Division of Gastroenterology and Liver Diseases, USC Research Center for Liver Diseases, Southern California Research Center for Alcoholic and Pancreatic Diseases & Cirrhosis, Keck School of Medicine USC, Los Angeles, CA, USA
Potential conflict of interest: Nothing to report.
Supported by NIH RO1AT1576, RO1AT004896 (M.L.M-C., S.C.L., J.M.M.), RO1DK051719 (S.C.L., J.M.M.), R01DK080010, R01DK15289 (C.W.), Spanish Plan Nacional I+D SAF 2011-29851 (J.M.M.), ETORTEK-2010 Gobierno Vasco (P.A., M.L.M.-C, J.M.M.), PI11/01588, Sanidad Gobierno Vasco 2008, Educación Gobierno Vasco 2011 (M.L.M.-C), 2012 (J.M.M.), Sanidad Gobierno Vasco 2012 (M.V.R.), Basque Goverment IT-336-10 (to P.A. and X.B.), Saiotek S-PE11UN033 (P.A.), BBVA Foundation (to J.M.M., R.H.F.). Ciberehd is funded by ISCiii.
Address reprint requests to: José M Mato, CIC bioGUNE, Parque Tecnológico de Bizkaia, 48160 Derio, Bizkaia, Spain. E-mail: firstname.lastname@example.org; fax: +34-944-0611301.
Methionine adenosyltransferase 1A (MAT1A) and glycine N-methyltransferase (GNMT) are the primary genes involved in hepatic S-adenosylmethionine (SAMe) synthesis and degradation, respectively. Mat1a ablation in mice induces a decrease in hepatic SAMe, activation of lipogenesis, inhibition of triglyceride (TG) release, and steatosis. Gnmt-deficient mice, despite showing a large increase in hepatic SAMe, also develop steatosis. We hypothesized that as an adaptive response to hepatic SAMe accumulation, phosphatidylcholine (PC) synthesis by way of the phosphatidylethanolamine (PE) N-methyltransferase (PEMT) pathway is stimulated in Gnmt−/− mice. We also propose that the excess PC thus generated is catabolized, leading to TG synthesis and steatosis by way of diglyceride (DG) generation. We observed that Gnmt−/− mice present with normal hepatic lipogenesis and increased TG release. We also observed that the flux from PE to PC is stimulated in the liver of Gnmt−/− mice and that this results in a reduction in PE content and a marked increase in DG and TG. Conversely, reduction of hepatic SAMe following the administration of a methionine-deficient diet reverted the flux from PE to PC of Gnmt−/− mice to that of wildtype animals and normalized DG and TG content preventing the development of steatosis. Gnmt−/− mice with an additional deletion of perilipin2, the predominant lipid droplet protein, maintain high SAMe levels, with a concurrent increased flux from PE to PC, but do not develop liver steatosis. Conclusion: These findings indicate that excess SAMe reroutes PE towards PC and TG synthesis and lipid sequestration. (Hepatology 2013;58:1296–1305)
Nonalcoholic fatty liver disease (NAFLD) is the most common liver disease in Western countries, frequently being associated with obesity, dyslipidemia, and insulin resistance, a group of disorders that constitute the metabolic syndrome. Although these aforementioned conditions predispose the individual to develop NAFLD, our understanding of the mechanisms by which fat accumulates in the liver is not fully understood. Decreased content of S-adenosylmethionine (SAMe), the major biological methyl donor, has been linked to the development of NAFLD in different experimental models of steatosis in rodents and in humans. For instance, deletion of methionine adenosyltransferase 1A (Mat1a), the principal gene involved in hepatic SAMe biosynthesis, leads to a chronic reduction in liver SAMe level and to the spontaneous development of NAFLD. The mechanisms linking SAMe with lipid homeostasis are not obvious at first glance. However, two recent publications have shed light on this process by showing: (1) that low hepatic SAMe reduces phosphatidylcholine (PC) content, leading to SREBP-1 activation and lipogenesis; and (2) that low liver SAMe disrupts very low density lipoprotein (VLDL) assembly, leading to the synthesis of small, lipid-poor VLDL particles, and to a decrease in the secretion of triglycerides (TG).
