5‐Aza‐2′‐deoxycytidine suppresses human renal carcinoma cell growth in a xenograft model via up‐regulation of the connexin 32 gene

The connexin (Cx) 32 gene, a member of the gap junction gene family, acts as a tumour suppressor gene in human renal cell carcinoma (RCC) and is down‐regulated by the hypermethylation of CpG islands in a promoter region of the Cx gene. The current study investigated whether the restoration of Cx32 silenced by hypermethylation in RCC by a DNA demethylating agent could be an effective treatment against RCC.


Introduction
Gap junctions are aqueous channels that connect the cytoplasm of contiguous cells, enabling the cells to directly share small metabolites by a process called gap junction intercellular communication (GJIC) or cell coupling. It has been shown that GJIC is implicated in the regulation of cell growth, mostly by negative growth control of cell proliferation (Mesnil, 2002). Correspondingly, disorders of GJIC are strongly associated with aberrant cell growth diseases, including cancer (Mehta et al., 1999;Evans and Martin, 2002;Trosko and Ruch, 2002;Gerido and White, 2004;Nakase and Naus, 2004;Severs et al., 2004;Mesnil et al., 2005). Owing to their frequent functional alteration in tumours, the genes of gap junction, connexin (Cx) genes, have been classified as tumour suppressor genes (Yamasaki and Naus, 1996). Previous reports have found that because of cell coupling, individual cancer cells can propagate celldeath signals into adjacent cells, which then also die by apoptosis, and that the bystander effect can enhance the cytotoxicity of anticancer agents in cancer cells Yamasaki and Katoh, 1998;Wilson et al., 2000). In addition to this beneficial effect, each Cx gene has a negative growth effect on cancer with a particular specificity, and the effect is independent of GJIC (Duflot-Dancer et al., 1997;Yano et al., 2001); that is, the Cx gene preferentially exerts a tumour-suppressive effect on these tumours from normal progenitor cells in which the particular Cx gene is naturally expressed. These reports suggest the idea of using a specific Cx gene to achieve efficient cancer control by direct effects on cell growth and by exerting a bystander effect on several cancer therapies.
Renal cell carcinoma (RCC) has a very poor prognosis, due in large part to the fact that of nearly 30% of all patients with localized disease, 40% ultimately develop distant metastases following removal of the primary tumour, and available chemotherapeutic agents are ineffective against RCC (Motzer et al., 1996). Therefore, management of metastatic RCC may rely on the establishment of new, potential therapies. In our recent studies, we showed that connexin 32 (Cx32) is prominently expressed in normal human renal epithelial cells, a progenitor cell of RCC, whereas, of Cx subtypes, Cx32 is specifically downregulated in several types of human RCC cell lines as well as cancerous regions of human kidneys . On the other hand, several malignant phenotypes of primary and metastatic RCC cells are drastically attenuated by Cx32 transfection (Fujimoto et al., 2004(Fujimoto et al., , 2005; for instance, Cx32 expression by transfection reduces the growth of primary RCC cells (Caki-2 cells) and metastatic RCC cells (Caki-1 cells) in nude mice, and the metastatic potential of RCC cells in a metastasis model of human RCC using severe combined immune deficiency mice (Fujimoto et al., 2005(Fujimoto et al., , 2007a. In particular, Cx32 expression drastically suppressed the development of metastatic RCC cells in nude mice (Fujimoto et al., 2005).
The strong suppressive effect of Cx32 in vivo on metastatic RCC cells mainly depends on the inhibition of vascular endothelial growth factor (VEGF) production from cells (Fujimoto et al., 2005). All tumours must undergo angiogenesis or neovascularization to acquire nutrients for continued growth and metastatic spread (Folkman, 1995), and VEGF is the most important inducer of angiogenesis (Folkman, 1995;Grunstein et al., 1999). The main target molecule of Cx32 in RCC cells is one that stimulates VEGF production. In our previous studies, we showed that c-Src is a key target in the regulation of VEGF production in metastatic RCC cells and that Cx32 actually reduces the production of VEGF by cells via inactivation of c-Src (Fujimoto et al., 2005;Yonezawa et al., 2005). c-Src is a member of the Src family of kinases, intracellular non-receptor tyrosine kinases, widely expressed in tissues and shown to play an important role in the regulation of cell adhesion, cell growth and differentiation (Stein et al., 1994;Brown and Cooper, 1996). In a range of human cancers, including RCC, the activation or overexpression of c-Src has been observed (Rosen et al., 1986;Yonezawa et al., 2005), and activated forms of c-Src are capable of transforming many different cell types, partly because of the inhibition of GJIC (Chang et al., 1985;Hirai and Varmus, 1990). In recent studies, there has been increasing evidence that Src plays an important role in tumour cell invasion, angiogenesis and chemoresistance by upregulating signalling molecules related to these events (Turkson et al., 1998;Sinbaldi et al., 2000). From these reports, it is suggested that c-Src is a potential therapeutic target in various types of cancers, including RCC. As mentioned, in addition to general GJICdependent growth control of cancer cells, Cx32 effectively suppresses the in vivo growth of metastatic RCC cells through inactivation of c-Src, so it seems that restoration of Cx32 expression by pharmacological treatment is a promising approach to establish a new effective therapy against metastatic RCC.
Methylation of CpG islands in the 5 0 regions of tumour suppressor genes is known to inhibit transcription, leading to silencing of the corresponding genes. We have reported that Cx32 is downregulated in RCC because of hypermethylation of CpG islands in the promoter region (Hirai et al., 2003;Yano et al., 2004). Also, we have confirmed that zebularine, a DNA-demethylating agent, induced re-expression of the Cx32 gene in RCC cells in vitro . These reports led us to speculate that treatment with a DNAdemethylating agent would induce the restoration of Cx32dependent tumour-suppressive effects in RCC in vivo. If so, this treatment approach may be an effective way to establish a new metastatic RCC therapy. To address this issue, we examined whether a DNA-demethylating agent, 5-aza-2 0deoxycytidine (5-aza-CdR), which is frequently utilized in human clinical trials (Mack, 2006), could restore Cx32driven tumour-suppressive effects in nude mice bearing tumours from Caki-1 cells, a representative human metastatic RCC cell line.

