YY1 Complex Promotes Quaking Expression Via Super-Enhancer Binding During EMT of Hepatocellular Carcinoma

Author Manuscript Published OnlineFirst on February 13, 2019; DOI: 10.1158/0008-5472.CAN-18-2238 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

YY1 complex promotes Quaking expression via super-enhancer binding during EMT of hepatocellular carcinoma

Jingxia Han*, 1,2,3 Jing Meng*,1,2,3 Shuang Chen*,2,3 Xiaorui Wang*,4 Shan Yin*,5 Qiang Zhang, 2,3 Huijuan Liu,2,3,4 Rong Qin,1,3 Zhongwei Li,1,3 Weilong Zhong,1 Chao Zhang,1 Heng Zhang,1,3 Yuanhao Tang,1 Tingting Lin,6 Wanfeng Gao,1 Xiaoyun Zhang,1 Lan Yang, 2,3 Yanrong Liu,2,3 Hong-gang Zhou,1 Tao Sun#,1,2,3 Cheng Yang#,1, 3 1 State Key Laboratory of Medicinal Chemical Biology and College of Pharmacy, Nankai University, Tianjin, China 2 Tianjin Key Laboratory for Evaluation of Pharmaceutical Property, Tianjin International Joint Academy of Biomedicine, Tianjin, China 3 Tianjin Key Laboratory of Molecular Drug Research, Tianjin International Joint Academy of Biomedicine, Tianjin, China 4 College of Life Science, Nankai University, Tianjin, China 5 OBiO Technology (Shanghai) Corp.,Ltd. 6 Tianjin Medical University Eye Hospital, School of Optometry and Ophthalmology, TMU, Tianjin Medical University Eye Institute, Tianjin, China *These authors contributed equally to this work. #Correspondence to: Tao Sun, Nankai University, Haihe River Education Park, 38 Tongyan Road, Tianjin 300350, China. Phone: 8613-5129-22691; E-mail: [email protected]; Cheng Yang, Nankai University, Haihe River Education Park, 38 Tongyan Road, Tianjin 300350, China. Phone: 8615-9013-51388, [email protected] Disclosure of Potential Conflicts of Interest: No potential conflicts of interest were disclosed.

Quaking (QKI) is an alternative splicing factor that can regulate circRNA formation in the progression of epithelial-mesenchymal transition, but the mechanism remains unclear. High expression of QKI is correlated with short survival time, metastasis, and high clinical stage and pathology grade in hepatocellular carcinoma (HCC). Here we report that transcription of the QKI gene was activated by the Yin-Yang 1 (YY1)/p65/p300 complex, in which YY1 bound to the super-enhancer and promoter of QKI, p65 combined with the promoter, and p300 served as a mediator to maintain the stability of the complex. This YY1/p65/p300 complex increased QKI expression to promote the malignancy of HCC as well an increased circRNA formation in vitro and in vivo. Hyperoside is one of several plant-derived flavonol glycoside compounds. Through virtual screening and antitumor activity analysis, we found that hyperoside inhibited QKI expression by targeting the YY1/p65/p300 complex. Overall, our study suggests that the regulatory mechanism of QKI depends on the YY1/p65/p300 complex and that it may serve as a potential target for treatment of HCC. Key words: YY1/p65/p300 complex; QKI; super-enhancer; EMT; HCC

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2019 American Association for Cancer Research. Author Manuscript Published OnlineFirst on February 13, 2019; DOI: 10.1158/0008-5472.CAN-18-2238 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Statement of significance Findings identify the YY1/p65/p300 complex as a regulator of QKI expression, identifying several potential therapeutic targets for the treatment of HCC.

Hepatocellular carcinoma (HCC) is one of the most prevalent and malignant tumors with a high mortality rate worldwide (1,2). Metastasis is the most important cause of mortality in HCC (3,4). Epithelial mesenchymal transition (EMT) plays a critical role in tumor progression (5,6), and its pathological activation during tumor development can lead to the metastasis of primary tumors (3,7,8). CircRNAs are purposefully synthesized and implicated in specific biological roles in EMT. Quaking (QKI) is a distinct functional protein that mediate alternative splicing, which belongs to the STAR family of the KH domain, containing RNA binding proteins. QKI regulate a wide range of genes via alternative splicing and circRNA formation during EMT (9). Thus, it displays a potential role in tumorigenesis (10). QKI is also associated with the development and progression of human cancer (11,12). However, its potential role in HCC and its regulatory mechanism in EMT have yet to be described. Yin-Yang 1 (YY1) is a transcription factor involved in cancer progression (13). YY1 can act as a transcriptional activator or a repressor in the regulation of gene expression (14,15). Extensive evidence indicates that YY1 is inversely correlated with E-cadherin expression and crucial in EMT and tumor cell metastasis (16,17). YY1 is also carcinogenic in various cancer types, such as breast and prostate cancer (18). YY1 is significantly upregulated in HCC tissues (1). In the present study, YY1 is highly expressed and positively correlated with QKI in HCC patients with a short survival time. YY1 binds to the super-enhancer and the promoter of QKI, p65 combines with the promoter, and p300 serves as a mediator, causing the formation of DNA loops and leading to the abnormal activation of QKI. Abnormally activated QKI causes the formation of circular RNA and the occurrence of EMT and tumor metastasis in HCC.

Cell culture All the cells were bought from the American Type Culture Collection and KeyGen

medium supplemented with 10% FBS and 1% penicillin streptomycin solution at

37 °C in 5% CO2 atmosphere. The HCC cell lines were used within less a year and tested for mycoplasma before thawing original stocks. All the cells were identified and periodically authenticated by morphologic inspection and biomarkers detection of hepatocellular carcinoma, growth curve analysis.

