Inhibition of CDK2 reduces EZH2 phosphorylation and reactivates ERα expression in high-grade serous ovarian carcinoma

Abstract

The cyclin-dependent kinase 2 (CDK2) inhibitor dinaciclib, a potential anti-cancer drug, has been tested in clinical trials and reported to suppress tumor initiating cells. Our recent study demonstrated that pharmacological inhibition of CDK2 or enhancer of zeste homolog 2 (EZH2) allows re-expression of ERα and converts triple-negative breast cancers (TNBC) to luminal ERα-positive, rendering TNBC cells targetable by tamoxifen. Like TNBC, EZH2 is also commonly overexpressed in ovarian cancers, and overexpression of cyclin E1 gene (CCNE1) and/or amplification of its associated kinase CDK2 gene is present in ovarian tumor specimens, both of which are associated with primary treatment resistance and poor outcome in high-grade serous ovarian cancer (HGSC). We determined whether inhibition of CDK2-mediated phosphorylation of EZH2 activates ERα expression in ERα-negative HGSOC cells, rendering them targetable by hormonal therapy. The specific CDK2 inhibitor repressed phosphorylation of EZH2 at T416, and in turn activated the expression of its downstream target ERα gene (ESR1). We tested the efficacy of the combination of CDK2 inhibitor and tamoxifen and found significant synergistic inhibition. We further demonstrated that CDK2 inhibitor is a more promising agent than EZH2 inhibitor in repressing TNBC and HGSOC due to a feedback increase in CDK2 activity by EZH2 inhibitor. Our results indicated that the combination treatment of CDK2 inhibitor and tamoxifen has the potential to benefit patients with ERα-negative HGSOC.

Introduction

EZH2 is the core component of the PRC2 complex and catalyzes H3K27 trimethylation to silence its target genes [1-4]. Studies have shown that EZH2 is one of the key epigenetic factors involved in maintaining self-renewal, pluripotency, and differentiation of embryonic and adult stem cells [2-6]. We previously showed that expression of EZH2 sustains a reversible and undifferentiated stem cell-like phenotype in breast cancer cells and contributes to their expansion [6-8]. It has also been reported that elevated expression of EZH2 correlates with poorly differentiated breast carcinomas and poor prognosis in many types of solid tumors [9]. Thus, EZH2 may serve as a therapeutic target whose inactivation de-represses its downstream genes [9,10]. Tazemetostat (EPZ-6348), a highly selective EZH2 inhibitor, is currently under evaluation in clinical trials for different types of cancers and has been submitted for FDA approval in July 2019. The FDA granted priority review to tazemetostat for the treatment of patients with metastatic or locally advanced epithelioid sarcoma who are not eligible for curative surgery and has been first approved in January 23, 2020.

Ovarian cancer is the second most lethal gynecological malignancy with the highest mortality rate and relapses in about 70% patients after primary surgery and chemotherapy [11-14]. High-grade serous ovarian carcinoma (HGSOC) is the most frequent type (> 70% of epithelial ovarian cancer) of ovarian cancers and accounts for the most tumor-associated mortalities [15,16]. The CCNE gene, which encodes cyclin E, is one of the most frequently amplified genes in HGSOC [17]. Cyclin E mainly coordinates with cyclin-dependent kinase 2 (CDK2) to facilitate G1/S cell cycle progression. In ovarian cancers, elevated CCNE1 is often correlated with higher CDK2 expression, and most of CCNE1-associated tumor promoting effects require the participation of CDK2 [18]. Therefore, CDK2 is a feasible target in HGSOC patients with high CDK2 expression. Currently, CDK2 inhibitor dinaciclib is under investigation in a phase 2/3 clinical trial, and preliminary results showed encouraging single-agent activity in patients with relapsed multiple myeloma [19]. In randomized phase II clinical trial for the patients with advanced breast cancer, dinaciclib monotherapy demonstrated some antitumor activity and was generally tolerated although the efficacy was not superior to capecitabine chemotherapy [20,21]. Thus, it is worthwhile to evaluate dinaciclib in selected patient populations with metastatic breast cancer and in combination with other agents. HGSOC is a highly heterogeneous disease with extremely poor prognosis and limited therapeutic options [11,16]. Because the ovaries are hormonal related organs and most HGSOCs are highly invasive, the development of novel therapeutic strategies remains an obstacle to overcome [11,16,22], which encouraged us to develop an effective combination therapy for this highly invasive cancer.

