Epigenetic changes, including DNA methylation and histone modification, play prominent roles in the oncogenesis of solid tumours and haematological malignancies (Esteller, 2008; Poreba et al., 2011; Yu, 2008). As such, many viral infections promote carcinogenesis via oncoviral protein-mediated interference with the host cell epigenetic machinery (Poreba et al., 2011). Epigenetic effector molecules targeted by oncogenic viruses therefore represent appealing targets for preventing and treating virus-induced malignancies (El-Araby et al., 2016).

DNA methylation is mediated by a complex of three protein subunits: DNA methyltransferase 1 (DNMT1), ubiquitin-like containing PHD and RING finger domain 1 (UHRF1), and ubiquitin-specific protease 7 (USP7) (Beck et al., 2018; Felle et al., 2011). UHRF1, also known as inverted CCAAT box-binding protein (ICBP90), epigenetically regulates DNA methylation by recruiting the methyltransferase DNMT1(Bostick et al., 2007; Sharif et al., 2007) and links DNA methylation and methylation maintenance following cell division (Bostick et al., 2007; Sheng et al., 2016). USP7, also known as HAUSP, is a member of the deubiquitinating enzyme family that prevents UHRF1 degradation and maintains DNA methylation by forming a trimeric complex with DNMT1 and UHRF1. USP7 promotes virus-induced epigenetic modification of host genomes, thereby regulating the life cycle of viruses such as herpesvirus and Epstein–Barr virus (Lindner, 2007). Various viruses, including human papilloma virus (HPV), induce USP7 expression and DNMT1 stabilisation. UHRF1 inhibits tumour suppressor genes (TSGs) through promoter methylation, allowing cancer cells to escape cell cycle arrest and apoptosis (Gronbaek et al., 2007). UHRF1 is overexpressed in several types of cancer, including bladder, prostate, lung, and cervical cancers (Alhosin et al., 2011; Babbio et al., 2012; Ge et al., 2016; Unoki et al., 2009; Wan et al., 2016).

Thioredoxin-interacting protein (TXNIP), also known as vitamin D3 up-regulated protein 1 (VDUP-1), is a key regulator of the redox scavenger system. This protein binds thioredoxin (Trx1), inhibiting its anti-oxidative function (Hong et al., 2016; Kaimul et al., 2007). Notably, TXNIP is frequently under-expressed in cancers lacking genetic mutations, including leukaemia and lymphoma (Erkeland et al., 2009). Hypermethylation of the TXNIP promoter and consequent loss or decrease of TXNIP expression were also recently identified in carcinogenesis (Dutta et al., 2005), and conversely, inhibition of tumour growth and apoptosis-induced cell cycle arrest (Dunn et al., 2010; Elgort et al., 2010; Han et al., 2003; Jeon et al., 2005; Liu and Min, 2002; Saitoh et al., 1998; Welsh et al., 2002) due to rebounding TXNIP expression were reported in renal cell carcinoma (RCC) (Jiao et al., 2019). These data suggest that TXNIP promoter methylation in cancer could represent an epigenetic marker for oncogenesis. However, despite evidence suggesting UHRF1 promotes cancer progression through TXNIP promoter methylation in RCC (Jiao et al., 2019), the epigenetic modulation of TXNIP expression by UHRF1 has not been studied in other cancers.

Here, we investigated the epigenetic regulation of TXNIP by UHRF1 in cervical cancer. To identify oncogenic markers, we constructed stable UHRF1-knockdown HeLa cells and performed cDNA microarray analysis. Whole-genome methylation assays showed that TXNIP was significantly overexpressed in UHRF1-knockdown cells and hypermethylated in cancer. Increased UHRF1 and decreased TXNIP protein levels were observed in cervical cancer tissue by western blot, immunohistochemistry, and Oncomine database analysis. UHRF1 binding to the TXNIP promoter was confirmed by chromatin immunoprecipitation (ChIP) and methylation was confirmed by hypermethylation pyrosequencing and methylation-specific polymerase chain reaction (PCR). We further showed that down-regulation of UHRF1 promoted cell cycle arrest and apoptosis, and overexpression of HPV E6/E7 induced ubiquitin-specific protease 7 (USP7), which promoted UHRF1 function, leading to TXNIP inhibition. Our studies suggest that TXNIP is a novel target of UHRF1 that is overexpressed in HPV-induced cervical cancer.

