MiR-873-5p acts as an epigenetic regulator in early stages of liver fibrosis and cirrhosis

Glycine N-methyltransferase (GNMT) is the most abundant methyltransferase in the liver and a master regulator of the transmethylation flux. GNMT downregulation leads to loss of liver function progressing to fibrosis, cirrhosis, and hepatocellular carcinoma. Moreover, GNMT deficiency aggravates cholestasis-induced fibrogenesis. To date, little is known about the mechanisms underlying downregulation of GNMT levels in hepatic fibrosis and cirrhosis. On this basis, microRNAs are epigenetic regulatory elements that play important roles in liver pathology. In this work, we aim to study the regulation of GNMT by microRNAs during liver fibrosis and cirrhosis. Luciferase assay on the 3ʹUTR-Gnmt was used to confirm in silico analysis showing that GNMT is potentially targeted by the microRNA miR-873-5p. Correlation between GNMT and miR-873-5p in human cholestasis and cirrhosis together with miR-873-5p inhibition in vivo in different mouse models of liver cholestasis and fibrosis [bile duct ligation and Mdr2 (Abcb4)-/- mouse] were then assessed. The analysis of liver tissue from cirrhotic and cholestatic patients, as well as from the animal models, showed that miR-873-5p inversely correlated with the expression of GNMT. Importantly, high circulating miR-873-5p was also detected in cholestastic and cirrhotic patients. Preclinical studies with anti-miR-873-5p treatment in bile duct ligation and Mdr2-/- mice recovered GNMT levels in association with ameliorated inflammation and fibrosis mainly by counteracting hepatocyte apoptosis and cholangiocyte proliferation. In conclusion, miR-873-5p emerges as a novel marker for liver fibrosis, cholestasis, and cirrhosis and therapeutic approaches based on anti-miR-873-5p may be effective treatments for liver fibrosis and cholestatic liver disease.


Introduction
Glycine N-methyltransferase (GNMT) is the most important and abundant S-adenosylmethionine (SAMe)dependent methyltransferase in the liver. GNMT is predominantly expressed in hepatocytes although it is also found in other cell types such as cholangiocytes, the epithelial cells of the bile duct [1][2][3] . Lack of GNMT induces undesired methylation reactions leading to proliferative, inflammatory, and profibrogenic responses that culminate in liver disease 4,5 . GNMT expression is reduced in different liver diseases including liver cirrhosis of diverse etiology 1,6 , chronic cholestatic liver disease 7 , hepatocellular carcinoma (HCC) 1 , and cholangiocarcinoma 2 . In agreement, Gnmt-deficient (Gnmt -/-) mice, characterized by elevated SAMe levels, develop liver fibrosis spontaneously at the age of 3 months and HCC at 8 months 8 . Indeed, an aberrant DNA methylation signature has been identified in the Gnmt -/mouse 8 , underscoring the importance of this protein as an epigenetic regulator. Although GNMT promoter hypermethylation causes GNMT downregulation in some HCC patients 9 , other mechanisms are likely to be involved in the regulation of this gene in liver diseases such as fibrosis or cirrhosis.
Hepatic fibrosis is the result of the wound-healing response of the liver to repeated injury that occurs in most types of chronic liver diseases. The injury can be caused by the accumulation of lipids in the liver, as occurs during non-alcoholic fatty liver disease (NAFLD), can be a result of a toxicant or viral insult, such as excessive alcohol consumption or hepatitis, respectively, or the accumulation of bile acids (BA), as in chronic liver cholestasis. The finding of potential molecular and pathway targets for reverting or halting the progression of liver fibrosis or cholestasis to cirrhosis and HCC is an emerging field.
MicroRNAs (miRNAs) are highly conserved, small noncoding RNAs that post-transcriptionally regulate gene expression 10 of essential biological processes, as well as cellular responses 11,12 . In the liver, miRNA signature has been implicated in NAFLD, cirrhosis, and liver cancer 13 . The biological significance and therapeutic potential of miRNAs in liver disease management is a rapidly growing field.
In this study, we show that liver fibrosis progression is associated with the repression of GNMT, which is targeted by miR-873-5p in two preclinical models of liver fibrosis and cholestasis: the bile duct ligation (BDL) and the Mdr2 (Abcb4)-deficient mice (Mdr2 -/-), respectively. Targeting miR-873-5p in these models resulted in a reduction of liver damage affecting mainly hepatocyte apoptosis and cholangiocyte proliferation, through a GNMT-dependent epigenomic mechanism. Of note, an inverse correlation between hepatic GNMT and miR-873-5p was identified in a cohort of cirrhotic patients with diverse etiology. Importantly, increased circulating miR-873-5p was also detected in serum samples from cirrhotic and cholestatic patients. A negative correlation was found between hepatic GNMT expression and serum levels of miR-873-5p in a cohort of cholestatic patients according with the fibrotic stage.
Overall, miR-873-5p emerges as a novel druggable target and a marker for liver fibrosis, cholestasis and cirrhosis.

