JMJD2A contributes to breast cancer progression through transcriptional repression of the tumor suppressor ARHI

Permission to use human tissue sections for research purposes was obtained and approved by an institutional review board at Huashan Hospital, Shanghai, China. All patients provided their full consent to participate in our study. For animal research, the protocol was approved by the Ethics Committee from Shanghai Medical College, Fudan University, China. And all efforts were made to minimize suffering.

Breast cancer cell lines MCF-7, T47D, SUM1315 and MDA-MB-231 were purchased from ATCC (Manassas, VA, USA) and cultured in DMEM (Hyclone, Logan, Utah, USA) supplemented with 10% FBS (Gibco, Los Angeles, CA, USA). Primary antibodies against JMJD2A (C37E5), E2F1 (3742), HDAC1 (5356) and HDAC3 (3949) were purchased from Cell Signaling Technology (Boston, MA, USA). Anti-estrogen receptor alpha (anti-ERα) (ab2746), anti-progesterone receptor (PR) (ab32085), anti-human epidermal growth factor receptor-2 (anti-HER2) (ab134182), anti-E2F4 (ab150360) and anti-ARHI (ab107051) primary antibodies were all obtained from Abcam (Cambridge, UK). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibody (sc-25778) was purchased from Santa Cruz (CA, USA). Lipofectamine 2000 was purchased from Invitrogen (Carlsbad, CA, USA). Trichostatin A (TSA) was purchased from Sigma Co. (St Louis, MO, USA). The chromatin immunoprecipitation (ChIP) kit was purchased from Upstate (a part of Millipore, Billerica, MA, USA). For the co-immunoprecipitation (Co-IP) assay, the protein G agarose beads and NP-40 lysis buffer were purchased from Beyotime Institute of Biotechnology (Nantong, China). Matrigel was purchased from BD Biosciences (San Jose, CA, USA). Cell counting kit-8 (cck-8) was purchased from Dojindo (Japan). And dual-luciferase reporter gene assay were from Promega (Madison, WI, USA).

JMJD2A expression plasmid was subscribed by Dr Ralf Janknecht from the Department of Biochemistry and Molecular Biology, Mayo Clinic College of Medicine and Dr Hsing-Jien Kung, Deputy Director and Director of Basic Science, UC Davis Cancer Center. JMJD2A (NM_014663.2) was PCR-amplified from the subscribed plasmids and cloned into pcDNA3.1 vetors (pcDNA3.1-JMJD2A). ARHI (AY890085.1) was amplified on mRNA from MCF-7 cells and cloned into pcDNA3.1 vetors (pcDNA3.1-ARHI) as per the molecular cloning manual. siRNA specific to JMJD2A was chemically synthesized by Qiagen Technology Co. Ltd (Valencia, CA, USA). siRNA was diluted to 20 μmol/L with RNase-free water. siRNA duplexes were synthesized as follows: sense: 5′-GAGUUAUCAACUCAAGAUA-3′, antisense:5′-UAUCUUGAGUUGAUAACUC-3′. The other two JMJD2A siRNAs were synthesized by GenePharma (Shanghai, China) with the sequences described previously [15]. siRNAs specific to E2F1 and E2F4 were also synthesized by GenePharma (Shanghai, China) with sequences verified and described previously [36, 37]. Scramble siRNA was used as negative control (NC) and sequences were as follows: sense: UUCUCCGAACGUGUCACGU, antisense: ACGUGACACGUUCGGAGAATT. Plasmids and siRNAs were transfected with Lipofectamine 2000 in accordance with the manufacturer’s instructions.

All the 155 cases of breast cancer and 30 cases of non-neoplastic tissues were retrieved from the Department of Pathology, Huashan Hospital, Shanghai, China, within 2012. All cases were diagnosed by two experienced pathologists without discrepancy. None of the patients had received chemotherapy or radiation therapy previously. The paraffin-embedded tissues were first stained with hematoxylin and eosin (HE) for histological examination. Subsequently, sections were subjected to antigen retrieval by heating the slides in a microwave at 100°C for 10 minutes in 0.1-M citric acid buffer (PH 6.0), and then incubated with corresponding antibodies at 4°C overnight. After secondary antibody incubation at room temperature for 1 h, the slides were developed in 0.05% diaminobenzidine containing 0.01% hydrogen peroxidase. For negative controls, specific antibodies were replaced with normal goat serum by co-incubation at 4°C overnight preceding the immunohistochemiscal staining procedure.