Materials and methods
Cell culture and treatment A representative human metastatic RCC cell line (Caki-1), obtained from ATCC (Manassas, VA, USA), was routinely maintained in McCoy's 5A medium supplemented with 10% fetal bovine serum and penicillin-streptomycin at 37 1C in an atmosphere of 5% CO 2 . The cells were plated and treated with 5-aza-CdR or zebularine at the indicated doses for 72 h. For combination treatment with short interfering RNA (siRNA) and 5-aza-CdR, the cells were pretreated with the latter for 48 h before starting the former treatment. The treatment period for PP2 (4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrrazolo[3,4-d]pyrimidine) was 48 h.

Cell viability assay
The cells were cultured on microtitre plates and treated with each chemical at the indicated doses for the indicated treatment periods. Cell viability was then determined using the Cell Proliferation Assay kit with WST-1 reagent (Sigma, Ishikari, Japan), according to the manufacturer's protocol.

Methylation-specific PCR
The chemical modification of cytosine to uracil by bisulphite treatment is a useful method to study DNA methylation. In this reaction, all cytosines are converted to uracil, but 5 0 -methylcytosines are resistant to modification and remain as cytosine. Gen Elute Mammalian Genomic DNA kit (Sigma) was used to extract genomic DNA from cell lines, as per the manufacturer's protocol. The bisulphite reaction was carried out on genomic DNA as follows (Herman et al., 1996): the bisulphite conversion reaction was carried out by incubating DNA with a 5-M bisulphite solution and 100 mM hydroquinone, pH 5.0, at 50 1C for 4 h. Surplus bisulphite was removed using the QIAEX II Gel Extraction kit (Qiagen, Tokyo, Japan). To estimate the methylation status of the promoter region in the Cx gene, bisulphite-treated genomic DNA was amplified using methylated-specific primers for Cx32 and unmethylated-specific primers for Cx32, as previously reported (Hirai et al., 2003). The PCR product (245 bp) was electrophoresed on 2% agarose gels and stained with GelStar Nucleic Acid Gel Stain (Takara).

