibitors (CBIs). Receptor-ligand binding experiments using WT and mutated ERβ support the presence of a second ligandinteraction position at the coactivator-binding site in ERβ, and direct binding experiments of ERβ and a coactivator peptide confirmed that these compounds act as CBIs. Our study is the first to propose that bisphenol derivatives act as CBIs, presenting critical insight for the future development of ER signaling-based drugs and their potential to function as endocrine disruptors.
Estrogen receptors (ERs) are members of the nuclear receptor family of transcription factors that directly bind to consensus nucleotide sequences to induce gene transcription. Forty-eight human nuclear receptors have been identified, including those for sex steroid hormones, glucocorticoids, retinoids, and vitamin D (1, 2), with many of these receptors recognized as therapeutic targets for a wide range of diseases (3). In particular, ERs are major drug targets for breast cancer (4) and menopausal disorders. Two ER isoforms exist, ERα and ERβ, that have high amino acid similarity in both the DNAbinding domains and ligand-binding domains (LBDs) (5). Many ERα and/or ERβ-associated gene disruption experiments have been reported (6). Female mice lacking ERα are infertile, whereas male mice exhibit decreased fertility (7). Disruption of ERα in female mice leads to hypoplastic uteri, and ERα-disrupted female mice do not respond to estradiol treatments. ERβ KO mice present with less-severe phenotypes than those with ERα KO, although ERβ-disrupted female mice are subfertile predominantly because of reduced ovarian efficiency (8). Moreover, ERα and ERβ double-KO mice show normal reproductive tract development during the prepubertal period. However, those animals present with similar features to ERα KO mice during adulthood. Furthermore, this diagnostic phenotype indicates that ERβ plays a role in oocyte progression in the postnatal ovary (9,10). Both ERα and ERβ are activated by endogenous estrogens; however, their expression patterns and actions are different (11), with each receptor assumed to have specific biological functions.
for endogenous ERs in MCF-7 cells (27), BPC was considered but ultimately not included in the list of in vitro endocrine disruptors by the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM) (NIH Publication No: 03-4503) in 2003. Historically, the designation of 2,2-bis(4-hydroxy-3-methylphenyl) propane (CAS No. 79-97-0, which does not have chlorine atoms) as BPC has led to some confusion in the literature; however, chlorinecontaining BPC has been detected in human breast milk (28).
ERα and/or ERβ are major targets of EDCs that interfere with their estrogen-responsive signaling pathways (29). Human ERα and ERβ have almost identical DNA-binding domains, differing by only two amino acids, and both receptors bind the same estrogen-response elements in transcriptional control regions. Although ERα and ERβ also have similar LBDs, they have some distinctive features in terms of ligand selectivity and target gene regulation (30). Endogenous estrogen, 17-β estradiol (E2), binds to ERα slightly stronger than to ERβ. Similarly, BPA binds ERα with higher affinity than ERβ, although its binding abilities are much weaker than those of E2. In contrast, BPAF and BPC display higher affinity for both ERα and ERβ than BPA, with a preference for ERβ over ERα binding. BPAF and BPC show antagonistic activity against ERβ in reporter gene assays using HeLa cells (31,32). BPAF and BPC show much stronger antagonist activity for ERβ than ERα, (32,33). While crystal structures have provided insight into ERα activation/inactivation mediated by BPAF and BPC binding (32,33), the structural changes induced by the strong antagonistic activity of BPAF and BPC against ERβ are not well established. Recently, we found that the bisphenol moiety is a privileged structure for ERα. Here, we describe the biphasic binding of BPAF and BPC to ERβ and propose a novel two-site binding model of the ERβ-BPC complex, based on the crystal structure of 4-hydroxytamoxifen (4OHT) bound to ERβ. This is the first study to mechanistically associate the antagonistic actions of EDCs with interactions at the coactivator-binding site, thereby providing insight into developing safer raw materials that do not exhibit endocrine-disrupting features.
To gain insight into the differences observed in ERβ and ERα binding, we compared the ligand-binding cavities in the deposited ERβ and ERα LBD crystal structures. The sizes of the canonical binding pockets were calculated for 45 ERα and 25 ERβ structures in their active conformations using MOE SiteFinder function, and the amino acid residues surrounding the bound ligands identified (Tables S2 and S3). The average ERβ pocket was smaller than for ERα (430.9 Å 3 and 369.3 Å 3 for ERα and ERβ, respectively; Fig. 2A). The typical ligandbinding pockets of each receptor in the active conformation are illustrated (Fig. 2, C and D). Moreover, the average size of the ligand-binding pocket in E2-bound ERα and ERβ structures was 419.4 Å 3 and 385.0 Å 3 , respectively, and in genistein-bound ERα and ERβ structures was 475.9 Å 3 and 375.8 Å 3 , respectively. Although these results suggested that ERα is able to accept larger ligands than ERβ, the amino acid residues surrounding the ligands differ slightly. Some of the smaller ligands fit more adequately into the ERβ than the ERα ligand-binding pocket.

