Liver-specific knockout of histone methyltransferase G9a impairs liver maturation and dysregulates inflammatory, cytoprotective, and drug-processing genes

Abstract

1. Methyltransferase G9a is essential for a key gene silencing mark, histone H3 dimethylation at lysine-9 (H3K9me2). Hepatic G9a expression is down-regulated by xenobiotics and diabetes. However, little is known about the role of G9a in liver. Thus, we generated mice with liver-specific knockout (Liv-KO) of G9a.

2. Adult G9a Liv-KO mice had marked loss of H3K9me2 proteins in liver, without overt liver injury or infiltration of inflammatory cells. However, G9a-null livers had ectopic induction of certain genes normally expressed in neural and immune systems. Additionally, G9a-null livers had moderate down-regulation of cytoprotective genes, markedly altered expression of certain important drug-processing genes, elevated endogenous reactive oxygen species, induction of ER stress marker Chop, but decreased glutathione and nuclear Nrf2. microRNA-383, a negative regulator of the PI3K/Akt pathway, was strongly induced in G9a Liv-KO mice. After LPS treatment, G9a Liv-KO mice had aggravated lipid peroxidation and proinflammatory response.

3. Taken together, the present study demonstrates that G9a regulates liver maturation by silencing neural and proinflammatory genes but maintaining/activating cytoprotective and drug-processing genes, in which the G9a/miR-383/PI3K/Akt/Nrf2(Chop) pathways may play important roles. G9a deficiency due to genetic polymorphism and/or environmental exposure may alter xenobiotic metabolism and aggravate inflammation and liver dysfunction.

Introduction

Epigenetic programming plays a dominant role in liver development. Various epigenetic enzymes are expressed at much higher levels in developing liver than the adult liver (Lu et al. 2012). Methylation of histone H3 at lysine-9 (H3K9me) is closely associated with inactive gene in the highly condensed heterochromatin (Shankar et al. 2013). Euchromatin is a lightly packed form of gene-rich chromatin, and is often under active transcription. Within euchromatin, H3K9me2 localizes specifically to silent domains and is dependent on the enzymatic activity of G9a (Rice et al. 2003). H3K9me2 candidate genes are enriched in many tightly controlled signaling and cell-type specific pathways (Miao et al. 2008). Deletion of histone methyltransferase G9a in mice markedly decreases H3K9me2 in euchromatin and causes embryonic lethality (Tachibana et al. 2007; Tachibana et al. 2002). A genome-wide study demonstrates that differentiated tissues have large H3K9me2 regions (Wen et al. 2009). These regions are highly conserved between humans and mice, and are differentiation-specific, covering only ~4% of genome in undifferentiated mouse embryonic stem (ES) cells, but ~31% in differentiated ES cells, and ~46% in the liver. These H3K9me2 modifications require G9a and are inversely related to mRNA expression (Wen et al. 2009). In addition to embryonic development, G9a is also important in the differentiation of brain, intestine, adipose tissues, and heart (Antignano et al. 2014; Maze et al. 2010; Papait et al. 2017; Schaefer et al. 2009; Wang et al. 2013).

Epigenetic modification is a key link between environmental exposure and human diseases (Mann 2014; Pogribny and Rusyn 2013; Stein 2012). DNA damage triggers degradation of G9a (Takahashi et al. 2012) in senescent cells, whereas G9a is recruited to chromatin in cancer cells after double-strand breaks to promote DNA damage repair and cell survival (Yang et al. 2017). The expression of G9a is altered by certain clinically important xenobiotics, including induction by arsenic, nickel, and chromium (Chervona et al. 2012) whereas down-regulation by nicotine (Chase and Sharma 2013), cocaine (Maze et al. 2010), and ethanol (Esfandiari et al. 2010). Additionally, via ejecting zinc from the zinc finger domain of G9a, clinically used ebselen, disulfiram, and cisplatin strongly inhibit G9a (Lenstra et al. 2018). Moreover, a common G9a SNP rs652888 is strongly associated with chronic hepatitis B virus (HBV) infection in humans (Jiang et al. 2015; Kim et al. 2013b). Thus, G9a may be in the forefront of modulating gene expression in response to environmental stressors.

In addition to histone methylation, G9a facilitates DNA methylation via interacting with DNA methyltransferase and DNA ligase 1 (Epsztejn-Litman et al. 2008; Esteve et al. 2006; Ferry et al. 2017). G9a also methylates important non-histone proteins such as p53 and C/EBPβ, inhibiting their transactivation activity (Huang et al. 2010; Pless et al. 2008). In addition to silencing transcription via its C-terminal SET domain, G9a can act as a co-activator of nuclear receptors, such as glucocorticoid receptor (GR) and androgen receptor (AR), to activate transcription via its N-terminal domain (Poulard et al. 2017; Shankar et al. 2013). Via these diverse mechanisms, G9a is important in various critical biological processes such as embryonic development, genomic imprint, lymphocyte and neuronal function, adipogenesis, tumorigenesis, and metastasis (Antignano et al. 2014; Maze et al. 2010; Schaefer et al. 2009; Shankar et al. 2013; Wang et al. 2013). However, little is known about the role of G9a in liver, the metabolic center.

Induction of G9a and H3K9me2 correlated with gene silencing in livers of mice deficient of a master regulator, hepatocyte nuclear factor 4α (Zhang et al. 2014). G9a is overexpressed in many human cancers (Shankar et al. 2013), and G9a contributes to survival of cancer cells by regulating key survival pathways, e.g. production of reactive oxygen species (ROS) and autophagic cell death (Kim et al. 2013a), serine-glycine synthesis (Ding et al. 2013), and angiogenesis (Ueda et al. 2014). Particularly, overexpression of G9a mediates the silencing of p21, a key tumor-suppressor (Duan et al. 2005; Kim et al. 2009; Nishio and Walsh 2004). G9a is overexpressed in liver cancer and promote liver cancer development in humans (Wei et al. 2017; Yokoyama et al. 2017).

