N-end Rule–Mediated Proteasomal Degradation of ATGL Promotes Lipid Storage

Obesity is a major risk factor for common diseases, such as nonalcoholic fatty liver disease, cardiovascular diseases, type 2 diabetes, and some cancers (1). The cause of obesity is complicated and includes factors such as excessive food intake, lack of physical activity, and genetic susceptibility; however, in general, the main reason is excessive lipid storage (2). At the cellular level, the balance of lipogenesis and lipolysis largely determines the lipid storage level. Reduced lipogenesis or elevated lipolysis has often been reported to be protective against the development of obesity and obesity-associated diseases (3,4).

ATGL is a rate-limiting executor of lipolysis, and its level or activity has been associated with various metabolic conditions (5–10). Loss-of-function mutations in human ATGL cause neutral lipid storage disease with myopathy (11). Atgl-deficient mice accumulate large amounts of lipid in their hearts, which causes cardiac dysfunction and premature death (12). Interestingly, adipose tissue–specific overexpression or deletion of ATGL appears to be beneficial. Mice with adipose ATGL overexpression were protected from diet-induced obesity and showed improved glucose homeostasis (13). Adipose-specific ATGL knockout mice had slightly increased body weight but exhibited improved glucose tolerance and hepatic insulin sensitivity (14). Pharmacological inhibition of ATGL by atglistatin has beneficial effects on high-fat diet (HFD)–induced obesity and hepatic steatosis (15). Therefore, the temporal and spatial regulation of ATGL level or activity appear to be critical for determining the physiological outcome.

ATGL protein is expressed at low levels in nonadipose tissues but is highly expressed in white adipose tissue (WAT) and brown adipose tissue (BAT) (16). Previous studies revealed that ATGL expression/activity can be regulated transcriptionally and posttranscriptionally (17–21). Numerous binding partners of ATGL have also been identified. α/β Hydrolase domain-containing 5 (ABHD5, also named CGI58) is a classic cofactor that directly binds to and activates ATGL (19). In contrast, ATGL activity can be inhibited by G0S2, which physically interacts with the N-terminal patatin domain of ATGL (20). In addition, other ATGL binding partners, such as UBXD8, PEDF, COP1, PEX2, and the Arf1 exchange factor GBF1, may be responsible for modulating the trafficking, localization, or protein level of ATGL (18,22–25). Despite significant advances in our knowledge of the control of ATGL transcription, localization, and activation, the complete set of in-vivo regulatory events for ATGL is far from clear.

In this study, we found that ATGL protein level is modulated by the N-end rule pathway E3 ligases UBR1 and UBR2. The N-end rule pathway is a proteolytic system in which certain N-terminal residues of short-lived proteins are recognized by a class of ubiquitin ligases to achieve proteasome-mediated degradation (26). We demonstrated that mice with a knock-in of the N-end rule–resistant mutation ATGL(F2A) (designated as Atgl^F2A/F2A) exhibit elevated lipolysis and are resistant to HFD-induced obesity and hepatic steatosis.

Atgl^F2A/F2A knock-in mice on the C57BL/6 background were generated by Biocytogen Pharmaceuticals Co., Ltd. (Beijing, China). All mice were housed in environmentally controlled conditions (temperature 22°C, 12:12-h light/dark cycle lights on at 0730 h). Male mice were used for all experiments. For HFD feeding, 8-week-old Atgl^+/+ and Atgl^F2A/F2A mice were housed individually and fed a HFD with 60% kcal fat (D12492; Research Diets, New Brunswick, NJ) for 16 weeks. A glucose tolerance test (GTT) and insulin sensitivity test (ITT) were performed after 16 weeks of HFD feeding. For hepatic knockdown of Ubr1, 8-week-old Atgl^+/+ and Atgl^F2A/F2A mice were injected intravenously with 3 × 10¹¹ genomic copies of AAV-TBG-shCon or AAV-TBG-shUbr1 adeno-associated viruses (AAVs), followed by feeding with a HFD with 60% kcal fat (Research Diets, New Brunswick, NJ) for 8 weeks. All mice were fasted for 5 to 6 h prior to euthanasia, except when otherwise indicated. All animal care and treatment procedures were approved by the Institutional Animal Care and Use Committee.

