The major finding of this study is that transgenic mice overexpressing PNPLA3I148M
in the liver recapitulate the fatty liver phenotype as well as other metabolic features associated with this allele in humans. Expression of the 148M allele in mice did not increase body weight or body fat content and was not associated with the adverse metabolic sequelae that often accompany fatty liver, such as insulin resistance or hypertriglyceridemia (2
). These results parallel the lack of association between the I148M polymorphism and BMI, insulin resistance, or plasma TAG levels observed in humans (7
) and suggest that the metabolic mechanisms by which the 148M allele promotes liver fat accumulation may be similar in mice and humans. Metabolic studies in the transgenic mice revealed that high level expression of PNPLA3I148M
in the liver, but not in adipose tissue, affected both hepatic TAG synthesis and catabolism. A surprising finding was that the PNPLA3I148M
transgenic mice have significantly increased fatty acid synthesis and an altered spectrum of TAG–fatty acids in the liver, with no evidence of insulin resistance.
The finding that hepatic TAG levels are normal in mice lacking (24
) or overexpressing wild-type PNPLA3 (Figure ) indicates that the I148M substitution does not cause steatosis by simply increasing or decreasing the normal activity of the enzyme. Metabolic studies performed in vivo and in primary cultured hepatocytes also support a more complex effect of the I148M substitution on hepatic TAG metabolism. Tracer studies provided evidence for increased synthesis and decreased hydrolysis of hepatic TAG in 148M transgenic mice. Changes in the gene expression profile and the fatty acid composition of hepatic TAG in the PNPLA3I148M
transgenic mice were consistent with increased de novo fatty acid synthesis in these animals. Taken together, our data indicate that the increased liver fat associated with the I148M substitution is the cumulative result of multiple changes in hepatic TAG metabolism.
Endogenous PNPLA3 is expressed primarily in adipose tissue in mice (27
), whereas humans express the gene at highest levels in the liver (28
). We failed to observe any consequence of overexpressing wild-type or mutant human PNPLA3 in adipose tissue, despite high levels of transgene expression. No changes were seen in morphology or function of either white or brown fat, and cold tolerance was preserved in the A-PNPLA3WT
animals. The increased liver fat content observed in the liver-specific PNPLA3I148M
transgenic mice, together with the lack of phenotype in A-PNPLA3I148M
mice, is consistent with the notion that the fatty liver phenotype associated with the 148M variant is due to the action of the allele in the liver, rather than adipose tissue.
The finding that PNPLA3I148M
causes liver fat accumulation when overexpressed in liver, but not in adipose tissue, is consistent with our previous observation that adenovirus-mediated expression of the 148M allele in liver results in increased hepatic fat content (22
). The increase in liver TAG levels in the PNPLA3I148M
transgenic mice was lower than that observed in mice infected with PNPLA3I148M
adenovirus, where hepatic TAG levels increased approximately 5-fold after just 3 days (22
). These data indicate that expression of the 148M allele results in a rapid change in hepatic TAG metabolism that quickly stabilizes at a new equilibrium.
Why do PNPLA3I148M
transgenic mice develop steatosis on a chow or sucrose diet but not on a high-fat diet? Ingestion of sucrose stimulates de novo synthesis of fatty acids (48
), whereas most of the hepatic fatty acids in livers of fat-fed mice are derived from circulating NEFAs. Perhaps PNPLA3 in hepatocytes is exposed preferentially to newly synthesized TAG and is shielded from fatty acids that enter the liver in lipoproteins or are synthesized from circulating NEFAs. Alternatively, PNPLA3 may function specifically under conditions of insulin-stimulated lipid anabolism. The finding that PNPLA3 is virtually absent from livers of fasting animals and is strongly upregulated both transcriptionally (23
) and posttranslationally (28
) by carbohydrate refeeding is consistent with the latter hypothesis.