The purpose of the present study was to determine the role of G9a in regulating hepatic gene expression, liver development, and liver pathophysiology. Results from our study of mice with liver-specific knockout (Liv-KO) of G9a showed that G9a deficiency in liver caused marked changes in hepatic epigenome and transcriptome, manifested by ectopic induction of certain genes normally expressed in neural and immune systems, but down-regulation of many cytoprotective genes and certain important drug-processing genes (DPGs). Moreover, when challenged with lipopolysaccharides (LPS), G9a Liv-KO mice had aggravated lipid peroxidation and proinflammatory response.

Materials and Methods

Generation and treatment of G9a Liv-KO mice

G9a Liv-KO mice (G9a fl/fl, Alb-cre/+) and wild-type littermates (G9a fl/fl, Alb-cre/-) were generated by crossing G9a floxed mice (Tachibana et al. 2007) with Alb-cre mice (Stock # 003574, Jackson Laboratory). Mice were fed rodent chow and allowed water and feed ad libitum. Liver and blood samples were collected from adult (10-week old) male G9a Liv-KO mice and the wild-type littermates (N=6 per group). Liver tissues were snap frozen in liquid nitrogen upon collection and stored at −80 ºC until use. For LPS study, adult (10-week old) male G9a Liv-KO mice and the wild-type littermates (N=6 per group) were sacrificed to collect liver and blood samples 16 h after ip injection of LPS (L3012, Sigma) 5 mg/kg (in saline). Core body temperature (rectal temperature) was measured 16 h after LPS treatment before tissue collection using a digital thermometer. To prepare serum samples, the clotted blood samples were centrifuged at 8000 rpm for 10 min and the resultant supernatants were stored at −80 ºC until use. All animals received humane care and all animal procedures in this study were approved by the Institutional Animal Care and Use Committee (IACUC) of the SUNY Upstate Medical University.

Western blot quantification of liver proteins

Liver nuclear extracts were prepared with a nuclear extract kit (Marligen Biosciences, Inc., Rockville, MD). Liver lysates were prepared by homogenization of liver samples with RIPA buffer. Proteins in the liver nuclear extracts or lysates were resolved in sodium dodecyl sulphate-polyacrylamide gel electrophoresis. Western blot quantification of G9a, nuclear factor, erythroid 2-like 2 (Nrf2), and histones in liver nuclear extracts was carried out with the primary antibodies as follows: G9a (#3306), H3K9me2 (D85B4), and H3K27me3 (C36B11) from Cell Signaling Technologies (Danvers, MA, USA), Nrf2 (ABE413), H3K9me1 (17–680), and H3K9me3 (05–1242) from Millipore (Billerica, MA, USA); Glyceraldehyde 3-phosphate dehydrogenase (Gapdh) (HPA040067) from Sigma-Aldrich (St. Louis, MO). Western blot of sequestosome 1 (Sqstm1/p62) and microtubule-associated protein 1 light chain 3 β (LC3B) in liver cytosol was conducted using the primary antibody against p62 (#5114) and LC3B (#2775) from Cell Signaling. Western blot quantification of glutathione S-transferase p1 (Gstp1) and lymphocyte antigen 6 complex, locus A (Ly6a) in liver lysates was carried out with the primary antibody against Gstp1 (#H00002950-D01P, Abnova, Taipei, Taiwan) and Ly6a (#ab109211, Abcam, Cambridge, MA). Primary antibodies were revealed with HRP-conjugated secondary antibodies (anti-rabbit IgG (W4011) or anti-goat IgG (V805A) from Promega) and ECL Western Blotting Substrate (W1015, Promega). ChemiDoc™ XRS+ System (Bio-Rad) and Image Lab 4.0 software (Bio-Rad) were used to capture signals and determine signal intensities.

Cytochrome P450 3A (Cyp3a) activity assay

Liver homogenates were prepared in 25 mM Tris-HCl (pH 7.4) containing 250 mM sucrose. After centrifugation of homogenates at 9,000g for 20 min, 200 μg of supernatant proteins (S9 fraction) were used to determine Cyp3a activity using P450-Glo™ Assay for CYP3A4 (V8901) and NADPH regeneration system (V9510) from Promega (Madison, WI). Mouse Cyp3a11 and human CYP3A4 share similar substrate specificity, and the P450-Glo™ Assay for CYP3A4 (V8901) has been used to determine mouse Cyp3a activity (Lee et al. 2013). CYP3A activity was described as the luminescence of the D-luciferin generated, with values of wild-type liver group set as 100%.

Determination of lipids in mouse liver and serum

Lipids from frozen liver tissue were prepared as described previously (Lu et al. 2011). The lipid pellets were dissolved in a mixture of 270 μl of isopropanol and 30 μl of Triton X-100. Triglycerides (TG) and total cholesterol (CHO) in liver and serum were determined using commercial triglyceride and cholesterol analytical kits with standards (Pointe Scientific, Canton, MI).

Quantification of reduced glutathione (GSH), endogenous reactive oxygen species (ROS), and lipid peroxidation in mouse liver

Liver homogenates were used to quantify GSH by Ellman’s reagent and a GSH standard curve (Ellman 1959). Hepatic levels of endogenous ROS were determined using 2’,7’-dichlorofluorescin diacetate as the fluorogenic probe and a standard curve of 2’,7’-dichlorofluorescein (Yoshimi et al. 2011). Lipid peroxidation was determined by quantifying thiobarbituric acid reactive substances (TBARS) using a standard curve of 1,1,3,3-Tetramethoxypropane.

Histopathological staining and analysis

Liver samples were preserved in neutral buffered formalin (10%) before use. Formalin-fixed tissues were embedded in paraffin, sectioned (5 μm), and stained with hematoxylin and eosin (H&E). H&E stained liver sections were evaluated by light microscopy at 200 X magnification for evidence of hepatocyte apoptosis, inflammation, and fibrosis using a Leica DMI3000 B Inverted Light Microscope System with Leica DFC450 C Digital Microscope Camera (Leica Microsystems, Inc.).

