|Home | About | Journals | Submit | Contact Us | Français|
Liver fibrosis is caused by chemicals or viral infection. The progression of liver fibrosis results in hepatocellular carcinogenesis in later stages. Recent studies have revealed the importance of DNA hypermethylation in the progression of liver fibrosis to hepatocellular carcinoma (HCC). However, the importance of DNA methylation in the early-stage liver fibrosis remains unclear.
To address this issue, we used a pathological mouse model of early-stage liver fibrosis that was induced by treatment with carbon tetrachloride (CCl4) for 2 weeks and performed a genome-wide analysis of DNA methylation status. This global analysis of DNA methylation was performed using a combination of methyl-binding protein (MBP)-based high throughput sequencing (MBP-seq) and bioinformatic tools, IPA and Oncomine. To confirm functional aspect of MBP-seq data, we complementary used biochemical methods, such as bisulfite modification and in-vitro-methylation assays.
The genome-wide analysis revealed that DNA methylation status was reduced throughout the genome because of CCl4 treatment in the early-stage liver fibrosis. Bioinformatic and biochemical analyses revealed that a gene associated with fibrosis, secreted phosphoprotein 1 (Spp1), which induces inflammation, was hypomethylated and its expression was up-regulated. These results suggest that DNA hypomethylation of the genes responsible for fibrosis may precede the onset of liver fibrosis. Moreover, Spp1 is also known to enhance tumor development. Using the web-based database, we revealed that Spp1 expression is increased in HCC.
Our study suggests that hypomethylation is crucial for the onset of and in the progression of liver fibrosis to HCC. The elucidation of this change in methylation status from the onset of fibrosis and subsequent progression to HCC may lead to a new clinical diagnosis.
Fibrosis is one of the most severe systemic diseases and is characterized by excessive accumulation of fibrous connective tissues, such as collagen, induced by acute or chronic injury . Fibroproliferative diseases occur throughout the body, including in the lungs, kidneys, and liver. The progression of fibrosis leads to the failure of the physiological functions of tissues. Liver fibrosis, in particular, has been extensively investigated because its progression results in hepatocellular carcinoma (HCC), which is the fifth most common cancer worldwide [2,3].
Currently, liver fibrosis is known to be a part of the dynamic process of continuous extracellular matrix (ECM) remodeling in chronic liver injury . In liver fibrosis, a liver injury activates the Kupffer cells--resident macrophages of the liver sinusoids--thereby inducing inflammation . This inflammatory response triggers the activation of hepatic stellate cells (HSCs), which play a key role in fibrogenesis by transdifferentiating into myofibroblasts . The proliferation of myofibroblasts and stimulation of ECM synthesis, ultimately results in liver fibrosis.
Recently, the progression of liver fibrosis has been reported to be associated with hypermethylation of DNA . HSC activation is inhibited by 5'-Azacytidine (5'-Aza), a DNA methylation inhibitor, resulting in the transdifferentiation of HSCs to myofibroblasts . Furthermore, in other tissues such as renal, several studies using heterozygous mice revealed that the DNA methyltransferase Dnmt1 and its inhibitor 5'-aza ameliorate renal fibrosis by inhibiting proliferation of myofibroblasts [8,9]. These results indicate the pivotal role of DNA hypermethylation in the progression of both liver and renal fibrosis. However, the importance of DNA methylation in early-stage liver fibrosis remains unclear.
