Demethylation at enhancer upregulates MCM2 and NUP37 expression predicting poor survival in hepatocellular carcinoma patients

Public bulk-RNA seq datasets TCGA-LIHC (https://​portal.​gdc.​cancer.​gov/​projects/​TCGA-LIHC), GSE57957 (https://​www.​ncbi.​nlm.​nih.​gov/​geo/​query/​acc.​cgi?​acc=​GSE57957), and E-MTAB-4171 (https://​www.​ebi.​ac.​uk/​arrayexpress/​experiments/​E-MTAB-4171/​), including 421 HCC samples and 104 paired nontumor liver samples, were downloaded for identification of differentially expressed genes (DEGs) at mRNA levels. Public proteomic dataset OEP000321 (https://​www.​biosino.​org/​node/​project/​detail/​OEP000321) including 159 pairs HCC and adjacent nontumor liver samples were downloaded from the National Omics Data Encyclopedia (NODE) for identification of differentially expressed proteins (DEPs). Moreover, public methylomic datasets E-MTAB-4169 (https://​www.​ebi.​ac.​uk/​arrayexpress/​experiments/​E-MTAB-4169/​), TCGA-LIHC (https://​portal.​gdc.​cancer.​gov/​projects/​TCGA-LIHC), and GSE56588 (https://​www.​ncbi.​nlm.​nih.​gov/​geo/​query/​acc.​cgi?​acc=​GSE56588), including 638 HCC samples, 93 paired nontumor liver samples, 10 normal liver samples and 9 cirrhotic liver samples, were analyzed for determining the methylation patterns of potentially identified biomarkers for HCC patients.

The R package “TCGAbiolinks” [12] was used to download and process the raw data of TCGA-LIHC datasets. Differentially expressed genes (DEGs) as well as differentially expressed proteins (DEPs) between the HCC and nontumor tissues were identified using the ‘limma’ R package [13] based on P value < 0.05 and |logFC| \(\ge\) 2 and 1, respectively. In addition, differential methylation regions (DMRs) were identified among CpG or non-CpG islands throughout the genome scale. The “minfi” and “limma” R packages [13, 14] were used for identification of DMRs with P value < 0.05 and |logFC| \(\ge\) 0.2.

We prospectively collected 10 HCC and adjacent tissue (peritumor) samples from primary HCC patients without receiving neoadjuvant treatment after curative surgical therapy from May to July 2020 at the First Affiliated Hospital of Zhejiang University, Hangzhou, China. The collected each fresh specimen was snap-frozen in liquid nitrogen, and then used for protein and RNA extraction, or stored at −80 °C for subsequent experiments. Moreover, two sets of tissue microarrays (TMAs), generated in-house, including paraffin-embedded tissues from 300 primary HCC patients who had undergone surgical resection from 2010 to 2018 in our hospital, were used for immunohistochemistry (IHC). All patients were followed-up routinely every 3 to 6 months after surgery. The clinicopathological information on these 300 cases of HCC patients was summarized in Additional file 1: Table S1. Written informed consents were obtained from each patient in this study. The experiment protocol for use of clinical samples was approved by Institutional Ethics Committee in First Affiliated Hospital of Zhejiang University.

Anti-MCM2 antibody (HPA031496), anti-NUP37 antibody (ab220675), anti-GAPDH antibody (ab181602), Goat Anti-Rabbit IgG H&L (ab205718) and DAB Substrate Kit (ab64238) were purchased from Sigma and Abcam, respectively. Human MCM2, NUP37 and β-actin primer sets were purchased from Gene Chem (Shanghai, CN). Decitabine (S1200) and CPI-455 (S8287) were purchased from Selleck (Shanghai, CN). Dimethyl sulfoxide (DMSO) was purchased from Sigma.