Generation of mammalian expression vectors and luciferase reporter vectors

A mammalian expression vector for C/EBP homologous protein (Chop), pcDNA3-mChop was generated by PCR cloning the Chop cDNA (NM_007837.3) into the HindIII/XbaI site of pcDNA3 using a cDNA library from mouse liver and primer pairs of TAAAAAAAGCTTAGTCATGGCAGCTGAGTCCCT (forward) and ATAAATTCTAGATTCATGC TTGGTGCAGGCTGA (reverse). The correctness of pcDNA3-mChop was verified by sequencing. The 2.2 kb mouse Cyp3a11 promoter (−2036~+165) was PCR cloned into the KpnI/NheI site of pGL3-Basic reporter vector to generate pGL3-Cyp3a11 Pro vector using mouse tail DNA as the template and primer pairs of TTGATTGGTACCTGAGGTGCTGGT TTCTCGACT (forward) and TATTAAGCTAGCACTTCTCTGTGTTCTCCCTAC (reverse). The pGL3- Sult1c2 reporter vector was generated by ligating a 82 bp fragment TTGCTCCTTCTAATTCGTTGAAGAATTGAGTTGGGATTTTGATGGGGATTGCATTGAATCTGTAGATTGCTTTTGGCAAGAT from mouse Sult1c2 promoter that contains the core element bound by CHOP-C/EBP heterodimers (underscored) (Ubeda et al. 1996) into the KpnI/XhoI site of pGL3-basic vector.

Cellular transfection and dual-luciferase assay

Human embryonic kidney 293 (HEK293, ATCC) cells were maintained in EMEM with 5% FBS. Cells were added to 96-well plates or 6-well plates and grown to ~80% confluence. To determine the role of induction of Chop in the regulation of Cyp3a11 and Sult1c2, plasmid DNA including pGL3 reporter vector, the pRL-CMV Renilla luciferase (as control for transfection efficiency), pcDNA3-HNF4A2 vector, pCMV-C/EBPα/β (gift from Dr. Magnus Nord, Karolinska Institute), and/or pcDNA3-Chop were complexed with Lipofectamine 2000 (Life Technologies, Grand Island, NY) and applied to individual wells, according to the manufacturer’s protocol. For dual-luciferase assay, transfected cells in the 96-well plates were lysed with passive lysis buffer (Promega) 24 h after transfection. Promoter activities of cell lysates were quantified by Dual-Glo™ luciferase assay (Promega) with the control values of pGL3 reporter versus pRL-CMV set at 1.0.

RNA isolation and real-time PCR quantification of mRNA

Total RNA from liver tissues or transfected HEK293 cells was extracted by using RNA STAT-60 (Tel-Test, Friendswood, TX, USA). cDNA was produced by the use of High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions. The resultant cDNA was used for real-time PCR quantification of mRNA using iQ™ SYBR® Green Supermix (Bio-Rad, Hercules, CA, USA) by MyiQ2™ Two-Color Real-Time PCR Detection System (Bio-Rad). The amount of mRNA was calculated using the comparative CT method, which determines the amount of target normalized to an endogenous reference, β-actin (for G9a Liv-KO mice) or Renilla luciferase (for transfected cells).

Quantification of microRNAs using real-time PCR

miRCURY LNA™ Universal RT microRNA PCR (Exiqon) was used to quantify microRNAs in RNA samples from livers of G9a Liv-KO mice. All PCR reagents and specific LNA-modified miR-383 PCR primer set were purchased from Exiqon. The relative expression of miR-383 was normalized by U6 rRNA with values of wild-type mice set at 1.0.

Bioinformatic analysis of DNA-binding of RNA polymerase II (Pol II) and H3K9me2 in miR-383 locus in mouse liver

The public genome-wide datasets of ChIP-seq of Pol2 (GSM722763) and H3K9me2 (GSM1711953) in wild-type mouse liver were retrieved from GEO DataSets and uploaded into the Integrative Genomics Viewer (IGV) software (Robinson et al. 2011) to visualize the DNA-binding of Pol II and H3K9me2 in the miR-383 locus in mouse liver.

Determination of blood levels of interleukin-6 (IL-6) and monocyte chemoattractant protein-1 (Mcp-1)

Serum levels of IL-6 and Mcp-1 in wild-type and G9a Liv-KO mice were quantified by Bio-Rad Bio-Plex Mouse Cytokine Group I 2-plex Assay kit using the BioRad BioPlex 200 Luminex Instrument. Serum levels of IL-6 and Mcp-1 were calculated with standard curves of IL-6 and Mcp-1 supplied in the kit.

Statistical analysis

All values are expressed as mean ± S.E. The student’s t-test was used to determine the statistical difference between G9a Liv-KO and wild-type samples (SigmaPlot 12.5). Statistical significance was set at p < 0.05.

Results

Generation of G9a Liv-KO mice

Using G9a floxed (Tachibana et al. 2007) and Alb-cre mice, we generated G9a Liv-KO mice (G9a fl/fl, Alb-cre/+). Western blot data confirmed the loss of G9a protein, substantial decrease of H3K9me2 and moderate decrease of H3K9me1, but little change in H3K9me3 in livers of adult male G9a Liv-KO mice (Fig. 1A), which is consistent with literature report on the role of G9a in regulating H3K9me2 and H3K9me1 (Tachibana et al. 2007). G9a and GLP have been shown to control the recruitment of polycomb repressive complex 2 and H3K27 trimethylation at genomic loci (Mozzetta et al. 2014). However, G9a deficiency had no effect on global levels of H3K27me3 in mouse livers (Fig. 1A). G9a deficiency did not cause any overt liver injury or infiltration of inflammatory cells (Fig. 1B), and there was no elevation in serum markers of liver injury, namely alanine transaminase and alkaline phosphatase in G9a Liv-KO mice (Fig. 1C). Additionally, the liver/body weight ratio remained unchanged in G9a Liv-KO mice (data not shown). However, G9a Liv-KO mice had 20% higher blood levels of total cholesterol (CHO) than wild-type (WT) mice (Fig. 1D), which suggests an impairment of CHO metabolism in G9a Liv-KO mice. Nevertheless, hepatic levels of cholesterol and triglycerides remained unchanged in G9a Liv-KO mice (data not shown).