The antisteatotic theory of SAMe has been challenged by the observation that deletion in mice of glycine N-methyltransferase (Gnmt), the main enzyme involved in hepatic SAMe catabolism, results in a marked increase in hepatic SAMe content and rapid NAFLD development. The Gnmt−/− mice show elevated serum aminotransferases at both 3 and 8 months of age. Histological examination of the livers of 3-month-old mutant mice showed steatosis and fibrosis, which were more pronounced in the livers of 8-month-old animals. At 8 months, Gnmt−/− mice also spontaneously developed multifocal hepatocellular carcinoma. This raises the question, How can SAMe be both anti- and prosteatotic at the same time? All mammalian cell types synthesize PC from choline and diglycerides (DG) via the CDP-choline pathway, but in hepatocytes PC is also synthesized by the sequential methylation of phosphatidylethanolamine (PE), a reaction catalyzed by the enzyme PE N-methyltransferase (PEMT). This reaction consumes three molecules of SAMe for each molecule of PC being formed. Herein we propose that as an adaptive response to the accumulation of liver SAMe, the synthesis of PC via PEMT is accelerated in Gnmt−/− mice, and that the excess PC generated is rerouted towards DG and TG synthesis and lipid sequestration (Fig. 1).
Consistent with this hypothesis, our present observations indicate that the flux from PE to PC is stimulated in the liver of Gnmt−/− mice, and that this produces a reduction in hepatic content of PE and a marked increase in DG and TG, with only a slight increase in hepatic PC. Conversely, reduction of hepatic SAMe level by feeding a methionine-deficient diet (MDD) reverted the flux from PE to PC of Gnmt−/− mice to that observed in wildtype (WT) animals and normalized the hepatic content of DG and TG, further confirming the steatotic effect of high SAMe concentrations. Importantly, Gnmt−/− mice with an additional deletion of perilipin2 (Plin2, previously known as Adfp or Adrp), a gene whose expression is induced upon Gnmt ablation, maintain high SAMe levels with a concurrent increased flux from PE to PC, but fail to develop liver steatosis. Plin2 is the predominant intracellular lipid droplet (LD) protein in hepatocytes, and a gene whose deletion protects against fatty liver.
Collectively, these findings indicate: (1) that SAMe regulates liver lipid homeostasis through a concerted collection of homeostatic actions that include: activation of lipogenesis and inhibition of TG secretion at low SAMe, and activation of TG synthesis via PEMT at high SAMe concentrations; and (2) that too much or too little SAMe can lead to an imbalance of these homeostatic actions and result in overt steatosis.
Materials and Methods
Three-month-old male Gnmt−/−, Plin2−/−, Gnmt−/−/Plin2−/− mice and their WT littermates were produced in the animal facility of bioGUNE. They were maintained on a rodent chow diet (Teklad Global, Diet 2018S), or an MDD (S8946-E020 EF AIN 76A 0,15% L-methionine, SSNIFF, Soest, Germany) for 21 days prior to being euthanized. Animal procedures were approved by the UPV/EHU and bioGUNE Animal Care and Use Committees.
Generation of Plin2−/− and Gnmt−/−/Plin2−/− Mice
Subjects consisted of male and female Plin2+/+, Plin2+/-, and Plin2−/− mice on a mixed 129SvEv/C57BL/6J background. Plin2 Gt(OST170322)Lex mutant mice (derived from OmniBank ES cell line OST170322) containing a gene trap vector inserted into the first intron of the Plin2 gene were obtained from the Texas A&M Institute for Genomic Medicine. Gnmt−/− mice were crossed to Plin2−/− mice to generate Gnmt−/−/Plin2−/− mice. A detailed description of the experimental procedure to generate Plin2−/− mice is provided as Supporting Information.
Hepatocytes were incubated with [3H]acetate (20 μM,20 μCi/mL), [3H]oleate (20 μM,2 μCi/mL) or [3H]ethanolamine (5 μCi/mL) as described. At the indicated times cells and medium were separated, lipids extracted, separated, and the label incorporated into lipids determined. A detailed description of the methods is provided as Supporting Information.
Ketone Bodies, Acid-Soluble Metabolites, and Glucose Measurements
Serum ketone bodies were quantified using a kit from Wako chemicals (Richmond, VA). Acid-soluble metabolites and glucose were measured as described[17, 18] and detailed in the Supporting Information section.
