EZH2 Alleviates Antituberculosis Drug-Induced Liver Injury in Mice by Reducing H3K27 Trimethylation at the Nrf2 Promoter


 BackgroundAnti-tuberculosis drug-induced liver injury (ADLI) limits the treatment of tuberculosis. The mechanisms underlying ADLI are unclear and there are no effective preventative measures to avoid this complication. MethodsIn this stuy, the protein contents of EZH2, Nrf2, NQO1 and HO-1 were detected by ELISA kit, while those of EZH2 and Nrf2 were determined by Western blot. The Chip experiment was used to detect the level of H3K27me3 in the Nrf2 promoter region.The liver were analyzed histopathologically in vivo using hematoxylin and eosin staining.ResultsHere we developed a murine model of ADLI that recapitulates liver injury in the human disease. Using this model, we investigated the potential involvement of the enhancer of zeste homolog 2 methyltransferase (EZH2), a histone methyltransferase which inhibits the transcriptional activation of the Nrf2-ARE oxidative stress pathway. Compared to controls, mice livers with ADLI showed decreased expression of EZH2 together with reduced H3K27me3 marks in the Nrf2 promoter. This was accompanied by increased expression of Nrf2 and its target genes NQO1 and HO-1. Liver injury in the mice with ADLI could be alleviated to an extent by in vivo delivery of siRNAs targeting EZH2, which further downregulated EZH2 expression and H3K27me3 levels in the Nrf2 promoter along with accompanying increases in Nrf2, NQO1 and HO-1 expression. ConclusionsTherefore, inhibiting EZH2 likely reduced liver damage in ADLI by enhancing this key anti-oxidative stress pathway. Our results establish a role for EZH2 in a mouse model of ADLI and furthermore provides valuable mechanistic insights into the development of ADLI pathology.


Introduction
A prominent obstacle which limits the treatment of tuberculosis has become anti-tuberculosis druginduced liver injury (ADLI) which is associated with commonly used treatments including isoniazid, rifampicin and pyrazinamide [1]. The etiology of ADLI is diverse and its pathogenesis is complex, but both direct and speci c hepatotoxicity of drugs and their metabolites are accompanied by in ammation and oxidative stress. However, the mechanisms underlying ADLI are unclear although some progress has been made in recent years in the area of epigenetic regulation.
Histone modi cations represent an important aspect of gene regulation. Here the reversible modi cation of lysine and arginine residues by methylation play key roles in regulating various processes including cell cycle, genomic stability, and nuclear structure. [2][3][4]. In ADLI, it has been shown that changes in histone deacetylase 1 (HDAC1) and P300 modulate endoplasmic reticulum stress (ERS) and affect isoniazid-induced hepatocellular injury [5]. SIRT1 which also plays a role in the deacetylation of histones can reduce the in ammatory response of isoniazid-induced liver-injured cells by reducing histone acetylation in the IL-6 promoter region [6]. It has been also been shown that Enhancer of zeste homolog 2 (EZH2) can regulate the expression of oxidative stress-related proteins in lung cancer through H3K27me3 modi cation. EZH2 is a speci c histone methyltransferase, which modi es the 27th lysine of histone H3 by trimethylation (H3K27me3), thus changing the chromosome structure, eventually leading to gene suppression or silencing [7]. However, up until now, there are no reports linking EZH2 expression and function with ADLI.
One of the prominent target genes repressed by EZH2 is the nuclear factor E2-related factor 2 (Nrf2), a key transcription factor regulating oxidative damage of cells [8]. Nrf2 together with antioxidant response elements (ARE) are regarded as the core regulatory pathway of the endogenous antioxidant system. When cells are stimulated by oxidative damage, Nrf2 translocates into the nucleus and binds to AREs to exert powerful anti-in ammatory, antioxidant and cytoprotective effects by regulating its downstream target genes, NADPH quinine oxidoreductase 1 (NQO1) and heme oxygenase-1 (HO-1) [9]. Notably, when liver injury occurs in mice, Nrf2 translocates to the cell nucleus to upregulate NQO1 and HO-1 expression.
Similarly, after curcumin treatment, NQO1 and HO-1 expressions in liver tissue of mice were upregulated, and liver tissue injury and splenomegaly were alleviated [10]. Therefore, we speculate that EZH2, Nrf2 and their downstream factors are involved in the process of ADLI. Moreover, the Nrf2 oxidative stress pathway may be a potential target for the prevention and treatment of ADLI.
In this study, we established a mouse model of ADLI by simulating the human antituberculosis treatment regimen which recapitulated the pathological changes in liver commonly seen in the human disease. Using this model, we investigated the regulation of EZH2 during ADLI together with the effects of EZH2 inhibition using siRNA treatment in vivo.