Figure 1.

(A) Western blot determination of hepatic protein expression; (B) liver histopathology (H&E staining); and blood levels of ALT and ALP (C) as well as total cholesterol (CHO) and triglycerides (TG) (D) in adult male mice with liver-specific knockout (Liv-KO) of G9a. N=6 for each group, mean ± SE. * p < 0.05 versus wild-type control.

Ectopic hepatic induction of genes in G9a Liv-KO mice

We used real-time PCR (RT-PCR) to determine changes in hepatic mRNA expression due to G9a deficiency (Fig. 2). Consistent with a known key role of G9a in gene silencing, G9a Liv-KO mice had marked induction of alpha-fetoprotein (Afp, ↑4.2 fold), a marker of fetal liver, which is also induced in G9a-null brain (Schaefer et al. 2009). G9a has been shown important in regulating the hematopoietic system. Accordingly, we found a remarkable induction of Ly6a (↑65 fold), which is widely used as a marker of hematopoietic stem cells and important in stem cell biology (Holmes and Stanford 2007). Additionally, G9a Liv-KO mice had marked induction of transmembrane protease, serine 4 (Tmprss4, ↑131 fold), absent in melanoma 2 (Aim2, ↑14 fold), and annexin A8 (Anxa8, ↑408 fold). Overexpression of Tmprss4 activates the JNK-AP-1 pathway (Min et al. 2014), whereas Aim2 is a cytosolic dsDNA-sensing inflammasome activated by bacterial, viral, and host DNA to trigger caspase-1 activation (Szabo and Csak 2012). Anxa8 recruits leukocytes to endothelial cells during inflammation (Poeter et al. 2014). Nevertheless, hepatic expression of key marker genes for inflammation, e.g. tumor necrosis factor α (TNFα), IL-1β, and IL-6 remained unchanged (data not shown), and there was no infiltration of inflammatory cells in G9a-null liver (Fig. 1B). Thus, G9a Liv-KO mice have no overt inflammation under unstressed conditions, but may have heightened response to proinflammatory stimuli.

Figure 2.

Hepatic mRNA expression of genes that are normally not or lowly expressed in adult male mice with liver-specific knockout (Liv-KO) of G9a. N=6 for each group, mean ± SE. * p < 0.05 versus wild-type control.

G9a has a critical role in regulating the basal activity and xenobiotic responses in the neurosystem (Chase and Sharma 2013; Maze et al. 2010; Schaefer et al. 2009; Subbanna et al. 2014; Sun et al. 2012). Consistently, we found that a group of genes important in the neurosystem were robustly elevated in G9a Liv-KO mice (Fig. 2), including: 1) signal transducing adaptor family member 1 (Stap-1, ↑25 fold), 2) p60-transcription-regulator-protein (p60TRP, ↑4.2 fold), 3) glutamate receptor, ionotropic, delta 1 (Grid1, ↑17 fold), 4) gamma-aminobutyric acid A receptor, beta 3 (Gabrb3, ↑7.5 fold), 5) G protein-coupled receptor 98 (Gpr98, ↑129 fold), 6) ATPase, Ca++ transporting, plasma membrane 2 (Atp2b2/Pmca2, ↑15 fold), 7) potassium channel, subfamily T, member 2 (Kcnt2, ↑5.8 fold); 8) RIC3 acetylcholine receptor chaperone (Ric3, ↑27 fold), 9) cholinergic receptor, nicotinic, alpha 4 (Chrna4, ↑1.6 fold); and 10) Chrnb2 (↑2 fold).

Induction of STAP1, a stem-cell-specific adaptor protein, promotes neurotoxic activation of microglia (Stoecker et al. 2009), whereas p60TRP is a novel transcriptional regulator of neural system (Heese 2013). Gpr98, a very large GPR, is expressed highly in developing neurosystem (McMillan et al. 2002). Kcnt2 is an ATP-sensitive potassium channel expressed highly in neurosystem (Bhattacharjee et al. 2003). Ric3 acts as a key molecular chaperone of nicotinic acetylcholine receptors (nAChRs) (Millar 2008). Atp2b2 (PMCA2) is a calcium pump that extrudes Ca²⁺ from the cytoplasm (Howard et al. 1994). PMCA2 controls calcium signaling by α7-containing nAChRs (Gomez-Varela et al. 2012). The marked induction of nAChRs as well as positive (Ric3) and negative (PMCA2) regulators of nAChRs suggest that hepatic nAChRs signaling might be altered in G9a Liv-KO mice.

Alteration in hepatic expression of cytoprotective genes in G9a Liv-KO mice

In contrast to ectopic induction of neural/immune genes, many key cytoprotective genes were down-regulated by G9a deficiency (Fig. 3A), including superoxide dismutase 2 (Sod2, ↓34%), catalase (Cat, ↓34%), epoxide hydrolase 1 (Epxh1, ↓40%), Ephx2 (↓50%), glutathione peroxidase-1 (Gpx1, ↓34%), glutamate cysteine ligase catalytic subunit (Gclc, ↓44%), Gclm (↓46%), glutathione S-transferase a3 (Gsta3, ↓38%), Gstp1 (↓91%), microsomal thioredoxin reductase 1/2 (Txnrd1/2, ↓34–43%), selenium-binding protein 2 (Selenbp2, ↓70%), and transferrin (Trf, ↓47%) whose reduction plays a key role in iron overload and iron toxicity (Brissot et al. 2012). Inhibition of G9a in cells markedly increases ROS via activation of NADPH oxidase (Nox) (Kim et al. 2013a). Interestingly, G9a Liv-KO mice had markedly higher Nox activator 1 (Noxa1, ↑5.2 fold), a Nox1 activator (Banfi et al. 2003) (Fig. 3A). Moreover, G9a Liv-KO mice had lower levels of GSH (↓20%) and elevated endogenous ROS (↑24%) (Fig. 3A). It is noteworthy that a moderate decrease in hepatic GSH levels usually does not cause hepatocyte injury.