Cell culture and treatment

HeLa (ATCC Cat. #CCL-002, HPV-18-positive), SiHa (ATCC Cat. #HTB-35, HPV-16-positive), and C-33A (ATCC Cat. #HTB-31, HPV-negative) cervical cancer cell lines, as well as retinal pigment epithelial (ATCC Cat. #CRL-2302, ARPE-19), and HPV16 E6/E7-transformed human epidermal keratinocyte (ATCC Cat. #CRL-2404, HEK001) cell lines, were purchased from the American Type Culture Collection (ATCC, USA). Human keratinocyte (Accegen Cat. #ABC-TC536S, HaCaT) was purchased from Accegen Biotechnology (USA). HeLa and HaCaT cells were maintained in Dulbecco’s modified Eagle medium (DMEM; Gibco BRL, USA). SiHa and C33A cells were maintained in Minimum Essential Medium (Gibco BRL). HEK001 cells were cultured in Keratinocyte-Serum-Free Medium (Gibco BRL), with 5 ng/ml human recombinant epidermal growth factor (EGF) and 2 mM L-glutamine (provided in kit), and ARPE-19 cells were grown in DMEM-Ham’s F12 media (Gibco BRL). All cell line lines were routinely tested for microbial contamination and monitored for mycoplasma contamination by PCR detection (Lookout Mycoplasma PCR Detection Kit, Cat. #MP0035-1KT; Sigma-Aldrich, USA) within 6 months of being thawed.

For 5-aza-deoxycytidine (5azadC) or Trichostatin A (TSA) treatment, HeLa cells were seeded in 100-mm dishes at 1 × 10⁶ cells/dish and adhered overnight at 37°C. Cells were then were treated with 0.1% DMSO as a control, 10 μM 5azadC, or 1 μM TSA for 24 h at 37°C.

Transfection of shRNA-expressing lentiviral particles for stable UHRF1 knockdown

Short hairpin (sh)RNA targeting UHRF1 (shUHRF1; sc-76805-v) and control shRNA (shCTL; sc-108080) were purchased from Santa Cruz Biotechnology (USA). For transfections, HeLa cells were grown to approximately 50% confluence in 24-well plates and infected with virus supernatants at a multiplicity of infection of 10, with 5 μg/ml Polybrene (sc-134220; Santa Cruz Biotechnology) overnight. Culture medium was then removed and replaced with 1 ml complete medium, without Polybrene, for 24 h, and HeLa cells stably expressing shRNA were selected by puromycin (10 μg/ml) treatment for two weeks.

Patient tumour samples

Human cervical tissues were collected from patients undergoing cervical biopsies and loop electrosurgical excision procedures. All patients were high-risk HPV16/18-positive and between 29 and 84 years of age (mean, 57.1 years). Tissue samples from normal subjects (n = 3) or cervical cancer patients (n = 22) were used for western blot analysis, pyrosequencing and methylation-specific PCR. Informed consent was obtained from all participants, and the study was approved by the Ethics Committee of Gyeongsang National University Hospital (IRB No.2014-10-024-001).

RNA isolation and microarray analysis

Global gene expression analyses using Affymetrix GeneChip^® Human Gene 2.0 ST oligonucleotide arrays were performed by the commercial microarray service Ebiogen (Korea). Total RNAs from shCTL HeLa cells and shUHRF1 HeLa cells (300 ng from each sample) was converted to double-stranded cDNA using random hexamers incorporating a T7 promoter and Fragmented cDNA was generated by manufacturer’s protocol (Affymetrix, USA). Fragmented end-labelled cDNA was hybridised to the array for 16 h at 45°C and 60 rpm, as described in the GeneChip Whole Transcript Sense Target Labeling Assay Manual (Affymetrix). The chip was scanned with a GeneChip Array Scanner 3000 7G (Affymetrix) and analysed using Affymetrix Command Console software, v1.1. Normalisation was performed with the Robust Multi-array Average (RMA) algorithm, implemented in Affymetrix Expression Console software, and graphs and heatmaps were prepared using the MeV program.