Hepatic TG Secretion Rate
To measure hepatic TG secretion, mice were injected with Poloxamer P-407 (Invitrogen, Carlsbad, CA) at 1 g/kg intraperitoneally as described. Prior to injection, and 2 and 6 hours after, blood samples were drawn, serum prepared, and TG concentrations determined. TG secretion rate was calculated from the difference in serum TG levels over the 6 hours following detergent injection.
Quantification of Lipids
Livers (300 mg) were homogenized and lipids extracted as described. TG were quantified using a kit (A. Menarini Diagnostics, Italy). PE, PC, and DG were separated by thin layer chromatography and quantified as described. Microsomes were isolated from liver samples (500 mg) and lipids extracted and quantified as detailed in the Supporting Information.
Liver lipid profiles were analyzed as described. Briefly, two separate UPLC-time-of-flight (TOF)-mass spectrometry (MS)-based platforms analyzing methanol and chloroform/methanol liver extracts were combined. Identified ion features in the methanol extract platform included nonesterified fatty acids (FA), acyl carnitines, bile acids, monoacylglycerophospholipids, monoetherglycerophospholipids, and oxidized FA. The chloroform/methanol extract platform provided coverage over glycerolipids, cholesteryl esters, sphingolipids, diacylglycerophospholipids, and acyl-ether-glycerophospholipids. Lipid nomenclature follows the LIPID MAPS convention, www.lipidmaps.org.
Data are represented as means ± standard error of the mean (SEM). Differences between groups were tested using Student t test. Significance was defined as P < 0.05.
GNMT Ablation Enhances TG Secretion
GNMT catalyzes the synthesis of sarcosine (methyl-glycine) from glycine, a reaction that consumes one molecule of SAMe for each molecule of sarcosine formed. Sarcosine has no known essential metabolic function and is demethylated, by mitochondrial sarcosine dehydrogenase, to regenerate glycine. The function of this futile cycle is to act as a cellular buffer that maintains a constant hepatic concentration of SAMe in order to avoid abnormal biological methylation reactions. Accordingly, Gnmt deletion results in an ∼40-fold elevation of hepatic SAMe and DNA hypermethylation. We reasoned that disruption of Gnmt would have no effect on hepatic lipogenesis, whereas it would enhance TG secretion. To prove this hypothesis, we assayed for de novo synthesis of TG in hepatocytes isolated from Gnmt−/− mice. This was accomplished by measuring the incorporation of [3H]acetate into TG (Fig. 2A), and in vivo hepatic TG secretion following inhibition of VLDL metabolism with poloxamer (P-407) (Fig. 2B). We also determined ketone bodies in serum (Fig. 2C) and the in vitro secretion of acid-soluble metabolites (Krebs cycle metabolites and ketones) (Fig. 2D), as a measure of FA β-oxidation. Whereas lipogenesis and FA β-oxidation were barely altered in hepatocytes from Gnmt−/− mice (Fig. 2A,C,D), the hepatic TG secretion rate in GNMT-depleted livers was elevated compared to livers from WT animals (Fig. 2B). Consistent with these studies, a comprehensive gene expression analysis showed that, overall, the expression of genes that supply NADPH and acyl-CoA for lipid synthesis was unaltered in mice without GNMT (Fig. 2E).
Despite the marked hepatic steatosis, mice without GNMT did not show insulin resistance or changes in serum FA concentrations (Supporting Fig. 1a,b). Depletion of GNMT in mice did not alter food intake or body weight (Supporting Fig. 1c,d). The greater liver weight in Gnmt−/− mice was not accompanied by differences in body weight, which may be explained by the reduced mass of the white adipose tissue in these animals (Supporting Fig. 1d-f).
