Potential conflict of interest: Nothing to report.
Supported by Canadian Institutes of Health Research Operating Grant 221920 (to G. A. M.).
Accumulation of cytoplasmic triacylglycerol (TG) underlies hepatic steatosis, a major cause of cirrhosis. The pathways of cytoplasmic TG metabolism are not well known in hepatocytes, but evidence suggests an important role in lipolysis for adipose triglyceride lipase (ATGL). We created mice with liver-specific inactivation of Pnpla2, the ATGL gene. These ATGLLKO mice had severe progressive periportal macrovesicular and pericentral microvesicular hepatic steatosis (73, 150, and 226 μmol TG/g liver at 4, 8, and 12 months, respectively). However, plasma levels of glucose, TG, and cholesterol were similar to those of controls. Fasting 3-hydroxybutyrate level was normal, but in thin sections of liver, beta oxidation of palmitate was decreased by one-third in ATGLLKO mice compared with controls. Tests of very low-density lipoprotein production, glucose, and insulin tolerance and gluconeogenesis from pyruvate were normal. Plasma alanine aminotransferase levels were elevated in ATGLLKO mice, but histological estimates of inflammation and fibrosis and messenger RNA (mRNA) levels of tumor necrosis factor-α and interleukin-6 were similar to or lower than those in controls. ATGLLKO cholangiocytes also showed cytoplasmic lipid droplets, demonstrating that ATGL is also a major lipase in cholangiocytes. There was a 50-fold reduction of hepatic diacylglycerol acyltransferase 2 mRNA level and a 2.7-fold increase of lipolysosomes in hepatocytes (P < 0.001), suggesting reduced TG synthesis and increased lysosomal degradation of TG as potential compensatory mechanisms. Conclusion: Compared with the hepatic steatosis of obesity and diabetes, steatosis in ATGL deficiency is well tolerated metabolically. ATGLLKO mice will be useful for studying the pathophysiology of hepatic steatosis. (HEPATOLOGY 2011;)
Nonalcoholic fatty liver is the most common chronic liver disease in the United States. Beginning as hepatic steatosis, it leads to fibrosis, cirrhosis, and hepatocarcinoma.1-5 Excessive energy consumption is a major cause of hepatic steatosis. In mouse studies,6, 7 fatty liver is typically induced by high fat and/or carbohydrate intake, dietary methionine restriction, or hormonal or immunological manipulation. Under these conditions, cytoplasmic triacylglycerol (TG) metabolism in the liver is not specifically modified. Instead, hepatic steatosis is one finding among several that occur in response to these systemic external stresses.
Despite the medical importance of hepatic steatosis, the pathways of cytoplasmic TG synthesis and degradation in hepatocytes remain unclear. They are best known in white adipose tissue.8, 9 TG synthesis and lipolysis are distinct pathways. Adipose triglyceride lipase (ATGL), a lipid droplet surface protein,10 is physiologically the main TG lipase of adipose tissue.11 Hormone-sensitive lipase (HSL) is the main diacylglycerol hydrolase.12 A monoglycerol lipase cleaves monoacylglycerols to glycerol and fatty acid (FA). Other lipases include adiponutrin, triglyceride hydrolase (TGH), and lysosomal acid lipase.
ATGL is highly expressed in white and brown adipose tissue, but also in muscle, heart, and liver. The main symptom of ATGL-deficient humans is lipid myopathy.13 Cardiomyopathy can also occur.14 The liver phenotype of ATGL-deficient patients has not been reported in detail. However, liver ATGL expression is decreased in hepatic steatosis patients.15 In mice, generalized ATGL deficiency causes TG deposition in multiple organs, including liver, with 50% mortality from lipid cardiomyopathy by 16 weeks.16 ATGL overexpression increased FA oxidation in mouse liver,17 reduced cellular TG in McA-RH7777 cells,17 and decreased hepatic TG.18 Conversely, mice deficient in other known neutral TG hydrolases, including TGH,19 HSL,20 and adiponutrin,21 do not have hepatic steatosis.
These observations suggest a possible inverse relationship between the expression of ATGL and hepatic TG content. If ATGL is the major cytoplasmic TG hydrolase in the liver, then isolated hepatic ATGL deficiency should cause steatosis. We created liver-specific ATGL-deficient mice and studied their long-term course.
Materials and Methods are described in the Supporting Information.