Plasmid construction and transfection pcDNA3.1-3×Flag-YY1, pcDNA3.1-3×Flag-p65, pDONR223-EP300, pLKD-U6-shRNA and pCD2.1-ciR plasmids were obtained from Youbio (Hunan, China), Obio technology (Shanghai, China), and Geneseed (Guangzhou, China). The QKI sequence was constructed into the pGL3 luciferase vectors (Promega, USA) containing the luciferase gene under the control of the SV40 promoter. siRNA was obtained from SANTA Genepharma (Shanghai, China). All of the plasmids and control vectors were transfected into the cells by using Lipofectamine 2000 (Invitrogen, USA). All of the constructs were prepared through PCR by using the appropriate primers (Table S1). The primers of the overexpression vector are provided in Table S2.

qRT-PCR Total RNA was extracted using TRIzol reagent (Invitrogen, USA) from cells and tumor tissues. cDNA synthesis was performed with Oligo(dT) or random primers by using a Quantscript RT kit (Tiangen, China). A SYBR RT PCR kit (Tiangen, China) was used for transcript quantification with specific primers. Expression levels were quantified using the 2 Ct -actin as an internal control. The primers are provided in Table S3 of the supplementary data.

Dual-luciferase reporter gene assay A pGL3 promoter vector containing different fragments of YY1 or p65 binding sites and YY1 or p65-overexpressed or QKI knockdown vectors were cotransfected into the HCC cells. After 48 h of transfection, luciferase activities were detected using a dual-luciferase reporter gene assay kit (Promega, USA) in accordance with the Renilla luciferase activity. All of the experiments were performed in triplicates.

Immunopurification and silver staining Lysates from PLC-PRF-5 cells expressing Flag-YY1 were prepared using 0.3% NP-40 lysis buffer (0.2 mM EDTA, 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.3% Nonidet P-40) containing the protease inhibitor cocktail (Roche). Anti- Flag Tag (L5) Affinity beads (Biolegend, USA) were incubated with the cell extracts for 12 h at 4 °C. After binding was completed, the beads were washed with cold 0.1% NP-40 lysis buffer (0.2 mM EDTA, 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.1% NP-40). Afterward, the Flag peptide (Sigma, USA) was applied to the beads to elute the Flag protein complex. The eluents were collected and visualized through 10% SDS-PAGE. Subsequently, silver staining was performed using a Fast Silver Stain kit (Beyotime, China). Distinct protein bands were retrieved and analyzed through LC MS/MS.

Immunoprecipitation and Western blot analysis In immunoprecipitation, 50 µL of 50% protein A/G agarose (Pierce, USA) was incubated with control or specific antibodies at 4 °C with constant rotation for 8 h. PLC-PRF-5 cell lysates were prepared by incubating the cells in 0.3% NP-40 lysis

buffer in the presence of protease inhibitor cocktails. Lysates were centrifuged at 12,000 rpm for 10 min at 4 °C and incubated with antibody-conjugated beads for additional 12 h. After incubation was performed, the beads were washed 5-6 times by using cold 0.1% NP-40 lysis buffer. The precipitated proteins were eluted from the beads by resuspending the beads in 2 × SDS-PAGE loading buffer and boiling for 10 min at 99 °C. The boiled immune complexes were subjected to SDS-PAGE and subsequent immunoblotting. Antibody against p65 (1:100, Cell Signaling Technology, USA), YY1 (1:100, Cell Signaling Technology, USA), p300 (1:100, Cell Signaling Technology, USA). The antibodies for western blot: QKI (1:1000, Affinity, USA), p300 (1:1000, Affinity, USA), E-cadherin (1:1000, Affinity, USA), Vimentin (1:1000, Affinity, USA) and GAPDH (1:5000, Affinity, USA).

FPLC chromatography PLC-PRF-5 cell extracts were applied to a Superdex 200 10/300 GL (GE Healthcare, USA) equilibrated with 1× phosphate-buffered saline (PBS). The column was eluted at a flow rate of 0.5 mL/min, and fractions were collected.

ChIP ChIP experiments were performed on 1 × 107 PLC-PRF-5 cells. The cells with different treatments were cross-linked using 1% formaldehyde (Sigma, USA) for 10 min, quenched with 0.25 M glycine, and washed with cold PBS. The cell pellet was resuspended in 1 mL of cell lysis buffer containing 1× Protease Inhibitor Cocktail II, incubated in ice for 15 min, dissociated by pipetting, and pelleted through centrifugation at 800 ×g for 5 min at 4 °C. The pellet was resuspended in 1 mL of nuclear lysis buffer. Effective sonication was confirmed through bioanalyzer analysis. The chromatin fraction was incubated with an anti-YY1 monoclonal antibody (1:100, Abcam, USA) at 4 °C overnight. Protein/DNA complexes were reversed cross-linked to obtain free DNA. DNA was extracted and used for PCR amplification with QKI-specific primers. The primers are provided in Table S3 of the supplemental data.

Cell invasion assays A cell invasion assay was performed using a 24-well Transwell chamber (BD, USA). At 48 h post-transfection, PLC-PFR-5 and cells were trypsinized and transferred to the Matrigel (BD, USA)-coated upper chamber in 100 of serum-free medium. FBS was added to the lower chamber as a chemoattractant. After 24 h, the invasive cells on the lower surface of the chamber were stained with 0.1% crystal violet and then counted.

Wound-healing assay At 24 h post-transfection, a straight scratch was created in the center of each well by using a micropipette tip. Cell migration was assessed by measuring the cell movement within the scratch in the well. The wound closure speed after 24 and 48 h was determined and normalized to length at 0 h. Each experiment was performed in

triplicate.

Immunofluorescence PLC-PRF-5 cells were fixed with 4% paraformaldehyde for 20 min at room temperature and blocked with 5% BSA for 30 min. The cells were incubated at room temperature with Vimentin (1:200, Affinity, USA) and E-cadherin antibodies (1:200, Affinity, USA) for 1 h and then with FITC- and TRITC-labeled secondary antibodies (1:200, Earthox LLC., USA) at room temperature for 1 h. For each step, the cells were washed twice with PBS for 5 min. The prepared specimens were counterstained with DAPI (Southern Biotechnology Associates, USA) for 2 min and observed under a confocal microscope (Nikon, Japan).