Here, we performed a global analysis of DNA methylation during the onset of liver fibrosis. A mouse model treated with carbon tetrachloride (CCl4) was used as a model of liver fibrosis. Analysis of the CCl4-treated mouse livers revealed symptoms of early-stage liver fibrosis. To analyze the genome-wide DNA methylation profile of this CCl4-induced early-stage liver fibrosis, we used a combination of methyl-binding protein (MBP)-based precipitation (MBP-IP) and high-throughput DNA sequencing (MBP-seq). This genome-wide analysis can reveal hypo- and hypermethylated sites (125 and 88, respectively) in the genomic DNA. Analyzing the MBP-seq data, we revealed that the DNA methylation status was reduced throughout the genome, and that the enhancer of secreted phosphoprotein 1 (Spp1), also known as osteopontin, was hypomethylated. Two bioinformatics tools, IPA and Oncomine, indicated that Spp1 is related to liver fibrosis and inflammation. Using biochemical methods, such as bisulfite modification and in-vitro-methylation assays, we confirmed that the hypomethylation of Spp1 enhancer up-regulates its mRNA. These results clearly indicate that hypomethylation of the genome may precede the onset of liver fibrosis. Moreover, Spp1 enhances tumor development. This suggests that hypomethylation during the early-stage liver fibrosis may be important in the development of the primary liver cancer HCC, which is an end-stage liver disease.
Five-week-old male C57BL/6 wild-type mice were purchased from CLEA Japan Inc. To induce liver fibrosis, 2 ml/kg CCl4 mixed with olive oil was intraperitoneally administered 3 times per week for 2 weeks. All animal husbandry and animal experiments were performed in accordance with the guidelines of the University of Tsukuba's Regulation of Animal Experiments Committee. Enzymatic activities of the serum proteins, alanine aminotranferease (Alt) and aspartate aminotransferase (Ast), were measured using the Fuji dri-chem 3000 analyzer (Fuji Film).
Mouse liver cell line, Hepa1-6 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS). Buffalo Rat liver cell line, BRL-3 A were maintained in Ham's F12 medium with 10% FBS. 3uM of the 5-Aza deoxy derivative, 5-dAza-C, were treated with BRL-3 A for 36hr, and then harvested for indicated experiments.
After 2 weeks with or without CCl4 treatment, liver tissues were fixed with formalin. Formalin-paraffin livers were cut into 5 μm thick. Hematoxylin and Eosin (H&E) staining, Masson's trichrome (MT) staining and Sirius red staining were performed according to the manufacturer's protocol.
The genomic DNA from each of 4 mouse livers treated with 2 ml/kg CCl4 for 2 weeks were pooled. The genomic DNA from 3 mouse livers treated with olive oil for 2 weeks was pooled and used as control. The genomic DNA purified from the CCl4-treated mouse liver specimens was randomly fragmented into 50-350 bp lengths as described in the SOLiD 5500xl fragment library protocol. The fragmented DNA was then subjected to the MethylMiner methylated DNA enrichment kit according to the manufacturer's protocol. These methylated fragments were eluted with 1000 mM NaCl and used to construct standard fragment libraries using a combination of adaptor ligation and nick translation (SOLiD Fragment Library Construction Kit, Invitrogen). Each DNA library was selected on the basis of size (inserts were approximately 500 bp) by AgencourtAMPure XP (Beckman) before PCR amplification, bead attachment, and emulsion PCR. Libraries were sequenced on a SOLiD 5500xl Analyzer (Applied Biosystems). The resulting tag sequences and quality files were mapped onto the mouse genome (NCBI Build 37, UCSC mm9) using Lifescope version 2.0 (Life Technologies), and peaks were detected using the Genomics Workbench version 4.7.2 (CLC Bio). Parameters for peak mapping and detection are described for details below. Gene annotation of sequence peaks was performed through the BioMart website version 0.7 (http://www.biomart.org). The chromosomal distribution of the MBP-seq peaks was described using R version 1.4 (http://www.r-project.org). UCSC Genome Browser (http://genome.ucsc.edu)  was used to obtain data on the epigenetic markers and to display epigenetic features around the target genome locus.
The bioinformatics tool IPA version 1.0 (http://www.ingenuity.com) was used for in silico analysis of the MBP-seq data in the context of known functions and pathways using the Ingenuity Pathways Knowledge Base as a reference set, filtering for molecules and relationships associated with physiological and pathological processes of the liver. For the in silico functional analysis, a right-tailed Fisher's exact test was used to calculate the p-value determining the probability that the hepatotoxic function assigned to that data set was owing to chance alone.
