MBD3 inhibits formation of liver cancer stem cells

Liver cancer cells can be reprogrammed into induced cancer stem cells (iCSCs) by exogenous expression of the reprogramming transcription factors Oct4, Sox2, Klf4 and c-Myc (OSKM). The nucleosome remodeling and deacetylase (NuRD) complex is essential for reprogramming somatic cells. In this study, we investigated the function of NuRD in the induction of liver CSCs. We showed that suppression of methyl-CpG binding domain protein 3 (MBD3), a core subunit of the NuRD repressor complex, together with OSKM transduction, induces conversion of liver cancer cells into stem-like cells. Expression of the transcription factor c-JUN is increased in MBD3-depleted iCSCs, and c-JUN activates endogenous pluripotent genes and regulates iCSC-related genes. These results indicate that MBD3/NuRD inhibits the induction of iCSCs, while c-JUN facilitates the generation of CSC-like properties. The iCSC reprogramming approach devised here provides a novel platform for dissection of the disordered signaling in liver CSCs. In addition, our results indicate that c-JUN may serve as a potential target for liver cancer therapy.


INTRODUCTION
Reprogramming of somatic cells by Yamanaka factors offers a new method to investigate cancer stem cells (CSCs). The reprogramming technique not only provides a way to induce CSCs from cancer cells, but also offers insight into the discovery of their disordered signaling pathways since many of the reprogramming factors are oncogenes. Therefore, reprogramming cancer cells to induced cancer stem cells (iCSCs) by Yamanaka factors profoundly affects cancer diagnosis and therapies.
Hepatocellular carcinoma (HCC) is the fifth leading cause of cancer-related death worldwide [1]. Liver cancer initiation and development resemble induction of induced pluripotent stem cells (iPSCs). HCC contains a subpopulation of stem-like cells expressing assorted stem cell markers [2]. Stem cell specific transcription factors, such as octamer-binding transcription factor 4 (Oct4) and Nanog are important regulators of CSCs maintenance and development, and play important roles in self-renewal, chemoresistance and invasion capacity [3,4].
Methyl-CpG binding domain protein 3 (MBD3) is an essential subunit of the Nucleosome Remodeling and Deacetylase (NuRD) complex, without which the complex is not assembled [5]. A previous study has shown that MBD3 inhibits induction of iPSCs [6], and MBD3 depletion results in deterministic and synchronized iPSCs reprogramming with nearly 100% efficiency [7]. However, another study has reported that MBD3 facilitates reprogramming in a context-dependent manner [8]. More studies are needed to investigate the specific role of MBD3/NuRD in somatic cell reprogramming. In addition, understanding the MBD3 involvement in iCSCs generation is incomplete. Therefore, more studies are needed to understand the function of MBD3 in somatic and cancer cell reprogramming.
In this study, we investigated the role of MBD3/ NuRD in liver iCSCs generation. Our results indicate that Research Paper www.impactjournals.com/oncotarget MBD3 inhibits the formation of liver CSCs. Furthermore, our results suggest that expression and activity of the transcription factor c-JUN are increased in iCSCs, and are essential for stemness and CSCs properties, indicating that c-JUN might serve as a target for liver cancer therapy.