SEM PLC-PRF-5 cells were fixed, dehydrated in acetone/isoamyl acetate (1:1), and dried with a gradient concentration of acetonitrile. Gold-coated cells were photographed through SEM (JEOL 6000, Japan).

Differential expression profiling PLC-PRF-5 cells were treated with YY1 or QKI overexpression lysed with TRIZOL Reagent (Invitrogen, USA). Microarray was performed on the basis of a high-throughput gene expression profile.

Proximity ligation assays (PLA) Proximity ligation assays was conducted to detect and quantify the amount of YY1-p65 protein complexes. The cells were fixed for 15 min with 200 µL of 4% PFA. Permeabilization was carried out for 10 min with 200 µL of 0.1% Triton X-100. The cover slips were blocked and incubated with 30-40 µL of diluted primary antibodies (Abs) for 90 min. YY1-p65 antibody pairs were used with a PLA staining

Bound antibodies were detected using TRITC-labeled probes and fluorescent microscope.

Xenograft tumor model Each BALB/c nude mouse in the control group was subcutaneously injected with approximately 1 × 106 PLC-PRF-5 and Hep3B cells. The other groups were injected with cells with stably expressed p65 or YY1 to the posterior flank. Each group (Ctrl, p65, YY1, p65+YY1, QKI, and p65+YY1+siQKI) consisted of five mice. Tumor-bearing mice were treated when they reached an average tumor size of 100 200 mm3. The p65+YY1+siQKI group was treated every 3 days with the intratumor injection of shQKI viral suspension (50 µL) and then continuously injected thrice. Tumor diameters were serially measured using a digital caliper every 3 days. Tumor volumes were calculated using the following equation: length × width2/2. On day 24, the mice were sacrificed. Tumor tissues were collected, fixed with 10% formalin, and

protein and RNA isolation. All of the animal experiments were performed in accordance with the approved protocols of the Institutional Animal Use and Care Committee.

Experimental metastasis assay The cells (2 × 105) were intravenously injected in the tail vein of BALB/c nude mice. Each group (Ctrl, p65, YY1, p65+YY1, QKI, and p65+YY1+siQKI) consisted of five mice. The YY1+p65+siQKI group was treated every 3 days with the intratumor injection of shQKI viral suspension (50 µL) and continuously injected thrice. After 8 weeks, the mice were sacrificed. The lungs were collected for metastasis analysis.

Orthotopic (intra-hepatic) injection of tumor cells Hep3B-Luc cells (1 × 106 cells /50 µL) were interhepatically injected into BALB/c nude mice. The group information and subsequent processing were the same as those in experimental metastasis assay. After 8 weeks, tumor burden was analyzed through bioluminescence. The mice were intraperitoneally injected with 150 mg/kg of luciferin (Caliper Life Sciences, MA) and imaged in accordance with the manufacturer s recommendations with tumors facing the camera by using an IVIS200 (Caliper Life Sciences).

Immunohistochemical assay and analysis The mouse tissues were incubated with xylene for deparaffinization and ethanol with decreased concentrations for rehydration. Afterward, 3% hydrogen peroxide was applied to block the endogenous peroxidase activity. The microwave antigen retrieval technique was utilized for antigen retrieval. After blocking was performed, the samples were incubated with the following primary antibodies at 4 °C overnight: QKI (1:200, Abcam, USA), YY1 (1:200, Abcam, USA), p65 (1:200, Affinity, USA), E-cadherin (1:200, Affinity, USA), and Vimentin (1:200, Affinity, USA). PBS displaced the primary antibody in the negative group. The secondary antibody was subsequently added using an HRP-polymer antimouse/rabbit IHC kit (Maixin Biotech, China) at room temperature for 1 h. The samples were developed with diaminobenzidine reagent, counterstained with hematoxylin, and mounted with permount33. The IHC score was calculated by multiplying the intensity (0=negative, 1=canary yellow, 2=claybank, and 3=brown) and the positive cell percentage scores (1=less than 25%, 2=25% 50%, 3=51% 75%, and 4=more than 75%).

Statistical analysis Statistical analyses were performed using GraphPad Prism 7 and SPSS v. 19. t-test, one-way ANOVA, Mann Whitney U- lation, and Kaplan Meier. P<0.05 was considered significant.

PLC-PRF-5 cells were subjected to ChIP analysis to explore the transcriptional regulation of YY1. The results revealed the presence of binding sites on the upstream of QKI transcriptional starting site (TSS). Enriched YY1 motifs are displayed in Fig. 1A (up). ChIP-polymerase chain reaction (PCR) analysis was carried out on PLC-PRF-5 cells by using specific antibodies against YY1, showing the occupancy of YY1 on the QKI promoter, which validated the ChIP-seq results (Fig. 1a, down). To assess the activity of YY1 on QKI transcription, we performed a dual-luciferase reporter assay by transfecting PGL3-promoter plasmid that contained the binding motif of YY1 alone or together with pcDNA3.1-3×Flag-YY1 plasmid into HCC cells. The results indicated that YY1 promoted the transcriptional activity of QKI (Fig. 1b). We performed pulldown experiments to determine the key interaction of YY1 in vivo. The experiment showed that YY1 was co-purified with a list of proteins, including p65 and p300 (Fig. 1c). A Venn diagram was generated to identify the shared proteins interacting with YY1, determine whether YY1 combined with other proteins to form a transcriptional complex, and consequently regulate the expression of QKI. A total of 82 proteins were found in the intersection of the protein interacting with YY1 by the FpClass database and proteins co-purified with YY1, which contained p300. YY1 was included in predicted transcription factors databases, which regulated QKI. Notably, the common genes from the three protein types included p65 (Fig. 1d). The interaction network of 87 red-marked proteins of Fig. 1d was shown (Fig. 1e). Additionally, the presence of p65 and p300 in the YY1-associated protein complex was confirmed through Western blot analysis on the column eluates (Fig. 1f). The results revealed that interactions occurred among YY1, p65, and p300. To further confirm the in vivo interactions among YY1, p65, and p300, we extracted total proteins from PLC-PRF-5 cells and performed coimmunoprecipitation with antibodies exploring the endogenous proteins (Fig. 1g). The results demonstrated that YY1 was efficiently coimmunoprecipitated with p65 and p300. Fast protein liquid chromatography (FPLC) experiments were subsequently performed using nuclear extracts. The elution patterns of p65 and p300 substantially overlapped with that of YY1 (Fig. 1h). Further statistical analysis on TCGA data showed that the expression levels of YY1, p65, and QKI in HCC tumor tissues were positively related between YY1 and QKI (r=0.6239, P<0.0001) and between p65 and QKI (r=0.3118, P<0.0001) (Fig. 1i).