DNA methylation analysis with Infinium MethylationEPIC BeadChip arrays

Genomic DNA was extracted from frozen cervical tissues and cervical cancer cell lines using NucleoSpin^® columns (Macherey-Nagel, Germany), and bisulfite converted with the EZ DNA Methylation Kit (Zymo Research, USA). Bisulfite-converted DNA (200 ng per sample) was used for Infinium MethylationEPIC arrays (Illumina, USA). Genomic DNA was amplified and hybridised to BeadChips from the Infinium MethylationEPIC BeadChip Kit (Illumina). Image intensities were extracted using Illumina iScan Control software, and graphs.

Oncomine database and GEO2R analysis

The Oncomine database (https://www.oncomine.org/) is an online data-mining platform, containing a collection of whole-genome microarray data, which is comprised of 715 datasets and 86,733 samples (Rhodes et al., 2004). Expression levels of UHRF1 mRNA (log2 median-centred intensity) in cervical cancer and normal cervix were retrieved from the Pyeon multi-cancer dataset (Pyeon et al., 2007) (GSE6791), and expression levels of TXNIP mRNA (log2 median-centred ratio) were obtained from the Biewenga dataset (Biewenga et al., 2008) (GSE7410).

Expression levels of TXNIP and USP7 mRNA were retrieved from public gene expression data in the Gene Expression Omnibus (GEO) (Barrett et al., 2013) database, using the GSE6791 (Pyeon et al., 2007) dataset. This contains expression profiles of HPV-positive and HPV-negative head/neck and cervical cancer from Affymetrix Human Genome U133 Plus 2.0 arrays (GEO accession GPL570). Only normal cervix (n = 8) and cervical cancer (n = 20) samples were extracted.

Bisulfite pyrosequencing

Bisulfite-modified genomic DNA was prepared using the EZ DNA Methylation-Lightning^TM Kit, according to manufacturer instructions. Bisulfite reactions were performed on 200 ng genomic DNA; reaction volume was adjusted to 20 μl with sterile water, and 130 μl CT Conversion Reagent (Zymo Research) was added. Reactions were performed in a thermal cycler (MJ Research; Bio-Rad, USA), using the following parameters: 8 min at 98°C and 60 min at 54°C; samples were stored at 4°C for up to 20 h.

DNA was purified using reagents provided by the EZ DNA Methylation-Lightning^TM Kit. Zymo-Spin ICTM Columns, containing 600 μl M-Binding Buffer, centrifuged at full speed for 30 s, flow-through was discarded, and columns were washed with 200 μl M-Wash Buffer, followed by a full-speed spin to remove buffer. After adding M-Desulphonation Buffer to each column (200 μl), they were incubated at room temperature (RT; 20°C-30°C) for 15-20 min. Columns were then washed again as described above. Converted genomic DNA was then eluted with 20 μl M-Elution Buffer, and DNA samples were stored at –20°C until further use.

Chromatin immunoprecipitation assays (ChIP)

ChIP assays were performed using the High-Sensitivity ChIP Kit (ab195913; Abcam, UK), according to manufacturer instructions. Antibodies used included the following: anti-UHRF1 (Epigentek, USA), anti-RNA polymerase II positive control (provided in the kit), or non-immune IgG negative control (provided in the kit). Primer sequences (Wei et al., 2017) are listed in Supplementary Table S1. ChIP-PCR assays were performed using the 2X FAST Q-PCR Master MIX SYBR (SMOBIO Technology, Taiwan). All data were normalised to negative control (IgG), and DNA binding enrichments were determined using the 2^{(IgG CT-TXNIP CT)} method.

Pyrosequencing analysis

Bisulfite pyrosequencing was used to assess methylation status of the TXNIP gene. Primers were designed using Pyrosequencing Assay Design Software (v2.0; Qiagen, Germany), and primer sequences are listed in Supplementary Table S2. PCR reactions were performed in a final volume of 20 μl, with 20 ng or more converted genomic DNA, PCR premixture (Enzynomics, Korea), 1 μl of 10 pmole/μl Primer-S, and 1 μl of 10 pmole/μl biotinylated-Primer-As. Amplification was conducted according to guidelines suggested by Pyrosequencing Assay Design Software.