shMBD3-iCSCs acquired pluripotency and properties of CSCs
We found that Esrrb, Dppa4, Sox2, Zfp42 and Klf5 were markedly increased in shMBD3-iCSCs, whereas they were almost unchanged in shCtrl-iCSCs and C3A-iCSCs ( Figure 2A). Immunofluorescence staining further demonstrated that Zfp42 was highly expressed in nucleus and cytoplasm in shMBD3-iCSCs, while it was expressed only in cytoplasm in C3A cells ( Figure 2B). The upregulation of these pluripotent genes indicated that shMBD3-iCSCs obtained pluripotency. Since CSCs could differentiate into various cell types [9], we investigated the iCSCs differentiation capacity. Dissociated iCSCs were suspended in iCSCs medium without bFGF for 21 days and then cultured in FBS containing medium for 7 days. We detected that shMBD3-iCSCs differentiated into a variety of cellular morphologies ( Figure 2C). Immunofluorescence staining showed that shMBD3-iCSCs had differentiated and expressed endoderm marker GATA4, ectoderm marker GFAP and mesoderm marker alpha-SMA, but shCtrl-iCSCs were unable to express these differentiation markers ( Figure 2D). Thus, shMBD3-iCSCs acquired pluripotency and were able to differentiate.
Next, we analyzed liver CSCs characteristics in iCSCs. We tested three liver CSCs markers CD133, CD44, and CD90 by flow cytometry analysis, which showed that shMBD3-iCSCs had higher levels of CD133 and CD44, while CD90 was not expressed in iCSCs ( Figure 3A). However, since merely a fraction of shMBD3-iCSCs expressed CD133 or CD44, MFI ratio showed only slight differences (Supplementary Figure  1A). Consistently, we found that CD44 was located in nucleus in shMBD3-iCSCs ( Figure 3B), indicating the CSCs properties [10]. We also analyzed the proliferation of iCSCs. The EdU cell proliferation assay showed that shMBD3-iCSCs had a lower proliferative state than shCtrl-iCSCs ( Figure 3C, 3D), suggesting the stem-like properties of shMBD3-iCSCs. The results from flow cytometry analysis using CFSE staining showed consistent properties (Supplementary Figure 1B). We also analyzed the epithelial-mesenchymal transition (EMT) markers in iCSCs, since previous studies showed that EMT-related transcription factors could induce a CD44-high/CD24low pattern on epithelial cells, which was associated with somatic cells obtaining stem cell and CSCs properties [11]. The results of qRT-PCR showed that mesenchymal marker N-cadherin was almost unchanged in iCSCs, and epithelial marker E-cadherin was slightly upregulated. Notably, mesenchymal markers Twist1, Slug and Vimentin were markedly increased in shMBD3-iCSCs ( Figure 3E), indicating that shMBD3-iCSCs obtained the potential of invasion and metastasis. We further investigated the invasion ability of iCSCs by wound-healing assay, which confirmed that shMBD3-iCSCs acquired invasion phenotype ( Figure 3F and Supplementary Figure 1C). Therefore, shMBD3-iCSCs acquired more CSCs properties.

Upregulated c-JUN in shMBD3-iCSCs activates pluripotent genes and induces EMT
The fact that repression of MBD3 upregulated expression of pluripotent genes and CSCs-related genes suggested that MBD3 might mediate reprogramming by modulating critical downstream genes, which could trigger upregulation of associated genes. Since the transcription factor c-JUN binds to Nanog promoter to express higher level of stem-like genes in human colorectal cancer [12], and MBD3 can repress c-Jun transcription in colon cancer cells [13], we analyzed c-JUN expression in iCSCs. Protein expression of c-JUN was increased in shMBD3-iCSCs ( Figure 4A). Consistently, STAT3 was phosphorylated in shMBD3-iCSCs ( Figure 4B), which can be induced by c-JUN/JNK pathway [14].
We then tested the regulatory role of c-JUN in the induction of pluripotent genes in shMBD3-iCSCs. Overexpression of c-JUN in shMBD3-iCSCs and shCtrl-iCSCs induced expression of pluripotent genes Esrrb, Dppa4, Zfp42, Klf5 and Sox2 ( Figure 4C). Inhibition of c-JUN by RNAi in shMBD3-iCSCs reduced the expression of these genes ( Figure 4D, 4E). Consistently, Twist1, Slug and Vimentin were upregulated by overexpressed c-JUN in shMBD3-iCSCs and shCtrl-iCSCs and downregulated by inhibition of c-JUN in shMBD3-iCSCs ( Figure 4F, 4G). However, we were unable to demonstrate a significant effect of c-JUN on CD44 expression, and could not detect c-JUN binding in close proximity to CD44 loci by ChIP-qPCR. These results suggest that c-JUN plays a positive regulatory role in shMBD3-iCSCs, in accordance with the fact that c-JUN is transcriptional activator instead of repressor.