Genome-wide studies have established that enhancers can be defined as DNA sequences that bind to H3K4me1 and H3K27ac, and they are located on the upstream of known TSSs (19-21). Promoters can be defined as DNA sequences that bind to H3K4me3 and POLR2A. Fig. 2a shows the human QKI genomic locus with ChIP-Seq data from the Cistrome Data Browser database for POLR2A, H3K4me3, H3K27ac, H3K4me1, YY1, p300, and p65. SE plays a critical role in gene expression (22). Therefore, we determined whether the expression of QKI is regulated by SE.

QKI SE is located at 163,826,086 163,833,600 on chromosome 6 (http://sea.edbc.org/). YY1 binds within this region. We found that YY1 bound to the SE and the promoter of QKI, p65 bound to the promoter, and p300 weakly bound to the promoter of QKI. To confirm the role of YY1 on QKI transcription, we cloned YY1 binding sites on the promoter (YBSs-pro) and YY1 binding sites on SE (YBSs-SE) into the pGL3 luciferase reporter. The results further confirmed that YY1 could bind to the SE and the promoter of QKI to improve the transcriptional activity of QKI (Fig. 2b). To identify the binding sites for YY1, p65, and p300, we cloned 12 deletion fragments of QKI into the pGL3 luciferase reporter. Notably, regions 5 and 8 exhibited the highest enhancer activity, and YY1 was overexpressed relative to that of the empty pGL3 luciferase reporter. Region 12 presented the highest enhancer activity with p65 overexpression. Conversely, p300 did not display a remarkable increase in enhancer activity (Fig. 2c). p300, a transcriptional coactivator that interacts with various transcription factors, is a coactivator of YY1 and p65 in QKI gene expression. SE is also implicated in gene expression (22). Therefore, we determined whether QKI expression is regulated by SE. We used a small molecule named JQ1, which can specifically disrupt SE (23). Luciferase assay showed that JQ1 could significantly reduce the pGL3 luciferase reporter luciferase activity that contained YY1 motif (Fig. 2d). Western blot analysis results demonstrated that JQ1 could reduce the QKI expression (Fig. 2e). Accordingly, the schematic showed the mechanism of QKI regulation by YY1 through the formation of a complex with p65 and p300 (Fig. 2f). Afterward, the effects of YY1 and p65 on QKI expression were verified. Western blot analysis results indicated that YY1 and p65 promoted the QKI expression level, and the QKI knockdown in YY1 or p65-overexpressing PLC-PRF-5 cells offset the promoting effect of YY1 and p65 on QKI expression (Fig. 2g). The silence of YY1 and p65 repressed the QKI expression (Fig. 2h). These results showed that YY1-p65-p300 promoted QKI transcriptional activity and gene expression by binding to QKI SE and promoter.

Western blot assay was conducted to explore whether YY1 and p65 regulate the EMT process by targeting QKI. The results showed that YY1 and p65 improved the expression of Vimentin and reduced the expression of E-cadherin. QKI knockdown in YY1 and p65-overexpressing PLC-PRF-5 cells offset the effects of YY1 and p65 on Vimentin and E-cadherin expression (Fig. 3a). The effects of YY1 and p65 on cell phenotypes were observed through scanning electron microscopy (SEM). Cell pseudopodia were increased, cell morphology was converted from an epithelial phenotype to a mesenchymal phenotype in YY1 or p65-overexpressing cells, and epithelioid was recovered after QKI depleted (Fig. 3b). Immunofluorescence analysis revealed that the E-cadherin expression significantly decreased, whereas the Vimentin expression increased when YY1, p65 or QKI was overexpressed. The QKI knockdown in YY1 and p65-overexpressing PLC-PRF-5 cells blocked the influence of YY1 and p65 on E-cadherin and Vimentin expression (Fig. 3c). Similarly, the QKI overexpression could also downregulate E-cadherin expression and upregulate

Vimentin expression. These results indicated that YY1 and p65 promoted EMT in HCC cells. Therefore, they might also intensify the promoting effects of QKI on migration and invasion. We selected two HCC cell lines, PLC-PRF-5 and Hep3B, to verify our hypothesis through wound-healing and invasion assays. In PLC-PRF-5 and Hep3B cells, migration (Fig. 3d) and invasion (Fig. 3e) abilities significantly increased in response to YY1 or p65-overexpressing treatment. QKI overexpression alone had similar effects. In QKI-siRNA-transfected cells, even with YY1 or p65 treatment, migration and invasion were maintained at basal levels, which were similar to that in nontreated cells. These results collectively implied that knockdown QKI could affect the YY1 or p65-induced phenotype of HCC metastasis. To assess the influence of YY1 on key biological functions, we analyzed the whole transcriptome in YY1-treated cells. We conducted gene set enrichment analysis (GSEA) by using differentially regulated genes. Differentially expressed genes were subjected to KEGG and GO analyses. After YY1 transfection was performed, VEGF, TGF- were upregulated. By contrast, p53, apoptosis, and cell death were downregulated (Figs. 3f and 3g). These results supported the theory that YY1 complex effectively enhanced EMT in HCC cells by promoting the QKI expression.