Docking analysis predicts BPC binding to the surface of ERβ
To investigate the contrasting actions of BPA derivatives as ERβ antagonists and ERα agonists, we performed docking simulations using the LBD of human ERβ and BPC, the strongest binder among the BPA derivatives examined using a competitive binding assay with [ 3 H]E2. Possible ligand-binding sites in 38 deposited ERβ crystal structures were identified using MOE SiteFinder, a program for binding-site analysis equipped in the Molecular Operating Environment (MOE). Canonical as well as putative binding sites were ranked according to propensity for ligand binding (PLB), a specific parameter in MOE SiteFinder (37). Consistently, the top five predicted sites in each structure were the canonical ligand-binding sites. Interestingly, an actual surface 4OHT-binding site close to the hydrophobic groove for the coactivator recognition surface of ERβ (PDB ID: 2FSZ) was ranked 11th in the PLB order. Moreover, this location was a predicted binding site on all antagonist-bound ERβ structures, based on PLB. Notably, this second site was not predicted as a binding site on over half of the agonist-bound structures (Table S6). These predictions suggest that ERβ antagonism induced by BPC and other BPA derivatives may be due to inhibition of coactivator recruitment. Next, we performed a docking simulation for ERβ LBD and BPC using both its canonical and second binding sites as target rooms. BPC was able to fit and bind in both rooms, with one of its chlorine atoms interacting with the tryptophan residue (Trp335) on helix 5 via halogen interaction (Fig. 3, A and B). The obtained model structure suggested that BPC binding to the second binding site prevented recruitment of coactivators for gene transcriptions, similar to 4OHT (Fig. 3, C and D). We hypothesized that the binding affinity of BPA derivatives to this coactivator binding site would correlate with antagonistic activity. To explore this notion, docking simulations were performed for each BPA derivatives (Fig. S3), and the free energy of ligand binding evaluated using a docking simulation and the GBVI/WSA dG scoring function (larger negative scores indicate more stable ligand/ receptor complexes) (38). Correlation of the GBVI/WSA dG scores with the extent of antagonism (reported as the % inhibition of 10 nM E2 induced transcriptional activity) revealed a linear relationship (correlation coefficient of -0.83), suggesting that inhibition of coactivator recruitment underlies the antagonism of ERβ by BPA derivatives (Fig. 3E).

Binding of the coactivator peptide is reduced by BPC
Ligand binding induces a conformation change in the ERβ LBD that facilitates its translocation to the cell nucleus and the subsequent recruitment of coactivator proteins. To explore the effects of BPC on ERβ activation, surface plasmon resonance experiments were performed to measure the direct binding of the coactivator peptide derived from human nuclear receptor coactivator 1, also known as steroid receptor coactivator (SRC1). Consistently, the E2 ligand increased SRC1 peptide binding to ERβ-LBD (K d 3.3 ± 0.6 μM and 9.1 ± 0.7 μM with and without E2, respectively; Fig. 3G). Notably, SRC1 peptide binding was reduced in the presence of BPC (K d 16.4 ± 0.9 μM; Fig. 3H).