Human HCC cell lines Huh7, Hep3B, HepG2, Bel-7402, HCC-LM3, normal liver cell line HL7702, and mouse hepatoma cell line Hep1-6 were purchased from The Cell Bank of Chinese Academy of Sciences (Shanghai, China) with authentication between 2018 and 2020. HL7702, Huh7, and Bel-7402 were cultured in high glucose DMEM (Gibco), other cells were cultured in MEM (Gibco) supplemented with 10% FBS, 100 μg/ml streptomycin (Gibco), and 100 U/mL penicillin (Gibco) at 37 °C in 5% CO2 atmosphere. All cell lines had been tested regularly for mycoplasma contamination. In addition, 3 × 10⁵ Huh7 cells per well were planted in 6-well cell culture plate and incubated for 48 h in high glucose DMEM supplemented with 10% FBS and the indicated concentration of decitabine or CPI-455. Decitabine was dissolved with phosphate buffered saline (PBS) and CPI-455 was dissolved with DMSO.

Cell viability was assessed using a CCK8 assay (Beyotime, Shanghai, CN). Briefly, Hep3B cells (1 × 10⁴ cells per well) were seeded into 96-well plates and incubated in antibiotic-free growth medium supplemented with decitabine, CPI-455, PBS or DMSO. After incubated for 24, and 48 h, fresh complete medium with CCK-8 (1:10) was added to each well, and the cells were incubated at 37 °C in 5% CO₂ atmosphere for 1 h. Finally, the absorbance at 450 nm was detected using a microplate reader (800 TS, BioTek, USA).

Soluble proteins from fresh HCC and paired peritumor tissue were obtained after following procedure: (1) manually grinding tissue into powder within 2 min using a clean mortar after immersed it into liquid nitrogen; (2) then collecting the tissue powder (5–10 mg) into a pre-cold Eppendorf tube containing 300 μl RIPA lysis buffer (Beyotime, Shanghai, CN) supplemented with 1 × protease inhibitor buffer (#5871, Cell Signaling Technology, CN) on ice; (3) constantly agitating the mixture on an orbital shaker at 4 °C atmosphere for 2 h; (4) After centrifugation at 16,000g for 20 min at 4 °C, the supernatants were collected. For normal liver cells and HCC cells, washing two times with ice-cold PBS, and lysis with 100 μl RIPA buffer supplemented with protease inhibitor cocktail for 30 min on ice. After centrifugation at 12,000g for 15 min at 4 °C, the supernatants were collected as the whole cellular protein extracts. Both tissue and cellular soluble protein extracts were quantified using the bicinchoninic acid protein assay kit (Beyotime, Shanghai, CN).

A total of 30 μg protein was resolved on SDS-PAGE (10% polyacrylamide), transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were blocked with 5% skim milk for 1 h at room temperature, then incubated with anti-MCM2 antibody (0.1 μg/ml), anti-NUP37 (1:2000) or anti-GAPDH antibody (1:5000) overnight at 4 °C. After 3 times washes of membranes in 1× tris-buffered saline with tween 20 (TBST), the membranes were incubated with anti-rabbit immunoglobulin G-horseradish peroxidase-linked secondary antibody (1:3000) for 2–3 h at 4 °C. The protein bands were visualized by a Chemiluminescence Dectection Kit for HRP according to the manufacture’s instruction and detected using ChemiScope 6100 Touch system (CliNX, CN).

Total RNA from human cells or tissues was extracted using Trizol reagent (Life Technologies) according to the instruction of the user manual. The concentration of RNA was determined by spectrophotometry at 260 nm using Nanodrop One (Thermo SCIENTIFIC). The purity of sample was determined by the 260/280 nm ratio. The detected ratio, ranging from 1.8 to 2.1, indicating that the extracted RNA was fine and suitable for subsequent experiments. cDNAs were generated from 1 μg RNA of each samples by using PrimeScript™ RT reagent Kit with gDNA Eraser (Takara, Japan). Real-time PCR (RT-PCR) analysis was performed using TB Green ™ Premix Ex Taq™ (Tli RNaseH Plus, Japan). A final volume of 20 μl mixture was amplified by RT-PCR using ABI QuantStudio-5 Real-Time PCR System (Applied Biosystems by Thermo Fisher Scientific). Reaction conditions were as follows: 95 °C for 30 s, followed by 40 cycles of denaturation at 95 °C for 5 s and annealing at 60 °C for 30 s. The relative mRNA expression was determined using the ΔΔCT method. The primer sequences used in real-time PCR are as follows: MCM2, forward 5’-CTACCAGCGTATCCGAATCCA-3’ and reverse 5’-CCTACAGCAACCTTGTTGTCCT-3’; NUP37, forward 5’-CTGCGTTTCGTGACCTTGTC-3’ and reverse 5’-TACACGTGCCAATGACCACA-3’; β-actin, forward 5’-CGACAGGATGCAGAACGAGA-3’ and reverse 5’-GACCCTGGATGTGACAGCTC-3’.