The web-based human cancer microarray database Oncomine (https://www.oncomine.com) was used to analyze the mRNA expression of target genes associated with HCC identified in three studies [11-13]. Details of standardized normalization techniques and statistical calculations can be found on the Oncomine website (https://www.oncomine.com) . In brief, Student's t-test was performed to generate a p-value indicative of the significance of an observation. The lower the p-value, the more confidence in the difference between the groups. The Student's t-test statistic provided for the Oncomine visualizations reflects the magnitude of the difference between groups. Fold change is the magnitude of difference between the primary class and the other control classes, shown on a linear scale. An over-expression fold change is designated with a positive number.
The purified genomic DNA from the mouse livers was randomly fragmented into 50-350 bp lengths as described in the SOLiD 5500xl fragment library protocol. The fragmented DNA was then subjected to the MethylMiner methylated DNA enrichment kit, according to the manufacturer's protocol (Invitrogen). qPCR was then performed to amplify and quantify fragments representative of the methylated genome using the Thermal Cycle Dice TP800 (TaKaRa) and SYBR Premix Ex Taq (TaKaRa). The primer sequences and genome locus used for MBP-IP are summarized in Additional file 1.
qPCR was performed as described previously . Tissues were homogenized in 1 ml of Sepazol and total RNA was extracted according to the manufacturer's instructions (Nacalai Tesque). cDNA was synthesized from total RNA using RevatraAce reverse transcriptase (TOYOBO) and oligodT primer. qPCR was performed to amplify and quantify fragments representative of the indicated mRNA expression using a Thermal Cycle Dice TP800 (TaKaRa) and SYBR Premix Ex Taq (TaKaRa). Cyclophilin and Gapdh was used as the normalization control (Additional file 2). The primer sequences for qPCR are summarized in Additional file 1.
The DNA fragments encoding Spp1 enhancer region (chr5:104846058-104846328) was amplified by PCR. The Spp1 enhancer fragment was methylated by the DNA methylase Sss1 (2U enzyme/10ug DNA), and then cloned into a pGL3-basic reporter plasmid. Each plasmid harboring methylated or unmethylated fragment was directly transfected into Hepa1-6 cells by Lipofectamine 2000 (Invitrogen), according to the manufacturer's instruction. Twenty-four hours after transfection, luciferase assays were performed by Dual-Luciferase Reporter 1000 Assay System (Promega), according to the manufacturer's instruction. phRG-TK (Promega) was used as a reference plasmid to normalize transfection efficiency.
The purified genomic DNA from the mouse livers were denatured in 0.3M NaOH for 20min at 37°C. Then 3.6N sodium bisulfite and 10mM hydroquinone solution were added. Samples were incubated in 1min at 95°C then 12hr at 50°C. Salts were removed using the Wizard DNA Clean-Up System (Promega) and desulfonated in 0.3M NaOH at room temperature for 5 min. Then, the Spp1 enhancer region was amplified by PCR and cloned into pGEM-T Easy Vector System (Promega). The inserts were sequenced to identify the methylated and unmethylated sites. The primer sequences used for amplification are summarized in Additional file 1.
Sequence data of CCl4-treated and control samples from this study has been deposited to the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra) under accession no. SRA048978 and SRA048984, respectively.
To develop early-stage liver fibrosis, mice were administered 2 ml/kg CCl4 for 2 weeks (see Methods for details). To determine whether this treatment induced liver fibrosis, we used blood samples to investigate two enzymatic activities of well-known fibrosis-indicative serum parameters--Alt and Ast. We found that CCl4 treatment increased enzymatic activities of both Alt and Ast (Figure (Figure1A).1A). Furthermore, we found that well-known fibrotic markers, such as αSma, Col1a2, and Timp1, were up-regulated in the CCl4-treated livers (Figure (Figure1B).1B). To confirm histopathological change in the livers following CCl4 treatment, we used MT and Sirius-red staining, which are commonly used to detect collagen accumulation in liver tissue. Although MT staining did not reveal histological changes, Sirius-red staining revealed an increased distribution of collagen (Figure (Figure1C).1C). Given that MT staining is known to detect continuous ECM whereas Sirius-red staining specifically detects collagen , these results suggest that 2 weeks of CCl4 treatment induces fiber production, but not fibroblast growth. This suggests that treatment with CCl4 induced the development of early-stage liver fibrosis.