The female sex hormone estrogen is mainly secreted by the ovaries and originated in granulosa cells. The ovarian surface epithelium is the main cell type of the ovaries. In epithelial ovarian cancers, the estrogen receptor (ER)-positive rate is about 43-81%, depending on the methodological criteria [11,15,22,23]. ERα belongs to members of the steroid receptor superfamily and plays a crucial physiological roles of the target organs, including genital tract, bone, mammary glands, adipose tissues, and brain, as transcriptional factors [24]. A large number of ovarian tumor tissue microarray analyses indicated that ERα expression is high in 60.2% and low or negative in 20.5% of HGSOC (n = 1691, The Cancer Genome Atlas) [11,25], which is similar to that (81%) in breast cancer, a disease which has been treated by hormone therapy for decades [26]. However, about 20-30% of patients with HGSOC who are ERα-negative show poor response to hormone therapy [27].

Tamoxifen (TAM) is not the first non-steroidal anti-estrogen treatment but has been widely used to treat women with luminal breast cancer before menopause [28]. TAM binds to ER instead of estrogen to block ERα signaling [29]. In a recent study, we demonstrated that CDK2 and EZH2 inhibitors can induce ERα re-expression in TNBC and that the combination therapy of CDK2 or EZH2 inhibitor and TAM may be a promising approach for TNBC [30]. Clinical studies indicated that patients with relapsed HGSOC and an ER histoscore > 200 who received endocrine therapies have significant greater benefits [31]. They also showed that the degree of ER expression is proportional to endocrine sensitivity in HGSOC [31]. The similarity in limited treatment options and poor prognosis between ERα-negative HGSOC and TNBC led us to determine whether blockade of the CDK2/EZH2 axis converts ERα negative to positive and renders HGSOC targetable by hormone therapy.

Methods

Cell line cultures and reagents

All cell lines were obtained from the ATCC (Manassas, VA) and independently validated by STR DNA fingerprinting at The University of Texas MD Anderson Cancer Center, and tests for mycoplasma contamination were negative. TNBC cell lines BT-549 and MDA-MB-231, and ovarian cancer cell lines SKOV3, OVCA433, CAOV3, DOV13, A2780, OVCA420 were cultured in DMEM (Hyclone) supplemented with 10% FBS and penicillin and streptomycin. All cells were incubated at 37ºC with 5% CO₂. EPZ-6438, 3-deazaneplanocin A hydrochloride, dinaciclib, and GSK343 were obtained from Selleckchem (Houston, TX). Antibodies against EZH2 and phospho-CDK2 were obtained from Cell Signaling (Danvers, MA). Antibodies against ERα (ab108398) was purchased from Abcam (Cambridge, MA).

Bioinformatics analysis

For genetic alternation analysis of EZH2 gene, the cbioportal (http://www.cbioportal.org) was queried for ‘EZH2’, and the frequency of alteration of EZH2 in related tumors is shown as a bar graph. To identify TCGA ovarian tumors with EZH2 gene overexpression (OE), we used data from normal ovarian tissue in TCGA and Yoshihara [32], and set up a cutoff at two standard deviation (SD) above median normal expression of EZH2. RNAseq dataset of Yoshihara includes a total 43 ovarian cancer samples and 10 normal ovarian tissues; TCGA includes a total 586 ovarian cancer samples and 8 normal tissues.

RNA extraction and real-time PCR

The HGSOC cells were treated with CDK2 inhibitor (CDK2i) or EZH2 inhibitor (EZH2i) for 3 days, and total RNA was extracted from the treated cell pellets with Trizol. The SuperScript III First Strand Synthesis System was used for RT-PCR (Invitrogen). Real-time PCR was performed using SYBR Green-base PCR (Qiagen). The primer sequences for the transcripts analyzed are as follows (5’ to 3’) (Table 2).