After obtaining gene targeting and germline transmission (Supporting Information and Supporting Fig. 1), we bred mice that were homozygous for the targeted allele and that also expressed a Cre recombinase transgene from the liver-specific albumin promoter (ATGLLKO mice). Liver DNA from ATGLLKO mice showed apparently complete excision of exon1 of the Pnpla2 (Fig. 1A), which encodes the start codon and catalytically essential residues of ATGL.11 Removal of this exon is predicted to completely inactivate ATGL. Liver ATGL messenger RNA (mRNA) levels were 1.6% of normal (Table 1). In ATGLLKO liver, ATGL protein was undetectable by way of western blotting (Fig. 1B), and cytoplasmic TG hydrolase activity was reduced by 65% compared with control liver (P < 0.01) (Fig. 1C).
Table 1. Selected Liver Transcripts
Samples were from 4-month-old mice fasted for 6 hours. Data are presented as the mean ± SEM (n = 6).
ATGLLKO Mice Have Marked Progressive Hepatic Steatosis.
Under standard conditions, viability was normal in ATGLLKO mice followed until 12 months. After a 6-hour fast in 3-month-old mice, plasma glucose, FA, cholesterol, and 3-hydroxybutyrate levels were similar to those of controls (Table 2). ATGLLKO mice had greater liver mass and three-fold higher TG content than controls at all ages studied (Fig. 2A-C). TG contents of heart (Fig. 2D) and skeletal muscle (data not shown) were normal, as were white and brown adipose tissue masses (data not shown). Histologically, periportal macrovesicular steatosis and pericentral microvesicular steatosis were observed in ATGLLKO livers (Fig. 3A-3D).
Table 2. Plasma Metabolites After a 6-Hour Fast
Results of tail vein blood taken from 12- to 13-week-old mice (n = 12). Data are presented as the mean ± SEM and are expressed in mmol/L.
7.84 ± 0.58
8.13 ± 0.50
0.53 ± 0.02
0.52 ± 0.05
0.46 ± 0.05
0.50 ± 0.04
1.51 ± 0.25
2.02 ± 0.23
0.06 ± 0.02
0.06 ± 0.02
ATGLLKO cholangiocytes also contained cytoplasmic lipid droplets (Fig. 3E), which were absent in controls. Plasma GGT levels were normal in ATGLLKO mice (data not shown).
ATGLLKO Mice Have Elevated Plasma Alanine Aminotransferase Levels but Normal Macrophage Numbers, Fibrosis, and Apoptosis.
ATGLLKO mice had higher plasma alanine aminotransferase (ALT) levels than controls (Fig. 4A) and a higher ALT/aspartate aminotransferase (AST) ratio (Fig. 4B). Histological examination of livers from 4-, 8-, and 12-month-old mice showed scattered foci of macrophage infiltration at 8 and 12 months to a similar extent in ATGLLKO and control livers (Fig. 4C,D). No signs of acute or chronic inflammation were present in ATGLLKO liver. Masson trichrome staining revealed no fibrosis (data not shown). Terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling staining showed normal counts of apoptotic cells at 8 months (Supporting Fig. 2) and 12 months (data not shown). In 4- and 8-month-old mouse livers, tumor necrosis factor α and interleukin-6 mRNAs were normal or decreased in ATGLLKO mice (Fig. 4E).
Similar Energy Homeostasis in ATGLLKO and Control Mice.
Insulin tolerance tests at 4 months of age were similar in ATGLLKO and control mice, both under normal diet (Fig. 5A) and high-fat diet (HFD) conditions (data not shown). Glucose tolerance tests were similar in normal diet–fed ATGLLKO and control mice at 4 (Fig. 5B), 8, and 12 months of age (Supporting Fig. 3A,B). In HFD-fed mice, there was no significant difference in glucose tolerance between ATGLLKO and control mice (data not shown). Gluconeogenesis from pyruvate was normal in ATGLLKO mice (Fig. 5C). Very low-density lipoprotein (VLDL) production, evaluated as the increase in plasma TG following injection of a lipoprotein lipase inhibitor (Fig. 5D) did not differ significantly between ATGLLKO mice and controls. Beta-adrenergic–stimulated in vivo adipose tissue lipolysis was normal in ATGLLKO mice (Fig. 5E).
Unlike constitutively ATGL-deficient mice,16 ATGLLKO mice tolerate prolonged fasting. Calorimetry showed no significant difference in oxygen consumption or respiratory exchange ratio (RER) between ATGLLKO mice and controls during a 48-hour fast (Fig. 6A,B). Heat production was also similar except at 48 hours, when it was lower in ATGLLKO mice than in controls (Fig. 6C). Measurements of activity were similar in ATGLLKO and normal mice (data not shown). After a 48-hour fast, plasma nonesterified FA levels were higher in ATGLLKO mice than in controls, but 3-hydroxybutyrate was as high in ATGLLKO mice as in controls (Table 3).
Table 3. Plasma Metabolites After a 48-Hour Fast
NA, not available.