Biphasic 4OHT binding indicative of two ERβ-binding sites
To further support the presence of a second ligand-binding site, competitive binding assays were performed using BPA, BPC, and BPAF and tritium-labeled 4OHT ([ 3 H]4OHT) (Fig. 4A). Notably, a biphasic dose-response curve was observed for BPC (18.1 nM and 2281 nM IC 50 ) that was not evident in the [ 3 H]E2 competitive analyses. Similarly, BPAF displayed a biphasic binding curve, albeit with weaker binding at both the high-and low-affinity sites than BPC. Moreover, 4OHT showed a biphasic curve, consistent with the 4OHT/ ERβ crystal structure (PDB: ID 2FSZ). In contrast, BPA, which did not elicit antagonistic activity, showed a sigmoidal curve indicative of a single ligand-binding site. Interestingly, the trifluorine substitution of the methyl groups in BPAF increased ERβ binding 50-fold compared with BPA. These results confirmed the presence of two distinguishable binding sites for BPC and BPAF on ERβ. In contrast, the typical sigmoidal curves seen in E2 competitive binding assays using [ 3 H]4OHT and [ 3 H]E2 are indicative of single ligand-binding site.

Trp335 is required for biphasic ligand binding
The docking simulations suggested that hydrophobic interactions between the BPA derivatives and the indole group of Trp335 were required for ERβ binding and identified a Figure 3. ERβ harbors two ligands in its LBD. A, two BPC bound to ERβ during the docking simulation. The canonical binding site is indicated in gray; the second binding site, located on the surface of the receptor, is shown in magenta. The activation helix, H12, is indicated in magenta. B, chlorine, a halogen atom of BPC, interacted with the Trp335side chain via halogen interaction in the second binding site. BPC and 4OHT are illustrated in blue and gray, respectively, in the stick model. C, superimposition of the calculated BPC-bound ERβ structure (blue) and its agonist form with the nuclear receptor coactivator 1, SRC1 (green, PDB ID: 3OLL). SRC1 is indicated as a red α-helix, H12 of its agonist form is indicated in purple, BPC is illustrated in blue, and 4OHT is shown in gray. BPC clashed with the amino acid residues on H12 in the ERβ agonist form; therefore, BPC prevented the ERβ activation. BPC and 4OHT disrupted the SRC1 binding due to steric hindrance of the amino acid residues shown in the red stick models. D, in ERβ-agonist form, amino acid residues surrounding Trp335 within 4.5 Å on H12 are shown in the purple stick model, while leucine residues on the SRC1 LXXLL motif are indicated via the red stick model. E and F, correlation of the calculated binding scores and inhibitory activity for ERβ. Inhibitory activity is defined as the ratio of chemicals inhibiting transcriptional activity induced by 10 nM E2. *p < 0.05, **p < 0.01, ***p < 0.001. G and H, dose response of SRC1 peptide binding to ERβ LBD in the presence of (G) 10 μM E2 or (H) 10 μM BPC. 4OHT, 4-hydroxytamoxifen; BPC, bisphenol C; E2, 17-β estradiol; ERs, estrogen receptors; LBD, ligand-binding domain; SRC1, steroid receptor coactivator. potential halogen interaction between the chlorine atom of BPC and the indole ring. To determine the contributions of these putative interaction to BPC binding, the corresponding tryptophan was mutated to alanine (A). Saturation binding assays revealed a typical sigmoidal dose-response curve and a K d of 23.1 nM for E2 against ERβ(W335A), indicating preservation of the canonical binding site (Fig. S4A).
Competitive binding assays confirmed two 4OHT-binding sites in ERβ, with K d values of 4.6 nM and 53.1 nM. In contrast, a single binding site was evident in ERβ(W335A) (K d 34.2 nM) (Fig. S4B). Similarly, the biphasic binding of BPC and BPAF was lost in the ERβ(W335A) mutant (Fig. 4, A and  B). The IC 50 values of 4OHT, BPC, and BPAF were 106 ± 51 nM, 691± 29 nM, and 1249 ± 579 nM, respectively. BPA illustrated a typical sigmoidal competitive dose-response curve against ERβ(W335A), similar to the result against ERβ. These results indicated that replacing Trp for Ala compromises the second 4OHT and BPA derivatives binding site on the surface of the ERβ LBD.