Table 2.

The primer sequences

ESR1	FORWARD	CTCTCCCACATCAGGCACA
	REVERSE	CTTTGGTCCGTCTCCTCCA
GAPDH	FORWARD	GAAGGTGAAGGTCGGAGTC
	REVERSE	GAAGATGGTGATGGGATTTTC
PGR	FORWARD	CTTAATCAACTAGGCGAGAG
	REVERSE	AAGCTCATCCAAGAATACTG
TFF1	FORWARD	GTCCCCTGGTGCTTCTATCC
	REVERSE	GAGTAGTCAAAGTCAGAGCAGTCAATCT
c-MYC	FORWARD	GAAGATGGTGATGGGATTTTC
	REVERSE	GAAGATGGTGATGGGATTTTC

Immunoblotting analysis

Whole-cell extracts were lysed in radioimmunoprecipitation buffer (RIPA, 10 mM Tris-HCl [pH 8.6], 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 140 mM NaCl, and 1× protease inhibitor; Complete Mini, Roche), which was freshly added before lysis. Total protein extracts (20-30 µg) from each sample were separated on 10% polyacrylamide gel and transferred onto a polyvinylidene difluoride membrane (Life Technologies). Antibodies against ERα, H2K27Me3, and phospho-CDK2 were incubated with the membranes overnight at 4°C after blocking in 5% non-fat milk for one hour followed by secondary anti-mouse-HRP or anti-rabbit-HRP (Cell Signaling) at room temperature for 1 hour. The membranes were washed with PBS-Triton buffer and imaged using ECL reagents (Bio-Rad Laboratories) in ImageQuant LAS 4010 (GE Healthcare).

MTT assay and calculation of combination index (CI) of drug synergy

For the MTT assay, SKOV3 (3,000 cells per well), OVCA433 (5,000 cells per well), and CAOV3 (4,000 cells per well) were seeded in a 96-well plate and cultured overnight before treatment. Cells were then treated with inhibitors either alone or in combination at the indicated concentrations for 3 days. MTT reagent was added into each well at a final concentration of 0.5 µg/ml and incubated for 30 min before formazan was extracted with DMSO. Cell survival was calculated using an absorbance at 590 nm and normalized to that of the untreated wells. The half maximal inhibitory concentration (IC₅₀) was calculated using MTT data and GraphPad Prism 8.0. Chou-Talalay combination index (CI) were calculated by using Compusyn software (www.combosyn.com) with cell survival data from the MTT assay. Drug combinations were designed using the constant ratio of 2 inhibitors according to the predictions by the Compusyn program [33,34]. Criteria for CI: CI ≤ 0.3, strong synergy; 0.3 < CI ≤ 0.85, synergy.

Tumor-sphere formation assay in 3D matrigel

Three-dimensional (3D) culture was performed as previously described [6,30]. Briefly, prechilled culture 24-well plates were coated with a thin layer of cold growth factor-reduced Engelbreth-Holm-Swarm extracellular matrix extract (EHS; Matrigel, BD Biosciences, San Jose, CA). The plates were then incubated for 30 min at 37°C to allow EHS gelation. MDA-MB-231 (5,000) cells were suspended in the culture medium containing 2.5% Matrigel, seeded on Matrigel-coated 24-well plate, and incubated at 37°C for 3 days followed by treatment with CDK2 and/or EZH2 inhibitor for 1 week. A representative area of MDA-MB-231 tumorsphere cultures was imaged. The number of spheres formed per field was quantified (n = 10).

Results

Previous studies showed that cyclin E/CDK2 phosphorylates EZH2 at T416 and activates EZH2 [6,35]. Phosphorylation of EZH2 at T435 and T487 by CDK1 and CDK2 epigenetically silences target genes in breast cancers [5,6,36,37]. Inhibition of CDK2 or EZH2 has been found to reverse epigenetic silencing and induces downstream gene such as ERα re-expression in TNBC [30]. In the present study, we determined whether inhibition of the CDK2/EZH2 signaling pathway by specific inhibitors induces ERα expression in ovarian cancer cells and tested the efficacy of combination treatment of CDK2i or EZH2i and tamoxifen (TAM) in vitro. We also explored the putative feedback mechanism of CDK2 activity triggered by EZH2i in TNBC and HGSOC cells.