Gnmt Ablation Activates the Flux from PE to PC Leading to the Accumulation of DG and TG
Based on the results depicted in Fig. 2, which demonstrate that Gnmt−/− mice have increased lipid secretion without affecting lipid synthesis or oxidation, it is not obvious how to explain the presence of fatty livers in these animals. We reasoned that an elevation of SAMe in Gnmt−/− mice would activate the flux from PE to PC via PEMT, which would lead to increased PC catabolism and the corresponding augmentation of hepatic DG and TG production (Fig. 2). To confirm this hypothesis, we measured the incorporation of [3H]ethanolamine into PE and PC in hepatocytes isolated from 3-month-old Gnmt−/− mice and calculated the radioactivity incorporated into PC as a percentage of the radiolabel incorporated into PC+PE (Fig. 3A). Because PC formed via PEMT primarily contains long-chain polyunsaturated FA (PUFA), such as docosahexaenoic acid (22:6n-3), whereas PC synthesized by the CDP-choline route do not, we also determined the PC(22:6n-3) to total PC ratio in GNMT-depleted and WT livers as a marker of hepatic PEMT activity. Given that PEMT activity is primarily located in the endoplasmic reticulum, we measured the content of PE and PC in whole liver microsomes (Fig. 3B,C). As shown in Fig. 3, high SAMe levels in Gnmt−/− hepatocytes associated with a 2.5-fold increase in the flux from PE to PC (P < 0.001) (Fig. 3A), and an increase in the PC(22:6)/PC ratio (from 0.18 ± 0.005 in WT to 0.25 ± 0.005 in GNMT-depleted livers, P = 3.23E-06). Also as predicted, the content of PE was reduced ∼2-fold in microsomes isolated from GNMT-depleted livers (P < 0.05), whereas the amount of PC was increased 2-fold (P < 0.05) (Fig. 3B,C). Measurement of whole liver PE and PC content revealed a 2-fold reduction in PE (P < 0.001) accompanied by a slight increase in PC (P < 0.05) in the Gnmt−/− mice as compared to WT animals (Fig. 3D,E). This suggests that once translocated from the microsomes to other cellular membranes, PC is rapidly catabolized and/or secreted in high-density lipoproteins (HDL). Accordingly, loss of GNMT increased hepatic DG and TG content (Fig. 4A,B), and HDL-PC levels (Supporting Fig. 2a).
Similar results were observed in 8-month-old Gnmt−/− mice, indicating that PE rerouting towards PC and TG synthesis is maintained during NAFLD progression (Supporting Fig. 3).
Importantly, feeding Gnmt−/− mice an MDD to reduce hepatic SAMe (Table 1), reverted the flux from PE to PC to that associated with WT animals (Fig. 3A), and restored normal PE, PC, DG, and TG levels (Figs. 3D,E, 4A,B), preventing steatosis (Supporting Fig. 2b). In hepatocytes isolated from Gnmt−/− mice, the inhibition of PEMT with 3-deazaadenosine (DZA) also resulted in decreased TG levels (Fig. 4C), These data strongly support the hypothesis that hepatic lipid accumulation in the absence of GNMT is best explained by the enhanced synthesis of PC via PEMT and the catabolism of these PC to TG. Finally, we observed that the incorporation of [3H]oleate into DG in hepatocytes from Gnmt−/− mice was increased 4-fold with respect to that found in WT animals (P < 0.0001) and that feeding an MDD normalized [3H]oleate incorporation into DG (Fig. 4D). Accordingly, Plin2, Cidec, Fitm1, and G0s2, which encode genes involved in lipid sequestration, and Scd1, an FA desaturase whose deletion protects form carbohydrate-induced steatosis, were upregulated in Gnmt−/− mice (Fig. 4E). Plin2, Cidec, and G0s2 are controlled by the prosteatotic transcription factor peroxisome proliferator-activated receptor γ (PPARγ), which was among the most prominently upregulated genes in mice without GNMT (Fig. 4E).
Table 1. Hepatic SAMe Content
SAMe ± SE, pmol/mg Liver Tissue
The values are from 3-month-old mice. Livers were collected and SAMe content determined as described. Data are from 4-6 animals per group. MDD, mice fed a methionine deficient diet for 3 weeks; WT, wildtype mice.