Animal model
Thirty-two 6-week-old SPF Kunming mice (Beijing Weitong Lihua Experimental Animal Technology Co., Ltd.), half male and half female, weighing 18-22g, male and female segregation, cagewere raised in the SPF Animal Laboratory of the Animal Experiment Center, North China University of Science and Technology. After 1 week of adaptive feeding, the mice were randomly divided into 4 groups of 8 mice. The control group (Control) was given a solvent stomach gavage, and the model group (ADLI) was given isoniazid 90mg/Kg.d+rifampicin 135mg/Kg.d+pyrazine Amide 315mg/Kg.d mixed gavage for 4 consecutive days. Based on the ALDI model group, the inhibitor group (Si-EZH2) was injected with 200nmol/Kg EZH2-siRNA (CAA CAC CCA ACA CAU AUA ATT, UUA UAU GUG UUG GGU GUU GTT) into the tail vein while the negative control group (Si-Ctrl) was injected with 200nmol/Kg EZH2-NC-siRNA (UUC UCC GAA CGU GUC ACG UTT, ACG UGA CAC GUU CGG AGA ATT) based on the control group. On the 5th day after treatment, blood was collected and serum separated to detect ALT and AST levels using a chemical method. Mice were then sacri ced by cervical dislocation method and the liver tissue was taken. Part of the liver tissue specimen was xed with 10% formaldehyde solution to observe pathological changes in the liver tissue.

RT-qPCR
The total RNA of liver tissue was extracted by the Trizol. Selecting RNA samples with the OD value detected by the microplate reader between 1.8 and 2.0. The total system of the reverse transcription system is 20 µl, of which Mix is 4 µl and the RNA mass is 1,000 ng, then the RNA volume is 1,000/RNA concentration, and the remaining system is supplemented to 20 µl with DEPC water. RT-qPCR was performed to detect EZH2, NRF2, NQO1, HO-1 mRNA levels using a standard protocol. The reaction system is 20 µl, Rox 0.4 µl, upstream and downstream primers 0.4 µl each, dH2O 6.8 µl, and cDNA 2 µl.

ELISA assays and Western blotting
Mouse liver tissue was lysed using RIPA buffer containing PMSF to extract protein. An ELISA kit was used to detect EZH2, Nrf2, NQO1 and HO-1 protein levels according to the manufacturer's instructions while Western blotting was used to detect EZH2 and Nrf2 proteins. The protein samples were separated by SDS-PAGE and transferred to PVDF membrane. The membrane was placed in a 5% skimmed milk powder shaker and sealed at room temperature. After being washed, the membrane was incubated overnight with EZH2 (1: 2,000) and Nrf2 (1: 2,000) primary antibodies 4. After the membrane was washed the next day, the secondary antibodies (1: 5,000) were incubated on the shaker at room temperature. They were developed and quanti ed by ECL color reagent in a gel imaging system.

ChIP assays
Chip assays were performed on mouse liver tissue using the ChIP-kit (Merck Millipore) according to the manufacturer's instructions. Chromatin was cross-linked with 1% formaldehyde and immunoprecipitations performed with either anti-EZH2 or anti-H3K27me3 antibodies, and the level of H3K27me3 in the Nrf2 promoter region was determined by qPCR. The primer sequences used are as follows F:5 '-CAG TGC TCC TAT GCG TGA-3', R:5 '-TCT GGG CGG CGA CTT TAT-3'.

Statistical analysis
Statistical analysis was performed with SPSS software (22.0; SPSS). In brief, the values are expressed as the mean ± standard deviation (SD). SNK test was used for comparisons between groups. P < 0.05 was considered statistically signi cant.