EZH2 and ERα expression in ovarian cancer cells

The classical ERα has been shown to originate from the ovarian surface epithelium although only a small number of ovarian cancers respond to anti-estrogen therapy [11,16,31]. A previous study showed that phosphorylation of EZH2 by CDK2 initiates the epigenetic silencing of downstream target genes, leading to the repression of ERα expression in vitro and in vivo [30]. To expand our study to ovarian cancer, we first sorted out the EZH2 expression and mutation in various cancers according to TCGA and Yoshihara database [32]. At first, the cBioPortal platform (http://www.cbioportal.org) was used to query for gene alternations of ‘EZH2’. Bar graph displayed the frequency of alteration of EZH2 in related tumors. EZH2 alternations including mutations, amplifications, deletions, fusions, and multiple mutations were found in many types of cancers (Figure 1A). The rate of frequency of EZH2 amplification (> 7%) in ovarian cancer was the highest among a panel of cancer types examined (Figure 1A, red arrows). To determine the expression levels of EZH2 in ovarian cancers, we analyzed two sets of RNAseq databases of TCGA and Yoshihara [32], and the results revealed EZH2 is overexpressed in 95% and 86% of patients with ovarian cancer (Table 1). The lists of patient samples of ovarian cancers with EZH2 overexpression are provided in Tables S1 and S2. Because activation of the EZH2-containing PRC2 complex serves as an initiator for epigenetic silencing of ERα expression, we next examined the expression of ERα in HGSOC cells. We at first performed Western blot analysis to screen the available ovarian cancer cell lines and selected those with silenced ERα gene expression. As shown in Figure 1B, compared with luminal breast cancer cell line T47D and TNBC cell line MDA-MB-231, the ERα protein levels in five of the seven HGSOC cell lines (SKOV3, A2780, DOV13, 2774C10 and CAOV3) examined were low or undetectable. Thus, we used those five cell lines to further examine ERα expression in response to the treatment of CDK2i and EZH2i.

Figure 1.

Expression of EZH2 in ovarian cancers. A. Alternations of EZH2 in human cancers. The alternations of ‘EZH2’ was queried using bioinformatics platform cbioportal (http://www.cbioportal.org). Bar graph showing the frequency of alteration of EZH2 in related tumors. Breast and ovarian cancers are highlighted with red arrows. B. The ERα proteins in the whole cell lysates of the ovarian cancer cell lines were immunoblotted with specific antibodies as indicated. T47D cell line was used as ERα positive and MDA-MB-231 cell line as ERα negative control.

Table 1.

Overexpression (OE) of EZH2 in ovarian cancer

Dataset	Normal	Tumor	Media N	Media T	OE Normal	OE Tumor
Yoshihara [32]	10	43	0.912	3.496	0	86%
TCGA	8	586	-1.044	1.876	0	95%

TCGA and Yoshihara RNAseq Dataset; OE, overexpression; N, normal; T, tumor.