At 48 h after transfection, cells in different treatment groups were harvested. The procedure was used as described previously [38, 39]. Antibodies against JMJD2A, ARHI, GAPDH, E2F1, and E2F4 were used to incubate target proteins overnight at 4°C. After the overnight incubation with the primary antibodies, membranes were washed and incubated with horseradish peroxidase (HRP)-labeled secondary antibody in Tris-buffered saline with Tween 20 (TBST) for 1 h. Immunoreactivity was detected with enhanced chemoluminescent autoradiography (ECL kit, Amersham, Pittsburgh, PA, USA), according to the manufacturer’s instruction. GAPDH was used as a loading control.

Total RNA of each group were extracted respectively using Trizol solution (Invitrogen) at 24 h after transfection. Fast-strand cDNA was generated from 1 μg of total RNA using the PrimeScript RT Master Mix Perfect Real Time (TaKaRa, Shiga, Japan). Real-time qPCR was performed in an ABI PRISM 7500 Real-Time System. A 10-fold dilution of each cDNA was amplified in a 50 μl volume, using the SYBR Premix Ex TaqTM Perfect Real Time (TaKaRa, Shiga, Japan). The primers used were as follows: JMJD2A: forward 5′-ATCCCAGTGCTAGGATAATGACC-3′, reverse 5′-ACTCTTTTGGAGGAACCCTTG-3′; ARHI: forward 5′-GATTACCGCGTCGTGGTAGTC-3′, reverse 5′-TCAATGGTCGGCAGGTACTCA-3′; GAPDH: forward, 5′-TGACGCTGGGGCTGGCATTG-3′, reverse 5′-GCTCTTGCTGGGGCTGGTGG-3′.

Primers were synthesized by Shanghai Daweike Biotechnology Co. Ltd (Shanghai, China). PCR cycle conditions were 95°C for 30 s, and 40 cycles of 95°C for 5 s and 60°C for 34 s. The amplification specificity was evaluated with melting curve analysis. Relative mRNA was determined by using the formula 2^-ΔCT (CT, cycle threshold) where, as described previously [40]:

HEK293T cells were seeded at a density of 2 × 10⁵/well in 24-well plates and co-transfected with indicated amounts of ARHI/luciferase reporter together with the full-length JMJD2A construct (JMJD2A-FL), the construct of JMJD2A with the substitution of histidine 188 by alanine (JMJD2A-H188A), or another construct of JMJD2A with the deletion of the tudor domains (JMJD2A-M867) [21, 41]. Renilla luciferase plasmid (pRL) was co-transfected as internal control. Thirty-six hours after transfection, cells were lysed using passive lysis buffer and assayed immediately for reporter and control gene activities with the dual-luciferase reporter gene assay using a Lumat LB 9507 luminometer (EG & G Berthold, Bad Wildbad, Germany). To determine the effect of HDACs, cells were treated with or without 100 nM trichostatin A (TSA) at 12 h after transfection. Each experiment was performed in triplicate, and the data represent three independent experiments after normalization to renilla activity.

Briefly, lysates were incubated with 4 μg of anti-JMJD2A antibody or normal rabbit IgG as a negative control. PCR amplification was performed using 1:100 dilution of input, an IgG negatively immunoprecipitated DNA and specific JMJD2A immunoprecipitated DNA. In general, samples were heated at 95°C for 3 minutes, followed by 31 cycles of 95°C for 30 s, 54°C for 30 s and 72°C for 20 s. After cycling, samples were incubated at 72°C for 10 minutes to permit completion of primers extension. Then PCR products were electrophoresed on a 2% agarose gel with ethidium bromide. The two pairs of ARHI primers (A1, A2) were as follows: A1 (−181 to 91) forward: 5′-TCGATTGTTGTAGATGCCAAG-3′, reverse: 5′-AGACTTACCTTTCTCGGAGGC-3′; A2 (−524 to −341), forward: 5′-TTTACCGGTCTTGCCACTAATG-3′, reverse: 5′-TCCAAAAGCAGTTTAATGCAGG-3′. GAPDH was used as a loading control (154 bp).

MDA-MB-231 cell lysates were obtained using NP-40 lysis buffer. Specific antibodies were used for immunoprecipitation as well as 20 μl of protein G agarose beads. The beads were washed in lysis buffer and boiled in 30 μl of SDS loading buffer; the entire sample was loaded on a SDS-polyacrylamide gel and processed by western blot. The membranes were immunoblotted with corresponding primary antibodies. Rabbit normal IgG was used as negative control.