Without repression of MBD3/NuRD, c-Jun directly activates pluripotent genes
To investigate whether MBD3 inhibits CSCs properties through c-JUN, we first analyzed expression of MBD3 and c-JUN. Mbd3 expression increased, while c-Jun expression decreased after reprogramming in C3A-shCtrl. In contrast, reprogramming of C3A-shMBD3 into shMBD3-iCSCs upregulated c-Jun expression ( Figure  5A), indicating an inverse correlation between MBD3 and c-JUN in iCSCs. Analysis of MBD3 binding to c-Jun promoter by ChIP revealed MBD3 binding to AP-1 site in the c-Jun promoter ( Figure 5B). In addition, we found that histone deacetylase 1 (HDAC1), another subunit in the NuRD complex, was bound to the same region in shCtrl-iCSCs ( Figure 5C), suggesting that the MBD3/ NuRD complex represses c-Jun transcription [13]. In contrast, ChIP of c-JUN on the AP-1 sites demonstrated that c-JUN was recruited more efficiently to the AP-1 sites in shMBD3-iCSCs ( Figure 5D). Since H3K9 acetylation increases transcription, we analyzed H3K9ac recruitment to c-Jun promoter. As shown in Figure 5E, the AP-1 sites in the c-Jun promoter exhibited higher level of H3K9 acetylation in shMBD3-iCSCs. Thus, without repression of MBD3/NuRD, c-JUN binds to AP-1 promoter sites and activates c-JUN transcription.   Next, we investigated whether c-JUN could bind to pluripotent genes directly as a transcriptional activator. We found a canonical c-JUN/AP-1-binding motif TGAXTCA in the promoter regions of many pluripotent genes induced by c-JUN ( Figure 6A). As shown in Figure 6B, c-JUN was recruited to AP-1 sites in Klf5 and Zfp42 promoters in shMBD3-iCSCs. We used luciferase reporter containing the AP-1 sites to confirm the activation of c-JUN on Klf5 and Zfp42 promoters. We found that c-JUN increased 14and 3-fold Klf5 and Zfp42 reporter activities, respectively ( Figure 6C). Considering that MBD3/NuRD can bind to AP-1 sites and repress c-Jun transcription, we analyzed MBD3 and HDAC1 recruitment to AP-1 sites in Klf5 and Zfp42 promoters. As shown in Figure 6D and 6E, MBD3 and HDAC1 indeed bound to Klf5 and Zfp42 promoters in shCtrl-iCSCs, suggesting that NuRD represses Klf5 and Zfp42 transcription. Together, our results indicate that c-JUN activates transcription of Klf5 and Zfp42, while NuRD represses transcription of these genes.