CDK2i reactivates ESR1 gene expression in ERα-negative ovarian cancer cell lines

The ESR1 gene encoding ERα is one of the well-identified EZH2 target genes [30]. To determine whether blockade of the CDK2/EZH2 axis by CDK2i reactivates ESR1 expression and thus renders cancer cells sensitive to TAM, we treated several ovarian cancer cell lines with CDK2i and performed quantitative real-time PCR assays to determine the mRNA levels of ERα. Because CDK2 is an upstream kinase that regulates EZH2 activity via site-specific phosphorylation at T416 [6,35,37], we treated ERα-low or -negative ovarian cancer cells with either CDK2i or EZH2i for 3 days followed by Western blot analysis of whole cell lysates. Among the ovarian cancer cell lines examined, only SKOV3, OVCA433 and CAOV3 re-expressed ERα in response to treatment with CDK2i dinaciclib (DINA; Figure 2A). Next, we examined the ERα and H3K27Me3 levels in CDK2i-sensitive HGSOC cells and showed that ERα was re-expressed with concomitant decrease in H3K27Me3 levels after treatment with CDK2i and EZH2i (Figure 2B). To determine whether CDK2i-induced ERα occurs at the transcriptional or translational level and whether the induced ERα is functional, we compared the mRNA levels of ESR1 and the well-defined ERα downstream target genes including C-MYC, TFF1, and PGR, in CDK2i-treated HGSOC cells by quantitative real-time PCR. The expression levels of ERα and its target genes were significantly increased after 3-day treatment with CDK2i (50 nM) and high concentration of EZH2i (GSK343, 10 µM) (Figuer 2C-F, P < 0.05 and 0.001). These results suggested that both CDK2 and EZH2 inhibitors block the CDK2/EZH2 axis to release the epigenetic repression of ESR1, which in turn upregulates the expression of its downstream target genes PGR, TFF1, and C-MYC. Blockade of the CDK2/EZH2 axis leads to reactivation of functional ERα, as indicated by upregulated ERα target genes PGR, TFF1, and c-Myc, and in turn renders ERα-negative HGSOC targetable by hormone therapy.

Figure 2.

CDK2 inhibitor induces expression of ERα in ERα-negative ovarian cancer cell lines. A. Selected ER-negative ovarian cancer cell lines were treated with 25 nM dinaciclib (DINA) for 3 days. ERα, EZH2, and H3K27Me3 protein levels in whole cell lysates were detected by Western blot analysis with specific antibodies against ERα after treatment with CDK2i (DINA). B. Western blot analysis of ERα, EZH2, and H3K27Me3 in HGSOC cells after treatment with 25 nM (DINA) or 10 µM GSK343 (GSK) for 72 hours. C-F. qPCR analyses of mRNA levels of ERα and its target genes in HGSOC cell lines after CDK2i or EZH2i treatment. Bar graphs of the mRNA levels of ERα and its downstream target genes PgR, TFF1, and c-Myc are shown as mean ± SD (n = 3). *P < 0.05, ***P < 0.01, and ****P < 0.001, Student’s t-test. ns, no significance.

TAM in combination with CDK2i or EZH2i induces synergistic inhibitory effects on HGSOC

TAM has been the gold standard of hormone therapy for ER-positive breast cancer for almost 40 years and is also used to treat ovarian cancer. Previously, we reported that inhibition of EZH2 phosphorylation by CDK2i induces ERα re-expression in TNBC and that the combination of CDK2i or EZH2i and TAM synergistically reduces TNBC proliferation in vitro and in vivo[30]. Accordingly, we asked whether this strategy also applies to HGSOC with low or negative ERα expression. To this end, HGSOC cells were treated with either CDK2i (dinaciclib, DINA) or EZH2i (GSK343, GSK) combined with TAM at different concentration ratios. The drug interactions between CDK2i or EZH2i and TAM were determined by MTT assay. Similar to our observations in TNBC, the combination of either CDK2i or EZH2i with TAM also synergistically inhibited proliferation of the ovarian cancer cell lines tested in vitro (Figure 3A-C; CI values < 0.8). Blockade of the EZH2/CDK2 axis with CDK2i or EZH2i reactivated ERα expression and sensitized HGSOC to hormone therapy, suggesting that this combination is effective in both TNBC and HGSOC.

Figure 3.

Synergistic effects of CDK2i (DINA) and EZH2i (GSK343) combined with tamoxifen in HGSOC. A-C. Ovarian cancer cells were pretreated with either CDK2i or EZH2i for 3 days followed by combination with tamoxifen or alone for 3 days. Cell viability was determined by MTT assays. The combination index (CI) values of CDK2i or EZH2i and tamoxifen were calculated using Compusyn software (www.combosyn.com).