Cells were seeded on 96-well plates at an initial density of 4 × 10³/well. At each monitored time point, cells of each well were stained with 10 μl CCK-8 (Dojindo, Japan) for 4 h at 37°C. Absorbance was measured using a synergy 2 multi-mode microplate reader (Bio Tek Instruments, Winooski, VT, USA) at 450 nm. All experiments were carried out in triplicate.

A wound-healing assay and Boyden chamber assay were performed as described previously [42]. Cells were plated on 6-well plates to form a confluent monolayer. Wounds made with sterile pipette tips were observed per 12 h. A migration assay was carried out using Boyden chambers (tissue culture-treated, 6.5-mm diameter, 8-μm pores, Transwell, Costar, Cambridge, MA, USA) containing polycarbonate membrane. For the invasion assay, 50 μl matrigel (BD Biosciences, San Jose, CA, USA) was used to mimic basement membrane. Briefly, 100 μl of 1 × 10⁶ cells in serum-free medium was added to the upper chamber and 600 μl of appropriate medium with 10% FBS was added to the lower chamber. Cells were incubated for 12 h. Migration cells on the under-surface of the membrane were fixed and stained with Giemsa for 10 minutes at room temperature. Photographs of five random regions were taken and the number of cells was counted to calculate the average number of migrated cells per plate.

Five-week-old female athymic nude mice (BALB/c^nu/nu) were used for the experiment. Cells stably expressing JMJD2A or both JMJD2A and ARHI were constructed as previously described [43]. Cells (1 × 10⁶) were injected subcutaneously into the mammary fat pad of the mice. Mice were randomized (n = 7 per group) and assigned to specific groups. Tumor diameters were measured twice a week and tumor volumes (TV) were calculated using the formula as described [44]:

On day 27 after tumor cell injection, the mice were sacrificed and the excised tumors were measured and weighed. Lung and liver metastatic nodules were also calculated.

All values are expressed as mean ± SD. The Student’s t-test and Spearman correlation analysis were used to evaluate the experimental data. P <0.05 was considered statistically significant.

To investigate the role of JMJD2A in breast cancer, we examined JMJD2A expression in 155 human breast cancer tissues and 30 non-neoplastic breast tissues. Histological examination was first carried out on non-neoplastic breast tissues and breast cancer tissues, respectively (Figure 1A and E). Then all the tissues were analyzed by immunohistochemistry using anti-JMJD2A antibody and graded based on staining intensity (Figure 1B,C,F,G,H and I). We found that JMJD2A protein mainly localized in the nuclei and the positive rate in breast cancer tissues (93%, 144/155, Figure 1D (b)) was significantly higher than that in non-neoplastic tissues (3%, 1/30, Figure 1D (a)) (P <0.001). Among 30 non-neoplastic tissues, only one sample (3%) presented as weakly positive for JMJD2A, whereas the other 29 samples (97%) were all JMJD2A-negative. In contrast, 11 (7%) out of 155 cases of breast cancer tissues were negative, 56 (36%) were weakly positive, 53 (34%) were moderately positive and 35 (23%) were strongly positive. These data strongly support that JMJD2A expression is significantly increased in breast cancer tissues. Additionally, we randomly selected two primary breast cancer tissues and performed western blot analysis. This revealed that breast cancer tissues displayed higher JMJD2A expression level than that in paired tumor-adjacent non-cancerous tissues in the two samples (Figure 2A). These results suggest that JMJD2A is highly expressed in breast cancer.

To determine whether the expression level of JMJD2A was associated with progression of breast cancer, we initially examined JMJD2A expression in four breast cancer cell lines, including two weakly metastatic cell lines (MCF-7 and T47D) and two highly metastatic cell lines (MDA-MB-231 and SUM1315). Compared with the weakly metastatic cell lines, highly metastatic cell lines had a higher expression level of JMJD2A (Figure 2B). We then carried out Spearman correlation analysis to determine the relationship between JMJD2A expression and clinical parameters (Table 1). Statistical analysis revealed that there was a significantly positive correlation of JMJD2A expression with tumor, node, metastasis (TNM) stage (P = 0.004, r = 0.227), and tumor size (P = 0.026, r = 0.179), but JMJD2A expression did not significantly correlate with lymph node metastasis (P = 0.102), or with age (P = 0.092). These results reveal that JMJD2A positively relates to tumor progression.