HepG2 cells (no. HB-8065), HeLa cells (no. CCL-2), and 3T3L1 preadipocytes (no. CL-173) were purchased from American Type Culture Collection (Manassas, VA). For transfection, a total of 100 pmol siRNA oligonucleotides was transfected into cells in six-well plates using Lipofectamine 3000 (Invitrogen, Waltham, MA).

The coding sequences of human ATGL and the ATGL(F2A) mutant were amplified by PCR and cloned into the adenovirus vector pADV-mCMV-MCS-3xFlag. Adenovirus was produced by transfection in 293A cells and purified via the cesium chloride gradient centrifugation method. AAV expressing shRNA against mouse Ubr1 was generated using the AAV vector pAAV-TBG-3Flag-miR30shUbr1-WPRE. The target sequence is 5′-CCCGTAAGATCCTTCATGA-3′. AAV was produced by transfection into 293T cells and purified via discontinuous iodixanol gradient (27). Viral genome titers were determined by quantitative RT-PCR.

All data are shown as mean ± SEM, except in Figs. 1D, 2D, and 3H, in which data are shown as violin plots with median. Statistical analyses were performed with the two-tailed, unpaired Student t test or ANOVA with post hoc Tukey multiple comparisons test using GraphPad.

Data sets and resources are available upon request.

To find new regulators of neutral lipid storage, we previously performed an RNA interference (RNAi) screen in Caenorhabditis elegans (28). We found that two proteasome component genes, pas-5 and pbs-4, caused a decreased lipid storage phenotype when knocked down (Fig. 1A, Supplementary Fig. 1A, and Supplementary Table 1). We then screened other proteasome components through RNAi. Knockdown of nearly all of the core components and a few regulatory components caused a similar reduced lipid storage phenotype (Supplementary Table 1). This suggests a general requirement for the proteasome for proper lipid storage. Meanwhile, in a genetic screen in Drosophila (29), we found that overexpression of the proteasome regulatory subunit Rpn2 dramatically increases lipid storage in Drosophila third instar salivary gland (Supplementary Fig. 1B). These findings suggest that proteasome activity promotes lipid storage in worms and flies.

We examined the effect of the proteasome inhibitor MG132 on lipid storage in worms and mammalian cells. Compared with the control, MG132 treatment reduced the fat content in worms in a dose-dependent manner (Fig. 1B). Similarly, the size and the number of lipid droplets, as well as triglyceride levels, were all decreased in MG132-treated HepG2 cells (Fig. 1C–F). Together, our results indicate that proteasome activity positively correlates with lipid storage in worm, fly, and mammalian cells.

Next, we explored the mechanism underlying the relationship between inhibition of proteasome activity and decreased lipid storage. The proteasome may be involved in the degradation of a negative regulator of lipogenesis or a positive regulator of lipolysis. Previous work showed that the level of ATGL is increased by proteasome inhibitor treatment (18). Therefore, we analyzed the link betweenproteasome activity, ATGL level, and lipid storage. MG132 treatment greatly increased the endogenous ATGL level in both normal and oleic acid (OA)–loaded cells (Fig. 1G). Interestingly, OA treatment also enhanced the protein levels of ATGL compared with the control (Fig. 1G). To test whether elevated ATGL is responsible for the MG132-mediated inhibition of lipid storage in OA-loaded cells, we knocked down ATGL in HepG2 cells. The reduced lipid storage caused by MG132 treatment was significantly suppressed by knockdown of ATGL (Fig. 1C–F and Supplementary Fig. 1C).

In agreement with previous findings (18,24), the level of ubiquitinylated ATGL was increased upon MG132 treatment (Fig. 1H). In contrast to MG132, treatment with bafilomycin A1, a lysosome H⁺-ATPase inhibitor, did not affect endogenous ATGL levels in cells with or without OA treatment (Fig. 1I). This indicates that ATGL is mainly degraded through the ubiquitin-proteasome pathway. Overall, these data suggest that proteasome activity regulates lipid storage, at least partially, through ATGL degradation.