To investigate the effect of CCl4 treatment on DNA methylation status in early-stage liver fibrosis, we attempted to profile the genome-wide DNA methylation of the CCl4-treated liver tissues. The genomic DNA was fragmented, and the methylated DNA fragment was precipitated by MBP. The methylated DNA fraction was then subjected to high-throughput sequencing using a SOLiD 5500xl. These sequence reads were then mapped onto the mouse genome (NCBI37/mm9) using the MethylMiner methylated DNA enrichment kit and Lifescope. We obtained 7,612,236 reads from the CCl4 sample, and 24,584,122 reads from the control . To find the significant peak, these data were further analyzed using the ChIP-seq tool in Genomics Workbench with the mouse genome annotated using the BioMart website. The peak-finding algorithm included the following four steps: 1) Calculate the null distribution of the background sequencing signal; 2) Scan the mappings to identify candidate peaks with a higher read count than expected from the null distribution; 3) Merge overlapping candidate peaks; 4) Refine the set of candidate peaks based on the count and the spatial distribution of forward and backward reads within the peaks. The estimate for the null distribution of coverage and the calculation of the false discovery rate (FDR) were based on the window size and maximum FDR (%) parameters. The window size specifies the width of the window that is used to count reads during estimation of the null distribution as well as during subsequent scanning for candidate peaks. Maximum FDR indicates the maximum proportion of false positive peaks that are acceptable among the called peaks. In this study, when only MBP samples were used, each negative binomial distribution was fitted to the counts from the low coverage regions. This distribution was used as the null distribution to obtain the number of windows with a particular count of reads expected in the absence of significant binding. By comparing the number of windows with the specific count we expected to observe under the null distribution and the number we actually observed in our data, we can calculate FDR for a given read count and window size as the "fraction of windows with a read count expected under the null distribution/fraction of windows with the observed read count". In this study, we set window size and FDR to 300 bp and 1%, respectively. We identified 125 and 88 peaks in the CCl4-treated and control samples, and summarized all peaks with CpG annotation and several statistic values, including FDR, normalized difference, and Wilcoxon filter p-value, in Additional file 3.
The pie chart in Figure Figure2A2A indicates the methylated sites in each genomic element, such as the transcription regulatory region including the promoter and the 3' flanking region, the gene body including the exon and intron, and the intergenic region (See Figure Figure2A2A legend for detail definition of these terms). Methylated sites in the gene body, the transcription regulatory region, and the intergenic region, were assigned (23.0/12.0%, 10.2/9.0%, and 66.0/79.0% in CCl4-treated/control samples, respectively) (Figure (Figure2A).2A). We found that almost methylated sites are specific to each sample (Additional file 3). Therefore, we considered CCl4-specific and control-specific methylated sites as CCl4-induced "hyper" and "hypo"methylation sites, respectively. The chromosomal distribution of the MBP peaks revealed that the number of control-specific methylation peaks was greater than that of CCl4-specific peaks in most chromosomes (Figure 2B-D). These results indicate that the onset of CCl4-induced early-stage liver fibrosis might be associated with global hypomethylation.
Then, the functions of the genes with methylated sites, which are assigned to the gene body and the transcription regulatory region, are classified using in silico analysis software, IPA (See for details Methods for IPA analysis, and Discussion for the functional aspect of methylated sites within intergenic region). We found that cirrhosis, fibrosis, and HCC were identified only in the control sample (Figure (Figure3A).3A). Previous studies have indicated that the progression of liver fibrosis leads to cirrhosis and eventually to HCC, a primary hepatic cancer . Furthermore, it has been known that aberrant regulation of gene expression is at the basis of many diseases, such as cancer . These results suggest that aberrant DNA methylation causes dysfunction of gene transcription and may affect the onset of fibrosis and its progression to post-fibrotic diseases.