Feedback regulation of the CDK2-EZH2 signaling axis in breast and ovarian cancers

CDK2 plays a crucial role in both cell cycle progression and apoptotic response, and cyclin E/CDK2 phosphorylates EZH2 at T416 [6,35,37]. We previously tested a CDK2 inhibitor, dinaciclib, and two EZH2 inhibitors, EPZ-6438 and GSK343, all of which have been evaluated in clinical trials and target different steps of the CDK2/EZH2 signaling axis. Interestingly, in previous study we found that the effectiveness and potency of these EZH2i and CDK2i inhibitors were different [30]. CDK2i inhibits EZH2 phosphorylation upstream of the CDK2-EZH2 axis and indirectly represses EZH2 activity [30]. In contrast, EZH2i directly binds to the SET domain, resulting in a decreased histone methyltransferase of EZH2 complex at the promoter of the target genes [38]. Accordingly, EZH2i should be more effective since it directly blocks the terminal regulator of the CDK2-EZH2 axis. To this end, ERα-negative cancer cells were treated with 5 μM EPZ-6438, an already FDA-approved drug, GSK343, or CDK2i dinaciclib (50 nM) for 3 days. Immunoblot analyses of the drug- and vehicle-treated cancer cells revealed that CDK2i was more effective than EZH2i in inducing ERα expression (Figure 4A, 50 nM dinacilib vs. 5 μM GSK343). Notably, as shown in Figure 4A, treatment with 5 μM EPZ-6438 did not activate ERα expression, suggesting that the efficacy of EPZ-6438 is not as potent as GSK343 in vitro. Additionally, treatment with 100 nM SNS032, a CDK2i that is also in clinical trial, significantly reactivated ERα expression in TNBC cells (Figure 4B). These observations are consistent with those reported in the previous studies [30]. Compared with GSK343, EPZ-6438 was less effective in ERα induction in vitro (Figure 4A, lane 2 vs lane 3). However, treatment with EPZ-6438 at higher concentrations (≥ 10 μM) also induced ERα expression in TNBC cells (Figure 4C). Moreover, using tumorsphere formation assay in 3D Matrigel culture, we examined the sensitivity of TNBC cells to the inhibitors. The results revealed that both EZH2i and CDK2i significantly reduced tumorsphere formation with CDK2i showing stronger inhibitory effect than did EZH2i in this functional assay (Figure 4D and 4E). To validate that CDK2i-induced ERα is dependent on EZH2, we knocked down EZH2 in MDA-MB-231 cells using shRNA against EZH2 (sh-EZH2). Indeed, compared with control sh-RNA against luciferase (sh-Luc) cells, ERα expression was significantly increased after depletion of EZH2 (Figure 5A, lanes 1 and 4). Intriguingly, CDK2i-induced ERα expression in sh-EZH2 cells was not significantly increased compared with that in sh-Luc control cells even with 100 nM of DINA (Figure 5A, lane 3 vs. 6; lane 4 vs. 6), suggesting that CDK2i-induced ERα expression is at least partially dependent on EZH2, and as a result, depletion of EZH2 attenuates CDK2-mediated reactivation of ERα expression. Next, we determined whether inhibition of EZH2 activity regulates CDK2 activity by a feedback mechanism in the signaling pathway. To this end, we tested the effects of three EZH2i, GSK343, 3-deazaneplanocin A hydrochloride (DZNEP), and EPZ-6438 on CDK2 activity in TNBC and HGSOC cell lines. MDA-MB-231 TNBC cells were treated with EZH2i for 72 hours, the ERα, EZH2 and pCDK2 protein levels in these cells were immunoblotted with specific antibody against activated CDK2, pCDK2(T160). EZH2i treatment significantly enhanced pCDK2 along with ERα expression in MDA-MB-231 cells compared with that in vehicle control and sh-EZH2 cells (Figure 5B, lanes 1 and 2). Notably, deletion of EZH2 significantly reduced pCDK2 (Figure 5B, lane 1 vs. 2), suggesting that EZH2 stabilizes pCDK2 in the CDK2/EZH2 complex. Similar results were observed in the HGSOC cell line SKOV3 after treatment with EZH2i (Figure 5C). These findings indicated that inhibition of EZH2 activity in the CDK2-EZH2 axis by the specific EZH2i led to a feedback CDK2 activation as indicated by increased pCDK2(T160), and this in turn compromises the inhibitory effects of EZH2i in the breast and ovarian cancer cells (Figure 5B and 5C). These findings revealed a new potential connection: incomplete inhibition of EZH2 activity could upregulate CDK2 phosphorylation via a feedback loop, which in turn increases EZH2 phosphorylation and thus undermines the effects of EZH2i (Figure 5D). Taken together, our findings suggested that the compromised inhibitory effects of EZH2i in cancer cells is partly attributed to a feedback loop of the CDK2-EZH2 signaling axis. Therefore, to achieve complete inhibition of EZH2 activity, a persistent higher dose of EZH2i is required.