W335A reduces ERβ transcription activity
Reporter assays revealed that E2-induced transcriptional activation was markedly reduced by the tryptophan to alanine substitution in ERβ (Fig. 4, C and D). Given that E2 binding ability was retained, this is consistent with reduced coactivator binding. Indeed, in the active conformation, Trp335 interacts with Leu491, Met494, and Leu495 on H12 (Fig. 4E). Supporting this notion, the SRC1 peptide bound poorly to ERβ(W335A), as measured by surface plasmon resonance experiments using Biacore T100 (Fig. S5). These results indicated that Trp335 on the ERβ coactivator-binding site plays an important role, not only in interacting with bisphenol derivatives but also in recruiting coactivators on the surface of ERβ by stabilizing H12 in its active conformation.

Discussion
Here, we report the ERβ transcriptional activities of BPA derivatives including BPC and BPAF using a combination of receptor binding and reporter assays. Of note, 18 derivatives bound ERβ with higher affinity than BPA. The binding abilities of these BPA derivatives are stronger than those of known environmental chemicals such as dichlorodiphenyltrichloroethane, nonylphenol, phytoestrogens, and dioxins (39). Unexpectedly, our results clearly showed that many BPA derivatives function as ERβ antagonists, contrasting with their previously reported ERα agonism. Docking simulations indicated that

Dual role of bisphenols as ERα agonists and ERβ antagonists
BPA derivatives bind to a second site located near the coactivator-binding site on the surface of ERβ-LBD that requires interactions with Trp335. Mutation of tryptophan to alanine led to the loss of this low-affinity binding site in ERβ. These results indicated that some BPA derivatives act as antagonists, although most of EDCs, including BPA, are assumed ER agonists. We previously reported that most of the BPA derivatives examined in this study act as weak agonists for ERα. The results obtained in this study demonstrate the importance of screening for both agonist and antagonist activity, especially against ERβ.
Several ERα-or ERβ-specific agonists have been reported, including propyl pyrazole triol that selectively binds to and transcriptionally activates ERα (42). The first chemical shown to function as an ERα agonist and ERβ antagonist is HPTE, a metabolite of the banned pesticide, methoxychlor [1,1,1trichloro-2,2-bis(4-methoxyphenyl)ethane] (43,44). Accumulated knowledge gained from protein crystal structures emphasize the importance of halogens in receptor-ligand interactions (45,46). We found that in addition to the halogen containing BPAF and BPC, many BPA derivatives display ERα agonist activities similar to HPTE. These results indicate the complexity of establishing the mechanisms of action of environmental chemicals that activate or suppress the physiological functions of one or more nuclear receptors. In particular, antagonist activities might be overlocked if both binding affinity and transcriptional activity are not determined, as environmental chemicals are typically categorized based on the ability to active ERs.
Recent studies have indicated the value of small molecules that bind to coactivator protein-binding sites on nuclear receptors (47). Coactivator-binding inhibitors (CBIs) have been developed for ERs, an androgen receptor, a progesterone receptor, a vitamin D receptor, a thyroid hormone receptor, a pregnane X receptor, a retinoid X receptor, and peroxisome proliferator-activated receptors (48)(49)(50)(51). This study is the first to conclude that EDCs can function as CBIs for ERβ, indicating the importance of assessing both agonist and antagonist activities of these chemicals.
In summary, we showed that tricyclic bisphenols elicit antagonistic activity against both ERα and ERβ. Our results also indicate that many next-generation bisphenols are agonists and antagonists of ERα and ERβ. Mutagenesis of an ERβ surface amino acid indicated that these next-generation bisphenols act as CBIs. While in silico docking analyses support this mechanism of action, future crystallographic studies will be required to provide more direct information on CBIs. This study highlights the mechanistic complexity of the next-generation of bisphenols acting as EDCs.