We next sought to identify the E3 ligases that might be responsible for ATGL ubiquitination and degradation. The candidate E3 ligase should fulfill at least two criteria: first, when mutated, it should have a lipid metabolism–related phenotype in vivo; and second, it should affect ATGL ubiquitination. Based on these two criteria, we examined the potential involvement of the N-end rule E3 ligase UBR1 (30). UBR is an important component of the N-end rule pathway, and it recognizes and binds to proteins bearing destabilizing N-terminal residues, leading to their ubiquitination and subsequent degradation (31). Importantly, ATGL bears a destabilizing N-end residue, phenylalanine (F), and this residue is conserved in vertebrates (Fig. 2A). In addition, a previous report has shown that Ubr1^−/− mice exhibit reduced adiposity (30).

We then examined whether the stability of ATGL is regulated by UBR1. UBR1 RNAi increased the ATGL protein level compared with the negative control (Fig. 2B). Similarly, the endogenous ATGL protein level was increased in HeLa cells with knockdown of UBR2, which belongs to UBR protein family and plays redundant roles (31) and was greatly increased in UBR1/UBR2 double RNAi cells compared with the control (Fig. 2B). Next, we tested whether UBR1 and UBR2 also affect ATGL-mediated lipolysis. ATGL RNAi led to increased lipid storage (Fig. 2C–F). In contrast, knockdown of UBR1 or UBR2 reduced the size and number of lipid droplets and lowered the level of triglyceride accumulation in ATGL RNAi HepG2 cells (Fig. 2C–F). This indicates that UBR1 and UBR2 affect ATGL-regulated lipid storage.

We also examined the physical interaction between the proteins. Both UBR1 and UBR2 were associated with ATGL-Flag (Fig. 2G and H). We further determined the lysine residues of ATGL that are ubiquitinated by UBR. Consistent with previous findings that the polyubiquitination signal of ATGL is located at the N terminus of the protein (18,24), the ubiquitination signal was detected in Flag-tagged full-length ATGL and a Flag-tagged N-terminal fragment (amino acids 1–160) (Fig. 2I and Supplementary Fig. 1D). Notably, the ubiquitination signal of both full-length and N-terminal ATGL was considerably decreased upon UBR1 knockdown (Fig. 2I and Supplementary Fig. 1D). We searched for UBR1-dependent ubiquitination sites in the N-terminal patatin domain of ATGL. There are six lysine (K) residues in that region, and UBR1 knockdown increased the ATGL protein level when lysine was mutated to arginine (R) at positions 68, 74, 78, 92, or 135 (Supplementary Fig. 1E–G). However, the protein and polyubiquitination levels of ATGL(K100R) did not respond to UBR1 knockdown (Fig. 2I and Supplementary Fig. 1G), suggesting that K100 is an UBR1-dependent ubiquitination site of ATGL.

Since OA treatment also increased ATGL protein levels (Fig. 1G), we then tested whether OA treatment affected the polyubiquitination levels of ATGL. The polyubiquitination levels of ATGL were slightly increased by treatment with OA or lipolytic inducer (Fig. 2J). Moreover, knockdown of UBR1 increased ATGL protein stability with or without OA treatment. Similar patterns were observed following the knockdown of COP1 and PEX2, two E3 ligases for ATGL, except that knockdown of PEX2 failed to decrease the polyubiquitination level of ATGL upon treatment with lipolytic inducer (Supplementary Fig. 2A and B). This may be due to lipolysis enhancing the protein level of PEX2 (24). In line with this, neither OA treatment nor knockdown of E3 ligases for ATGL significantly affected the polyubiquitination levels of ATGL(K100R) (Supplementary Fig. 2C and D), which provides further evidence that K100 is important for its protein stability. Collectively, these results indicate that the stability of ATGL can be regulated by UBR1 and UBR2.

To investigate whether the N-terminal phenylalanine residue of ATGL is important for its stability, we compared the stability of wild-type (WT) ATGL in HeLa cells with two ATGL mutants, ATGL(F2A) and ATGL(F2V), in which the N-terminal destabilizing residue phenylalanine (F) of ATGL was mutated to the stabilizing residue alanine (A) or valine (V). Western blot results showed that the ATGL(F2A) and ATGL(F2V) mutants are more stable than ATGL(WT) (Fig. 3A and B). We then explored the contribution of the destabilizing phenylalanine residue to the ubiquitination of ATGL. The level of ubiquitinated ATGL(F2A) was lower than that of ATGL(WT) in the presence of MG132 (Fig. 3C). Furthermore, knockdown of UBR1 and UBR2 enhanced the level of ATGL(WT) protein but not the ATGL(F2A) mutant (Fig. 3D). This suggests that the N-terminal phenylalanine residue of ATGL is important for UBR1- and UBR2-regulated protein stability.