Single-strand DNA template was prepared from 18 μl of biotinylated PCR products using Streptavidin Sepharose^® High Performance beads (Amersham Biosciences, UK), following the PSQ 96 sample preparation guidelines using multichannel pipets. Sequencing reactions were performed with 15 pmol of the respective sequencing primer and run on a PyroMark ID system with the PyroMark Gold Reagents Kit (Qiagen), according to manufacturer instructions and without further optimisation. TXNIP methylation level is calculated as the average of the proportion of C (%) at the position 1 CpG sites.

Methylation-specific PCR (MSP)

Bisulfite-treated DNA (500 ng) was amplified with primers specific for methylated and unmethylated TXNIP promoter. MSP primers were designed by the MethPrimer program (http://www.urogene.org/methprimer/), and all primer sequences are listed Supplementary Table S3. Reactions were performed in a thermocycler using the following parameters: 94°C for 2 min, followed by 40 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s, and a final extension at 72°C for 5 min. Bands were detected using the iBright Imaging System (Thermo Fisher Scientific, USA).

Western blot analysis

Cells and tissues were homogenised in RIPA buffer (Thermo Fisher Scientific) with protease inhibitor cocktail (Sigma-Aldrich). Protein concentrations were determined with the Pierce Bicinchoninic Acid (BCA) Protein Assay Kit (Thermo Fisher Scientific). Proteins were separated by SDS-PAGE on 8%-15% gels and transferred to nitrocellulose membranes (Millipore, USA). Blot images were captured on a RAS-4000 image reader (Fujifilm, Japan). Nuclear and cytoplasmic protein fractions were separated with the Nuclear/Cytosol Fraction Kit (Biovision, USA); density values were normalised to β-actin and analysed using Image J software. All antibodies are listed in Supplementary Table S4.

Immunocytochemistry

Tissue sections and cells were fixed on gelatine-coated slides, deparaffinised in xylene, and rehydrated in graded alcohol solution. Endogenous peroxidase activity was inhibited by incubation for 30 min in 0.3% H2O2 in 0.01 M Tris, and non-specific binding was reduced by blocking in 5% serum. Samples were incubated with primary antibody overnight at 4°C, and then with specific fluorescence-conjugated secondary IgG for 1 h in a light-protected chamber at RT. Sections were counterstained with 4,6-diamidino-2-phenylindole dihydrochloride (DAPI; Invitrogen, USA) and mounted. All images were obtained using a BX51-DSU fluorescence microscope (Olympus, Japan).

Terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) staining

HeLa cells stably expressing shCTL or shUHRF1 were seeded into 24-well plates at 2 × 10⁴ cells per well and grown for 16 h. Apoptotic cells were measured using the In Situ Cell Death Detection Kit (Sigma-Aldrich), stained with TMR red (Roche Applied Science, Germany) for 15 min, and mounted on slides using ProLong Gold Antifade Mountant (Invitrogen), with DAPI nuclear staining. All images were obtained using a BX51-DSU fluorescence microscope (Olympus).

Cell proliferation assay

HeLa cells stably expressing shCTL or shUHRF1 were seeded at 1.0 × 10⁴ cells/well in 96-well plates and treated with TQ (0, 15, 30, or 60 µM) for 24 h. MTT solution (2 mg/ml) was then added to each well, and plates were incubated at 37°C for 2 h. The resulting formazan crystals were dissolved in dimethyl sulfoxide, and absorbance was measured at 570 nm using a microplate reader (Tecan, Switzerland).

Cell cycle and apoptosis analysis

Cells were collected and seeded into 6-well plates at 1 × 10⁵ cells per well and cultured for 16 h. These were treated with 30 µM TQ for 24 h, trypsinised, fixed with cold 90% ethanol, and incubated 1 h at 4°C. Cells were then pelleted, resuspended in 1 ml PBS, containing 1 mg/ml propidium iodide (PI) and 1 mg/ml RNase A, and incubated at 37°C for 30 min. Cell cycle (Sub-G1) and apoptosis were determined by flow cytometry, according to manufacturer protocols (FACScan; BD Biosciences, USA), and data were analysed using CXP 2.2 software.