Immunohistochemical staining was performed as described previously [15]. After anti-NUP37 antibody (1:200) and anti-MCM2 antibody (1:200) were incubated with tissue sections and TMAs overnight at 4 °C, the biotinylated secondary antibody (1:300) was applied before visualization using DAB substrate. The IHC reaction result was calculated by multiplying a percentage score and staining intensity score. The percentage score was used to describe the estimated percentage of positively stained neoplastic cells (0: 0–5%; 1: 6–25%; 2: 26–50%; 3: 51–75%; 4: > 75%). The intensity score described the estimated staining degree (0: negative; 1: weak; 2: moderate; 3: strong). Both the percentage score and staining score were evaluated independently by two pathologists who were blinded to the research design. Any discrepancy wound be resolved by a consensus among these two pathologists after a re-assessment. The NUP37 score and MCM2 score were used to stratify patients into low and high groups according to the optimal cut-off value.

For validation of the methylation pattern of MCM2 enhancer region and NUP37 promoter region, genomic DNA from frozen HCC tissues and paired nontumor tissues samples (n = 6) were extracted using Blood and Tissue Kit (#69504, Qiagen). The concentration of genomic DNA from each sample was determined using NANO Quant infinite M200PRO (TECAN) and a total of 500 ng genomic DNA (A260/A280 ratios ranging from 1.8 to 2.0) from each sample was used for bisulfite DNA conversion using the EpiTect Bisulfite Kit (#59104, Qiagen) following the manufacturer’s handbook for user. After cleanup of bisulfite converted genomic DNA, we amplified the target sequence using ABI 9700 PCR System and performed quantitative pyrosequencing using the PyroMark Q96 ID (QIAGEN). The genomic DNA methylation percentage was calculated with using Pyro Q—CpG software (Biotage). The primer sequences used in PCR amplification and pyrosequencing, as well as the chromosomal location of each target sequence in this study were summarized in Additional file 1: Table S2. Pyrosequencing was performed using the platforms from Geneland Biotech Co., Ltd (Shanghai, CN).

To establish subcutaneous implantation model that was used to assess the impact of decitabine on HCC progression in vivo, 3 × 10⁵ Hep1-6 cells (per mouse) in the logarithmic phase, resuspended in 100 μl PBS, were injected into the flanks of 5-week-old male C57BL/6 J mice (Model Animal Research Centre of Nanjing University, Nanjing, CN). Tumor growth was monitored daily using a digital caliper. Upon tumor reached 150 to 300 mm³, mice were randomized into control (PBS) or decitabine (3 mg/kg, 3 times weekly) groups with each group involving 5 mice. The mice were administrated with decitabine or PBS by intraperitoneal injection. After the treatment, each mouse body weight was monitored daily using an electronic balance. Mice were sacrificed according to the Institutional Animal Care & Use Committee (IACUC) protocol and subcutaneous tumor lesions were completely excised with using the dissection instruments in a mouse surgical kit. We calculated the tumor volume as follows: Volume = length × (width)² × π/6.