ATGL is highly expressed in adipocytes, and ATGL-mediated lipolysis is essential for providing free fatty acid (FFA) for energy production during fasting (16). We then tested whether the ATGL(F2A) mutant affects lipolysis and lipid storage in 3T3L1 adipocytes. In basal state 3T3L1 adipocytes, levels of lipolysis, assessed by FFA and glycerol release, were comparable in cells expressing ATGL(WT) and ATGL(F2A), except that glycerol release was increased in ATGL(F2A) cells at 4 h (Fig. 3E and F). Stimulation of lipolysis with isoproterenol led to increased FFA and glycerol release, and this enhancement was greater in ATGL(F2A) cells compared with ATGL(WT) cells (Fig. 3E and F and Supplementary Fig. 2E). The N-end rule residue substitution did not affect the binding of ATGL with CGI58, an activator of ATGL, or with G0S2, an inhibitor of ATGL (Supplementary Fig. 2F and G). Accordingly, lipid droplet size and number and triglyceride accumulation were decreased in stimulated ATGL(F2A)-expressing cells compared with ATGL(WT)-expressing cells (Fig. 3G–J). In sum, the N-end rule residue affects the stability of ATGL and the level of ATGL-mediated lipolysis.

We next sought to reveal the effect of stabilized ATGL in mice. We generated mice with knock-in of the ATGL(F2A) mutation, designated as Atgl^F2A/F2A (Supplementary Fig. 3A and B). Supplementary Table 2 shows the metabolic profiles of Atgl^+/+ and Atgl^F2A/F2A mice fed a chow diet.

In gonadal WAT and BAT, the ATGL level was increased 3-fold and 1.5-fold, respectively, in Atgl^F2A/F2A mice compared with control mice (Fig. 4A and B). Similar to adipose tissue, the ATGL protein level was elevated in muscle from Atgl^F2A/F2A mice (Supplementary Fig. 3C). Along with the increased protein level, ATGL(F2A) caused enhanced lipolysis ex vivo (Fig. 4C and D). When mice were fed a HFD for 8 weeks, ATGL(F2A) led to enhanced FFA release from labeled triglyceride compared with control mice (Fig. 4E).

Previous studies reported that overexpression of ATGL leads to activation of peroxisome proliferator–activated receptor α (PPARα) signaling and fatty acid β-oxidation (FAO) (32–34). Similarly, expression levels of genes related to PPARα signaling and FAO were higher in the gonadal WAT from the HFD-fed Atgl^F2A/F2A mice compared with the control (Fig. 4F). Expression levels of genes involved in lipogenesis and lipolysis were not significantly affected (Fig. 4F). Moreover, direct measurement of FAO using [³H]palmitate showed that FAO was enhanced in the WAT of HFD-fed Atgl^F2A/F2A mice (Fig. 4G). The development of obesity is associated with stereotypical changes in adipose tissue expression of inflammatory genes. We then examined the adipose inflammation in HFD-fed mice. The adipose inflammation was not significantly affected in HFD-fed Atgl^F2A/F2A mice (Supplementary Fig. 3D). Overall, these data suggest that Atgl^F2A/F2A mice have increased levels of ATGL protein and triglyceride hydrolase activity in adipose tissue.

Next, we examined the effect of ATGL(F2A) on the development of obesity. Atgl^+/+ and Atgl^F2A/F2A mice were pair-fed a HFD for 16 weeks starting from 8 weeks of age. We used pair-feeding to ensure similar food intake by these two groups (Supplementary Fig. 3E), because it has been reported that loss of Atgl or pharmacological ATGL inhibition affects food intake (15,35). The plasma parameters are shown in Supplementary Table 3.

The body weight gain of Atgl^F2A/F2A mice was less than that of Atgl^+/+ mice (Fig. 5A). Glucose tolerance and insulin sensitivity were improved in HFD-fed Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Fig. 5B–D). To determine the effect of ATGL(F2A) on energy balance, we measured oxygen consumption (O₂), carbon dioxide (CO₂) production, and energy expenditure. These levels were higher in HFD-fed Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Fig. 5E and F and Supplementary Fig. 3F and G).