MiR-873-5p inversely correlates with GNMT expression in liver fibrosis and cholestasis
The inverse association between GNMT and miR-873-5p expression was confirmed in different preclinical models of liver fibrosis. Administration of carbon tetrachloride (CCl 4 ) for 6 weeks induced a reduction in Gnmt expression, associated with the induction of miR-873-5p (Suppl. Fig. 1C). Additionally, in two models of cholestatic liver disease induced by BDL after 3 and 7 days of surgery 14 and the Mdr2 -/mouse, a model of inflammationinduced cholestatic liver injury, fibrosis, and cancer 15 , we found an inverse correlation between GNMT levels and miR-873-5p expression (Fig. 1c, d).
Likewise, an inverse correlation between hepatic GNMT expression and miR-873-5p was identified in a group of 16 cirrhotic patients of diverse etiology (Suppl. Table 1, Fig. 2a). These results prompted us to assay circulating miR-873-5p levels in cirrhotic patients. Blood samples were collected at the screening visit in 35 cirrhotic patients diagnosed with cirrhosis or advanced scaring (F4) with diverse etiology and levels of circulating miR-873-5p were evaluated and compared to nine healthy controls. Higher miR-873-5p levels in serum were observed in cirrhotic patients (Fig. 2b). Subsequently, we analyzed circulating miR-873-5p levels in a cohort of cholestatic patients primary biliary cholangitis (PBC)/ primary sclerosing cholangitis (PSC) (n = 41) and compared to healthy patients with antimitochondrial antibodies (AMA + ) but without abnormal liver test (Suppl .  Table II). In these patients, whose circulating miR-873-5p levels are increased relative to healthy controls, a positive correlation was also found with fibrosis grade (Fig. 2c). Furthermore, a reduction of hepatic GNMT expression was associated with the fibrosis grade in a cohort of 64 liver cholestasis patients (Suppl . Table III and Fig. 2d). Finally, a significant correlation between reduced expression of hepatic GNMT and increased levels of miR-873-5p in the serum of a cohort of patients with early and advance cholestatic liver disease was detected in 19 paired  Fig. 1B, D, E. Data normalized as fold change vs. control. Error bars represent the means ± SEM. Statistical significance was determined by the Student's t-test or ANOVA when more than two groups were compared. p < 0.05*; p < 0.01**; p < 0.001*** ). (f) Correlation between circulating miR-873-5p and hepatic GNMT with fibrotic stage in cholestatic patients, respectively. Data normalized as fold change vs. control. Error bars represent the means ± SEM. Statistical significance was determined by the Student's t-test or ANOVA when more than two groups were compared. p < 0.05*; p < 0.01**; p < 0.001*** samples (Fig. 2e). Fibrotic stage positively correlated with miR-873-5p and negatively correlated with GNMT levels (Fig. 2f). Overall, miR-873-5p expression is augmented and inversely correlated with hepatic GNMT expression in cirrhosis and cholestatic disease.
Anti-miR-873-5p attenuates liver injury after bile duct ligation The potential benefit of targeting miR-873-5p was evaluated in the BDL animal model of cholestasis. Thus, we performed BDL surgery for one week in Gnmt wildtype (WT) mice and injected anti-miR-873-5p or a miR-Ctrl i.v. at day 3 and 5, corresponding to the initial phase of hepatic fibrogenesis 14 ( Fig. 3a and Suppl Fig. 2A) and sacrificed animals at 7 days after BDL. MiR-873-5p inhibition ( Fig. 3b) resulted in GNMT restoration levels ( Fig. 3c). No effect on miR-873-5p and Gnmt mRNA levels was detected in other tissues where Gnmt is usually expressed, such as pancreas and kidney, after miR-873-5p inhibition (data not shown). Anti-miR-873-5p treatment was associated with reduced serum transaminases (ALT/ AST), caspase 3 activity and PARP cleavage as well as parenchymal disruption as detected by H&E staining, all readouts of cell death ( Fig. 3d-f).
RNA sequencing analysis further retrieved a completely different gene ontology (GO) profile in the BDL model after miR-873-5p inhibition. During this initial phase of hepatic fibrogenesis, biological processes related to cell proliferation, natural killer cell activation, and immune response processes were downregulated in anti-miR-873-5p vs. miR-Ctrl mice. On the other hand, anti-miR-873-5p-treated-mice showed increased gene expression related to biological processes implicated in steroid and fatty acid metabolism and in the respiratory electron transport chain compared to miR-ctrl mice (Supp. Fig. 3A and Supp. Table IV and V).
Of relevance, the treatment of Gnmt -/mice after 3 and 5 days of BDL with anti-miR-873-5p failed to counteract both inflammation and fibrosis (Supp Fig. 4a), indicating that Gnmt is one of the principal targets that mediates the anti-miR-873-5p effect in vivo, even though miR-873-5p could affect many different mRNAs besides Gnmt.
Overall, these data show that miR-873-5p inhibition is hepatoprotective and reduces cell death, affects ductular reaction and prevents the generation of liver inflammation and fibrosis after BDL, which is suggested to take place by enhancing GNMT expression.