44.7 ± 18.6
65.1 ± 7.4
WT-Gnmt + MDD
56.4 ± 7.8
41.7 ± 4.3
38.7 ± 5.8
2,452 ± 158
Gnmt−/− + MDD
74.4 ± 12.6
1,667 ± 408
Deletion of Plin2 in Gnmt−/− Mice Inhibits Lipid Sequestration and Promotes Gluconeogenesis
Increased TG and the formation of LD are the two key features of fatty liver. LDs are intracellular storage places that protect neutral lipids, such as TG, DG, and cholesteryl esters, from unregulated degradation. Upon specific stimulation, TG undergoes a cycle of lipolysis and re-esterification prior to being assembled into VLDL. Fatty acids are also released from LD in order to be either oxidized and generate energy, or reutilized for the synthesis of cellular membranes. Perilipin (PAT)-domain proteins form a conserved family of proteins that are localized at the surface of neutral LD. PAT-proteins not only stabilize LD, but also protect their neutral lipids from degradation.[12, 13] PLIN2, a member of the PAT-family, is the predominant neutral LD protein in hepatocytes.[12, 13] PE methylation via PEMT has been shown to promote LD formation and stability in adipocytes and adipose tissue. Importantly, we previously noticed that the hepatic expression of Plin2 was increased following Gnmt ablation. We reasoned that increased PC synthesis via PEMT would promote LD biosynthesis and steatosis in Gnmt−/− livers. To prove this, we created double knockout mice by crossing Plin2−/− mice with Gnmt−/− mice to produce a novel Gnmt−/−/Plin2−/− double knockout mouse model, in which we determined the hepatic SAMe content (Table 1) and the levels of PE and PC (Fig. 5). Furthermore, we also determined the hepatic content of DG and TG (Fig. 5). As shown in Table 1, Plin2 deletion had no effect on hepatic SAMe concentration, as the double knockout mice showed a 40-fold elevation (P < 0.0001) in SAMe, which was similar to that observed in the Gnmt−/− animals. Consistent with this, total liver PE content was reduced 2-fold (P < 0.01) in Gnmt−/−/Plin2−/− mice, whereas PC levels remained normal (Fig. 5A,B), suggesting that PC was rapidly catabolized just as in the Gnmt−/− animals. In contrast to the situation in Gnmt−/− mice, while DG levels in the double knockout mice were significantly elevated (P < 0.01), the TG content actually underwent a 2-fold reduction (P < 0.05) (Fig. 5C,D). As expected, Gnmt−/−/Plin2−/− mice failed to develop hepatic steatosis (Fig. 5E) despite having high hepatic SAMe concentration (Table 1) and reduced PE levels (Fig. 5). Inhibition of lipid sequestration in Gnmt−/− mice decreased lipogenesis, had a minor effect on secretion of acid-soluble metabolites, decreased serum ketone bodies, yet maintained a higher hepatic TG secretion rate (Fig. 6A-C,E). The finding that the concentration of acid-soluble metabolites did not change, whereas serum ketone bodies were reduced, suggests that acetyl-CoA generated via β-oxidation is driven towards the Krebs cycle and gluconeogenesis (Fig. 6B,C). Accordingly, glucose production in the absence or presence of the precursors lactate/pyruvate and glycerol was increased in hepatocytes without GNMT and PLIN2 (Fig. 6D).
Identification of SAMe-Regulated Lipids
In keeping with the lipid tracing studies, a comprehensive lipidomic analysis of livers from control diet Gnmt−/−, MDD-treated Gnmt−/−, and Gnmt−/−/Plin2−/− mice was performed and compared with that of their corresponding WT animals. Increased SAMe is characterized by a marked remodeling of lipid composition (Fig. 7, Supporting Table 1). These changes included an increase in TGs that are rich in PUFA(18:2, 20:2, 20:4, 22:4, 22:5, 22:6), of DG such as DG(18:1+18:1), DG(16:0+20:4), and DG(16:0+18:1), of ceramides such as Cer(d18:1/18:0), and of free unsaturated FA (UFA)(16:1n-x, 18:1n-9, 20:3n-3, and 22:4n-6); and a marked decrease in PE rich in PUFA, and of a variety of sphingomyelins such as SM(d18:1/22:0), SM(d18:1/21:0), and SM(d17:1/22:0). We found that, after MDD treatment, Gnmt−/− mice revealed a lipidomic signature that resembled the signature presented by WT mice (Fig. 7, Supporting Table 1). Gnmt−/−/Plin2−/− mice showed also a reversal of the TG and UFA signatures, although the profile of the DG, PE, and sphingomyelins in these animals resembled the lipidomic signature of Gnmt−/− mice, and the PC signature was only partially reverted (Fig. 7, Supporting Table 1). These results are consistent with the observation that these double knockout mice maintain high SAMe levels with a concurrent increased flux from PE to PC, but do not develop liver steatosis.