The attenuated body weight gain in HFD-fed Atgl^F2A/F2A mice led us to examine adiposity. The liver weight and the weights of WAT and BAT were decreased in Atgl^F2A/F2A mice compared with Atgl^+/+ mice when fed a HFD (Fig. 5G and Supplementary Fig. 3H). The sizes of adipocytes in WAT and BAT were also decreased in adipose tissue sections from HFD-fed Atgl^F2A/F2A mice compared with control mice (Fig. 5H–J).

The reduced liver weight in HFD-fed Atgl^F2A/F2A mice led us to further analyze the effect of ATGL(F2A) in liver. Hepatic triglyceride and total cholesterol levels and hepatic lipid droplet accumulation were significantly decreased in HFD-fed Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Fig. 5K and L). The ALT level, which indicates liver damage, was decreased in HFD-fed Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Fig. 5M). ATGL protein levels were increased in the liver of Atgl^F2A/F2A mice compared with control mice (Supplementary Fig. 3I). To further determine whether reduced lipid accumulation in the liver of Atgl^F2A/F2A mice was attributable to enhanced lipid degradation, we measured the FAO level. Expression levels of genes involved in FAO and fatty acid transport were significantly increased (Fig. 5N). The FAO level was significantly enhanced in the liver of HFD-fed Atgl^F2A/F2A mice compared with control (Fig. 5O). Taken together, these results suggest that Atgl^F2A/F2A mice, which carry a stabilizing N-terminal amino acid substitution, are resistant to HFD-induced obesity and hepatic steatosis.

We then examined the physiological effect of UBR1-mediated ATGL degradation in mice. We knocked down Ubr-1 (AAV-TBG-shUbr1) in the liver of Atgl^+/+ or Atgl^F2A/F2A mice. The thyroxine-binding globulin (TBG) promoter ensures gene knockdown in the liver. Control animals received AAV-TBG-shCon. Hepatic knockdown of Ubr1 did not affect hepatic triglyceride levels in fed or fasted mice on a chow diet (Supplementary Fig. 4A). We also fed the animals with a HFD for 8 weeks. As expected, HFD-fed Atgl^F2A/F2A mice showed attenuated body weight gain, less hepatic lipid accumulation, and decreased plasma ALT levels compared with HFD-fed Atgl^+/+ mice (Fig. 6A–C and Supplementary Fig. 4B). These beneficial effects were not affected by hepatic knockdown of Ubr1 in HFD-fed Atgl^F2A/F2A mice (Fig. 6A–C and Supplementary Fig. 4B), which suggests that UBR1-mediated ATGL degradation is blunted in Atgl^F2A/F2A mice. Nevertheless, hepatic knockdown of Ubr1 caused reductions in hepatic lipid accumulation and plasma ALT levels in HFD-fed Atgl^+/+ mice (Fig. 6A–C and Supplementary Fig. 4B).

We then tested the energy balance and glucose homeostasis in HFD-fed Atgl^+/+ and Atgl^F2A/F2A mice with or without knockdown of hepatic Ubr1. The VO₂, VCO₂, and energy expenditure were significantly enhanced in Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Supplementary Fig. 4C–E). Knockdown of Ubr1 did not affect their levels irrespective of genotypes (Supplementary Fig. 4C–E). In line with previous findings, HFD-fed Atgl^F2A/F2A mice showed improved glucose homeostasis and insulin sensitivity compared with Atgl^+/+ mice (Fig. 6D–F and Supplementary Fig. 4F). Although knockdown of Ubr1 improved glucose homeostasis in AAV-TBG-shUbr1–treated Atgl^+/+ mice compared with AAV-TBG-shCon–treated Atgl^+/+ mice, it had no effects on glucose homeostasis in Atgl^F2A/F2A mice (Fig. 6D–F and Supplementary Fig. 4F). To examine the activity of the insulin pathway, HFD-fed Atgl^+/+ and Atgl^F2A/F2A mice were injected with 1 unit/kg insulin intraperitoneally. p-AKT(Ser473) and p-AKT(Thr308) levels were enhanced in the liver of Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Fig. 6G). Importantly, knockdown of hepatic Ubr1 did not affect the activity of the insulin pathway in the liver of Atgl^F2A/F2A mice (Fig. 6G). Moreover, knockdown of hepatic Ubr1 did not affect the activity of the insulin pathway in the muscle and WAT in both genotypes (Supplementary Fig. 4G and H). Taken together, these data suggest that the beneficial effects of the Atgl^F2A mutation on HFD-induced hepatic steatosis and glucose homeostasis in mice are not affected by knockdown of hepatic Ubr1.