All statistical analyses were performed with using R version 4.0.3 for Windows 64-bit. Data are presented as arithmetic mean ± Standard Deviation (SD). ANOVA, chi-square test, fisher's exact test, Student’s t test, or Wilcoxon test was used to determine the statistical significance according to variable type and distribution. The “surv_cutpoint ()” function, encoded in the “survminer” R package (https://​cran.​r-project.​org/​web/​packages/​survminer/​index.​html) was used to determine the optimal cut-off point for continuous variables that correspond to the survival of HCC patients. Overall survival (OS) analysis was performed using the Kaplan–Meier method, and survival differences were determined by log-rank test encoded in “Survival” R package (https://​cran.​r-project.​org/​web/​packages/​survival/​index.​html). Multivariate survival analysis was performed using the Cox proportional hazards regression model. Spearman or Pearson correlation analyses were performed to determine the relationship between NUP37 and MCM2 expression, MCM2 or NUP37 mRNA expression and DNA methylation levels, respectively. Differences or correlations with p values < 0.05 were considered statistically significant.

Briefly, a total of 1915 differentially expressed genes (DEGs) were identified between the HCC and nontumor tissues by mining the transcriptomic dataset (Fig. 1A). Out of those, 1730 genes were upregulated and 185 were downregulated in HCC. In the proteomic dataset, 38 significantly upregulated and 330 downregulated proteins were identified in HCC compared to nontumor (Fig. 1A). Additionally, 5779 upregulated and 46,906 downregulated methylated cytosine-phosphate-guanine (CpG) or non-CpG islands were present throughout the HCC genomes (Fig. 1A). DNA strand elongation, ECM proteoglycans, and glycolysis were the top three biological processes enriched by the identified DEGs matched with DEPs (Fig. 1B). After assessing the results of reactome pathways enrichment analysis with ClueGO version 2.5.6. (Fig. 1B) and a comprehensive literature review, we subsequently focused on MCM2 and NUP37 that significantly enriched in DNA strand elongation and glycolysis process, respectively.

Results from bulk RNA-seq dataset of TCGA showed that MCM2 and NUP37 mRNA levels in HCC tissues are significantly upregulated in comparison with that in nontumor tissues (Fig. 2A). Similarly, the MCM2 and NUP37 protein levels in HCC tissues are significantly higher than that in peri-tumor tissues from NODE proteomics dataset (Fig. 2B). Consistently with the expression profiles in omics datasets of HCC, validation by western blot (Fig. 2C, D) and RT-PCR (Fig. 2E) in 10 pairs of human HCC tissue samples showed that MCM2 and NUP37 were significantly upregulated in HCC relative to the paired peri-tumor samples. H&E staining was performed to assess the histopathology type of each HCC tissue and paired peri-tumor tissue (Additional file 1: Fig. S1). Similar with the findings from our internal validation dataset, results from two external validation datasets also showed that MCM2 and NUP37 mRNA are significantly overexpressed in HCC relative to the paired nontumor tissue samples (Fig. 2F, G).

We also assessed the expression profiles of MCM2 and NUP37 in 5 different hepatoma cell lines and one normal liver cell line (HL7702). Results from western blot showed that MCM2 protein was highly expressed in HL7702, Huh7, Bel-7402 and HCC-LM3 cells compared with that in HepG2 or Hep3B cells (Fig. 2H). Meanwhile, NUP37 protein was determined in all the indicated cells (Fig. 2H). Results from RT-PCR showed that MCM2 mRNA expression levels were higher in the indicated hepatoma cell lines than that in normal liver HL7702 cells (Fig. 2I). In comparison with the expression levels in HL7702 cells, NUP37 mRNA levels were upregulated in the indicated hepatoma cell lines except for HepG2 cells (Fig. 2J).

IHC staining of HCC tissue samples showed that MCM2 and NUP37 were mainly expressed in nucleoplasm of carcinoma cells (Fig. 3A, B). MCM2 mRNA expression levels were significantly correlated with NUP37 mRNA expression levels in patients with HCC from TCGA dataset (Fig. 3C) and ArrayExpress dataset (Fig. 3D). Similarly, MCM2 protein expression was significantly correlated with NUP37 protein expression in HCC patients from NODE dataset (Fig. 3E). This significantly positively relationship between MCM2 and NUP37 expression profile was further validated by IHC staining in our internal HCC cohort (Fig. 3F).