Statistical analysis

Data are expressed as the mean ± SEM. Statistical significance was determined using the two-tailed Student’s unpaired t-test for comparison of two groups and ANOVA to compare multiple treatment groups (GraphPad Prism 5; GraphPad Software, USA). P values < 0.05 were considered statistically significant.

TXNIP is induced in shUHRF1 HeLa cells and highly methylated in cervical cancer

To identify down-stream target genes of UHRF1, an important regulator of CpG methylation, we constructed stable UHRF1-knockdown HeLa cells using shRNA. We then performed cDNA microarray analysis and identified 28 genes up-regulated >2-fold in shUHRF1 HeLa cells compared to control cells expressing shCTL (Supplementary Table S5). Among these, TXNIP shows the largest increase in response to shUHRF1 HeLa cells (Fig. 1A).

We next utilised the Illumina Infinium MethylationEPIC method to perform whole-genome CpG methylation analysis in both cervical cancer and normal cervix tissue, as well as in cervical cancer cell lines (HeLa, SiHa, Caski). We found that among the common hypermethylated genes identified (Supplementary Tables S6 and S7, Supplementary Fig. S1), TXNIP showed the most significant levels of differential methylation in both cervical cancer tissue and cervical cancer cell lines compared to normal cervix tissue (Figs. 1B and 1C). These results suggest that TXNIP may be a down-stream target of UHRF1 in cervical cancer.

Expression of UHRF1 and TXNIP is inversely correlated in cervical cancer

To better understand the relationship between UHRF1 and TXNIP, their expression levels were measured in cervical cancer tissue and cervical cancer cell lines by western blotting. We found that UHRF1 was significantly increased in cervical cancer tissue (Fig. 2A) and cell lines (Fig. 2B) compared to respective controls. In contrast, TXNIP was decreased in cervical cancer compared to normal cervix tissue (Fig. 2A), as well as in cervical cancer lines versus control cell lines (Fig. 2B). Immunohistochemical staining for UHRF1 and TXNIP in normal cervix and cervical cancer tissue similarly showed increased expression of UHRF1 and decreased expression of TXNIP in cervical cancer tissue compared to normal cervix tissue (Figs. 2C and 2D), thereby confirming that expression levels of these proteins were inversely correlated in cervical cancer. Additionally, we found that Ki-67 co-localised with UHRF1 in cervical cancer tissue, but not in the normal cervix (Figs. 2C and 2D).

We also analysed expression of UHRF1 and TXNIP in several datasets containing cervical cancer samples retrieved from the Oncomine database. Consistent with our expression data, we found that UHRF1 mRNA was up-regulated in cervical cancer, whereas TXNIP mRNA was down-regulated compared to normal tissue (Figs. 2E and 2F), providing further evidence that expression of UHRF1 and TXNIP is inversely correlated in cervical cancer.

UHRF1 binds the TXNIP promoter and down-regulates TXNIP expression

UHRF1 binds at the inverted CCAAT domain in gene promoters and down-regulates expression by promoting methylation (Unoki et al., 2004). We therefore hypothesised that UHRF1 regulates TXNIP expression by binding to the inverted CCAAT domain in its promoter and inducing DNA methylation. To test this, we performed ChIP assays using anti-UHRF1 and negative control IgG antibodies in HeLa cells, with three pairs of primers specific to the region 1000 bp upstream of the TXNIP TSS (Figs. 3A and 3B, Supplementary Table S1). We detected UHRF1 binding in the region –780 to –491 bp upstream of the transcription start site (TSS), which contains an inverted CCAAT domain (Fig. 3C). We then compared UHRF1 binding to this region in HeLa cells versus two normal cell lines (HaCaT and ARPE-19). As expected, we measured higher UHRF1 binding and lower expression of TXNIP in HeLa cells compared to the other cell lines tested (Fig. 3D). These results suggest that UHRF1 binds to the inverted CCAAT domain in the TXNIP promoter and inhibits gene expression via CpG methylation.