DISCUSSION
In this study, we have identified a suppressor role of MBD3/NuRD in generation of stemness and CSCs properties in iCSCs. In addition, we have found that c-JUN is increased in iCSCs when MBD3 is inhibited, resulting in increased expression of CSCs-related genes.
Our results are in accordance with recent studies that indicated the inhibitory role of MBD3 in reprogramming somatic cells into IPSC [6,7], although other study suggested that MBD3 facilitates generation of IPSC [8]. It seems that the reprogramming conditions could influence the role of MBD3, but the signaling pathways and transcription factors involved remain elusive. Our data indicates inhibitory role of MBD3 in CSCs properties. In addition, our results suggest that NuRD represses c-Jun transcription directly, while c-JUN could activate transcription of pluripotent genes. Although the reprogramming mechanisms of somatic cells and cancer We also found that c-JUN facilitates expression of pluripotent genes, such as Klf5, Zfp42, Dppa4 and Esrrb, which is consistent with previous observations that c-JUN plays a pivotal role in the maintenance of self-renewal in cancer stem cells [15,16]. Moreover, Klf5 is essential for oncogenesis and can strengthen drug resistance in CSCs [17][18][19]. Zfp42 belongs to Myc-centered regulatory network, which is active in various cancers [20]. In our experiments, depletion of MBD3 enabled c-JUN to activate Klf5 and Zfp42 transcription in shMBD3-iCSCs, whereas MBD3/NuRD could bind to the AP-1 sites in Klf5 and Zfp42 promoters.
Our results are in apparent disagreement with the recent study by Liu et al., that suggested an inhibitory role for c-JUN in reprogramming [21]. There are several differences between our study and the study by Liu et al, including the choice of cells and reprogramming conditions. In contrast to our study, Liu et al. used mouse embryonic fibroblasts (MEFs) to generate iPSCs. In liver cancer, c-JUN forms a positive feedback loop with pluripotent genes to expedite cancer stemness [22]. However, the signaling network of c-Jun may be significantly different in MEFs [23]. Moreover, NuRD can interact with pluripotent genes and c-Jun in various ways [13,[24][25][26], and thus depletion of MBD3 can influence the interaction between c-Jun and pluripotent genes. As a member of the AP-1 family, c-JUN can dimerize with other AP-1 members to generate diverse transcription complexes. Thus, more information about the involvement of other AP-1 members, such as JunB, c-Fos, Fra1 and Fra2, in shMBD3-iCSCs is needed.
Previous studies have shown that NuRD represses transcription by binding to methylated DNA, mediating heterochromatin formation, and transcriptional silencing [27,28]. NuRD can also repress expression of pluripotent genes by opposing transcriptional activators. A previous study has shown that NuRD is recruited to Nanog by the transcriptional repressor Zfp281, and inhibits expression of pluripotency transcription factors, such as Klf4, Klf5, Zfp42, and Tbx3 in ESCs [24]. Therefore, NuRD might repress Klf5, Zfp42 and other pluripotent genes directly or indirectly; more data are needed to dissect the mechanism. In the future, we will investigate the network between MBD3 and c-JUN and their regulatory mechanisms in iCSCs.
CD133 and CD44 are CSCs markers, which are involved in invasion, metastasis and chemotherapy resistance [29][30][31][32]. In our study, we have found that CD44 is highly expressed in the nucleus in shMBD3-iCSCs. It has been reported that nuclear translocation of CD44 can reprogram colon cancer cells into CSCs [10]. In addition, CD44 is required for stemness properties and might induce EMT [33]. In somatic cells reprogramming, c-JUN upregulates CD44 and induces EMT [21]. We showed that the induced EMT-related genes in shMBD3-iCSCs could be activated by c-JUN, but we were unable to detect any influence of c-JUN on CD44. Our results suggest that c-JUN activates CSCs-related genes in liver cancer stem cells, but more studies are needed to understand the network of MBD3, c-JUN, EMT and CD44.
In summary, we have demonstrated that MBD3/ NuRD inhibits formation of liver iCSCs. In addition, we have shown that MBD3 suppression induces c-JUN, resulting in the induction of pluripotent genes in iCSCs. Therefore, our study provides a novel way to establish liver CSCs to study the mechanisms of liver cancer development. Furthermore, our study indicates that c-JUN serves as a potential therapeutic target for liver cancer.

Lentiviral infection
The day before transfection, 293T cells were seeded at 8×10 6 cells per 100 mm dish. On the next day, retroviral vectors were introduced into 293T cells using VigoFect transfection reagent (Vigorous Biotechnology, China) according to the manufacturer's recommendations. 24 hours after transduction, the medium was replaced. After 24 h, virus-containing supernatants derived from these 293T cultures were filtered through a 0.45 mm cellulose acetate filter. The viruses were then concentrated by ultracentrifugation and transduced into C3A cells in medium containing 10 ng/ml polybrene. Target cells were incubated in the virus/polybrene-containing supernatants for 4 h to overnight. After infection, the cells were replaced in 10 ml fresh medium. For stably transfected C3A, three days after infection, we added G418 at a final concentration of 0.6 to 0.8 mg/ml for 2 weeks. G418resistant cells were further cultured in the presence of 0.4 mg/mL G418.

In vitro differentiation of iCSCs
Adherent iCSCs colonies were dissociated into clumps with Accutase (Life Technologies, New York). The produced clumps were transferred to low attachment culture dishes in iCSCs medium without bFGF. After suspension culture for 21 days, iCSCs were passaged onto gelatin-coated dishes and incubated in DMEM containing 10% fetal bovine for another 7 days.

Luciferase assay
Transfections were carried out in 293T cells using VigoFect transfection reagent (Vigorous Biotechnology, China). Cells that were transfected with the expression vectors pHBLV-c-JUN and the reporter vectors were harvested 48 h after transfection. Reporter activities were measured by using the dual-luciferase reporter assay system (Promega). Each assay was performed in duplicate, and all results are shown as mean ± SD for at least three independent assays.

EdU cell proliferation assay
EdU cell proliferation assay was performed using EdU HTS Kit 488 (SIGMA-ALDRICH, St. Louis). Cells were exposed to EdU for 30 min. www.impactjournals.com/oncotarget

Statistical analysis
All data are expressed as the mean ± SD from at least three independent experiments. The Student's twotailed t-test with significance set at p<0.05 was used to determine statistical significance for each assay. Error bars indicate standard deviation.