Immunoblot analysis
After each treatment, the cells were lysed in Cell Lysis/ Extraction Reagent with Protease Inhibitor Cocktail and Phosphatase Inhibitor Cocktails 1 and 2 (Sigma), and 15 mg protein extract from each sample was loaded onto 10% SDS-polyacrylamide gel. After electrophoresis, the proteins were transferred onto nitrocellulose membranes. The blots were incubated with each primary antibody. Each immunoreactive band was detected with an ECL system (Amersham, Buckinghamshire, UK) and the cooled CCD camera-linked Cool Saver System (Atto). Molecular sizing was performed with the Rainbow MW marker (Amersham). Protein concentration was determined by the DC Protein Assay System (Bio-Rad, Hercules, CA, USA).
Antitumour effect of 5-aza-CdR on a mouse xenograft model All animal experiments were conducted under approval from the Institutional Animal Care and Use Committee. Small fragments of Caki-1 tumour (2 Â 2 Â 2 mm) were transplanted subcutaneously into the right flank region of 4-to 5-week-old female athymic CAnN.Cg-Foxn1 nu /CrlCrlj mice (Charles River Laboratories Japan Inc., Yokohama, Japan). After 10 days, the mice received an i.p. injection of 5-aza-CdR (2.5 mg per kg body weight), dissolved in phosphate-buffered saline (PBS) or PBS only (control), once a week for 4 consecutive weeks. The mice were weighed and the tumour was measured every 3 or 4 days. Tumour volume was calculated as (p/6) Â large diameter Â (small diameter) 2 . On day 36, tumours were carefully removed after death and stocked at À80 1C for isolation of RNA or ELISA. Part of the tumours were immediately fixed in 20% formalin and embedded in paraffin for histopathological examination.
Additional studies were performed using Cx32 siRNA. Nine days after tumour transplantation, Caki-1 tumourbearing mice were treated with intratumoral injection of Cx32 siRNA or NSsiRNA at a dose of 10 mg in a total volume of 50 ml PBS twice a week for 3 consecutive weeks, in addition to 5-aza-CdR.

Protein extraction and ELISA
The tumours were lysed in CelLytic-MT (Sigma), and the lysate was cleared by centrifugation. The tumour contents of VEGF were determined with a Quantikine Human VEGF Immunoassay kit (R&D Systems, Minneapolis, MN, USA), according to the manufacturer's instructions. The tumour protein content was measured using the DC RC protein assay system (Bio-Rad).

Histopathological analysis
The tumour samples were subjected to histopathological examination. Paraffin sections (4 mm thick) were routinely stained with haematoxylin and eosin. To assess tumour cell proliferation and death, immunohistochemical staining for Ki67 antigen (DakoCytomation, Kyoto, Japan) was performed. Briefly, after deparaffinization and rehydration, the sections were immersed in 0.3% hydrogen peroxide to quench intrinsic peroxidase activity. The diluted antibodies were then added to the sections and incubated at 37 1C for 1 h. The labelled antigen was visualized with the Histofine kit (Nichirei Pharmaceutical, Tokyo, Japan), followed by reaction with 3,3 0 -diaminobenzidine. Finally, the sections were counterstained with haematoxylin. Ki67-positive cells were determined by counting about 500 nuclei in randomly selected microscopic fields, and the Ki67 labelling index was expressed as the ratio of Ki67-positive cells to total cells.

Statistical analysis
Data are expressed as the mean±s.e. and analysed by one-way ANOVA followed by Dunnett's t-test or Student's t-test only. Po0.05 was considered as showing a significant difference between means.