Figure 3.

Hepatic mRNA expression of genes important in reactive oxygen species (A), xenobiotic metabolism (B), cholesterol/bile acid homeostasis (C), and sulfation (D) in adult male mice with liver-specific knockout (Liv-KO) of G9a. N=6 for each group, mean ± SE. * p < 0.05 versus wild-type control.

Alteration in hepatic expression of drug-processing genes in G9a Liv-KO mice

G9a Liv-KO mice had marked down-regulation of key DPGs cytochrome P450 3a11 (Cyp3a11, ↓50%), Cyp2c37 (↓83%), Cyp2d9 (↓52%), UDP glucuronosyltransferase 2a3 (Ugt2a3, ↓36%), Ugt2b34 (↓33%), and trend of lower Ugt1a1, but induction of Cyp2b10 (↑1.2 fold) and Ugt1a9 (↑1.5 fold) (Fig. 3B). Importantly, the 50% lower Cyp3a11 mRNA was associated with a similar 53% lower hepatic Cyp3a activity in G9a Liv-KO mice (Fig. 3B).

Alteration in hepatic expression of genes important in cholesterol and bile acid (BA) metabolism in G9a Liv-KO mice

To understand the mechanism of hypercholesterolemia in G9a Liv-KO mice, we determined mRNA expression of key genes in CHO and bile acid (BA) metabolism (Fig. 3C). G9a Liv-KO mice had 63% higher HMG-CoA reductase (Hmgcr), the rate limiting enzyme for CHO biosynthesis, but 44% and 40% lower Cyp27a1 and Cyp8b1, key enzymes for BA biosynthesis (Chiang 2009) (Fig. 3C). Moreover, G9a Liv-KO mice had 45% lower Abcg5, key transporter for biliary efflux of CHO. Thus, increased hepatic biosynthesis but decreased hepatic catabolism of CHO might be the underlying mechanism of elevated blood cholesterol in G9a Liv-KO mice. Cysteine sulfinic acid decarboxylase (Csad) is the key enzyme for the biosynthesis of taurine (Park et al. 2014), whereas bile acid-CoA:amino acid N-acyltransferase (Baat) is the key enzyme for the conjugation of BA with taurine or glycine (Pellicoro et al. 2007). G9a is important for expression of serine hydroxymethyltransferase 2 (mitochondrial) (Shmt2), a key enzyme for serine-glycine synthesis (Ding et al. 2013). We found that G9a Liv-KO mice had 63, 36, and 36% lower Csad, Baat, and Shmt2, respectively (Fig. 3C) than WT mice. Thus, major classical pathways in BA conjugation, namely taurine and glycine conjugation (Chiang 2013), might be impaired in G9a Liv-KO mice.

In contrast, G9a deficiency markedly induced most sulfotransferases (Sults), including Sult1a1 (↑52%), Sult1b1 (↑84%), Sult1c2 (↑3.9 fold), Sult1e1 (↑18 fold), Sult2a7 (↑58 fold), and sulfate transporter Slc13a4 (↑33 fold) (Fig. 3D). In contrast, hepatic mRNA expression of other Sult2a family members remained unchanged in G9a Liv-KO mice (data not shown).

Alteration in hepatic expression of key transcription factors in G9a Liv-KO mice

To elucidate the mechanism of the marked alteration in hepatic mRNA gene expression in G9a Liv-KO mice, we determined hepatic mRNA expression of key transcriptional regulators (Fig. 4). We found that G9a deficiency did not affect hepatic mRNA expression of Glp, the heterodimerization parter of G9a, or master regulators of hepatic basal gene expression, namely hepatocyte nuclear factor 1α (Hnf1α), Hnf4α, CCAAT/enhancer binding protein α (C/ebpα), C/ebpβ, Gr, liver receptor homolog-1 (Lrh-1), Retinoid X receptor α (Rxrα), liver X receptor α (Lxrα), or peroxisome proliferator-activated receptor α (Pparα) (Fig. 4A). Interestingly, G9a Liv-KO mice had higher hepatic expression of Chop, a marker of endoplasmic reticulum (ER) stress, but lower expression of a group of stress/xenobiotic sensors namely the BA receptor farnesoid X receptor (Fxr, ↓46%), an orphan nuclear receptor small heterodimer partner (Shp, ↓46%), aryl hydrocarbon receptor (AhR, ↓49%), constitutive androstane receptor (Car, ↓46%), and the pregnane X receptor (Pxr, ↓36%) (Fig. 4B). Additionally, G9a Liv-KO mice had 116% higher estrogen receptor α (Esr1) but 73% lower androgen receptor (Ar) (Fig. 4B). The contribution of decreased hepatic mRNA expression of these hormone/xenobiotic receptors to the altered hepatic expression of DPGs warrants further investigation.

Figure 4.

Hepatic mRNA expression of transcriptional regulators that are important in the regulation of basal expression (A) and gene induction (B) in adult male mice with liver-specific knockout (Liv-KO) of G9a. N=6 for each group, mean ± SE. * p < 0.05 versus wild-type control.

Alteration in hepatic protein expression in G9a Liv-KO mice

To verify whether the marked alterations in hepatic mRNA expression translated into alterations in the corresponding proteins, we used Western blot to determine hepatic protein levels. We found that the substantial up- and down-regulation of Ly6a (Fig. 2) and Gstp1 (Fig. 3A) mRNAs in G9a Liv-KO mice corresponded to the marked increase and decrease of Ly6a and Gstpi proteins, respectively (Fig. 5). Surprisingly, nuclear Nrf2 protein markedly decreased in G9a Liv-KO mice (Fig. 5), although Nrf2 mRNA remained unchanged (Fig. 4B). In G9a Liv-KO mice, although Keap1 mRNA was unchanged, p62 mRNA decreased and LC3B mRNA tended to be induced (Fig. 4B). Interestingly, protein levels of the cytosolic autophagic marker LC3-II tended to increase, but that of p62 markedly decreased in G9a Liv-KO mice (Fig. 5).