We subsequently analyzed the relationship of JMJD2A expression with tumor regulators including estrogen receptor (ER)α, progesterone receptor (PR), HER2 and ARHI. Our results showed that ERα and PR were stained mainly in the nuclei and HER2 was localized in the cell membrane (data not shown), whereas ARHI was mainly in the cell membrane and partially in the cytoplasm (Figure 3B and D). Spearman correlation analysis showed that expression of JMJD2A was inversely correlated with expression of ARHI (P = 0.016, r = −0.194, Table 2). However we did not observe a significant association between JMJD2A immunoreactivity and other tested clinicopathological parameters, including ERα (P = 0.778), PR (P = 0.147) and HER2 (P = 0.742). This result implies that JMJD2A may be negatively linked with the tumor suppressor ARHI.

To verify the negative relationship between JMJD2A and ARHI, we firstly performed western blot analysis using human breast cancer tissues and paired tumor-adjacent non-cancerous tissues. As expected, ARHI was downregulated in JMJD2A-overexpressed cancer tissues and upregulated in tumor-adjacent tissues (Additional file 1: Figure S1). We then carried out IHC staining on serial sections with both anti-JMJD2A and ARHI antibodies in non-neoplastic breast tissues and breast cancer tissues. Our results showed that JMJD2A-negative staining sections were ARHI-positive both in non-neoplastic tissues (Figure 3A and B) and cancer tissues (Figure 3C and D), whereas ARHI-negative staining sections were JMJD2A-positive in cancer tissues (Figure 3E and F). Taken together, JMJD2A expression inversely correlated with ARHI expression in both non-neoplastic tissues and breast cancer tissues.

To further investigate the role of JMJD2A in ARHI expression, we carried out western blot analysis in vitro with weakly-metastatic cell line MCF-7 and highly metastatic cell line MDA-MB-231. siRNA specific against JMJD2A (siRNA group) and JMJD2A expression plasmid (pcDNA3.1-JMJD2A (JMJD2A)) were transfected into both cell lines. Transfection efficiency was approximately 71.3% (data not shown). Our results showed that siRNA-induced JMJD2A silencing caused a significant increase in ARHI expression whereas overexpression of JMJD2A led to a decrease in ARHI expression both in MDA-MB-231 cells (Figure 4A) and MCF-7 cells (Figure 4C). Two additional siRNAs against JMJD2A also elucidated the negative relationship between JMJD2A and ARHI expression at protein level (Additional file 2: Figure S2). Consistently, real-time qPCR analyses indicated that mRNA of ARHI was elevated in the siRNA group and reduced in the JMJD2A group (Figure 4B and D), suggesting that JMJD2A negatively regulated ARHI expression at transcriptional level. Recent evidence shows that JMJD2A could interact with HDACs and pRb [21], and E2F-HDAC complex functions as a transcriptional repressor of ARHI[33]. In fact, two binding sites of transcription factors at the ARHI promoter were reported including the site bound by E2Fs (from −524 to −341, indicated as the A2 site) and another site directly or indirectly bound by HDACs (from −181 to 91, indicated as the A1 site) [32, 33]. Thus we hypothesized that JMJD2A might bind to the ARHI promoter by interaction with E2Fs (mainly E2F1 and E2F4) and HDACs (mainly HDAC1 and HDAC3). We first assessed whether JMJD2A could immunoprecipitate the ARHI promoter. DNA immunoprecipitated by antibodies was then amplified by PCR with two pairs of primers specific for the respective binding sites. As expected, no DNA fragment was detected when normal IgG was used. In contrast, DNA fragments immunoprecipitated by JMJD2A antibody yielded a 183-base product that spanned nucleotides (−524 to −341) (Figure 4E), suggesting that JMJD2A could directly or indirectly bind to the binding site of E2Fs. Moreover, we could also amplify the DNA fragment (−181 to 91) to observe a 270-base product (Figure 4E), suggesting that JMJD2A might form a complex with HDACs and/or other transcription factors. Therefore, JMJD2A could bind to the promoter of ARHI to inhibit the expression of ARHI.