Figure 4.

A comparison of the inhibitory effects between EZH2i and CDK2i on ERα induction and tumor sphere formation. TNBC cell line MDA-MB-231 cells were treated with 50 nM dinaciclib (DINA) and 5 μM EPZ-6438 (EPZ) and GSK343 (GSK) for 72 hours. ERα protein levels in the inhibitor-treated cell lysates were detected by an immunoblotting analysis with specific antibodies against ERα. Actin was used as the loading control. (B) MDA-MB-231 cells were treated with 5 μM GSK343 (GSK) and 100 nM CDK2i SNS032 (SNS) as described in (A). (C). EZH2i-induced ERα expression in TNBCs was dose-dependent. TNBC cell line SUM-149 cells were treated with EPZ-6438 (EPZ) at different concentration for 72 hours. ERα protein levels were immunoblotted as described in (A), and T47D cell lysate was used for ERα-positive control. (D, E) The inhibitory effects of CDK2i and EZH2i on tumor sphere formation. MDA-MB-231 (5,000) cells in 2.5% Matrigel were seeded in Matrigel-coated 24-well plate and incubated at 37°C for 3 days followed by treatment with CDK2 and EZH2 inhibitors for 1 week. Representative images (D) and quantitation (E) of the spheres are shown. ***, P < 0.001; ****, P < 0.0001, Student’s t-test. Scale bar, 200 μm.

Figure 5.

EZH2i enhances CDK2 activity by feedback loop in breast and ovarian cancer cells. (A) MDA-MB-231 control cells (sh-Luc) and EZH2-Knockdown cells (sh-EZH2) were treated with different doses of dinaciclib (DINA) for 72 hours. ERα and EZH2 proteins in the whole cell lysates were separated on 10% SDS- polyacrylamide gel and immunoblotted with indicated specific antibodies. (B) MDA-MB-231 control and sh-EZH2 cells were treated with 5 µM GSK343 (GSK), 5 µM 3-deazaneplanocin A hydrochloride (DZNEP) and 20 μM EPZ-6438 (EPZ) for 3 days. Immunoblotting analysis of the treated cells were performed with indicated specific antibodies. (C) SKOV3 cells were treated with different doses of DINA or GSK and analyzed as described in (A). (D) A model illustrating the CDK2-EZH2 axis feedback loop. CDK2-mediated phosphorylation of EZH2 at T416 activates EZH2 to silence target genes, and higher EZH2 activity triggers a negative feedback inhibition of CDK2 activation to maintain homeostasis of the signaling axis. CDK2i and EZH2i block the CDK2-EZH2 signaling axis and thus repress the EZH2 target genes. Incomplete inhibition of EZH2 results in feedback activation of CDK2, compromising EZH2i efficacy.

Discussion

Our data demonstrated that similar to TNBC, blockade of the CDK2/EZH2 axis by CDK2i or EZH2i reactivated ERα expression in HGSOCs, rendering the cancer cells sensitive to hormonal therapy. In addition, partial inhibition of EZH2 activity by EZH2i compromised the inhibitory effects of EZH2i on both of TNBC and HGSOC via a feedback mechanism.