Using IPA, we then re-analyzed the annotated genes in the promoters, which is important for gene transcription. Intriguingly, liver fibrosis was detected as the predominant hepatotoxic function in the control sample (p-value = 1.06 × 10-3-6.37 × 10-3) (Figure (Figure3B).3B). Moreover, none of the liver diseases detected in control samples were detected in the CCl4-treated samples. IPA re-analysis revealed that this fibrotic function in the control sample was associated with one gene--Spp1.
In IPA, the p-value is calculated using a right-tailed Fisher's exact test to assess the probability that the association between the focal molecules in the experiment and a given function is owing to random chance. Smaller the p-value, lesser is the probability of random association. In general, p-values < 0.05 indicate a statistically significant, nonrandom association. Therefore, these results suggest that CCl4-induced hypomethylation of a regulatory region, such as the promoter or enhancer of Spp1, may result in the onset of liver fibrosis and its progression to post-fibrotic diseases such as cirrhosis and HCC through an increase in Spp1 expression.
The hypomethylation peak was located approximately 18 kbp upstream of the transcription start site (TSS) of Spp1 (Figure (Figure4A,4A, red). We placed this MBP-seq peak into the UCSC genome browser and found that the hypomethylation site upstream of Spp1 has several chromatin features identified from studies by other groups, including: 1) mono- and trimethylation of histone 3 lysine 4 in the liver (H3K4 me1 and H3K4 me3) (Figure (Figure4A,4A, orange) ; 2) the binding of RNA polymerase II and related acetyltransferace p300 in the liver and in the MEL leukemia cell line (Figure (Figure4A,4A, light blue) ; and 3) it is a DNase I hotspot in the liver (Figure (Figure4A,4A, green) . Previous studies have demonstrated that these epigenetic features are observed in the enhancer region [20,23-25], and that hypomethylation of an enhancer predominantly induces mRNA expression [26-28]. Thus, these epigenetic annotations lead to the hypothesis that this hypomethylated site may function as an enhancer that regulates Spp1 expression.
To confirm this hypothesis, we performed biochemical assays. Considering with the technical limitation of MBD-IP assay, a bisulfite modification assay was carried out to verify at a single-locus-based resolution (See Methods for details). Complementary using these assays, we confirmed that the site annotated by the sequence database was actually hypomethylated by CCl4 treatment (Figure (Figure4B4B and and4C).4C). We then examined whether the methylation on the upstream of Spp1 TSS will affect transcriptional activity. Luciferase assay were performed using in vitro methylated upstream of Spp1 TSS. This in-vitro-methylation assay revealed that methylation on upstream of Spp1 TSS down-regulated its mRNA levels (Figure (Figure5A).5A). We then performed a qPCR assay and revealed that Spp1 expression was increased by treatment with CCl4 (Figure (Figure5B)5B) and 5-dAza-C (Figure (Figure5C),5C), both of which induce hypomethylation upstream of Spp1 TSS. These results allow us to conclude that the upstream of Spp1 TSS functions as enhancer of Spp1. These results suggest that CCl4 treatment might up-regulate Spp1 expression in early-stage liver fibrosis through hypomethylation of the Spp1 enhancer.
In silico analysis and functional validation of hypomethylation suggest that Spp1 expression might cause fibrosis-induced liver diseases, such as those shown in Figures Figures3A3A and and3B;3B; this hypothesis is in agreement with previous studies . Thus, to obtain functional insight into Spp1 in the end-stage liver disease HCC, we analyzed the mRNA level using the human cancer microarray database Oncomine. Analysis of studies by Mas et al.  , Chen et al.  , and Wurmbach et al.  revealed that Spp1 expression was higher in HCC than in normal liver tissues (Figure (Figure6),6), suggesting that Spp1 might play a crucial role in HCC development as well as in the onset of fibrosis.