Results of tail vein blood taken from 12- to 13-week-old mice (n = 12). Data are presented as the mean ± SEM and are expressed in mmol/L. **P ≤ 0.01.
5.82 ± 0.68
5.05 ± 0.68
1.21 ± 0.14
1.76 ± 0.09**
0.59 ± 0.06
0.66 ± 0.09
0.96 ± 0.11
1.08 ± 0.19
ATGLLKO Livers Have Low Levels of Peroxisome Proliferator-Activated Receptor α, Carnitine Palmitoyltransferase-1α, and Diacylglycerol Acyltransferase-2 mRNAs.
In ATGLLKO liver, mRNA levels of transcription factors related to FA and energy metabolism showed a marked reduction in peroxisome proliferator-activated receptor α (PPARα) level (Table 1). Despite the normal fasting 3-hydroxybutyrate level in ATGLLKO mice, carnitine palmitoyltransferase-1α (CPT-1α) mRNA was markedly decreased (Table 1). mRNA levels of liver lipases other than ATGL were normal (Table 1). TGH protein levels were also normal (Supporting Fig. 4A). HSL is minimally expressed in the liver and is undetectable in western blots of ATGLLKO or control mice (data not shown). In contrast, diacylglycerol acyltransferase-2 (DGAT2) mRNA was markedly decreased, whereas that of DGAT1 was mildly increased. The level of microsomal triglyceride transfer protein, which is essential for VLDL synthesis, was normal (Supporting Fig. 4B).
Decreased FA Oxidation in ATGLLKO Livers.
In liver slices, measures of FA oxidation were approximately one-third lower in ATGLLKO than in control tissue (Fig. 6D).
ATGLLKO Liver Ultrastructure Reveals Increased Numbers of Lipolysosomes.
On electron microscopy, ATGLLKO liver showed increased size and number of hepatocyte lipid droplets. Mitochondria appeared normal. Lipolysosomes were more abundant in ATGLLKO than in control hepatocytes (Fig. 7A-C). mRNA levels for atg5, atg7, and LC3-II were similar in ATGLLKO mice and controls (Table 1). Western blotting revealed normal levels of the lysosomal membrane protein LAMP2 (Supporting Fig. 5). Counts of lipolysosomes in ultrastructural sections of hepatocytes revealed a three-fold increase in ATGLLKO hepatocytes versus littermate controls (P < 0.001) (Fig. 7D).
ATGLLKO mice have marked hepatic steatosis at all ages studied but are healthy otherwise. Gene targeting in ATGLLKO liver appears to be complete (Fig. 1) and tissue specific. In ATGLLKO mice, cardiac TG content, adipose lipolysis, fasting tolerance, white and brown adipose weights and viability until at least 1 year of age are all normal. Each of these parameters is strikingly abnormal in constitutive ATGL knockout mice, which die of cardiomyopathy at approximately 4 months, the age of the youngest cohort described in this article.16, 22 During preparation of this manuscript, Ong et al.18 demonstrated that adenoviral-mediated ATGL knockdown causes detectable hepatic steatosis within 1 week. Our results support and extend these groups' findings, providing the first description of the long-term course of a primary hepatic steatosis.
Hepatic ATGL deficiency increased liver TG content approximately three-fold at all ages studied. The levels of steatosis observed in ATGL deficiency were greater than all but the most severe, chronic forms of HFD-induced steatosis.21, 23-25 The steatosis of ATGLLKO mice is concentrated in the periportal and central zones, suggesting that ATGL exerts its greatest effect in these regions. Intriguingly, ATGLLKO cholangiocytes also accumulate excess cytoplasmic TG. This unique change was well-tolerated, with normal gamma-glutamyltransferase levels and lack of periductal inflammation or fibrosis. Of note, ATGL deficiency is expected to be present in ATGLLKO cholangiocytes and hepatocytes, because both arise from a common precursor that expresses albumin,26, 27 allowing gene excision in both cell types. These observations strongly suggest that ATGL is physiologically the main cytoplasmic TG hydrolase of both hepatocytes and cholangiocytes.
The elevation of plasma ALT levels in ATGLLKO mice was similar to or less than that observed at similar levels of steatosis in mice with diet-induced obesity.28, 29 In contrast, the inflammatory and fibrotic responses in ATGLLKO liver were mild and less than those reported for similar degrees of steatosis in diet-induced obesity.28, 30-33
Insulin and glucose tolerances were normal in ATGLLKO mice, showing that overall body energy homeostasis is preserved despite hepatic ATGL deficiency. In contrast, constitutive ATGL-deficient mice have increased insulin sensitivity compared with controls.16 This finding has been attributed to enhanced insulin sensitivity in muscle.34 The lack of insulin sensitivity of ATGLLKO mice is consistent with this finding, suggesting that the insulin sensitivity of ATGL−/− mice is not of hepatic origin.