EZH2 activity is regulated by phosphorylation/dephosphorylation, leading to alternation in trimethylation of H3K27 and subsequently epigenetic silencing or de-repressing of the target genes [2,3,5,6,36]. Accumulating evidence demonstrates that EZH2 is aberrantly overexpressed in different types of solid tumors and is associated with metastasis and progression [39,40]. We previously reported that phospho-EZH2 binds to the promoter of ERα to repress the expression of ERα in TNBC and that blocking phosphorylation of EZH2 at T416 with CDK2i reactivates ERα expression, thereby rendering cancer cells targetable by tamoxifen in TNBC [30]. In the current study, we observed that similar to TNBC, blockade of the CDK2/EZH2 axis in ERα-negative HGSOC cell lines with specific inhibitors led to the induction of ERα expression, rendering HGSOC targetable by hormone therapy. It has been reported that about 81% of HGSOC express ERα [11]. The rationale for endocrine maintenance therapy for HGSOC, which is partially driven by the estrogen pathway, is based on high ERα expression. The expression levels of ERα are correlated with endocrine sensitivity in HGSOC [26]. Hormone therapy for patients with HGSOC results in a significant benefit, particularly in those with relapsed, ER positive diseases [31]. The development of an approach to convert ERα-negative to positive in HGSOC provides an alternative choice of combination treatment for those patients, and thus has important clinical significance. Our observations in vitro revealed that ERα and its downstream target genes were re-activated in response to CDK2/EZH2 axis blockade with specific inhibitors in ovarian cancer cell lines SKOV3, OVCA433, and CAOV3. However, the CDK2i-treated ovarian cancer cell lines DOV13 and A2780 failed to reactivate ERα expression (Figure 2A). The underlying mechanism of this failure remains to be further investigated.

Our findings indicated that the magnitude of the CDK2i and EZH2i-induced inhibitory effects on tumor growth in vitro and in vivo was significantly different (e.g., 25-100 nM dinaciclib vs. 2.5 µM EPZ-6438 or 5 µM GSK343) (Figure 4). CDK2i was much more efficient than EZH2i at the doses tested (Figure 4) [19,30,38]. It has been reported that CDKs have a critical role in cell cycle progression and that they govern a large number of targets for novel anticancer drug development [41,42]. A previous study demonstrated that CDK2i represses the phosphorylation of EZH2 and inhibits tumor cell growth [43]. To identify the underlying mechanism of the difference in efficacy between CDK2i and EZH2i in the cancer cells, we examined the CDK2 activity in the EZH2i-treated TNBC and HGSOC cells. An increase of phospho-CDK2 at T160 (pCDK2) suggested that inhibition of EZH2 activity triggered an upregulation of CDK2 phosphorylation, which in turn increased EZH2 activation (Figure 5D). This homeostatic mechanism maintains the activation of EZH2 in the loop and impairs the efficiency of EZH2i. The phenomenon also implies that full blockade of the CDK2-EZH2 axis by targeting EZH2 requires a high dose of EZH2i (Figure 5D). Indeed, previous in vivo studies demonstrated more than 300 mg/kg of EPZ-6438 and 2-8 mg/kg of dinaciclib had the same potency in inhibiting tumor growth [30,38,44]. The putative homeostatic feedback loop might at least partly explain the insensitivity of tumors to EZH2i compared with CDK2i. However, the underlying mechanism of feedback regulation on the CDK2-EZH2 signaling pathway remains to be further investigated. According to our results and those from previous reports, CDK2i not only directly inhibits cell cycle but also inactivates the oncogenic EZH2 activity, leading to the release of epigenetic repression in target genes [6,18,30]. Moreover, CDK2i more potently blocked the CDK2/EZH2 axis than did EZH2i in inhibiting phosphorylation of EZH2 and in exerting a cytotoxic effect on tumor cells without feedback up-regulation of EZH2 activity (Figure 5D).

In summary, this study provides additional evidence that ERα expression can be induced in HGSOC and TNBC treated with CDK2i and EZH2i. The combination therapy with CDK2i and tamoxifen demonstrated a synergistic inhibitory effect on the HGSOC cell proliferation and may be an alternative treatment approach for HGSOC patients with ERα-negative disease. Further in vivo studies are needed before it can be evaluated in clinical trials.

Disclosure of conflict of interest

None.