Analysis of hepatic gene expression levels showed that Ubr1 deficiency downregulated the expression of Pparg and its target Fabp4 in Atgl^+/+ mice, but this effect was blunted in Atgl^F2A/F2A mice (Fig. 6H). A similar pattern was shown by other genes involved in lipogenesis (Supplementary Fig. 4I). The expression levels of genes involved in FAO and lipolysis were enhanced in Atgl^F2A/F2A mice compared with Atgl^+/+ mice (Fig. 6H and Supplementary Fig. 4I). Knockdown of Ubr1 also increased the expression levels of FAO genes in Atgl^+/+ mice, but it caused no further upregulation of these genes in Atgl^F2A/F2A mice (Fig. 6H and Supplementary Fig. 4I). Next, we tested whether UBR1 regulates ATGL levels in vivo. ATGL protein levels were elevated by knockdown of Ubr1 in Atgl^+/+ mice and were not affected by knockdown of Ubr1 in Atgl^F2A/F2A mice (Fig. 6I). Accordingly, the polyubiquitination levels of ATGL were lowered by knockdown of hepatic Ubr1 in Atgl^+/+ mice (Fig. 6J). PNPLA3 and PNPLA4, which are PNPLA family members, also contain N-terminal destabilized residues. PNPLA4, but not PNPLA3, was regulated by UBR1 in HeLa cells (Supplementary Fig. 4J and K). Together, these data suggest that in the HFD condition, hepatic knockdown of Ubr1 reduces lipogenesis and increases FAO. Moreover, the phenotypic similarity of Atgl^F2A/F2A mice with or without Ubr1 knockdown indicates that the N-end rule–mediated degradation of ATGL by UBR1 occurs in vivo.

We further examined the correlation between UBR1 and ATGL levels in obese mice. Consistent with previous reports (36), ATGL protein levels were downregulated in ob/ob mice. Interestingly, UBR1 levels were upregulated in ob/ob mice, which suggests a negative correlation between UBR1 and ATGL levels (Fig. 6K). Together, these results demonstrate that the N-end rule-mediated proteasomal degradation of ATGL regulates hepatic lipid metabolism and insulin sensitivity.

In this study, we found that ATGL, which possesses a typical destabilizing N-terminal residue, is regulated through the N-end rule pathway. Knockdown of the E3 ligases UBR1 and UBR2, or treatment with a proteasome inhibitor, elevates the ATGL level and reduces lipid storage. Importantly, stabilized ATGL [ATGL(F2A)] has beneficial effects on HFD-induced obesity and associated hepatic steatosis in mice.

Based on our results and previous findings (18,24), inhibition of proteasome activity or RNAi of proteasome components results in reduced lipid storage in C. elegans, Drosophila, and cultured mammalian cells. The proteasomal regulation of lipid storage occurs at least partially through ATGL degradation. Previous studies on ATGL protein levels used N-terminal tagged ATGL, thus possibly masking the N-end rule regulation of this protein (18,25). The N-end rule regulation of ATGL is apparently not the only mechanism that regulates ubiquitination or degradation of ATGL because the ATGL(F2A) protein can still be ubiquitinated (Fig. 3C). E3 ubiquitin ligase COP1 and PEX2 also target ATGL for proteasomal degradation (18,24).

Our study showed that knockdown of UBR or treatment with proteasome inhibitor can reduce lipid storage in the absence of ATGL in OA-loaded HepG2 cells (Figs. 1C–F and 2C–F). This suggests that other factors involved in lipolysis or lipogenesis can also be involved in UBR1- or proteasome inhibitor–mediated lipid metabolism. In fact, UBR1 has been shown to degrade lipid droplet proteins in yeast (37). A recent study identified PLIN2 as a substrate of UBR1 in mice (38).