To determine the clinical relevance of MCM2 or NUP37 overexpression in human HCC, we firstly investigated the correlation between tumor stage (TNM stage) and MCM2 or NUP37 overexpression using TCGA samples. The mRNA expression profiles of MCM2 or NUP37 stratified by the tumor stage showed that MCM2 (Stage II vs. Stage I, p < 0.05; Stage III vs. Stage I, p < 0.001) or NUP37 (Stage II vs. Stage I, p < 0.05; Stage III vs. Stage I, p < 0.05) expression was significantly increased in advanced tumor stage (Fig. 4A, B).

Univariate survival analysis was performed to assess the impact of MCM2 or NUP37 expression on the survival of HCC patients from 3 large independent HCC cohorts (n = 820). Based on the optimal cut-off value of 7.623 (log₂ expression) for MCM2 mRNA, HCC patients from TCGA cohort were stratified into two groups. Kaplan–Meier survival curve showed that HCC patients with high MCM2 mRNA levels had worse OS (median OS, 24.1 (13.8 -NA) months vs. 71.0 (51.0–104.0) months; log-rank, 4.124; p < 0.0001, log-rank test; Fig. 4C). Similarly, OS curve showed that patients with high NUP37 mRNA levels (log₂ expression = 6.914) had worse OS (median OS, 25.2 (14.2 -NA) months vs. 60.6 (50.3–94.6) months; p = 0.0015, log-rank test; Fig. 4D). HCC patients from NODE proteomic cohort were divided into two groups according to an optimal threshold value of -0.047 (unshared log ratio), and Kaplan–Meier survival curve showed that HCC patients with high MCM2 protein levels had worse OS (median OS, 33.9 (29.8–37.9) months vs. 70.7 (64.4–76.9) months; p = 0.00066, log-rank test; Fig. 4C). Similarly, Kaplan–Meier survival curve showed that patients with high NUP37 protein levels (unshared log ratio \(\ge\) 0.124) had poor OS (median OS, 51.8 (43.4–60.3) months vs. 59.4 (53.7–65.0) months; p = 0.0059, log-rank test; Fig. 4D).

Notably, these findings were further validated by analysis of our in-house HCC cohort (n = 300). Based on the score of MCM2 or NUP37 determined by IHC, these 300 cases of HCC patients were classified into MCM2 or NUP37 high and low groups. OS analysis showed that HCC patients with MCM2 or NUP37 high expression had worse OS (median OS, 37.3 (29.2–45.3) months vs. 59.5 (53.4–65.6) months; p = 0.00019, log-rank test; Fig. 4C; median OS, 35.4 (27.9–42.8) months vs. 60.6 (54.2–67.0) months; p < 0.0001, log-rank test; Fig. 4D), respectively. Moreover, Cox univariate survival analysis showed that HCC patients with tumor size > 5.6 cm, advance tumor stage (T4/T3/T2), AFP > 25 ng/ml, ALT > 25U/L, or AST > 60U/L significantly associated with worse OS (Fig. 4E).

Correlation analyses showed that there was no significant relationship between MCM2 or NUP37 expression and the comorbidities of HCC patients (Table 1). Meanwhile, correlation analyses of MCM2 or NUP37 expression with clinicopathological features revealed a significant association between NUP37 overexpression and tumor metastasis (Table 2). In addition, multivariate survival analysis showed that tumor size > 5.6 cm, advanced tumor stage, ALT > 25U/L, MCM2 or NUP37 overexpression were independent risk factors for worse OS of HCC patients (Fig. 4F).