MiR-873-5p directly regulates hepatocyte apoptosis and cholangiocyte proliferation
In the pathology of liver fibrosis and in cholestatic liver injury different types of hepatic cells including hepatocytes, cholangiocytes, inflammatory macrophages (Kupffer cells) 16 and Natural Killer 7,17-19 cells, and hepatic stellate cells (HSC) 20 are known to mediate different effects, regulating the development and the progression of the disease. During initiation of fibrogenesis, hepatocyte apoptosis and cholangiocyte proliferation are the primary events driving disease progression 14 . Although GNMT in the liver is known to be mainly found in hepatocytes, it has been described to be expressed in other cell types such as cholangiocytes, where its expression is associated with a differentiated and a non-proliferative phenotype of these cells 2 . Thus, we have analyzed the targeting of Gnmt by miR-873-5p in these cells and its potential contribution to cholestatic liver injury.
Firstly, primary isolated murine hepatocytes were cultured with the toxic bile acid deoxycholic acid (DCA, 100 µM) for 2 h to induce hepatocyte apoptosis, an in vitro cell model that mimics in vivo BDL. Under these circumstances, DCA treatment was associated with reduced Gnmt expression together with the induction of miR-873-5p and hepatocyte apoptosis, suggesting the importance of the regulation of GNMT-miR-873-5p axis in hepatocytes (Fig. 4a). The potential role of miR-873-5p in the induction of BA-induced apoptosis in hepatocytes through the repression of Gnmt was studied by inhibiting miR-873-5p directly in these cells. MiR-873-5p inhibition in hepatocytes induced GNMT levels (Fig. 4b), resulting in attenuation of BA-induced apoptosis measured by Caspase 3 activity, TUNEL assay and JNK phosphorylation (Fig. 4c, d). Moreover, miR-873-5p inhibition increased the expression of different genes related to BAinduced apoptosis [B-Cell CLL/Lymphoma 2 (Bcl2) and the Hepatocyte nuclear factor (Hnf) Hnf1α and Hnf4α] and, more importantly, genes implicated in BA metabolism and regulation, such as the master regulator of BA metabolism Fxr and different BA exporters, such as Bsep, Abcg5 and some Mdr/Mrp (Fig. 4e).
Overall, these data suggest that anti-miR-873-5p has broad effects in liver cells affecting fibrosis progression, mainly mediating anti-apoptotic activity in hepatocytes and anti-proliferative effect in cholangiocytes contributing to reduce liver injury and fibrosis progression.
Anti-miR-873-5p prevents liver injury in Mdr2 -/mice Mice lacking the ATP-binding cassette ABCB4 protein encoded by the Mdr2 gene (Mdr2 -/-) provide a model for the study of cholestasis in the context of chronic inflammation as a result of increased BA accumulation 15 . The increase in miR-873-5p and the low expression levels of GNMT at the different stages of hepatic fibrogenesis in the Mdr2 -/mice (Fig. 1d) prompted us to investigate the impact that anti-miR-873-5p treatment could have in this model. Four-month old Mdr2 -/mice, with advanced fibrosis and biliary proliferation (Suppl Fig. 6a), were treated once a week with anti-miR-873-5p during 4 weeks (Fig. 5a). Anti-miR-873-5p treatment rescued GNMT (Fig. 5b, c) and reduced ALT/AST transaminases, total serum BA levels (Fig. 5d, e), cholangiocytes proliferation by CK19, hepatocyte death assessed by caspase 3 activity and hepatic ammonia content 21 (Fig. 5f). Moreover, Mdr2 -/mice treated with anti-miR-873-5p showed lower cell death receptor Dr5 and higher biliary acids transporters of the ABC superfamily Mdr1, Mrp3, and Mrp5 (Suppl Fig. 6B), consistent with the reduction of BA content, BA-induced apoptosis and markers related to cholangiocyte activation (Suppl Fig. 6B). These changes were accompanied by a reduced inflammatory response as determined by F4/80 and Il-6, Ccl1 and iNos expression, as well as less fibrosis evaluated through Sirius red staining and the levels of profibrogenic (Tgfb, Timp1, and Timp2) genes ( Fig. 5f and Suppl Fig. 6b). Altogether, these data suggest that anti-miR-873-5p targeting in the Mdr2-/-cholestasis animal model reduces cholangiocyte proliferation, increases the export of bile acids and ameliorates BA-induced hepatocyte cell death and inflammatory response.