The Atgl^F2A/F2A mice reported in this study presumably represent a whole-body gain of function of ATGL. These mice also provide us with an opportunity to study the relationship between ATGL protein stability and organismal physiological function. Atgl^F2A/F2A mice show improved GTT and ITT results, elevated energy expenditure when fed a HFD, and resistance to HFD-induced obesity and hepatic steatosis. These beneficial effects appear similar to those in G0S2^−/− and adipose-specific ATGL overexpression (ap2-desnutrin) mice (13,39). The common features among these mouse models are elevated lipolysis in adipose tissue and reduced triglyceride accumulation in liver upon HFD feeding.

The elevated flux of fatty acids from adipose tissue can result in triglyceride accumulation in other peripheral tissues, such as liver. The decreased triglyceride accumulation in the liver of HFD-fed Atgl^F2A/F2A mice may be due to decreased FFA release from adipose tissue or increased triglyceride degradation in liver. Although the ATGL level is apparently enhanced in the adipose tissue of Atgl^F2A/F2A mice, the change of plasma FFA level is modest upon HFD feeding (Fig. 4A and Supplementary Table 3). The enhanced FAO and PPARα signaling in the adipose tissue of Atgl^F2A/F2A mice may dampen the FFA release from adipose tissue (Fig. 4F and G). Similarly, plasma FFA levels were only slightly higher in ap2-desnutrin mice compared with control mice, which was in part due to elevated FAO within adipose tissue (13). In contrast, elevated hepatic ATGL levels and enhanced FAO in the liver of Atgl^F2A/F2A mice may account for attenuated HFD-induced hepatic steatosis. In addition, enhanced energy expenditure, improved insulin sensitivity, and attenuated HFD-induced body weight gain may also contribute to the beneficial effect in the liver.

Glucose tolerance and insulin sensitivity are improved in HFD-fed Atgl^F2A/F2A mice. It has been reported that lipotoxicity is a causal factor for insulin resistance. It is plausible that reduced lipid accumulation in liver and decreased adiposity relieve the burden of HFD-induced lipid overload, thus improving glucose tolerance and insulin sensitivity. In line with that, both G0S2^−/− and ap2-desnutrin mice showed improved glucose homeostasis upon HFD feeding. Notably, the beneficial effect in G0S2^−/− and Atgl^F2A/F2A mice results from the action of ATGL in both liver and adipose tissue, while the beneficial effect in ap2-desnutrin mice is predominantly due to the action of ATGL in adipose tissue. We have observed that knockdown of hepatic Ubr1 improved the activity of the insulin pathway (Fig. 6D–G). UBR1 may directly regulate components of the insulin signaling pathway. Alternatively, it may regulate hepatic lipids, such as diacylglycerol or ceramide, which in turn affect hepatic insulin signaling.

ATGL apparently has dual effects on metabolism and physiology. Tissue-specific knockout or overexpression of ATGL appears to have beneficial effects on glucose metabolism in mice (13,14,40). In humans, both whole-body loss of function and gain of function of ATGL result in deleterious effects. On one hand, loss of ATGL results in neutral lipid storage disease with life-threating myopathy (11). On the other hand, gain of ATGL function in patients with a C-terminal mutation in PLIN1 is associated with a dominant partial lipodystrophy with severe dyslipidemia and insulin resistance (41). We cannot rule out the possibility that the deleterious effects in patients with the PLIN1 C-terminal truncation could be caused by a combination of both gain of function of ATGL and partial loss of function of PLIN1. Nevertheless, these results indicate that maintaining a suitable level of ATGL in vivo appears to be essential for sustaining healthy physiological conditions in humans. In summary, our findings suggest that the level and the site of ATGL upregulation are probably critical to determining the outcomes of ATGL manipulation.

Acknowledgments. The authors thank Drs. H. Yang (University of New South Wales), J. Liu (Mayo Clinic), C. Yang (Yunnan University), and J. Speakman, S. Bao, and Z. Xu (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences) for providing reagents and helpful discussions.

Funding. This research was supported by the National Natural Science Foundation of China grants 32230044 and 91954207, and the Ministry of Science and Technology of China grant 2018YFA0506902.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.X., Z.L., J.Z., S.C., and W.W. conducted the experiments and analyzed the data. Z.L. contributed to C. elegans experiments. J.X. and Z.L. contributed to cell experiments. J.X., Z.L., J.Z., S.C., and X.Z. contributed to mouse experiments. W.W. contributed to Western blotting. M.Z. contributed to the identification of UBR1. J.X., Z.L., and X.H. wrote the article. X.H. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.