To assess the effects of QKI on key biological functions, we conducted a whole-genome expression spectrum analysis. Total RNA was extracted from the PLC-PRF-5 cells transfected with QKI, and the whole genome expression chip was analyzed. To determine the biological functions and pathways affected by differentially regulated genes, we subjected data to GO analysis. Fig. 4a shows that cell proliferation, migration, angiogenesis, cellular response to VEGF stimulus, and NIK/NF-kappaB-related biological processes were upregulated after QKI transfection. On the contrary, apoptosis and adhesion biological processes were downregulated. The upregulated genes associated with proliferation-, migration-, and angiogenesis-related biological processes after QKI treatment are presented in Fig. 4b. To examine the relationship among circRNA expression and HCC, YY1, or QKI expression, we searched the whole genome expression spectrum data and compared the difference in circRNA expression between each group through cluster analysis (Figs. 4c, 4d, and 4e). We analyzed the common circRNAs with expression levels that were largely affected relatively by upregulation with YY1 or QKI overexpression. We also analyzed the common circRNAs with expression levels highly affected by upregulation with YY1 or QKI overexpression (Fig. 4f). circRNAs exhibited a significant overlap in the two groups. To examine the high-order properties of inter-ceRNA signaling, we identified the optimal ceRNA pairs and constructed the optimal ceRNA regulatory network (Fig. 4g). Our results showed that YY1 increased the QKI expression to promote abundant circRNA formation. YY1 participated in promoting the occurrence and development in HCC by targeting QKI through the proposed ceRNA mechanism. We constructed a circRNA-miRNA-target gene

network to visualize their interactions. circRNAs adsorbed miRNAs and promoted oncogene expression. In the network, 14 miRNAs ranked relatively high, and 162 of the most likely target genes of these miRNAs were collected. Among these miRNAs, miR-6860 and miR-296 were the top predicted miRNA targets of mRNA. The expression of MAPK signaling-pathway-related genes, namely, CACNB3, AKT2, ARRB1, CACNA1H, ELK4, FGF11, DUSP3, CACNB1, and FGF22, increased. We screened six circRNAs that were upregulated evidently with YY1 and QKI overexpression. We conducted qRT-PCR to further verify the influences of YY1 and QKI on circRNA expression (Fig. 4h). The results demonstrated that six circRNAs improved at different levels in response to YY1 overexpression. In QKI knockdown-treated cells, circRNA expression was maintained at basal levels even with YY1 overexpression. YY1 indirectly controlled the expression of circRNAs by regulating the expression of QKI. We also detected that QKI caused the formation of circ-0008150, which adsorbs miRNAs (miR-615-5p) targeted Vimentin to decrease Vimentin expression (Fig. S1a-c). circ-0007821, which can absorb miR-381-3p targeted Zeb1(Fig. S2a-c), was also increased under QKI overexpressed cell. The whole-genome expression spectrum analysis showed that EMT-related transcriptional factors, especially Zeb1, was increased in QKI overexpressed cells (Fig. S3). The results showed that QKI promotes Vimentin expression by upregulated circ-0008150 absorbing miR-615-5p. Meanwhile, QKI indirectly inhibits E-cadherin expression by upregulated Zeb1, which regulated by circ-0007821 absorbing miR-381-3p. The effects of circ0008150, circ0007821, and circ0029308 on cell invasion were also assessed. The results indicated that circ0008150 and circ0007821 increased cell invasion ability. (Fig. 4i).

To gain further insights into the effects of the YY1 complex targeting QKI on tumor growth, EMT, and metastasis, we used subcutaneous PLC-PRF-5 and Hep3B tumor models. The results showed that p65, YY1 or QKI increased tumor volume and knockdown QKI under p65 and YY1 overexpressed cells inhibits the promotion effect on tumor growth (Fig. 5a). The number of metastatic lung nodules increased in overexpression of YY1, p65, or QKI groups. QKI interference maintained the number of metastatic lung nodules at the same level as that of the control group (Fig. 5b). We measured the expression levels of QKI, YY1, p65, E-cadherin, and Vimentin of the tumors through immunohistochemistry (IHC). High YY1 and p65 expression levels significantly increased the expression of QKI and Vimentin but inhibited the expression of E-cadherin. QKI knockdown counteracted the effects of YY1 and p65 on E-cadherin and Vimentin expression but did not affect YY1 and p65 expression (Fig. 5c and 5d). QKI expression was positively correlated with YY1 and p65 in tumor tissues (Fig. 5e). Furthermore, the circ0008150 and circ0007821 expression in each group was detected through qRT-PCR (Fig. 5f). High expression levels of YY1 and p65 resulted in increased expression levels of circRNAs. To examine whether QKI contributed to YY1 complex-promoted metastasis in vivo,

we conducted experimental metastasis model by intravenous injection and divided the same groups as previously described. The result showed that the average number of metastatic lung nodules was significantly increased in the highly expressed YY1 or p65 groups compared with the control or QKI knockdown group (Figs. 5g and 5h). In intrahepatic injection model, every BALB/c mouse were intrahepatically implanted with 1 × 106 Hep3B-Luc cells. After 8 weeks, tumor burden was analyzed through bioluminescence imaging. The result showed that YY1, p65, and QKI could promote the intrahepatic metastasis of HCC. The knockdown of QKI on the base of YY1 and p65 overexpression maintained the metastasis at the same level as that of the control group (Fig. 5i). These findings confirmed that YY1, p65, and QKI could expedite the intrahepatic metastasis.

Clinical data analysis was carried out to further explore the role of YY1 complex and its target gene QKI in HCC. Cancer transcription analysis on TCGA samples from UALCAN database (http://ualcan.path.uab.edu/index.html) showed that YY1 was highly expressed in HCC compared with that in normal liver tissues (Fig. S4a). Survival analysis on the expression of YY1 showed that the high level of YY1 (P < 0.0001) was associated with poor overall survival in patients with HCC (Fig. S4b). YY1 expression was positively correlated with the clinical stage and pathological grade (Figs. S4c, 4d). Further statistical analysis showed that patients with high expression of p65 and YY1 in TCGA database had worse prognosis than other patients (Fig. 6a). These data were consistent with the roles of YY1 and p65 in accelerating HCC development. The representative IHC images of normal liver tissues and HCC for QKI expression levels are shown in Fig. 6b. QKI was highly expressed in HCC compared with that in normal liver tissues (Fig. 6c). The QKI expression was positively correlated with clinical stage and pathological grade (Figs. 6d and 6e). The Kaplan-Meier survival curve revealed that the increased QKI expression indicated poor survival in patients with HCC (Fig. 6f). High YY1 and QKI expression levels were associated with short overall survival. This result was also observed in p65 and QKI (Fig. 6g). These data were consistent with the role of the YY1 complex in promoting the QKI expression to facilitate HCC progression.