Figure 5.

Hepatic protein expression of representative genes in adult male mice with liver-specific knockout (Liv-KO) of G9a. N=3 pooled samples for each group. Two liver samples from each group with equal protein amount were pooled as one pooled samples for Western blot quantification of protein expression.

Hepatic induction of miR-383 in G9a Liv-KO mice

The marked decrease of nuclear Nrf2 protein without changes in Nrf2 mRNA in G9a Liv-KO mice suggest that Nrf2 protein expression or nuclear translocation may be affected. Thus, we screened hepatic microRNA expression in G9a Liv-KO mice by microarray and found that miR-383 was strongly induced in G9a Liv-KO mice (data not shown). ChIP-seq data showed that the peak of H3K9m2 in the miR-383 locus was associated with diminished binding of Pol II (Fig. 6A). Results of real-time PCR confirmed that miR-383 expression was 9.7 fold higher in G9a Liv-KO mice than WT mice (Fig. 6B).

Figure 6.

(A) ChIP-sequencing of RNA polymerase II (Pol II) and histone H3 dimethylation at lysine 9 (K9me2) in the miR-383 locus in mouse liver. DNA-binding of Pol II and K9me2 in mouse liver determined by ChIP-sequencing were retrieved from the public database of GEO DataSets and visualized in the IGV software. (B) Effect of G9a deficiency on hepatic expression of miR-383 in adult male mice with liver-specific knockout (Liv-KO) of G9a. N=6 for each group, mean ± SE. * p < 0.05 versus wild-type control.

Differential effects of ER stressor Chop on the promoter activities of Cyp3a11 and Sult1c2

Co-transfection of HEK293 cells with HNF4α and C/EBPα markedly increased the luciferase reporter activities of Cyp3a11 promoter by 5 fold. C/EBPα alone or C/EBPα plus Chop did not affect Cyp3a11 promoter (data not shown). Addition of Chop caused 47% decrease in the Cyp3a11 promoter activity co-activated by HNF4α and C/EBPα (Fig. 7A). In contrast, C/EBPβ alone had no effect on Sult1c2 reporter activity, whereas co-transfection of HEK293 cells with C/EBPβ and Chop increased Sult1c2 reporter activity by 42% (Fig. 7B).

Figure 7.

Regulation of mouse Cyp3a11 (A) and Sult1c2 (B) promoter by Chop in HEK293 cells. HEK293 cells were transfected with firefly luciferase reporter vector for Cyp3a11/Sult1c2, pRL-CMV, as well as expression vectors for HNF4α, C/EBPα/β, and/or Chop. Transfected cells were lysed 24 h after transfection for dual-luciferase assay of reporter activities. Y-axis represents Cyp3a11 or Sult1c2 reporter activities normalized to pRL-CMV, with the control value set as 1.0. N=4 for each group, mean ± SE. * p < 0.05 versus control. $ p < 0.05 versus C/EBP group.

Alteration in hepatic gene expression in G9a Liv-KO mice treated with LPS

G9a is required for endotoxin tolerance in cells via mediating the silencing of proinflammatory genes by the anti-inflammatory NF-kB RelB subunit (Chen et al. 2009). Thus, we hypothesized that G9a deficiency will aggravate proinflammatory response to LPS. Indeed, 16 h after LPS (5 mg/kg ip) challenge (Fig. 8), compared to WT mice, livers of G9a Liv-KO mice had: 1) much higher mRNA expression of many proinflammatory genes, including Aim2 (14 fold↑), IL-6 (1.9 fold↑), intercellular adhesion molecule 1 (Icam-1, 2.2 fold↑), inducible NO synthase (iNOS, 2.7 fold↑), Mcp-1 (4.6 fold↑), IFN-γ-inducible protein 10 (IP-10/CXCL10, 15 fold↑), and trend of higher IL-1b and chemokine (C-X-C motif) ligand 2 (Cxcl2) (Fig. 8A); 2) much higher proinflammatory transcription factors, including c-Jun (1.7 fold↑), Stat1 (1.3 fold↑), interferon regulatory factor 1 (Irf1, 1.2 fold↑), activating transcription factor 4 (Atf4, 1.4 fold↑), and trend of higher Chop (1.5 fold↑, p=0.07) (Fig. 8A); 3) markedly higher Tmprss4 but lower Gstp1, genes that activate and repress JNK, respectively (Fig. 8B), and higher growth arrest and DNA damage-inducible protein 45β (Gadd45β, 4.2 fold↑) and A20 (2.5 fold↑), classical NF-kB target genes (De Smaele et al. 2001); 4) higher LC3B (1.5 fold↑); and 5) trend of lower Gclc and Csad (Fig. 8B).

Figure 8.

Hepatic mRNA gene expression (A & B), hepatic levels of GSH and MDA (C), as well as blood levels of IL-6 (D) and Mcp-1 (E) in adult male mice with liver-specific knockout (Liv-KO) of G9a 16 h after LPS treatment (5 mg/kg ip). N=6 for each group, mean ± SE. * p < 0.05 versus corresponding wild-type control.

Additionally, LPS-treated G9a Liv-KO mice had trend of lower GSH levels, but significantly higher lipid peroxidation, determined by hepatic levels of malondialdehyde (MDA, 54%↑) (Fig. 8C). Nevertheless, blood levels of alanine aminotransferase remained similar between LPS-treated wild-type and G9a Liv-KO mice (data not shown). Consistent with higher hepatic induction of IL-6 and Mcp-1 mRNAs, LPS-treated G9a Liv-KO mice had much higher blood levels of IL-6 (24 fold↑, Fig. 8D) and Mcp-1 (20 fold↑, p = 0.07, Fig. 8E) than wild-type mice. High blood levels of IL-6 play the key role in the development of hypothermia during severe sepsis (Remick et al. 2005). Sixteen hours after LPS (5 mg/kg ip) challenge, WT mice had largely normal body temperature (37.6 ± 0.1 °C), whereas G9a Liv-KO mice developed prominent hypothermia (33.8 ± 0.3 °C). Thus, G9a in hepatocytes may play an important role in limiting the activation of NF-kB and JNK by LPS and the resultant increase of proinflammatory genes (e.g. IP-10, Mcp-1, Icam-1, iNOS) and lipid peroxidation.