Results
Effects of 5-aza-CdR on cell growth, Cx32 expression and Src activation in vitro To estimate whether a representative DNA-demethylating agent (5-aza-CdR) could suppress the in vitro growth of Caki-1 cells based on restoration of Cx32 expression, the cells were treated with 5-aza-CdR. Treatment with 5-aza-CdR significantly inhibited cell proliferation by 40% compared with the control (Figure 1a). Similarly, the treatment increased Cx32 mRNA levels in a dose-dependent manner (Figure 1b), and methylation-specific PCR analysis showed clearly that the treatment induced demethylation of a promoter region in the Cx32 gene ( Figure 1c).
As previously reported (Fujimoto et al., 2005), we confirmed that enforced expression of Cx32 induced the inactivation of c-Src in Caki-1 cells and that the inactivation of c-Src led to negative growth control of RCC cells (Figures  2a and b). Similarly, 5-aza-CdR induced the inactivation of c-Src in cells, whereas Cx32 siRNA reversed 5-aza-CdRinduced Cx32 expression and activated c-Src (Figures 2a and  c). These results indicate that 5-aza-CdR exerted negative growth control of cells via the inactivation of c-Src by restoration of Cx32 expression.
Effects of 5-aza-CdR in a murine xenograft model We next examined whether 5-aza-CdR treatment could restore Cx32 expression in vivo. On day 36 after tumour implantation, tumour volume and weight in the 5-aza-CdR treatment group were significantly reduced to 29 and 27% compared with the control, respectively (Figures 3a-c). 5-Aza-CdR treatment had no influence on body weight compared with the control group (Figure 3d). In addition, none of the mice died from the treatment during the course of the experiment (data not shown). Cx32 mRNA expression in tumour tissues from the 5-aza-CdR-treated group was drastically increased compared with that from the control group (Figure 4a), and, in densitometric analysis, the level of Cx32 mRNA showed about a sevenfold increase by the treatment. In contrast, 5-aza-CdR treatment in tumour tissues reduced mRNA and protein levels of VEGF (the major molecule necessary for angiogenesis) (Figures 4b and c). Levels of mRNA and protein in the treatment group were reduced to about 10 and 25% of those in the control group,  Figure 5a) and 5-aza-CdR-treated (Figure 5b) tumour tissues were basically identical under routine microscopic observation. Immunohistochemical examination showed that the proliferative activity of tumour cells indicated as Ki67-positive cell number was higher in the control than in the 5-aza-CdRtreated samples (Figures 5c and d). The ratio of Ki67-positive cells to total cells in the former and latter samples was 32.5 and 7.3%, respectively. These results demonstrate that 5-aza-CdR-induced expression of Cx32 was closely associated with negative growth control of Caki-1 cells in this xenograft model.
Effects of Cx32 siRNA on tumour development of 5-aza-CdR-treated mice To confirm a significant role of Cx32 in 5-aza-CdR-dependent growth control of Caki-1 cells in the xenograft model, we evaluated whether intratumoral injection of Cx32 siRNA effectively abolished the suppressive effect of 5-aza-CdR. As Restoration of methylated Cx32 in RCC H Hagiwara et al shown in Figure 6, about 40-50% of the 5-aza-CdRdependent suppressive effect in the xenograft model was reversed by Cx32 siRNA treatment. Also, we confirmed silencing of the Cx32 gene by RT-PCR. This result strongly supports the above observation that restoration of Cx32 expression by 5-aza-CdR treatment plays an important role in suppressing the development of Caki-1 tumours in the xenograft model.