Establishment of the ALDI mouse model
To examine if the treatment regime with isoniazid, rifampicin and pyrazine resulted in characteristic features of ALDI, we compared the results of liver function tests between the control and treated mice.
Indeed, there were signi cant increases serum levels of ALT and AST treated mice in comparison with the Control group (Fig. 1). Moreover, the pathological analysis of the liver in the ADLI group, showed obvious disordered arrangement of the liver cells, loose cytoplasm, destruction of hepatic lobules, and darker staining of some nuclei (Fig. 2). This contrasted the pathology of the liver in the control group where cells were normal in morphology, radiating from the central vein, with complete hepatic lobules and clear intercellular boundaries. Together these data con rmed that the ADLI mouse model reproduced two of the main characteristics of ALDI observed in the human disease.
We then considered the effect of the siRNA treatments. Notably, the liver function of ADLI mice was signi cantly improved after siRNA inhibition of EZH2 gene expression (P < 0.05, for ALT and AST), while the control siRNA inhibitor had no effect on liver function parameters (Fig. 1). Regarding to the liver histopathological results, a disordered cell arrangement and hepatic lobular structure were also found in si-EZH2 group but the extent was less than that in the ADLI group (Fig. 2). The liver pathology of si-Cntrl group appeared normal and similar to the Control group. Therefore, these data indicated that inhibition of EZH2 expression can improve the degree of liver injury to a certain extent in the ALDI model.

EZH2 is involved in ADLI
It was important to establish whether the levels of EZH2 changed in liver hepatocytes in response to ALDI and to con rm the effects of the si-EZH2 treatment. Examination of the EZH2 levels in the liver tissues of mice in the respective groups showed that EZH2 mRNA and protein levels in the ADLI group were decreased compared with the control group (Fig. 3). Furthermore, assessing the effects of the siRNA treatments showed that the levels of EZH2 mRNA and protein in the si-EZH2 group was reduced compared with the ADLI group (P < 0.05), whereas the si-Cntrl group levels of EZH2 were similar to the Control group. Collectively these data suggested that EZH2 was involved in the process of ADLI and also con rmed that the si-EZH2 treatment can further suppress EZH2 levels in the liver.
Inhibition of EZH2 up-regulates the expression of Nrf2 pathway to reduce the degree of liver injury As per the Introduction, EZH2 is a H3K27me3 methyltransferase that modi es and represses Nrf2 gene expression through association with its promoter. Consistent with the changes in EZH2, H3K27me3 levels in the Nrf2 promoter region were prominently down-regulated in ADLI liver extracts compared with the Control group (Fig. 4). The addition of si-EZH2 treatment further decreased these levels. These nding establish that inhibition of EZH2 reduces the H3K27Me3 levels in the Nrf2 promoter region and suggests up-regulation of the Nrf2 pathway may help alleviate the degree of liver injury.
EZH2 is correlated with the expression of NRF2 and its downstream effectors in ADLI mouse liver tissue Nrf2 has shown to be a key factor in regulating oxidative stress and its expression is subject to epigenetic suppression by EZH2. In accordance with the changes in EZH2 expression, the levels of Nrf2 were signi cantly upregulated in ADLI mice. In addition, NQO-1 and HO-1, genes regulated by Nrf2, were also up-regulated in ADLI (Fig. 5). Together this indicated activation of the Nrf2 pathway during the process of ADLI in the mouse model. Correlation analysis showed that EZH2 mRNA levels were negatively correlated with Nrf2, NQO-1 and HO-1 mRNA levels (r= -0.574, -0.461, -0.537, all P < 0.05), EZH2 protein levels were also negatively correlated with Nrf2, NQO-1 and HO-1 protein levels (r= -0.624, -0.425, -0.549, all P < 0.05). Thus, EZH2 was correlated with Nrf2 and its downstream target genes. Furthermore, as anticipated, inhibiting the expression of EZH2 in ADLI mice resulted in up-regulation of Nrf2, NQO-1 and HO-1 at both mRNA and protein levels (all P < 0.05).