It has recently been suggested that PC made via PEMT may be an important source of hepatic TG. Herein we investigated the role of SAMe on TG homeostasis and found that activation of PEMT by an excess of hepatic SAMe leads to increased TG synthesis and ultimately to liver steatosis. For these studies we used Gnmt−/− mice, a model we developed which is characterized by the high hepatic content of SAMe, and the rapid development of fatty liver. We initially observed that hepatocytes from Gnmt−/− animals maintained on a normal diet present with normal lipogenesis and FA β-oxidation but with increased TG secretion. This is consistent with recent observations showing that low SAMe stimulates de novo lipogenesis in the liver, while impairing VLDL assembly and TG secretion. We subsequently observed that the flux from PE to PC via PEMT was markedly stimulated in Gnmt−/− hepatocytes, which is in perfect agreement with the observation that this model is characterized by a 40-fold increase in hepatic SAMe concentration. Concurrent with PEMT's cellular localization, we observed a reduction in microsomal PE concomitant with an increase in PC in the Gnmt−/− mice. Whereas the mass of PE in whole liver was reduced 2-fold, PC content was only slightly increased, suggesting an increase in PC catabolism and/or secretion in HDL in mutant mice. Accordingly, both the mass of hepatic DG and TG and the serum HDL-PC levels increased in Gnmt−/− mice compared to WT animals. These results are consistent with recent findings showing that membrane PC can be an important precursor of liver TG under normal physiological conditions.
Administration of an MDD to Gnmt−/− mice normalized hepatic SAMe content, which is fully consistent with our understanding of the function of GNMT in methionine catabolism. More important, in addition to lowering hepatic SAMe, placing the Gnmt−/− mice on an MDD also served to restore normal hepatic lipid metabolism, indicating that SAMe is the rate-limiting substrate linking PE with TG via PC and DG. The observation that PEMT inhibition with DZA in Gnmt−/− hepatocytes resulted in decreased TG levels further supports the role of PEMT in this process. These results demonstrate for the first time that PEMT is an unexpected source of hepatic TG, particularly in cases of SAMe excess. We conclude that stimulation of the PEMT pathway induced by high SAMe can explain hepatic steatosis in Gnmt mutant mice. This finding is clearly relevant to human disease, as children with GNMT mutations were recently identified to suffer from liver injury.[29, 30] Although it is not yet known if PC made via PEMT is quantitatively an important source of hepatic TG under normal physiological conditions, there are situations other than Gnmt deletion that can augment hepatic SAMe content, such as an increase of dietary methionine or cystathionine β-synthase (CBS) ablation. Interestingly, Cbs−/− mice exhibit severe liver injury, steatosis, and fibrosis.
In a similar manner, we also investigated the role of PLIN2 on a background of high SAMe. For these studies we generated a novel, double Gnmt−/−/Plin2−/− knockout mouse model that is characterized by high hepatic SAMe and low PE content, much like Gnmt−/− mice, that in contrast did not develop fatty liver. Consistent with previous findings,[12, 13] knockout of Plin2 in GNMT-depleted livers decreased lipogenesis and increased TG secretion. Plin2 ablation increased gluconeogenesis, indicating that crosstalk exists between lipid synthesis and sequestration and glucose metabolism. Since PE methylation has been shown to promote LD formation, these results support a model where increased PEMT activity induces both TG synthesis and its accumulation into newly formed LD.
Collectively, these observations, taken in light of previous findings,[5, 6] demonstrate that SAMe regulates liver lipid homeostasis through a concerted series of homeostatic actions that include: activation of lipogenesis and inhibition of TG secretion at low SAMe, and activation of TG synthesis via PEMT at high SAMe. This cascade of events goes a long way towards explaining why a chronic imbalance in hepatic SAMe synthesis, or catabolism, is capable of inducing NAFLD.
We thank Virginia Gutiérrez de Juan and Begoña Rodríguez-Iruretagoyena for technical support and Azucena Castro for discussion; and human and technical support from Unidad de formación e investigación UFI11/20, University of Basque Country.