Discussion
Human metastatic RCC is known to show severe chemoresistance against several types of chemotherapeutic agents and has a poor prognosis in most cases. Therefore, it is a priority to establish a new therapy against RCC. In our previous studies, we reported that Cx32 reduced several malignant phenotypes of primary and metastatic RCC cell lines, mainly due to the regulation of c-Src signalling (Yano et al., 2006). Also, we have demonstrated that upregulation of c-Src during tumour progression in kidneys contributes to the acquisition of resistance against cancer therapy (Yonezawa et al., 2005). From these reports, we hypothesized that restoration of Cx32-dependent tumour suppression by pharmacological treatment could lead to the establishment of a new effective therapeutic strategy for RCC. The present study was undertaken to examine this possibility. Here, we report that 5-aza-CdR significantly suppressed the growth of human metastatic RCC cells (Caki-1 cells) in a xenograft model, mainly because of the restoration of Cx32 gene expression.
In this study, we demonstrated that 5-aza-CdR reinduced Cx32 expression in Caki-1 cells in vivo and that restoration of the Cx gene was closely associated with the suppression of RCC cells. Furthermore, to confirm the contribution of Cx32 to the negative growth control of RCC cells in vivo, we planned to induce silencing of the Cx32 gene using siRNA in 5-aza-CdR-treated xenograft model mice. As a result, the intratumoral injection of Cx32 siRNA to 5-aza-CdR-treated nude mice clearly demonstrated that Cx32, restored by 5-aza-CdR treatment, dominantly contributed to the suppression of tumour development in nude mice inoculated with Caki-1 cells. From these results, we concluded that restoration of Cx32 by 5-aza-CdR was absolutely required for the in vivo negative growth control of RCC cells driven by the agent. However, some parts of 5-aza-CdR-dependent suppressive effects were not affected by Cx32 siRNA, even though siRNA almost totally suppressed the expression of the Cx32 gene. From this observation, it was assumed that 5-aza-CdR upregulated other tumour suppressor genes related to cell growth control. To determine the tumour suppressor gene(s), we examined mRNA levels of deathassociated protein kinase (DAPK), the Ras association domain family 1A (RASSF1A), von Hippel-Lindau (VHL) and p16 INK4a , which are well known for suppression due to hypermethylation in human RCC (Sanz-Casla et al., 2003;Alleman et al., 2004;Tokinaga et al., 2004;Christoph et al., 2006). In Caki-1 tumours, mRNA levels of DAPK, RASSF1A and VHL, determined by RT-PCR, were not changed by 5-aza-CdR treatment, suggesting that the expressions of these genes were not suppressed by hypermethylation in Caki-1 tumours. In contrast, we confirmed that p16 INK4a mRNA expression was not observed in Caki-1 tumours and that 5-aza-CdR treatment induced p16 INK4a expression (data not shown). As p16 INK4a prevents the entry of cells into S phase in the cell cycle by antagonizing the activity of cyclin D-dependent kinases (Sherr and Roberts, 2004), the expression of p16 INK4 may contribute to some 5-aza-CdR-dependent suppressive effects on the development of Caki-1 tumours in nude mice.
In our previous study, enforced expression of the Cx32 gene effectively suppressed the growth of primary and metastatic RCC cells in vivo as well as in vitro because of the inactivation of c-Src (Fujimoto et al., 2005). In this study, we Suppressive effect of Cx32 siRNA on Cx32 mRNA expression and the effect of Cx32 siRNA on tumour growth in 5-aza-CdR-treated nude mice. Mice bearing Caki-1 tumours were treated with intratumoral injection of Cx32 siRNA or NSsiRNA on days 9, 14, 18, 21, 25 and 28 after cell inoculation. 5-Aza-CdR was also administered by i.p. injection on days 9, 18 and 28. Tumour volumes were measured in each indicated period. Knockdown of Cx32 by siRNA was confirmed by RT-PCR. Control-vehicle, PBS treatment; Aza-vehicle, 5-aza-CdR and PBS treatment; Aza-NSsiRNA, 5-aza-CdR and NSsiRNA treatment; Aza-siRNA, 5-aza-CdR and Cx32 siRNA treatment. Each value is the mean of five determinations; vertical lines indicate s.e.mean. *Po0.05 compared to Aza-vehicle. 5-Aza-CdR, 5-aza-2 0 -deoxycytidine; Cx32, connexin 32; NSsiRNA, nonspecific short interfering RNA; PBS, phosphate-buffered saline; RT, reverse transcription; siRNA, short interfering RNA.