UHRF1 knockdown inhibits UHRF1 binding to TXNIP promoter and enhances TXNIP expression through promoter demethylation

We next measured UHRF1 binding to the TXNIP promoter in HeLa cells stably expressing shUHRF1 and shCTL and found that shRNA-mediated UHRF1 silencing decreased UHRF1 binding to the TXNIP promoter and increased TXNIP expression (Fig. 4A). Using pyrosequencing, we then analysed TXNIP promoter methylation in shCTL and UHRF1-knockdown HeLa cells, revealing increased methylation at the CpG site located –780 to –491 bp upstream of the TXNIP TSS in shCTL versus shUHRF1 cells (Fig. 4B). We also designed primer sets specific for methylated or unmethylated TXNIP promoter and performed MSP on shUHRF1 and shCTL cells. Consistent with pyrosequencing data, the methylated DNA-specific product was increased in shCTL cells relative to shUHRF1 cells, whereas the band specific for unmethylated DNA was increased in shUHRF1 HeLa cells (Fig. 4C). We further tested the effect of the DNA methylation inhibitor 5azadC on TXNIP promoter methylation status in HeLa cells and found that treatment with 5azadC decreased levels of methylated TXNIP (Fig. 4D). We then measured TXNIP expression in HeLa cells co-treated with 5azadC and TSA, a known histone deacetylase (HDAC) inhibitor, and found that TXNIP levels were significantly increased in HeLa cells after 5azadC and TSA co-treatment (Fig. 4E). Collectively, these data indicate shRNA-mediated UHRF1 knockdown decreases binding at the TXNIP promoter, resulting in decreased TXNIP promoter methylation and increased TXNIP expression.

The TXNIP promoter is highly methylated in human cervical cancer

We further used pyrosequencing analysis to measure levels of CpG methylation at the TXNIP promoter in cervical cancer samples compared to normal cervix tissue. We detected increased methylation in early-stage cancer relative to healthy controls, although the difference was not statistically significant (Fig. 4F). Notably, however, levels of TXNIP promoter CpG methylation were found to be significantly increased in advanced-stage cervical cancer compared to normal cervix tissue (Fig. 4F). MSP analysis further showed increased hypermethylation in cervical cancer relative to normal cervix tissue (Fig. 4G). These data therefore demonstrate that the TXNIP promoter is highly methylated in cervical cancer, particularly in late-state disease.

TXNIP protein induced by UHRF1 knockdown translocates to the nucleus and regulates apoptosis through cell cycle arrest in HeLa cells

UHRF1 is involved in regulation of the cell cycle and apoptosis (Zhang et al., 2018). Therefore, we measured the effect of stable UHRF1 knockdown on these cellular processes in HeLa cells. We first confirmed that TXNIP expression was significantly decreased in shUHRF1-expressing HeLa cells compared to shCTL cells (Fig. 5A). To further validate these results, we also measured TXNIP expression in HeLa cells transfected with small interfering (si)RNA targeting UHRF1 or control siRNA. As expected, we detected increased TXNIP expression in UHRF1-knockdown cells compared to control cells (Supplementary Fig. S2). We then measured cellular localisation of UHRF1 and TXNIP in fractionated cells and found that UHRF1 was primarily in the nuclear fraction, whereas TXNIP was localised in both the cytosol and nucleus (Fig. 5B). As expected, in UHRF1-knockdown cells, nuclear UHRF1 expression was decreased, and TXNIP expression increased in both the nuclear and cytosol fractions. Immunocytochemistry also showed that significantly nuclear UHRF1 expression was decreased and TXNIP expression increased in both the nuclear and cytosol of UHRF1-knockdown cells (Figs. 5C and 5D).