The detailed binding mode of the YY1 complex is shown in Fig. 7a. YY1 and p300 had a hydrogen bond interaction (Asp1625-Arg314), p65 and p300 had one hydrogen bond interaction (Ser1232-Asn1196), YY1 and p65 had three hydrogen bond interactions (Gly96-Thr348, Ser61-Arg371, and Ser39-Thr356). According to the structures of YY1, p65, and p300, lead compounds were selected from the Traditional Chinese Medicine database through virtual screening. Hyperoside obtained the most excellent docking score with the YY1 complex (Fig. 7b). Hyperoside is one of the flavonol glycoside compounds from natural plant. The binding mode of hyperoside inhibited the YY1 complex (Fig. 7c). To determine the effect of hyperoside on the

YY1 complex, we performed Co-IP with an antibody against YY1, followed by IB with anti-YY1, p65, and p300 (Fig. 7d). The results demonstrated that the interaction between YY1 and p65 or p300 was weakened after 50 µM hyperoside treatmentment. In addition, in situ proximity ligation assays (PLA) was performed by using YY1 and p65 primary antibodies and species-specific PLA probes. The results also revealed that YY1 and p65 interactions were reduced under hyperoside treatment, however, QKI could not restore the reduced interaction between YY1 and p65 under hyperoside treatment (Fig. 7e). Western blot analysis further demonstrated that hyperoside could decrease the QKI expression to inhibit EMT and did not affect YY1 and p65 expression. While, the overexpression of QKI could mitigate the effects of hyperoside (Fig. 7f). We also investigated the influence of hyperoside on migration and invasion. The hyperoside treatment of the PLC-PRF-5 cells decreased cell migration and invasion, but the overexpression of QKI could counteract this inhibitory effect (Fig. 7g and 7h). qRT-PCR results showed that hyperoside suppressed the level of circ-0008150 and circ-0007821, and QKI restored the expression to normal levels (Fig. 7i). Collectively, these experiments supported the notion that hyperoside suppressed QKI and circRNAs expression to inhibit EMT. The nude mice that were subcutaneously injected with PLC-PRF-5 cells underwent hyperoside treatment to verify the effect of hyperoside against HCC. Hyperoside treatment inhibited the tumor growth in a dose-dependent manner, and the overexpression of QKI could mitigate the effects of hyperoside (Fig. 7j). The survival curve revealed that the survival times of hyperoside treatment groups were longer than those of the control group (Fig. 7k). Metastatic lung nodules also decreased after hyperoside treatment (Fig. 7l). The QKI overexpression could counteract the therapeutic effect of hyperoside. We further measured the QKI, E-cadherin, and Vimentin expression levels in tumor tissues through IHC (Fig. 7m). The results showed that hyperoside treatment was associated with increased the E-cadherin expression and decreased the QKI and Vimentin expression levels compared with those of the control group. In the hyperoside+QKI group, E-cadherin and Vimentin expression were similar to that of the control group. The measurement of the circRNA expression in tumors as determined through qRT-PCR indicated that hyperoside efficiently inhibited circ-0008150 and circ-0007821 expression. circRNA expression returned to normal in the hyp+QKI group (Fig. 7n). Overall, these experiments indicated that hyperoside could block EMT and metastasis by targeting the YY1 complex at the interface of the YY1 complex, and the overexpression of QKI could mitigate the effects of hyperoside.

YY1 is functionally diverse because it is a transcription promoter and transcriptional repressor. YY1 also participates in EMT and contributes to cancer progression (13,24-26). YY1 stimulates the expression of the transcription factor Snail(27), which in turn causes cells to undergo EMT by directly repressing metastasis-suppressor gene products, such as E-cadherin (28) and claudins (29), and

inducing metastasis-inducer gene products, such as Vimentin (30). High YY1 expression suggests the worsening condition of malignant tumors and poor prognosis. In pancreatic ductal adenocarcinoma, YY1 suppresses cell invasion and metastasis by downregulating MMP10 (31). In melanoma, YY1 contributes to cell proliferation, cell cycle progression, migration, and cell invasion (32). Here, we found that YY1 is highly expressed in HCC cells, thereby playing a crucial role in EMT program. YY1 overexpression in HCC may be involved in the regulation of progression from cirrhosis to HCC and malignant progression (33). In our results, YY1 expression is also positively correlated with the clinical stage and pathological grade of patients with HCC. These findings indicated that YY1 served as a functional biomarker in HCC progression. YY1 can form transcriptional complexes with transcription factors and a homologous dimer to exert transcriptional activation effects (34). YY1 dimer is combined with the promoter and enhancer to form an enhancer promoter loop, causing an abnormal gene transcription (35). YY1 combined with p300 triggers histone acetylation, which causes gene activation by facilitating the binding of RNA polymerase (36). YY1 can form transcriptional complexes with transcription factors and a homologous dimer to exert a transcriptional activation effect(37). Our results showed that YY1 combined with the super-enhancer and the promoter of QKI in the YY1-p65-p300 complex to trigger EMT. The mechanism is different from that of the homologous dimer of YY1, but the function is consistent. In this YY1-p65-p300, whether YY1 binds to two DNA sites in the form of a monomer or a homologous dimer still needs further verification. In our study, transcriptome analysis showed that VEGF-, TGF- -, and Wnt-related signaling pathways were upregulated with QKI overexpression. By contrast, p53- and apoptosis-related signaling pathways were downregulated. QKI is associated with adult cancers potentially through its regulation of microRNA functionality (38). For example, QKI functions as a tumor suppressor in colon cancer (39). In our study, high QKI expression was correlated with poor prognosis in HCC. YY1 promoted the expression of QKI by forming a transcriptional complex with p65 and p300. YY1 accelerated the expression of circRNAs by targeting QKI and caused abundant circRNA formation. These processes promoted tumor progression in part by its ability to induce EMT (40-42). QKI affects pre-mRNA splicing, mRNA turnover, and translation in many cancers. QKI is sufficient and necessary to mediate plasticity between epithelial and mesenchymal states and drive EMT-associated alternative splicing by promoting exon skip, even in the absence of EMT inducers (43). Simon J. Conn showed that QKI dynamically regulates abundant circRNA formation by the alternative splicing factor during EMT. The abundance of circRNA is dependent on intronic QKI binding motifs. CircRNAs are purposefully synthesized and implicated in specific biological roles in EMT (9). QKI may regulate a wide range of genes via alternative splicing and circRNA formation during EMT. In our study, differential expression profiling analysis demonstrated that QKI caused the formation of EMT-related circRNAs, such as circ-0008150 and circ-0007821. Circ-0008150, which adsorbed miR-615-5p and targeted Vimentin, thereby decreasing its expression level. CircRNA-0007821, which