MiR-873-5p regulation results in GNMT-dependent epigenomic modulation DNA (CpG) methylation acts as a blueprint for global alterations in the epigenome that drive liver injury 22 among them hepatic fibrogenesis [23][24][25][26] . BDL and Mdr2 -/mice under anti-miR-873-5p treatment showed improved hepatic SAMe metabolism with a decrease in the ratio SAMe/S-adenosylhomocysteine (SAH) consistent with the re-expression of GNMT (Table 1). This effect in the methylation flux correlated with a reduction in the global DNA methylation content as detected after miR-873-5p inhibition (Fig. 6a). DNA methylation is mediated by two groups of DNA methyltransferases (DNMTs), DNMT1 that is implicated in maintenance of DNA methylation and DNMT3a/3b that are responsible for de novo DNA methylation 27 . According to the changes in DNA methylation levels in the liver, anti-miR-873-5p treatment reduced Dnmt3a expression in both fibrotic animal models (Fig. 6b). Indeed, the inhibition of miR-873-5p during BDL and in Mdr2 -/mice increased the expression of other genes of the methionine cycle besides Gnmt, such as Methionine adenosyltransferase (Mat1a) and S-adenosylhomocysteine hydrolase (Sahh) (Suppl. Fig. 7A).
Different genes could be regulated by methylation of their CpG islands. The activity of STAT3, a key player in liver inflammation, was repressed in BDL under miR-873-5p inhibition (Fig. 3g), correlating with the re-expression of its inhibitor Protein inhibitor of activated STAT (Pias) Pias 1/3 (Fig. 6c), that has been associated to its promoter  (N = 4). Statistical significance was determined by the Student's t-test or ANOVA when more than two groups were compared. p < 0.05*; p < 0.01**; p < 0.001***  N = 4). Data normalized as fold change vs. control. Error bars represent the means ± SEM. Statistical significance was determined by the Student's t-test or ANOVA when more than two groups were compared. p < 0.05*; p < 0.01**; p < 0.001*** methylation status 28 . Similarly, the promoter regulation of RAS-association domain family (RASSF) (inhibitors of RAS signaling), has been reported to be susceptible of hypermethylation in the absence of GNMT 8 and in early HCCs 28 . We have found that the expressions of RASSF genes (Rassf1a/Rassf4a) are upregulated in Mdr2 -/mice after blockade of miR-873-5p (Fig. 6c).
Regulation of cholangiocyte proliferation and BA metabolism have both been described to be under epigenetic control 29,30 . SRY-Box 17 (SOX17) regulates the phenotype of normal human cholangiocytes acting as a tumor suppressor in cholangiocarcinoma, occurring its downregulation through DNA methylation 31 . Under anti-miR-873-5p treatment, Sox17 expression was significantly upregulated in BDL and Mdr2 -/mice (Fig. 6d). Finally, inhibiting miR-873-5p also restored the expression of genes methylated under pathological conditions, such as Abcg5 and Bsep, as well as the nuclear receptors Shp and Fxr, implicated in BA transport and metabolism 29 (Fig. 6e).
Besides DNA, SAMe can methylate histones at different residues. Histone modification is also involved in establishing patterns of gene repression during liver injury. Methylation of histones and DNA are linked to the activation of HSCs 32 . Importantly, miR-873-5p inhibition significantly reduced the protein levels of EZH2, a profibrogenic histone lysine methyltransferase, after BDL and in Mdr2 -/mice (Fig. 6f). Also, the mRNA level of the histone methyltransferase Ash1 is reduced after anti-miR-873-5p treatment, correlating with the decrease in the methylation level of its target H3K4, which has been related with TGFβ induction in fibrosis (Suppl. Fig. 6B) 33 .
Overall, these data suggest that the prompt recuperation of hepatic GNMT expression through anti-miR-873-5p treatment could lead to the restoration of SAMe metabolism and the epigenetic regulation of specific genes that are implicated directly in BA homeostasis, cholangiocyte proliferation and inflammatory and fibrogenic pathways (Fig. 7).