Trx1, a negative regulator of TXNIP, is involved in cellular redox balance and cell cycle control in cancer and is regulated by UHRF1. We therefore examined whether UHRF1 causes cell cycle arrest through modulation of Thioredoxin (expression. We detected decreased expression of Trx1 and increased p27 expression in shUHRF1-expressing HeLa cells by western blot analysis (Fig. 5E), suggesting UHRF1 knockdown induces cell cycle arrest by increasing expression of TXNIP through p27. We then performed TUNEL assays to measure apoptosis and found that TUNEL-positive cells were increased by 32% in UHRF1-knockdown cells compared to controls. These results suggest that TXNIP induces cell death through UHRF1 (Fig. 5F).

UHRF1 antagonist thymoquinone induces apoptosis and cell cycle arrest in HeLa cells

To further confirm that UHRF1 regulates TXNIP, we treated HeLa cells with various concentrations of a highly selective UHRF1 antagonist, thymoquinone (TQ), and determined cell viability using MTT assays. From these experiments, the half-maximal inhibitory concentration (IC50) value of TQ for HeLa cells was estimated to be 30 µM (Fig. 6A). We then measured TXNIP expression in TQ-treated HeLa cells and found that treatment strongly enhanced the levels of TXNIP expression (Fig. 6B).

We next measured apoptosis in HeLa cells treated with 30 µM TQ for 24 h by FACS analysis and TUNEL staining. We found that apoptotic cells, in the SubG1 phase, were increased 32% in TQ-treated cells compared to controls (Fig. 6C). The number of TUNEL-positive cells also increased by 28% in response to TQ treatment compared to no-treatment control (Fig. 6D). Consistent with these results, western blot analysis showed that similar to UHRF1 knockdown, TQ treatment significantly reduced levels of Trx1 and increases p27 expression compared to untreated cells (Fig. 6E). Additionally, we observed significantly increased levels of cleaved PARP, a marker for apoptosis, in TQ-treated HeLa cells versus no-treatment control (Fig. 6F). Thus, our data indicate that TXNIP induces cell death through cell cycle arrest by UHRF1 inhibition.

HPV E6/E7 induces UHRF1 expression through USP7

USP7 regulates stability of UHRF1, thereby maintaining DNA methylation, and its expression is induced by viral proteins (Beck et al., 2018; Felle et al., 2011). Therefore, we tested whether HPV promotes USP7 expression and if HPV-induced USP7 can regulate TXNIP expression through up-regulation of UHRF1. Western blot analysis revealed that expression of USP7 was indeed increased in cervical cancer tissues compared to normal cervix (Fig. 7A). We then used the online tool GEO2R to analyse USP7 expression in cervical cancer and normal cervix samples from existing datasets and found that expression of USP7 was increased in cervical cancer compared to normal cervix (Fig. 7B).

To directly test the effect of HPV protein expression on USP7, we overexpressed HPV16 E6/E7 in two normal cell lines (HaCaT, ARPE-19). We found that expression of USP7, DNMT1, and UHRF1 was increased in HPV16 E6/E7-overexpressing cells, whereas TXNIP expression was decreased (Figs. 7C and 7D). This suggests that the vial protein stabilises DNMT1 and UHRF1 via USP7, thereby increasing methylation activity and decreasing TXNIP expression. We further measured expression of these proteins in HEK001 cells, which over-express HPV16 E6/E7, transfected with siRNA targeting E6/E7 (siE6/E7). As predicted, we detected decreased expression of USP7, UHRF1, and DNMT1, and increased TXNIP expression in siE6/E7-transfected cells relative to controls (Fig. 7E). Collectively, these data indicate that HPV E6/E7 induces expression of USP7, which promotes TXNIP methylation through the epigenetic regulator UHRF1, leading to enhanced cell proliferation and decreased apoptosis in cervical cancer (Fig. 7F).