Restoration of methylated Cx32 in RCC H Hagiwara et al observed that 5-aza-CdR-induced Cx32 expression exerted negative growth control of Caki-1 cells via the inactivation of c-Src, and this observation was confirmed by Cx32 siRNA and Src inhibitor experiments. As mentioned, it is well known that c-Src plays an important role in tumour cell proliferation, survival, invasion and angiogenesis (Turkson et al., 1998). Of the Src-regulated signal molecules, the signal transducer and activator 3 is considered key to inducing the Src-dependent appearance of malignancy in several types of cancers (Gray et al., 2005). We have observed that the inactivation of signal transducer and activator 3 based on the inhibition of c-Src leads to the suppression of cell growth by the inhibition of cyclin D required for G1/S progression in the cell cycle, survival by decreased Bcl-xL as a representative antiapoptotic molecule and angiogenesis by the downregulation of VEGF as an inducer of angiogenesis in RCC cells (Yonezawa et al., 2005). We also found that 5-aza-CdR significantly reduced the production of VEGF by Caki-1 tumours in nude mice, indicating that the restoration of Cx32 expression by this agent might suppress the development of Caki-1 tumours in vivo via the inactivation of c-Src/ signal transducer and activator 3 signalling. However, the inhibition of c-Src by Cx32 transfection or Src inhibitors in Caki-1 cells induces a cytostatic effect but not apoptotic cell death, maybe due to higher activity of c-Src (Yonezawa et al., 2005;Fujimoto et al., 2005Fujimoto et al., , 2006. In contrast, the same treatment by Src inhibitors can cause apoptotic cell death in other RCC cells such as ACHN cells (Yonezawa et al., 2005). Of the RCC cells utilized in the study, Caki-1 cells were established from human metastatic RCC and had higher c-Src activity, and other RCC cells were established from human primary RCC cells and had lower c-Src activity (Yonezawa et al., 2005). These observations suggest that the increase of c-Src activity during tumour progression in the kidney contributes to the acquisition of resistance against the induction of apoptosis. The available chemotherapeutic agents, such as vinblastine (VBL), are ineffective against metastatic RCC (Torti, 1983); thus, overcoming treatment-resistant malignancy is necessary to establish an effective therapy for metastatic RCC.
The significant advances in understanding the tumoursuppressive effect of Cx32 on metastatic RCC might lead to the development of therapies. We recently reported that transfection of the Cx32 gene into Caki-1 cells potentiated VBL-induced cytotoxicity in vivo as well as in vitro. The enhancing effect of Cx32 depends on the bystander killing effect based on Cx-driven GJIC and the inactivation of c-Src in a GJIC-independent manner (Sato et al., 1983(Sato et al., , 2007b. In particular, Cx32 enhanced VBL accumulation in RCC cells by the suppression of P-glycoprotein, partly because of the inactivation of c-Src. As P-glycoprotein appears to function as an energy-dependent transport pump capable of decreasing the level of VBL in RCC cells (Ambudkar et al., 1999), the downregulation of P-glycoprotein by Cx32 expression may directly contribute to the enhancement of VBL-induced cytotoxicity in RCC cells. Collectively, it seems that the combination of the bystander effect and c-Src inhibition exerted by Cx32 was one of the most effective procedures to enhance VBL-induced cytotoxicity in metastatic RCC. If this is confirmed, 5-aza-CdR-induced restoration of Cx32 expression could be an effective procedure to improve chemotherapy against metastatic RCC.
In summary, 5-aza-CdR treatment significantly suppressed the development of Caki-1 tumours in a xenograft model via the restoration of Cx32 expression caused by demethylation in the promoter region. The present data suggest that Cx32 is a potential target for the treatment of human metastatic RCC and that treatment with 5-aza-CdR is a promising therapy against RCC.