It remains possible that the large increase in hepatic TAG levels that occurred in the fat-fed mice (4-fold increase) obscured the differences between the lines. Arguing against this interpretation is the prior observation that hepatic TAG content was increased when the I148M variant was expressed at high levels in the ob/ob
mice, in which the liver fat levels are appreciably higher than are those of the fat-fed mice in this study (22
PNPLA3 transgene expression had no detectable effect on 2 of the major pathways used by the liver to remove TAG: VLDL secretion and fatty acid oxidation. TAG hydrolysis (measured as glycerol release) was decreased in hepatocytes from PNPLA3I148M
transgenic mice. This decrease in lipolysis in hepatocytes is consistent with the reduction in TAG hydrolysis by the purified recombinant 148M isoform we observed previously in vitro (22
). However, liver TAG levels are not increased in Pnpla3–/–
mice, as would be expected if PNPLA3 was rate limiting for hepatic TAG hydrolysis (24
). Thus, the increased liver fat associated with hepatic expression of the 148M allele cannot be due simply to a loss of normal enzyme activity. The 148M allele may have an indirect effect and inhibit lipolysis mediated by other enzymes, such as ATGL, possibly by sequestering a required cofactor or by altering the composition of the lipid droplet.
Nor can the increased hepatic TAG associated with the variant allele be due to an increase in the normal enzyme activity, since we did not observe any increase in hepatic TAG in the mice expressing very high levels of the wild-type enzyme. Kumari et al. (26
) reported that a PNPLA3-trigger factor fusion protein catalyzes the acylation of LPA to form PA, a key step in the glycerophosphate pathway for TAG biosynthesis, and that the 148M substitution increases that activity. This gain of function could account for the liver steatosis associated with the 148M allele. We have confirmed that recombinant PNPLA3 has LPAAT activity in vitro when expressed as a fusion protein with trigger factor (data not shown), though we do not observe an increase in LPAAT activity when the recombinant protein is expressed in cultured cells (23
). Our data are not compatible with a model in which expression of the PNPLA3I148M
variant causes hepatic steatosis simply by increasing hepatic LPAAT activity. We failed to observe a significant, reproducible increase in LPAAT activity in either the membranes or the lipid droplets isolated from the PNPLA3 transgenic mice (Figure E). Nor did we observe an increase in hepatic PA formation in vivo, as determined by tritium incorporation from 3
O (Figure ). Although liver LPA levels were decreased in the transgenic mice (Figure A), the decrease was similar in PNPLA3WT
transgenic and PNPLA3I148M
transgenic animals, yet the PNPLA3WT
transgenic animals did not have increased hepatic TAG levels. Liver TAG levels in Pnpla3–/–
mice are similar to those of WT animals, despite an approximately 50% reduction in LPAAT activity (26
). Moreover, overexpression of wild-type PNPLA3 did not increase hepatic TAG concentrations, as would be expected if the enzyme promoted TAG biosynthesis through increased LPAAT activity.
Direct measurements of TAG biosynthesis using glycerol, acetate, and oleate did not reveal differences in the rate of synthesis of hepatic TAG in hepatocytes from the 3 mouse lines. Previous studies have shown that primary hepatocytes from mice lose their responsiveness to insulin/SREBP-1–mediated stimulation of fatty acid and TAG biosynthesis within few hours of being in culture (Jay Horton, personal communication). Since the effect of the transgene is most apparent under conditions in which SREBP-1c is activated (high-sucrose diets), the SREBP pathway may play an important role in 148M-induced steatosis. Accordingly, we examined the 2 major pathways of TAG biosynthesis in intact mice using glycerol and tritiated water as tracers. 3
H-labeled glycerol was incorporated into hepatic TAG at similar rates in all 3 lines. In contrast to these results, incorporation of tritium from 3
O into TAG was significantly increased in the PNPLA3I148M
transgenic animals, although no differences were found in the incorporation of tritium into PA or DAG. A possible explanation for these seemingly different results is that the 2 tracers label different pathways. Glycerol is converted to glycerophosphate before it is incorporated into LPA, PA, DAG, and ultimately TAG (the glycerophosphate pathway). Tritiated water is incorporated into fatty acids and thus is incorporated into TAG via 2 pathways: the glycerophosphate pathway and by reesterification of monoglycerides with acyl-CoAs (the monoglyceride pathway). The increased incorporation of 3
O, but not glycerol, into hepatic TAG in the PNPLA3I148M
transgenic mice may reflect increased biosynthesis of hepatic TAG via the monoglyceride pathway. The finding that MGAT1
mRNA levels were increased in the PNPLA3I148M
transgenic mice is consistent with this possibility. Humans with NAFLD have recently been reported to have increased hepatic MGAT activity (51
Levels of mRNA encoding the lipid droplet–associated protein CIDEC, which were increased in the livers of PNPLA3I148M
transgenic mice, are also upregulated in other mouse models of fatty liver disease (52
). Expression CIDEC is associated with an increase in the size of cytosolic lipid droplets in various cell types (53
). Whether the increased expression of CIDEC contributes to the increased lipid droplet size in this animal model will require additional studies.