ERβ expression and purification
The LBD of ERβ (amino acids 263-530) was expressed as a GST-fused protein for receptor-binding assays. Human ERβ cDNA was obtained from OriGene Technologies. The cDNA of ERβ-LBD was amplified using PCR and subcloned into a pGEX-6p-1 expression vector. The expression of GST-fused ERβ-LBD was induced by 1 mM IPTG in Escherichia coli BL21α at 16 C for overnight. The resulting crude protein was affinity-purified using Glutathione-Sepharose 4B (Cytiva), followed by gel filtration using a Sephadex G-10 column (Cytiva).

Radioligand-binding assay
Radioligand-binding assays for ERβ and ERβ(W335A) were performed mainly according to a previously reported method (31,34). Saturation binding assays were conducted with [ 3 H] E2 or [ 3 H]4OHT using GST-ERβ-LBD or GST-ERβ(W335A)-LBD to evaluate the binding ability of radiolabeled compounds. The reaction mixtures of each LBD (20 ng) and a series of concentrations of [ 3 H]E2 (0.01-10 nM) or [ 3 H]4OHT (0.1-30 nM) were incubated in a total volume of 100 μl of the binding buffer (10 mM Tris-buffered saline (pH 7.4), 1 mM EGTA, 1 mM EDTA, 10% glycerol, 0.5 mM PMSF, 0.2 mM leupeptin, and 1 mM sodium vanadate (V)) at 20 C for 2 h, to analyze total binding. Corresponding reaction mixtures, containing 10 μM nonlabeled E2 or 4OHT, were incubated to detect each nonspecific binding. [ 3 H]E2 or [ 3 H]4OHT-specific binding was evaluated by subtracting the obtained radioactivity values of total binding from the those of nonspecific binding. After successive incubation with 100 μl of 0.4% dextran-coated charcoal (DCC) (Sigma-Aldrich) in PBS (pH 7.4) on ice for 10 min, free radioligands bound to DCC were removed using a vacuum filtration system with a 96-well filtration plate (Mul-tiScreenHTS HV, 0.45-mm pore size, Merck KGaA) for the bound/free separation. The radioactivity of each eluent was measured using a liquid scintillation counter (LS6500; Beckman Coulter) and Clear-sol I (Nacalai Tesque Inc). Calculated specific binding of [ 3 H]E2 was assessed using Scatchard plot analysis (52). Competitive binding assays were performed to evaluate the binding ability of each test compound using [ 3 H]E2, for a library screening or detailed BPA binding assay. Each compound was dissolved in dimethyl sulfoxide to prepare a 1.0 mM stock solution and further diluted to prepare serial dilutions (10 −12 M to 10 −5 M) in the binding buffer. To assess their binding abilities, each compound was incubated with GST-ERβ-LBD or GST-ERβ(W335A)-LBD (20 ng) and radiolabeled ligand (5 nM of [ 3 H]E2 or 5 nM of [ 3 H]4OHT, final concentration) for 2 h at 20 C. Bound/free separation was performed as described above, and the radioactivity was determined using a MicroBeta microplate counter (PerkinElmer Inc). The IC 50 value of each test compound was calculated from the dose-response curves generated via nonlinear regression analysis using Prism software (GraphPad Software Inc).