Discussion
GNMT is considered a major regulator of the transmethylation flux in the liver controlling liver homeostasis and health 4,5 . A decrease in GNMT levels occurs in NAFLD, at early stages of fibrosis, cirrhosis, and HCC 1,6,7,34,35 . Data highlighting that attenuation of GNMT was found at early stages of liver disease, where fibrosis is still not prominent, points to GNMT deficiency as a contributor/mediator of fibrogenesis rather than a consequence. Even though GNMT repression in some primary tumors has been associated with aberrant promoter methylation 9 , hypermethylation does not seem to be the unique mechanism for GNMT regulation, and it may coexist with other regulatory pathways 9 .
miRNAs regulate a wide variety of biological functions in the liver and are directly implicated in inflammation, cirrhosis, and malignant transformation 13 . These small molecules are, therefore, emerging as viable therapeutic tools. Three independent unbiased approaches revealed miR-873-5p as a potential microRNA targeting GNMT. In here, we have shown that miR-873-5p is modulated in several in vitro and in vivo models of liver fibrosis where GNMT levels are modified. Furthermore, in the clinical setting, we have found high levels of circulating miR-873-5p in the serum of cirrhotic patients as well as a negative correlation between hepatic GNMT and miR-873-5p in human cirrhosis and cholestasis. Altogether, this evidence allowed us to identify for the first time GNMT as a direct target of miR-873-5p.
Although liver fibrosis was historically thought to be an irreversible process, the understanding of its molecular pathogenesis has stimulated the development of novel antifibrotic therapies, still in an experimental phase 36 . Increased miR-873-5p levels have been observed in    Fig. 7C. Data normalized as fold change vs. control. Error bars represent the means ± SEM. Statistical significance was determined by the Student's t-test or ANOVA when more than two groups were compared. p < 0.05*; p < 0.01**; p < 0.001*** hepatocytes and in proliferative cholangiocytes under cholestatic/fibrogenic situations. Anti-miR-873-5p treatment in vitro decreased the BA-induced hepatocyte cell death as well as cholangiocyte proliferation. Importantly, anti-miR-873-5p treatment in Gnmt wild-type mouse submitted to BDL surgery blocked the overwhelming inflammatory profile associated to fibrosis. Of relevance, no effects under miR-873-5p inhibition were detected in the Gnmt -/mice, indicating that GNMT activity is the hub of the regulatory network modulated by miR-873-5p.
In the Mdr2 -/cholestatic mouse model inhibition of miR-873-5p for only 4 weeks starting at the time at which an overt inflammation, cholestasis, and severe fibrosis is observed (4 months old mice), resulted in reduced BArelated cell death and cholangiocyte proliferation, with mild blunted inflammation and decreased fibrosis related markers.
DNA methylation and hydroxymethylation are epigenetic mechanisms that modulate gene expression and play a critical role in the development of liver fibrosis 22,24-26 . We have identified that miR-873-5p changes methyltransferase signaling and DNA methylation state during BDL-mediated hepatic injury and in the chronic model of Mdr2 -/mice. Anti-miR-873-5p therapy restores the transmethylation flux to normal levels in both animal models of liver damage. This miR-873-5p-dependent modulation of the methylome reduces global DNA methylation by decreasing Dnmt3a expression that functions as a de novo methyltransferase 24 . DNA methylation changes that