Purified recombinant PNPLA3 has been found to have 5 enzymatic activities: TAG, DAG, and MAG hydrolysis (22
) as well as acyl-CoA thioesterase (23
) and LPAAT activity (26
). These different activities are not equally affected by the I148M substitution. In in vitro assays, the I148M substitution results in a substantial loss of TAG, MAG, and DAG hydrolytic activity (23
); a modest reduction in acyl-CoA thioesterase activity (23
); and an increase in LPAAT activity (26
). Considered in isolation, none of these activities alone can explain all of the changes in TAG metabolism observed in the PNPLA3I148M
transgenic mice. Therefore, some of the metabolic changes observed in these animals are likely to be secondary rather than direct consequences of altered PNPLA3 activity.
The spectrum of TAG–fatty acids in the livers of both transgenic lines of mice provides further evidence that PNPLA3 expression affects hepatic TAG composition as well as quantity. The TAG–fatty acid profiles of the PNPLA3I148M
transgenic mice (and to a lesser extent the PNPLA3WT
transgenic mice) differ significantly from those of wild-type mice and are qualitatively similar to those observed when SREBP-1c is overexpressed in liver (54
). The hepatic TAG–fatty acid profile of the PNPLA3I148M
transgenic mice is therefore consistent with the increased fatty acid synthesis we observed in these animals (Figure ). The proportion of long-chain polyunsaturated fatty acids in hepatic TAG was significantly reduced in the transgenic mice (Figure B), suggesting that PNPLA3 may play a role in remodeling TAG during the postprandial period. Since long-chain polyunsaturated fatty acids are powerful inhibitors of SREBP-1c (55
), expression of the mutant PNPLA3 transgene may lead to increased SREBP-1c activity by lowering polyunsaturated fatty acid levels. The finding that multiple mRNAs of SREBP-1c target genes are increased in the livers of PNPLA3I148M
transgenic mice is consistent with this notion.
Simple hepatic steatosis appears to be benign, but a subset of individuals with steatosis develop an inflammatory condition (steatohepatitis) that can progress to cirrhosis (2
). The reason why some individuals are more susceptible to the adverse sequelae of increased hepatic TAG levels is not known. The PNPLA3I148M
variant has been shown to increase the risk of progressive liver disease, including steatohepatitis and cirrhosis (11
). Whether the increased risk of progressive liver disease conferred by the PNPLA3I148M
variant results directly from the action of the mutant protein, or is a consequence of the increased risk of steatosis conferred by the allele, is not known. Studies in the PNPLA3I148M
transgenic mice may help to address this question and potentially elucidate the factors that mediate the transition from benign steatosis to clinically significant liver disease.
A potential limitation of this model is that the transgene is human and is expressed at higher levels than the endogenous mouse protein. The human and mouse PNPLA3 proteins share 75% sequence identity within the patatin domain but are less similar in the C-terminal half of the proteins. Despite these differences, the wild-type and mutant human proteins appear to traffic normally to the lipid droplet when expressed in the mouse liver (Supplemental Figure 1). We are in the process of developing mice in which the I148M variant has been knocked into the endogenous gene. It remains possible that the fatty liver phenotype requires higher level expression or sequences specific to the human protein.
The relationship between the I148M polymorphism in PNPLA3 and NAFLD is one of several hundred genetic associations revealed by genome-wide association studies. Whereas these associations can identify genes that contribute to disease processes, in almost all cases the phenotypic effect attributable to the variation is small and the mechanistic basis of the association remains to be defined. The modest phenotypic effects of common sequence variations indicate either that the variation has limited impact on gene function or that the gene is not a major determinant of the phenotype. The finding that neither genetic ablation nor chronic, high level expression of wild-type PNPLA3 affects liver fat content in mice suggests that variation in the normal activity of PNPLA3 is not a major determinant of liver fat content and that the association with fatty liver disease is peculiar to 148M allele. The development of an animal model that recapitulates the metabolic phenotype associated with the allele in humans provides a new in vivo system in which to elucidate the mechanistic relationship between this variant and NAFLD.