Discussion

The present study demonstrates that liver-specific knockout of G9a markedly decreases hepatic levels of H3K9me2 and H3K9me1 and substantially alters hepatic transcriptome without significant liver injury or inflammation. However, G9a Liv-KO mice have elevated blood cholesterol as well as hepatic ROS and the ER stress marker Chop under unstressed condition. Moreover, G9a Liv-KO mice have aggravated inflammatory response and lipid peroxidation after LPS challenge.

G9a Liv-KO mice have marked ectopic induction of certain immune and neural genes, which is consistent with data from mice with T-cell- and brain-specific knockout of G9a (Lehnertz et al. 2010; Schaefer et al. 2009). One underlying mechanism of such ectopic induction in G9a Liv-KO mice is likely due to the loss of H3K9me2 (Fig. 1A), a prominent epigenetic signature for gene silencing. The remarkable induction of both mRNA and protein expression of Ly6a, a key marker of hematopoietic stem cells (Holmes and Stanford 2007), in the absence of infiltration of inflammatory cells, in combination with the marked induction of Afp, the marker of fetal liver, clearly indicate that the G9a-null livers have impairment in maturation during development. Thus, G9a plays an important role in the maturation of liver, which is consistent with the marked increase of H3K9me2 during liver maturation (Wen et al. 2009).

The present study demonstrates that G9a deficiency in mouse liver causes induction of Hmgcr and moderate hypercholesterolemia. Interestingly, inhibition of G9a by BIX-01294 has been shown to cause induction of cholesterol biosynthetic pathway in mouse and human pancreatic cells, via decreasing the suppressing mark H3K9me2 and increasing the activating mark H3K4me3 in the promoter of Hmgcr (Kubicek et al. 2012). Thus, the induction of Hmgcr in G9a Liv-KO mice may be due to decreased H3K9me2. Conversely, BIX-01294 treatment did not affect the cholesterol synthetic pathway in human hepatoma HepG2 cells (Kubicek et al. 2012). G9a is induced in liver cancer (Wei et al. 2017). The difference in the regulation of HMGCR by G9a inhibition in mouse hepatocytes and HepG2 cells might be due to difference in the magnitude of G9a inhibition, difference between normal hepatocytes and hepatoma cells, and/or specifies differences in HMGCR regulation between mice and humans.

G9a Liv-KO mice have down-regulation of many cytoprotective genes, most of which are Nrf2-target genes. Surprisingly, hepatic protein, but not mRNA levels of Nrf2 are markedly decreased in G9a Liv-KO mice, suggesting an important role of G9a in posttranscriptional regulation of Nrf2. Under normal conditions, Keap1 controls nuclear levels of Nrf2 and activation of Nrf2-target cytoprotective genes by binding to Nrf2 in the cytosol and promotes the degradation of Nrf2 (Suzuki et al. 2013). Nrf2 induces p62, a cargo receptor for autophagy, and p62 is required for long-term activation of Nrf2 by inactivating Keap1 (Jain et al. 2010; Komatsu et al. 2010; Lau et al. 2010). LC3B competes with Keap1 in binding to p62 and LC3B causes p62 degradation via autophagy (Jain et al. 2010); whereas knockdown of LC3B increases p62 protein in cells (Maruyama et al. 2014). Previous in vitro study demonstrates that G9a knockdown induces LC3B and autophagy (Artal-Martinez de Narvajas et al. 2013), the present data demonstrate that LC3B mRNA and protein expression tend to be higher in unstressed G9a Liv-KO mice, and hepatic mRNA expression of LC3B is significantly higher in LPS-treated G9a Liv-KO mice than wild-type mice. Thus, the potential role of the LC3B-p62 pathway in the decrease in nuclear Nrf2 and down-regulation of cytoprotective genes in G9a Liv-KO mice warrants further investigation.

The present study demonstrates a novel key role of G9a in suppressing miR-383 expression in mouse liver. Interestingly, G9a overexpression and miR-383 down-regulation are associated with the progression of liver cancer (Chen et al. 2016; Wei et al. 2017; Yokoyama et al. 2017). miR-383 suppresses insulin-like growth factor 1 receptor and the PI3K-Akt pathway (He et al. 2013; Teng et al. 2017). The PI3K-Akt pathway is important in the activation of Nrf2 by promoting its nuclear translocation (Nakaso et al. 2003). Additionally, PI3K-Akt inactivation induces CHOP expression in endoplasmic reticulum-stressed cells (Hyoda et al. 2006). Interestingly, when writing this manuscript, it is reported that G9a is down-regulated in livers of the diabetic db/db mice and high-fat diet-fed mice, and knockdown of G9a in hepatic cells result in downregulation of insulin receptor and p-Akt (Xue et al. 2018). Thus, the marked induction of miR-383 in G9a Liv-KO mice may mediate the down-regulation of Nrf2 and induction of Chop via G9a/miR-383/PI3K-Akt/Nrf2 and G9a/miR-383/PI3K-Akt/Chop pathways.

The present study demonstrates a novel role of G9a in maintaining hepatic expression of Gstp1/p2 which plays a key role in the detoxification of certain xenobiotics and suppression of the JNK stress signaling (Tew and Townsend 2011). Gstp1 is a Nrf2 target gene and also transactivated by androgen (Ikeda et al. 2002). G9a can act as a co-activator for androgen and estrogen receptors (Shankar et al. 2013). Interestingly, androgen receptor is markedly down-regulated in G9a Liv-KO mice (Fig. 4B). Thus, decreases in the transactivation by both Nrf2 and androgen receptor may explain the dramatic down-regulation of Gstp1, in contrast to the more moderate down-regulation of other Nrf2-target genes (Fig. 3A), in G9a Liv-KO mice.