could absorb miR-381-3p targeted Zeb1, also increased in QKI-overexpressing cells. Consequently, Zeb1 increased to inhibit E-cadherin expression. The overexpression of QKI also promoted the expression of other EMT-related transcriptional factors, such as Twist and Snail families, which decreased E-cadherin by combining E-box and promoting Vimentin expression. YY1 silencing downregulated QKI and circRNA expression. Many studies have shown that the expression profiles of miRNAs and circRNAs are abnormal in many cancer types, and many of them have focused on their epigenetic regulation in cancer development (2,44). The abnormal expression of circRNAs is commonly observed in various cancer types (21). Certain circRNAs present EMT-related functions; thus, they may affect mesenchymal cell properties, such as migration, invasion, and propensity for cancer to metastasis. QKI regulated hundreds of alternative splicing targets and the formation of circRNAs to exert pleiotropic effects, such as increasing cell migration and invasion and promoting EMT. Hyperoside, one of the flavonol glycoside compounds from natural plant, specifically inhibits the formation of the YY1/p65/p300 complex by targeting the complex interaction interface and subsequently depresses the transcription of QKI and the formation of downstream circRNA. Therefore, hyperoside may inhibit HCC EMT. In conclusion, our studies displayed the regulatory mechanism of YY1 in the QKI expression in HCC. Specifically, a novel mechanism presented that the YY1/p65/p300 complex regulated the QKI expression by binding to the SE and the promoter to promote abundant circRNAs formation and promote EMT in HCC. Thus, YY1 was a potential biomarker of HCC, thereby improving the accelerated effects of QKI in HCC tumorigenesis in vitro and in vivo. YY1 could also be used as an independent predictor of metastasis and survival in HCC. In brief, the elucidation of the roles and regulatory mechanisms of the YY1/p65/p300 complex might facilitate the treatments for the suppression of cancer cell metastasis.

National Natural Science Funds of China (Grant nos. 81572838, 81872374, 81402973, 81703581, 81871972), Tianjin Science and Technology Project (Grant nos. 15PTGCCX00140, 18PTSYJC00060), Chinese National Major Scientific and Technological Special Project for "Significant New Drugs Development" (Grant nos. 2018ZX09736-005, SQ2018ZX090201), The National Key Research and Development Program of China (Grant no. 2018YFA0507203), Postdoctoral support scheme for innovative talents (Grant no. BX20180150), China Postdoctoral Science Foundation (Grant no. 2018M640228), The Fundamental Research Funds for the Central Universities, Nankai University and Natural Science Foundation for Young Scholar of Tianjin (Grant no. 16JCQNJC12300).

Fig. 1 YY1 is a key transcription factor for QKI expression in PLC-PRF-5 cells. (a) ChIP-seq analysis on the DNA-binding motifs of YY1 on the promoters of QKI (up). ChIP experiments on the QKI promoter by using antibodies against YY1 in PLC-PRF-5 cells. (b) Dual-luciferase reporter assay in four HCC cells: the promoter activities of QKI were significantly increased because of overexpressed YY1. (c) Pulldown analysis on YY1-associated proteins. Cellular extracts from Flag-YY1-expressing PLC-PRF-5 cells were subjected to affinity purification with anti-Flag affinity columns and eluted with Flag peptide. Elutes were resolved through SDS-PAGE and silver stained. (d) Venn diagram of transcription factors on the upstream of QKI prediction by using Chipbase database, proteins interacting with YY1 by using FPclass database, and proteins interacting with YY1 by performing pulldown analysis. (e) Protein-protein interaction network of 87 red-marked proteins in the intersection of Fig. 1d. (f) Immunoblot for YY1, p65, and p300 by using pulldown products. (g) Co-IP of endogenous YY1, p65, and p300 was performed in PLC-PRF-5 cells. (h) FPLC experiments in PLC-PRF-5 cells. Chromatographic elution profiles and IB analysis of the chromatographic fractions with antibodies against the indicated proteins are shown. Equal volumes from each fraction were analyzed, and the elution positions of the calibration proteins with known molecular masses (kDa) are indicated. (i) Co-expression analyses on YY1, p65, and QKI in TCGA Database. Data are presented as means ± SD. *P<0.05, **P<0.01.

Fig. 2 YY1 enhances QKI gene transcription and translation through the formation of a complex with p65 and p300. (a) Human QKI genomic locus with ChIP-Seq data from the Cistrome Data Browser database. (b) Reporter assay constructs; summary of luciferase promoter and enhancer assays of YY1 binding sites (YBSs-Pro and YBSs-SE) in PLC-PRF-5 cells. (c) QKI promoter deletion cloned upstream of the luciferase gene in the pGL3 luciferase reporter. Activity values were normalized to Renilla luciferase activity. (d) Luciferase activities with JQ1 treatment. (e) Western blot analysis detecting QKI expression with JQ1 treatment. (f) Schematic of the mechanism through which YY1/p65/p300 regulates QKI. (g) QKI expression measured through Western blot analysis in PLC-PRF-5 cells transfected with YY1 or p65 alone or co-transfected with QKI siRNA. (h) QKI expression measured by using Western blot analysis with the transfection of YY1 or p65 siRNA.