Luciferase reporter gene assay
Transcriptional activities of ERβ and ERβ(W335A) were measured as previously reported previously (31,34). HeLa cells were maintained in Eagle's minimum essential medium (Nissui Pharmaceutical Co, Ltd) supplemented with DCC-treated fetal bovine serum (10%, v/v) at 37 C under 5% CO 2 . To evaluate agonistic activity, HeLa cells were seeded at a density of 5 × 10 5 cells per 60-mm dish and cultured for 24 h, followed by transfection of the reporter plasmid (3 μg, pGL4.23/3×ERE) and each expression plasmid (1 μg, pcDNA3.1/ERβ or pcDNA3.1/ERβ(W335A)) using Lipofectamine LTX with Plus Reagent (Thermo Fisher Scientific, Inc), according to the manufacturer's instructions. After incubation for 24 h, cells were harvested and seeded onto 96-well plates at 5 × 10 4 cells/ well, and then treated with a series of the test compounds (10 −12 M to 10 −5 M, final concentration) diluted with 1% bovine serum albumin/PBS (v/v). After a 24-h incubation, luciferase activity was measured using the ONE-Glo Luciferase Assay System (Promega Co) on an EnSpire multimode plate reader (PerkinElmer, Inc). To analyze antagonistic activity, serial concentrations of test compounds (10 −12 M to 10 −5 M) were treated in the presence of 10 nM E2, which normally induces full transcriptional activity levels in transiently expressed ERβ.

Docking simulation of each antagonist onto the ERβ LBD
Three-dimensional coordinates of the compounds were obtained from the Cambridge Structural Database (CSD-Core, The Cambridge Crystallographic Data Centre). Ligand IDs of compounds utilized for docking simulations are summarized in Table S7. For the compounds with no corresponding entry in the CSD System, 3D coordinates were constructed in silico using Gaussian 16 (Gaussian, Inc), with the basis set of 6-31G. Docking simulations for the ligand-ERβ complex were performed using a Dock functions in the MOE package (Chemical Computing Group); the free energy of each complex was evaluated according to its GBVI/WSA dG score (38).
Ligand-binding cavity volumes of the deposited crystal structures were analyzed and calculated using the MOE SiteFinder function in MOE.

Binding analysis of ERβ LBD and SRC1 peptide by surface plasmon resonance
The anti-GST antibody was immobilized on a Sensor Chip CM5 (Cytiva) using Amine Coupling kit (Cytiva) and GST Capture kit (Cytiva) according to the manufacturer's instruction for Biacore T100 instrument (Cytiva). The binding of SRC1 peptide (amino acids 685-697; ERHKILHRLLQEG) to the ERβ-LBD was analyzed by capturing GST-ERβ-LBD on the sensor chip and injecting SRC1 peptide with E2 or BPC. The peptide was synthesized using the ABI 433A peptide synthesizer (Applied Biosystems) by the solid-phase method with Fmoc chemistry. GST-ERβ-LBD (50 μg/ml) was incubated with 10 μM E2 or 10 μM BPC for 1 h and captured at 25 C with a flow rate of 5 μl/min on the sensor chip. Binding between SRC1 peptide and ERβ-LBD was analyzed using HBS-EP+ buffer (0.01 M Hepes, pH 7.4, 0.15 M NaCl, 3 mM EDTA, and 0.05% (w/v) Surfactant P20) as a running buffer under the following conditions: contact time 120 s, flow rate 30 μl/min, and dissociation time 180 s. The sensor chip was recovered by 10 mM Gly-HCl (pH 2.0) with a flow rate of 20 μl/min and a contact time of 120 s. The data obtained were analyzed using the Biacore T100 evaluation software.

Statistical analysis
Significance of the data between experimental groups was determined using unpaired t-tests. Data are presented as the mean ± SD, and p values are summarized in supplementary tables.

Data availability
All data needed to evaluate the conclusions in the article are present in the article and/or the supporting information.
Supporting information-This article contains supporting information.
Acknowledgments-We appreciate R.T. Yu and A.R. Atkins (Salk Institute for Biological Studies) for helpful suggestions and discussions, and L. Ong and C. Brondos for administrative assistance. We appreciate Y. Shimohigashi (Kyushu University) for providing the chemical library and X. Liu (Kyushu University) for providing the ERβ-mutated plasmid. We thank the RIKEN BRC Cell Bank and Cell Resource Center for Biomedical Research, Institute of Development, Aging and Cancer, Tohoku University, for providing the HeLa cells. This work was supported in part by a grant from Izumi Science and Technology Foundation.