Discussion
ADLI is the most common and harmful adverse drug reaction since it can lead to liver failure and endanger life [11]. Therefore, cessation of treatment is essential for some patients which can greatly affect the management of their disease. On this basis, understanding and preventing ADLI is essential to overcome this limitation of current tuberculosis treatments. Clinicians usually stop tuberculosis treatment when liver damage occurs during tuberculosis treatment and give liver protection drugs at the same time. But it is worth noting that all drugs are broken down, transformed, metabolized and detoxi ed by the liver, and liver protection drugs are no exception. Long-term excessive use of hepatoprotective drugs will undoubtedly increase the burden on the liver. At the same time, the accumulation of drugs in the body for a long time will also cause liver damage. Therefore, it is necessary for society to nd new treatments to protect the liver. Research progress in the eld of epigenetics together with the use of drugs and molecular technologies to modify epigenetics provides new opportunities for treating many diseases. In this study we showed that EZH2 expression levels were downregulated in a mouse liver tissue in a mouse model of ADLI. Inhibiting the expression of EZH2 levels by RNAi reduced the H3K27me3 mark in the Nrf2 promoter, leading to improved levels of anti-oxidative stress, and ultimately achieving the goal of reducing ADLI.
EZH2 is the catalytic subunit of polycomb-repressive complex 2 (PRC2) which promotes transcriptional silencing by modi cation of histones and regulation of chromatin structure. EZH2 is known to be differentially expressed in various disease states [12]. For example, EZH2 expression in the peripheral blood of patients with severe acute stage dengue fever is down-regulated compared with non-severe patients, and its levels can be used as a biomarker to predict disease severity at an early stage [13]. Downregulation of EZH2 can also promote steatosis and upregulate in ammatory genes to accelerating the development of nonalcoholic fatty liver disease [14]. Our results in the mouse model of ADLI showed that the expression of EZH2 mRNA and protein were also down-regulated in liver. Since the combination of the rst-line anti-tuberculosis drugs causes relevant epigenetic changes including the decreased expression of EZH2, we concluded that EZH2 participates in ADLI and may be closely related to liver injury.
Under homeostatic circumstances, the Nrf2 pathway is suppressed by sequestration of Nrf2 with Keap1. However, stress stimuli such as electronophilic substances or oxidative stress factors, cause Nrf2 to dissociate from Keap1 and to translocation from the cytoplasm to nucleus. Nuclear Nrf2 binds to gene promoters containing AREs, resulting in transcriptional activation of downstream antioxidant stress factors such as NQO1 and HO-1 [15]. In this manner, the Nrf2-ARE pathway acts cytoprotectively to improve the body's ability to resist harmful chemical toxicity. Therefore, the Nrf2-ARE pathway may be a potential target for the treatment of ADLI.
Previous studies have shown that the curcumin derivative F10 increased Nrf2 expression levels together with related phase II detoxi cation genes in prostate cancer cells [16]. Here F10 effectively reduced H3K27me3 levels at the Nrf2 promoter to activate the Nrf2-ARE pathway. Similarly, in prostate cancer, Yang et al. found that Nrf2 promoter histone trimethylation was also reduced after treatment with coral alkyd. Instructively, the expression levels of Nrf2, NQO1 and HO-1 were all increased in prostate cancer cells, but the same phenomenon did not occur when the Nrf2 gene was knocked out [17]. In our study, speci c siRNAs were introduced into mice to inhibit the expression level of EZH2 and observe the effects on the Nrf2-ARE pathway. The results showed that the Nrf2 pathway was upregulated in mouse liver tissue in the ADLI group. After inhibiting EZH2, the Nrf2 pathway was further up-regulated, and the degree of liver damage was partially alleviated. These results suggested that accentuating the activation of the Nrf2 pathway by inhibiting EZH2 in mice could improve liver injury caused by anti-tuberculosis drugs.
Animal models have many advantages but there are some limitations to be considered. Firstly, we need to address the implications of the study in humans. Studies akin to the ADLI animal model are not practical in humans but further follow-up in vitro studies using relevant human cell line models could be readily employed. Here both gene knock-down or over-expression studies would be helpful to clarify the mechanism of ADLI disclosed involving EZH2 regulation. Additionally, we altered the expression level of EZH2 gene in mice using RNAi technology but our ndings could also be extended to include treatments that not only inhibit EZH2 expression but also up-regulate EZH2.

Conclusion
In summary, our study revealed that EZH2 liver expression is down-regulated in a mouse model of ADLI. Changes in the Nrf2 pathway were observed in ADLI consistent with the epigenetic regulation of Nrf2 and its cytoprotective gene targets. Furthermore, inhibition of EZH2 expression improved the degree of liver damage in the ADLI model, suggesting that the effects of EZH2 on ADLI are regulated through H3K27me3-dependent modi cation of the Nrf2 promoter region and upregulation of the Nrf2 oxidative stress pathway. This study provides a theoretical basis for understanding the pathogenesis of ADLI and provides a possible therapeutic target for the treatment of ADLI, which has potential clinical signi cance.

Abbreviations ADLI
Anti-tuberculosis drug-induced liver injury Availability of data and materials The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.