occur in the livers of mice under anti-miR- Fig. 7 MiR-873-5p targets GNMT expression in the liver mediating global epigenomic changes and contributing to liver fibrosis progression 873-5p treatment are associated with altered expression of specific genes. Indeed, in BDL mouse, anti-miR-873-5p restores Pias 1,3 levels, genes highly sensible to methylation changes and regulators of STAT3 phosphorylation with the concomitant regression of the inflammatory response. The reduction in STAT3 activation observed in BDL mouse after anti-miR-873-5p treatment, could also have an important effect in the amelioration of fibrosis, due to its direct role in promoting HSC, the major fibrogenic cell type 37 . In Mdr2 -/mice, the upregulation of the RASSF tumor suppressor, also susceptible to epigenetic modification linked to miR-873-5p treatment, blocks the Ras/Raf/MEK/ERK pathway, related to tumor formation 38 .
Hepatic fibrosis is also related to methylation changes in BA metabolism and cholangiocyte proliferation 29 . BAs are highly cytotoxic, their synthesis, transport, and pool size are tightly regulated under physiological conditions. FXR, implicated in every step of BA homeostasis and susceptible to be regulated by methylation 29 , was upregulated under anti-miR-873-5p treatment. FXR has been shown to induce BSEP and MRP2 in hepatocytes 39 . Importantly, the inhibition of miR-873-5p counteracts the repression of BSEP and the basolateral drug transporters MRP3 and MRP4 in our cholestatic models. Thus, the stimulation of basolateral BA efflux may be an important protective response against the cholestatic liver injury detected in BDL and Mdr2 -/that could also be regulated by DNA methylation of different genes. Regarding cholangiocyte proliferation, SOX17 promoter was previously found to be hypermethylated and downregulated in cholangiocarcinoma 31 . In addition, downregulation of SOX17 is implicated in cholangiocyte damage and fibrosis 40 . MiR-873-5p therapy is able to re-express Sox17 in both cholestatic animal models potentially regulating cholangiocyte proliferation.
Finally, the activation of HSC to myofibroblast is a key process in liver fibrosis mediated by epigenetic mechanism 41 . We now provide evidence that SAMe flux recovery mediated by anti-miR-873-5p treatment and GNMT rescue reduced EZH2 levels. This histone methyltransferase is one of the major regulators for profibrogenic factors 42 . Moreover, other signatures of histone methylation are observed in the BDL model, including changes in the histone methyltransferase Ash1, correlating with methylation of the histone H3K4, which has been related with TGFβ induction of fibrosis 33 .Therefore, it is tempting to hypothesize that changes in global methylation seen after GNMT recovery could exert a negative feedback in the transdifferentiation and activation of HSC.
Our results show that miR-873-5p inhibition directly regulates in vivo BA metabolism, cholangiocyte proliferation and activation and hepatocyte cell death. Modulation of hepatocyte cell death by inhibiting miR-873-5p further reduces inflammatory responses and ameliorates the fibrotic phenotype in diverse liver injury mouse models, possibly associated with epigenetic changes (Fig. 7). Indeed, there is an emerging field in the screening of epi-drugs in the fibrosis disease.
Summing up, the significant inverse correlation between hepatic GNMT and miR-873-5p expression both in cirrhotic and cholestatic patients targeting the miR-873-5p/GNMT axis may provide a novel therapeutic approach to treat liver fibrosis.