To investigate the epigenetics mechanisms upregulating MCM2 expression in HCC patients, we analyzed the methylation profiles of MCM2 gene sequences. Bioinformatic analysis using online ensemble project showed that MCM2 promoter region has some CpG islands (Fig. 5A), which can potentially silence gene expression by DNA methyltransferases-mediated methylation. Nine promoter-associated CpG sites (Fig. 5B) were identified with using “IlluminaHumanMethylation450kanno.ilmn12.hg19” R package (http://​bioconductor.​org/​packages/​IlluminaHumanMet​hylation450kanno​.​ilmn12.​hg19/​). By integrative analysis of TCGA bulk RNA-seq and methylomic datasets of HCC patients, we found that DNA methylation levels of MCM2 enhancer (cg08889930) was significantly negatively correlated with MCM2 mRNA expression (Pearson correlation test; r = − 0.548, p = 2.57e−33, Fig. 5C, D). This significant relationship was also observed in both one external dataset (Pearson correlation test; r = − 0.652, p = 9.52e−05; Fig. 5E) and our internal validation dataset (Pearson correlation test; r = − 0.85, p < 0.001; Fig. 5F).

To determine the direct role of DNA methylation in MCM2 expression, decitabine was used to inhibit DNA methyltransferase in vitro. Results from RT-PCR showed that MCM2 mRNA expression levels in Huh7 cells significantly increased after decitabine treatment (Fig. 5G). To investigate whether MCM2 overexpression can be silenced by inhibition of histone demethylation, CPI-455, an inhibitor of histone lysine demethylase 5 (KDM5), was used to suppress the demethylation of histone H3 on lysine 4 (H3K4) in Huh7 cells. Interestingly, results from RT-PCR showed that MCM2 expression significantly downregulated after CPI-455 therapy (Fig. 5H). These findings were further validated by western blot (Fig. 5I, J), which indicating that inhibition of DNA methyltransferase can induce MCM2 expression via downregulating methylation levels of MCM2 gene. Furthermore, inhibition of histone demethylase can significantly attenuate MCM2 expression by inhibition of KDM5 activity.

To precisely decipher DNA methylation profile of MCM2 promoter region, we analyzed the methylation patterns of CpG islands or CpG island shores in HCC as compared with adjacent nontumor tissue samples from 3 independent HCC datasets with same sequencing platform (Illumina Infinium HumanMethylation450 BeadChip). Results from E-MTAB-4169 showed that there are 4 significantly differential methylation regions between HCC samples and paired adjacent samples (Fig. 6A). Of these 4 regions, the methylation levels of cg08889930 and cg15057726 sites were significantly lower in HCC samples (n = 43) than in nontumor samples (n = 43). Interestingly, in comparison with other CpG sites in promoter region, cg08889930, an enhancer of MCM2 gene, located at CpG island shores, with a mean betavalue > 0.6, was obviously hypermethylated in both tumor and paired nontumor samples (Figs. 5B,  6A). The obvious demethylation patterns of cg08889930 occurred in HCC relative to paired nontumor samples was further confirmed by using another two external validation datasets (Fig. 6B, C) and our internal validation dataset (Additional file 1: Fig. S2A). To assess the impact of aberrant methylation of cg08889930 on the prognosis of HCC patients, HCC patients from TCGA cohort were stratified into two groups according to the optimal threshold value (β-value = 0.556). Kaplan–Meier survival curve showed that HCC patients with lower methylation degree of cg08889930 had significant poor OS (median OS, 15.1 (10.0–28.7) months vs. 81.9 (56.5–104.2) months; p < 0.0001, log-rank test; Additional file 1: Fig. S2).

Together, these findings indicated that demethylation at enhancer of MCM2 was a common event in HCC patients, which was responsible for MCM2 overexpression in HCC by regulating MCM2 transcription at genome-wide.