To uncover the detailed mechanism of JMJD2A binding to the ARHI promoter, we first investigated the interactions of JMJD2A, E2Fs and HDACs. As shown in Figure 5A, JMJD2A specifically co-immunoprecipitated E2F1, E2F4, HDAC1 and HDAC3. In fact, HDAC1 and HDAC3 were previously reported to immunoprecipitate JMJD2A [21]. E2F1 and E2F4, respectively, also immunoprecipitated JMJD2A (Figure 5B and C), suggesting that E2Fs and HDACs could form a complex with JMJD2A. We then synthesized specific siRNAs against E2F1 and E2F4 to determine whether E2Fs are required for JMJD2A binding to the ARHI promoter. Two siRNAs were proved to work efficiently (Figure 6A). Subsequent ChIP assay showed that band density was lower when either siE2F1 or siE2F4 was transfected. And after co-transfection of both siRNAs, band density was further decreased (Figure 6B), suggesting that JMJD2A binding with the A2 site (−524 to −341) in the ARHI promoter required the involvements of E2Fs. Tudor domain of JMJD2A is critical for protein interaction and the JmjC domain is pivotal for enzymatic activity. We constructed two mutants of JMJD2A expression plasmid, namely JMJD2A-M867 and JMJD2A-H188A, respectively and performed a dual luciferase reporter assay. Our results showed that JMJD2A-H188A could inhibit the ARHI promoter activity as did by JMJD2A-FL whereas JMJD2A-M867 partially lost the inhibition effect (Figure 6C), indicating that the Tudor domains, but not the catalytic activity, of JMJD2A might be required for JMJD2A-mediated repression of promoter activity. Furthermore, TSA treatment could impair JMJD2A-mediated inhibition of the ARHI promoter activity, corroborating the conclusion that JMJD2A-mediated repression requires HDACs. Overall, JMJD2A binding to the ARHI promoter requires the involvement of E2Fs and HDACs.

To reinforce the link between ARHI and JMJD2A, we investigated the role of ARHI during JMJD2A overexpression-mediated tumor progression. To this end, the inhibitory effects of ARHI on cell proliferation, migration, and invasion were first confirmed in vitro and in vivo (Additional file 3: Figure S3). Next, cell proliferation was detected in MCF-7 cells and MDA-MB-231 cells transfected with JMJD2A alone or along with ARHI. We found that ARHI significantly suppressed JMJD2A-induced tumor cell hyper-proliferation (Figure 7A and E). The proliferative rate was dampened by approximately 40% in MDA-MB-231 cells and 33% in MCF-7 cells on day 4, when ARHI was co-transfected. The wound-healing assay also showed that JMJD2A-mediated motility was inhibited by overexpression of ARHI (Figure 7B and F). Similarly, compared with the JMJD2A group, both migration and invasion abilities were also reduced after co-transfection of JMJD2A and ARHI plasmids (Figure 7C and G). In the migration assay, ARHI reduced migratory cells by 43.5% in MDA-MB-231 cells (Figure 7D) and 18.5% in MCF-7 cells (Figure 7H) compared with the JMJD2A group. In the invasion assay, ARHI decreased invasive cells by 47.8% in MDA-MB-231 cells (Figure 7D) and 30.8% in MCF-7 cells (Figure 7H) compared with the JMJD2A group.

To further investigate the role of ARHI on JMJD2A-induced tumorigenesis in vivo, we used a female athymic mice model of breast cancer. Tumor diameters were measured regularly to calculate tumor volumes (Figure 8A). Volume measurement showed that tumors displayed a smaller size in volume in each monitored time point when ARHI was co-transfected (Figure 8A). Tumors were excised on day 27 and also weighed (Figure 8B and C). Consistently, compared with the average weight when JMJD2A was overexpressed alone, a lower mass in weight was also observed when ARHI was overexpressed together with JMJD2A (Figure 8B and C). Actually, compared with JMJD2A group, tumor growth was slowed down by nearly 50% on day 27 when ARHI expression plasmid was transfected along with JMJD2A expression plasmid (Figure 8C). During dissection, we observed metastatic nodules in lung and liver. Nodules in each group of mice were counted to calculate average nodule numbers. Our results showed that average metastatic nodules in liver in each group were significantly different (Figure 8D, *P <0.05 versus the JMJD2A group). This was in accordance with our results showing that ARHI suppressed JMJD2A overexpression-induced tumor migration and invasion in vitro. However, we did not observe a significant difference in the average number of metastatic nodules in lung (data not shown). In all, ARHI could invert JMJD2A-induced tumor progression.