Despite the marked steatosis of ATGLLKO mice, the mainstreams of hepatocyte FA flux were preserved. The normal fasting oxygen consumption, RER, heat production, and fasting tolerance in ATGLLKO mice, and their normal level of 3-hydroxybutyrate after a 48-hour fast, demonstrate that substantial rates of mitochondrial beta oxidation and ketogenesis are possible in ATGLLKO mice. Furthermore, their hepatic mitochondrial ultrastructure is normal. In addition, gluconeogenesis, which is fueled by reducing equivalents from FA oxidation,35 was normal in ATGLLKO mice. This contrasts with the low levels of PPARα and CPT-1α mRNAs, which are predicted to reduce FA oxidative capacity. Direct measurement of FA oxidation in liver slices showed 31% less carbon dioxide production in ATGLLKO than in normal liver (Fig. 6D), consistent with a reduced capacity for FA oxidation in ATGLLKO liver. The residual oxidative capacity of ATGLLKO liver appears adequate to meet most physiological demands, including 48-hour fasting.
VLDL production is the other main fate of FA in hepatocytes. It appeared to be normal in ATGLLKO mice (Fig. 5D). Plasma TG concentrations were normal in fed and fasted ATGLLKO mice (Tables 2 and 3). Levels of microsomal triglyceride transfer protein and TGH, two microsomal proteins implicated in VLDL production, were normal (Supporting Fig. 4). In addition, following injection of a lipoprotein lipase inhibitor, plasma TG levels increased at similar rates in ATGLLKO mice and controls (Fig. 5D). Of note, a similar dissociation of hepatic steatosis and metabolic abnormalities has also been observed in mice that overexpress DGAT2, which develop steatosis but are protected from the metabolic changes associated with HFD-induced obesity.36
Our findings are highly complementary to and extend those of Ong et al.18 That group studied mice 7 days after adenoviral-mediated knockdown of ATGL, whereas we studied chronic genetic ATGL deficiency. In each model, increased liver TG content, decreased TG hydrolase activity, lower rates of FA oxidation, and similar VLDL secretion were found in ATGL-deficient mice compared with controls. The severity of steatosis varied six- to seven-fold from the level reported by Ong et al. (0.22 mg/g protein, ≈3% fat by weight) to >200 μmol/g (≈20% fat) in our 12-month-old ATGLLKO mice (Fig. 2C).
Although the steatosis of ATGLLKO mice was impressive, if ATGL mediates the only pathway of cytoplasmic TG hydrolysis, and if other pathways of TG metabolism are unchanged, an even more severe steatosis would by predicted. This suggests that compensatory changes occur in other pathways than cytoplasmic lipolysis. As discussed above, VLDL production was not increased and the capacity for beta oxidation was reduced. Thus, neither of these pathways is likely to limit the severity of steatosis in ATGLLKO liver. Two other processes, or a combination of both, may explain this attenuation: (1) reduction of TG synthesis and/or (2) non–ATGL-dependent TG degradation. Two observations are consistent with reduced TG synthesis. Fasting FFA levels were increased in ATGLLKO mice (Table 3), which we speculate may reflect reduced liver uptake of FAs for TG synthesis. In addition, the marked reduction of DGAT2 mRNA (Table 1) may reflect a decreased capacity for TG synthesis, because DGAT2 is thought to mediate the main reaction of cytoplasmic TG synthesis.37
Esterases other than ATGL may also contribute to TG hydrolysis in the liver. HSL is another cytoplasmic lipase, but it is difficult to muster evidence for a major HSL-dependent effect. HSL is expressed at very low levels in liver and is mainly a diacylglycerol hydrolase.12 HSL-deficient mice do not have constitutive hepatic steatosis.20 An alternative pathway might involve internalization and degradation of cytoplasmic TG in lysosomes. Lysosomal acid lipase can cleave TG, diacylglycerols, and monoacylglycerols.38 Autophagy and lysosomal TG degradation have been shown to intensify during fasting and under HFD conditions.39 The observation of abundant lipolysosomes in ATGLLKO hepatocytes suggests that lysosomal TG metabolism may be a factor in reducing steatosis.
In ATGLLKO mice, hepatic steatosis is the primary and direct result of liver ATGL deficiency. Therefore, it clearly differs from the complex multisystemic conditions in which fatty liver occurs clinically and in most experimental situations. ATGLLKO mice may be helpful in distinguishing the effects of hepatic steatosis itself from those of the extrahepatic changes that accompany most models of hepatic steatosis.
We thank Natalie Patey for help with histology, Josée Marie Dubbé for technical assistance with electron microscopy, and André Tremblay for interesting discussions.