Similarly, CpG islands in NUP37 promoter region (Fig. 7A) are capable of silence NUP37 expression via DNA methyltransferases-mediated methylation. Meanwhile, a total of 6 CpG sites, including 4 promoter-associated CpG sites and 3 promoter-unassociated CpG sites, were identified using “IlluminaHumanMethylation450kanno.ilmn12.hg19” R package (Fig. 7B). To determine the correlation of DNA methylation levels with NUP37 mRNA expression, we performed an integrative analysis of TCGA bulk RNA-seq and methylomic datasets of HCC patients. However, no moderate/strong correlations were determined between NUP37 mRNA expression and the methylation degree of CpG islands/non-CpG islands (Fig. 7C). We hypothesized that the demethylation patterns of cg08889930 indirectly regulates NUP37 expression since NUP37 shares a similar expression pattern with MCM2 at both mRNA and protein levels. Then, by analysis of the relationship between NUP37 mRNA expression and the methylation level of cg0889930, we found that DNA methylation degree of cg08889930 was significantly negatively correlated with NUP37 mRNA expression (Pearson correlation test; r = − 0.328, p = 1.14e−11, Fig. 7D). Notably, this finding was finally verified by using one external validation dataset (Pearson correlation test; r = − 0.619, p < 1e−05; Fig. 7E) and our internal validation dataset (Pearson correlation test; r = − 0.763, p = 0.011; Fig. 7F).

To investigate the impact of DNA methylation on NUP37 expression, decitabine was used to treat HCC cells by inhibition of DNA methyltransferase in vitro. Results from RT-PCR showed that NUP37 mRNA expression levels in Huh7 cells significantly increased after treatment with decitabine (Fig. 7G). To further investigate whether NUP37 expression can be downregulated by inhibition of histone demethylation, CPI-455 was used to treat Huh7 cells. Results from RT-PCR showed that NUP37 expression significantly downregulated after CPI-455 therapy (Fig. 7H). These findings were further validated by western blot (Fig. 7I, J). Taken together, our results showed that inhibition of DNA methyltransferase increased NUP37 expression. Whereas inhibition of histone demethylase decreased NUP37 expression by inhibiting KDM5 activity.

To get a more precise understanding of DNA methylation patterns of NUP37 gene at genome-wide scale, we also analyzed the methylation profiles of CpG sites in HCC as compared with adjacent nontumor tissue samples from 3 independent HCC datasets with same sequencing platform. Results from E-MTAB-4169 showed that the methylation levels of the indicated CpG sites in HCC samples (n = 43) is similar with that in paired adjacent nontumor samples (Fig. 8A). Consistently with the findings from E-MTAB-4169 dataset, results from another two external validation datasets showed that there are no significantly differential methylation sites between clinical HCC samples and adjacent nontumor samples (Fig. 8B, C). Furthermore, results from our internal validation dataset showed that the methylation pattern of cg03502446 island within NUP37 promoter region is similar between HCC and adjacent nontumor tissues (Additional file 1: Fig. S3B). Interestingly, the methylated percentage of CpG sites of cg24737639 and NUP37 promoter region are significantly lower in HCC tissues than in paired nontumor tissues (Additional file 1: Fig. S3C, D). Importantly, results from both external and internal validation datasets showed that the mean beta-value of NUP37 promoter region was less than 0.2, which indicates that a hypomethylation pattern of NUP37 promoter occurred in both HCC and nontumor tissue samples (Fig. 8). Together, these findings suggested that the possibility of other methylation sites at genome-wide beyond NUP37 location-scale are indirectly responsible for NUP37 mRNA overexpression in HCC patients.

Cell viability assays showed that DNA methyltransferase (DNMT) inhibitor decitabine promoted HCC cells proliferation, whereas KDM5 inhibitor CPI-455 inhibited HCC cell proliferation (Fig. 9A). To investigate the effect of decitabine on HCC progression in vivo, we treated tumor-bearing C57BL/6 J mice with decitabine by intraperitoneal injection. We observed that subcutaneous tumor volumes were significantly increased in decitabine treatment group compared to control group (Fig. 9B). Mouse body weight and mouse survival rate were markedly decreased in decitabine group relative to controls (Fig. 9C, D). In addition, MCM2 and NUP37 expression were increased in decitabine-treated group relative to the controls (Fig. 9E). These findings indicated that decitabine can promote HCC progression via upregulating methylation-induced oncogene MCM2 and NUP37 expression. Due to CPI-455 is water-insoluble, we therefore did not assess its treatment effect in vivo.