Fig. 3 YY1 and p65 promote EMT, migration, and invasion by targeting QKI. (a) Expression of E-cadherin and Vimentin transfected with YY1, p65, and QKI or co-transfected with QKI siRNA. (b) Morphological changes in PLC-PRF-5 cells transfected with YY1, p65, and QKI or co-transfected with QKI siRNA observed with SEM. (c) Representative immunofluorescence images of E-cadherin and Vimentin of PLC-PRF-5 cells treated in the same manner as in the wound-healing assay. (d) Migration of PLC-PRF-5 cells transfected with YY1, p65, and QKI or co-transfected with QKI siRNA. (e) Cell invasion assays performed in PLC-PRF-5 cells transfected with YY1, p65, and QKI or co-transfected with QKI siRNA. (f) GSEA and KEGG analyses of the enriched expression of YY1 target gene sets after YY1 overexpression in PLC-PRF-5 cells (red, YY1; blue, control). (g) GSEA and GO analyses on the enriched expression of YY1 target gene sets after YY1 overexpression (red, YY1; blue, control). Data are presented as means ± SD. *P<0.05, **P<0.01.

Fig. 4 QKI is related to EMT progression, and it promotes the generation of circRNAs. (a) Signaling pathways promoted and suppressed by QKI in HCC cells. (b) Upregulated genes associated with proliferation, migration, and angiogenesis in PLC-PRF-5 cells after QKI treatment analyzed by using DAVID. (c e) Heatmap of identified circRNAs with changes between groups. (f) Venn diagram of the upregulated circRNAs with YY1 or QKI overexpression. (g) Graphical view of the circRNA-miRNA-mRNA network of the upregulated circRNAs after QKI and YY1 overexpression. Purple circles, red circles, blue circles, and edges correspond to circRNAs, miRNAs, mRNAs, and direct interaction links, respectively. (h) qRT-PCR of six circRNAs performed in PLC-PRF-5 cells. (i) Cell invasion assays performed in PLC-PRF-5 cells transfected with circRNAs.

Fig. 5 QKI expression is required for the YY1 complex to promote the metastasis and malignancy of HCC. (a) Comparison of PLC-PRF-5 and Hep3B tumor growth in mice were measured starting 6 days after implantation. All of the groups were euthanized on day 24. (b) Metastatic lung nodules were counted in each group. (c) QKI, p65, YY1, E-cadherin, and Vimentin expression levels in the tumor tissues of p65, YY1, p65+YY1, QKI and p65+YY1+siQKI groups analyzed through IHC staining. (d) IHC staining index of QKI, YY1, p65, E-cadherin, and Vimentin in HCC tissues. (e) Correlation analysis between p65 and QKI (P=0.0012) and YY1 and QKI (P=0.0048). (f) qRT-PCR analysis of circ-0008150 and circ-0007821 levels in tumor tissues. (g) Representative figures of HE staining on the serial sections of metastatic tumors in the lungs. (h) Metastatic lung nodules were counted in each group. (i) Representative images of in vitro BLI. Quantification the signal intensity in each group. Quantified values are shown in Total flux (photons/second). Data are presented as means ± SD. *P<0.05, **P<0.01.

Fig. 6 QKI is upregulated by YY1 complex, and it can promote the malignant progression of HCC. (a) Survival analysis of TCGA samples for the relationship between survival time of HCC patients and p65 and YY1 expression levels. (b) Representative QKI expression images of normal liver tissues and HCC samples. (c) QKI expression level in HCC tumors and normal liver tissues based on TCGA dataset. (d) Analysis on the QKI expression levels in TCGA LIHC samples based on an individual clinical stage. (e) Analysis on the QKI expression levels in TCGA HCC samples based on an individual pathological grade. (f) Kaplan-Meier plots of the overall survival rate of HCC patients with QKI low or high expression levels (P=0.0001). (g) Kaplan-Meier plots of the overall survival rate of HCC patients with YY1/QKI low or high levels (left), and with p65/QKI low or high levels (right) (P<0.0001). Data are presented as means ± SD. *P<0.05, **P<0.01.

Fig. 7 Hyperoside inhibits the EMT and metastasis of HCC by targeting YY1 complex. (a) Detailed binding mode of YY1, p65, and p300. (b) Prediction score of docking between small-molecule drugs and YY1 complex. (c) The structure of hyperoside (left). Predicted binding modes of hyperoside (green) and YY1/p65/p300 complex (right). (d) Co-IP of endogenous YY1, p65, and p300 was performed in PLC-PRF-5 cells treated with 120 µM hyperoside. (e) The effect of hyperoside on YY1 and p65 interaction assessed by using Duolink proximity assay. (f) Western blot with antibody against p65, YY1, QKI E-cadherin and Vimentin under different treatments. Western blots were also probed for GAPDH as a loading control. (g-h) Migration and invasion of PLC-PRF-5 cells with different treatments. (i) qRT-PCR analysis on circ-0008150 and circ-0007821 levels under hyperoside and QKI treatment. (j) Representative images and tumor growth curves of nude mice in four groups. (k) Survival curve of nude mice in four groups. (l) Metastatic lung nodules were counted in each group. (m) IHC staining index of QKI, E-cadherin, and Vimentin in HCC tissues in different groups. (n) Expression of circ-0008150 and circ-0007821 analyzed by using qRT-PCR. Data are presented as means ± SD. *P<0.05, **P<0.01.

YY1 complex promotes Quaking expression via super-enhancer binding during EMT of hepatocellular carcinoma

Jingxia Han, Jing Meng, Xiaorui Wang, et al.

Cancer Res Published OnlineFirst February 13, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-18-2238

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2019/02/13/0008-5472.CAN-18-2238.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet been Manuscript edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2019/02/13/0008-5472.CAN-18-2238. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.