Human studies
All the studies were performed in agreement with the Declaration of Helsinki, and with local and national laws. The Human Ethics Committee of Valdecilla Hospital and of the University of Navarra approved the study procedures.

Cirrhotic patients Cohort 1
Sixteen liver samples from cirrhotic patients (grade F4) of diverse etiologies were included in this study. The characteristics of these patients are summarized in Suppl. Table I. Healthy liver samples (from individuals with normal or minimal changes in the liver) were collected at surgery of digestive tumors or from percutaneous liver biopsy performed because of mild alterations of liver function. All samples were obtained from the Biobank of the University of Navarra (Pamplona, Spain).

Cohort 2
Thirty-five serum samples from cirrhotic patients (grade F4) (45.7% women and 54.3% men, mean age 63.9 ± 13.4 years) were included in this study from patients recruited ate the Marqués de Valdecilla University Hospital (MVUH, Santander, Spain). Cirrhosis diagnosis was established by clinical or histological data. Etiology was varied: 43% hepatitis C, 31% alcoholic, 17% autoimmune hepatitis, 6% primary biliary cholangitis and 3% hepatitis B. Serum samples (n = 9; five women and four men, mean age 32.4 ± 6.4 years) from healthy volunteer subjects were used as controls and recruited through the Gastroenterology and Hepatology Department of MVUH, Santander, Spain.

Cholestatic patients Cohort 3
Forty-one serum samples from cholestatic patients (PBC = 36; PSC = 5) with different grade of fibrosis (early = 34; advanced = 7) were included in the study (90% women and 10% men, mean age 59.5 ± 12.2 years). PBC/ PSC diagnosis was established in MVUH and based on clinical and biochemical data, immunological markers, imaging, liver histology, and exclusion of other possible causes of liver injury. Healthy human serum samples (n = 13) (92% women and 8% men, mean age 45.5 ± 10.5) from subjects with AMA in serum and without abnormal liver test were included and provided by the Gastroenterology and Hepatology Department MVUH, Santander, Spain. Characteristics of these patients are summarized in Suppl. Table II.

Cohort 4
Sixty-four liver samples from patients with cholestatic liver disease, PBC (n = 60) and PSC (n = 4) were included in this study (90% women and 10% men, mean age 51.64 ± 10.65 years). PBC/PSC diagnosis, established in Marqués de Valdecilla University Hospital (MVUH, Santander, Spain), was based on clinical and biochemical data, immunological markers, imaging, liver histology, and exclusion of other possible causes of liver injury. Patients were classified for early or advance fibrosis based on histological analysis and according to Ludwig classification stage (Grade I-IV). Characteristics of these patients are summarized in Suppl. Table III. Healthy human liver samples (n = 12; 42% women and 58% men, mean age 63.16 ± 13.01 years) from organ-transplant donor were used as controls and recruited through the Gastroenterology and Hepatology Department of MVUH, Santander, Spain.

In vivo miR-873-5p inhibition after BDL
Three-month-old male Gnmt WT and Gnmt -/mice were injected in the tail vein with miRIDIAN microRNA Hairpin Inhibitor anti-miR-873-5p or miR-Control (60 µg/mouse) (Dharmacon, USA) at 3 days after BDL and repeated on day 5 using Invivofectamine 3.0, following the manufacturer's instructions (Invitrogen, USA). Mice were sacrificed at day 7, blood withdrawn and livers removed and snap frozen in liquid nitrogen or fixed in formalin for subsequent analysis.
In vivo miR-873-5p inhibition in Mdr2 -/mice Four-month-old male Mdr2 -/mice were injected in the tail vein with miRIDIAN anti-miR-873-5p or miR-Control (60 µg/mouse) once a week during 4 weeks using Invivofectamine 3.0, following the manufacturer's instructions (Invitrogen, USA). Animals were then sacrificed, blood withdrawn and livers were removed and snap frozen in liquid nitrogen or fixed in formalin.

In vitro miR-873-5p inhibition
Mouse primary hepatocytes were isolated from male Gnmt WT mice via collagenase perfusion as described 44 . All adhered cells were maintained in MEM with 10% fetal bovine serum (FBS). NHC cholangiocytes and primary mouse hepatocytes were transfected with anti-miR-873-5p or a miR-Control using DharmaFECT transfection reagent (Dharmacon) following the manufacturer's procedure.

SAMe and SAH measurement
Hepatic SAMe and SAH were determined by Liquidchromatography/mass spectrometry (LC/MS) using a Waters ACQUITY-UPLC system coupled to a Waters Micromass LCT Premier Mass Spectrometer equipped with a Lockspray ionization source as described previously 45 .

Global DNA methylation measurement
Global DNA methylation (5mC) analyses were performed following the method previously described 46 .

Bile acids (BA) measurement
Total bile acids in serum of Mdr2 -/mice were measured using the Total Bile Acid Assay Kit (Cell Biolabs, Inc, USA) following the manufacturer's instructions.

Statistical analysis
Data are represented as mean ± SEM. mRNA/RNA data is normalized as fold change vs. control. Statistical significance was determined by the Student's t-test when two groups were compared and when comparing three groups one-way analysis of variance (ANOVA) was used. A p < 0.05 was considered significant.