In this study, we show that UHRF1 down-regulated TXNIP expression in cervical cancer by promoting hypermethylation at the TXNIP promoter, thereby contributing to UHRF1-mediated carcinogenesis. UHRF1 is involved in maintaining DNA methylation, and overexpression of UHRF1 induces oncogenesis through DNA methylation and down-regulation of TSGs in numerous types of cancer, including cervical cancer (Alhosin et al., 2011; Babbio et al., 2012; Ge et al., 2016; Unoki et al., 2009). A previous study reported that UHRF1 is up-regulated by the HPV oncogenic E6/E7 proteins and, in turn, down-regulates UbcH8 expression to inhibit apoptosis in cervical cancer (Zhang et al., 2018). However, despite these findings, the function and down-stream targets of UHRF1 in cervical cancer remain unclear. Here, we constructed a stable UHRF1-knockdown HeLa cell line and performed cDNA microarray analysis. From this analysis, we identified TXNIP as a novel down-stream target of UHRF1. We then performed whole-genome methylation profiling to identify genes that are hypermethylated in both cervical cancer tissues and cervical cancer cell lines compared to normal cervix tissue (Supplementary Tables S6 and S7, respectively). Among these genes, TXNIP was identified as the most highly differentially hypermethylated gene in cervical cancer tissue and cell lines compared to normal controls.

TXNIP is classified as a TSG, and accordingly, its expression is decreased in many cancer types (Baldan et al., 2015; Morrison et al., 2014; Nie et al., 2015). Recently, it was reported that UHRF1 regulates TXNIP expression through DNA methylation in RCC (Jiao et al., 2019). Therefore, we hypothesised that UHRF1-induced methylation may similarly modulate TXNIP to induce cervical carcinogenesis. Here, we detected elevated expression of UHRF1 and decreased TXNIP expression in cervical cancer tissue compared to normal cervix by both western blot and immunohistochemistry, indicating an inverse correlation between expression of these proteins. A similar expression pattern was also detected in datasets from the Oncomine database, providing additional evidence that TXNIP is negatively regulated by UHRF1.

Using ChIP analysis, we further identified the UHRF1-binding site in a region –780 to –479 bp upstream of the TXNIP TSS. Notably, this region contains an inverted CCAAT domain—the known UHRF1-binding motif (Unoki et al., 2004). Jiao et al. (2019) previously reported UHRF1-binding domains in the TXNIP promoter in regions –945 to –826 bp and –610 to –329 bp upstream of the TSS; however, they did not detect inverted CCAAT domains in these regions. Here, we also performed pyrosequencing and MSP in UHRF1-knockdown and shCTL cells, as well as in cervical cancer and normal tissue, which clearly show that UHRF1 induces hypermethylation in the CpG island of the TXNIP promoter region. Interestingly, a negative correlation between TXNIP expression and patient clinicopathological features, such as tumour differentiation status, was observed in RCC (Jiao et al., 2019), suggesting TXNIP expression may have similar prognostic value in cervical cancer.

We further found that down-regulation of UHRF1 by shRNA or treatment with UHRF1 antagonist, TQ, increased TXNIP expression, induced cell cycle arrest, and increased levels of apoptosis, further suggesting that epigenetic modification of TXNIP by UHRF1 is involved in cervical carcinogenesis. It has been shown that expression of USP7, which regulates stability of UHRF1 for maintaining DNA methylation, is induced by proteins from numerous viruses (Felle et al., 2011). Thus, we predicted that the HPV E6/E7 oncoproteins might induce expression of USP7, which in turn, promotes UHRF1 activity and consequently, down-regulates TXNIP expression. Indeed, we found that HPV E6/E7 induced USP7 in cervical cancer, leading to hypermethylated TXNIP through the action of the epigenetic regulator UHRF1, consistent with previous findings (Jiao et al., 2019).

In summary, we propose a model whereby HPV E6/E7 oncoproteins induce cell proliferation and decreased apoptosis through UHRF1-mediated TXNIP promoter methylation and down-regulation of TXNIP expression in cervical cancer (Fig. 7F). These data suggest that TXNIP could represent a possible therapeutic target for cervical cancer, as well as other types of virus-related cancer.

This research was supported by National Research Foundation of Korea (NRF-2015R1A5A2008833 and NRF-2019R1F1A1058581) and Ministry of Agriculture, Food and Rural Affairs (117082-03).

M.J.K. and W.S.C. designed and wrote the manuscript. M.J.K. performed the experiments. H.J.L, M.Y.C., Y.S.K., and S.S.K. gave technical support and conceptual advice. J.K.S. interpreted the data. W.S.C. and Y.S.K. edited the manuscript. All authors participated in review of the manuscript.

The authors have no potential conflicts of interest to disclose.