The DNA methylation status of well-known genes associated with fibrosis progression, such as Rasal1, Fli1, and Thy1, has been reported to increase along with fibrosis progression, which induces proliferation of fibroblasts and the production of collagen [9,30,31]. Furthermore, 5'-Aza, an inhibitor of DNA methylation, reportedly attenuates the progression of renal and liver fibrosis in vivo and in vitro [7,8]. These reports indicate the importance of hypermethylation of the genome in the progression of liver fibrosis.
In contrast to these previous reports, our analysis revealed that DNA methylation status was significantly reduced in CCl4-induced early-stage liver fibrosis. Detailed analyses revealed that in early-stage liver fibrosis, the Spp1 enhancer was hypomethylated and Spp1 expression was up-regulated. Previous findings from genetically engineered Spp1-knockout mice have shown that Spp1 plays an important role in both progression and reduction of liver injury and fibrosis [32,33]. Although the precise mechanism underlying the role of Spp1 in fibrosis remains unknown, Spp1 is a pivotal cytokine/chemokine generated by the Kupffer cells in response to liver damage  that induces inflammation , which is a contributing factor in liver fibrosis. Our results indicate that an epigenetic alteration, DNA hypomethylation of the Spp1 enhancer, may precede the up-regulation of Spp1 expression and induce the onset of CCl4-induced early-stage liver fibrosis. Spp1 is also a known enhancer of tumor development and metastasis . Using the Oncomine microarray database, we demonstrated that Spp1 expression level is increased in HCC. It has been documented in primary gastric cancers, that major phenotypic change in cancer-associated myofibroblasts is a global reduction in DNA methylation . This may also be indicated in liver, the importance of hypomethylation of the Spp1 enhancer in the progression of HCC.
Unlike Spp1, we found that almost DNA hypomethylation induced in early fibrosis was assigned in intergenic regions (Figure (Figure2A).2A). Recently, it has been reported that DNA hypomethylation causes genomic instability and alteration of gene transcription in several human cancers (i.e. colorectal and prostatic adenocarcinoma, breast cancer, intestinal type-gastric carcinoma, and HCC) [37,38]. These results indicate that DNA hypomethylation in intergenic regions may trigger the progression of cancer through genomic instability of cancer-related genes in addition to transcriptional regulation of those genes in the onset of liver fibrosis. On the other hand, 5'-Aza treatment successfully reduces cancer in mammals, involving human, suggesting that carcinogenesis results from DNA hypermethylation as well as DNA hypomethylation [38-40]. Therefore, how these distinct DNA modifications commonly progress cancer is most important study for epigenetic effects on diseases in the future.
Although hypermethylation occurs during the progression of liver fibrosis, our results indicate the importance of hypomethylation in the onset of liver fibrosis. According to our results as well as other reports, it appears that DNA methylation status may change from hypo- to hypermethylation during the progression of liver fibrosis. Thus, hypomethylation in early-stage liver fibrosis may contribute to the onset and/or development of HCC.
α-Sma: α-Smooth muscle actin; Alt: Alanine aminotransferase; Ast: Aspartate aminotransferase; CCl4: Carbon tetrachloride; Col1a2: Collagen, type I, alpha 2; ECM: Extracellular matrix; Fli1: Friend leukemia integration 1; HCC: Hepatocellular carcinoma; Rasal1: RAS protein activator like 1; Spp1: Secreted phosphoprotein 1; Thy1: Thymocyte differentiation antigen 1; Timp1: Tissue inhibitor of metalloproteinase 1
The authors declare that they have no competing interests.
YK and NI performed the animal experiments and biochemical analyses. TW performed the high-throughput sequencing and bioinformatic analysis. WO and CY supported these analyses. All authors wrote, read, and approved the final manuscript.
The pre-publication history for this paper can be accessed here:
Table of primer sets used in this study. These qPCR primer sets were used in Figures 1B, 5B and C.
mRNA levels of reference genes. Cyclophilin were used to normalize genes in Figure 1B and 5B. Gapdh were used to normalize genes in Figure 5C.
Statistics and annotation of MBP-seq peaks. This table was generated by Genomics Workbench, and used in Figures 2, 3, and 4A.