Certain important drug-processing genes down-regulated in G9a Liv-KO mice are not directly regulated by Nrf2. Cyp3a is the most important Phase-I DPG whereas Ugt1a1 is essential for glucoronidation of bilirubin and xenobiotics. Both the mRNA expression of Cyp3a11 and Cyp3a activity are decreased in G9a Liv-KO mice (Fig. 3B). Hepatic mRNA expression of Car and Pxr, key regulators of Cyp3a, are down-regulated in G9a Liv-KO mice. Currently, the mechanism of selective down-regulation of xenobiotic sensors Car, Pxr, and AhR remains unknown. C/EBPs play important role in hepatic expression of CYP3A, Ugt1a1 (Lee et al. 1997) and many other DPGs. G9a Liv-KO mice have hepatic induction of Chop, a repressive member of the C/EBP family which can inhibit the binding of C/EBPα and C/EBPβ to the C/EBP target genes (Ron and Habener 1992) in liver. The present study demonstrates that Chop decreases the promoter activities of Cyp3a11 co-activated by HNF4α and C/EBPα to the magnitude similar to the down-regulation of Cyp3a11 in G9a Liv-KO mice. Thus, induction of Chop may play a key role in down-regulation of Cyp3a11 and certain other DPGs in G9a Liv-KO mice. In this regard, overexpression of C/EBPβ prevents the down-regulation of CYP3A4 and PXR in human hepatocytes induced by ER stressors brefeldin A and thapsigargin (Vachirayonsti et al. 2015).

The present study demonstrates a novel role of G9a in regulating hepatic expression of sulfotransferases. Hepatic mRNA expression of most Sults and the sulfate transporter Slc13a4 are markedly induced by G9a deficiency. Chop has a dual role in gene regulation: in addition to the suppression of C/EBP target genes by preventing their binding to the classical C/EBP binding sites, the Chop-C/EBP heterodimer can bind to novel sequences that contain a unique core element PuPuPuTGCAAT(A/C)CCC to activate gene expression (Maier et al. 2014). Interestingly, we found such core element for Chop-C/EBP heterodimer in the promoters of Sult1c2 and Sult2a7, two genes markedly induced in G9a Liv-KO mice, and Chop cooperates with C/EBPβ to activate the reporter that contains the core element for Chop-C/EBP heterodimer in the Sult1c2 promoter (Fig. 7B). Interestingly, induction of SULT1C2 is common in liver diseases and chemical exposure of hepatocytes in humans (Grinberg et al. 2014). In addition to ER stress, Chop can be induced by oxidative stress mediated by AP-1 (Guyton et al. 1996), whereas Nrf2 acts as a direct transcriptional repressor at the CHOP promoter (Zong et al. 2012). Cyps and Ugts are located in the ER, whereas Sults are located in the cytosol. During ER and oxidative stresses, down-regulation of Cyps and Ugts but induction of Sults might be an adaptive protective mechanism to decrease the metabolic burden of the stressed ER, and such changes of DPGs may markedly alter the metabolism and toxicity of certain xenobiotics.

The present study demonstrates a novel role of G9a in the attenuation of hepatic inflammatory response induced by LPS. The markedly higher induction of IP-10/CXCL10 by LPS in the G9a Liv-KO mice is consistent with a reported key role of G9a in silencing CXCL10 gene expression in fibroblasts (Coward et al. 2018). Overexpression of Tmprss4, a type II transmembrane serine protease, is associated viral infections and cancers. Mechanistically, Tmprss4 is linked with activation of JNK, NF-kB, and STAT3 (Jianwei et al. 2018; Jin et al. 2016; Min et al. 2014), key regulators of inflammatory responses. Thus, overexpression of Tmprss4 may play an important role in the aggravated proinflammatory response to LPS in G9a Liv-KO mice. The numbers of T helper 17 (Th17) and Foxp3+ regulatory T cells (Treg) increase in liver and serum from patients with various acute and chronic liver injury, and they promote inflammation and fibrosis during non-alcoholic steatohepatitis, alcoholic liver disease, as well as HBV and HCV infection (Harley et al. 2014; Pellicoro et al. 2014; Tan et al. 2013; Zhao et al. 2014). After pathogen infection, G9a-deficient Th cells fail to develop into protective Th2 cells (Lehnertz et al. 2010), and G9a deficiency promotes the differentiation of T cells into Th17 and Foxp3+ Treg cells in mice, with marked induction of IL-17a (Antignano et al. 2014). Interestingly, our results indicate that the common G9a SNP rs652888 which is strongly associated with chronic HBV infection in humans may markedly decrease the expression of functional G9a protein (Fig. S1), suggesting that G9a deficiency due to SNP rs652888 may contribute to chronic HBV infection and disease progression in HBV patients.

Taken together, the present study clearly demonstrates a novel important role for G9a in regulating epigenome and transcriptome in liver. G9a is dispensable for the differentiation but important for the maturation of liver in mice. G9a deficiency induces certain immune and neural genes but down-regulates certain key cytoprotective genes and profoundly alters DPG expression in mouse liver. Loss of H3K9me2, a key silencing epigenetic signature and reduced Nrf2 signaling, but induction of miR-383 and Chop and activation of NF-kB and JNK might be key underlying mechanisms. However, the effects of G9a deficiency on non-histone targets, such as GR and AR may also play important roles. The expression of G9a is altered by many clinically important xenobiotics, and hepatic down-regulation of G9a by ethanol is more profound in mice deficient for glutathione synthesis (Esfandiari et al. 2010). Thus, G9a deficiency due to genetic polymorphism and/or exposure to environmental stressors may increase the susceptibility to infectious and metabolic diseases, whereas overexpression of G9a may provide liver cancer the survival advantage and resistance to anticancer drugs. Future studies on how G9a deficiency affects infectious, inflammatory, and metabolic liver diseases as well as hepatic metabolism and disposition of xenobiotics are warranted.