Rab2A regulates the progression of nonalcoholic fatty liver disease downstream of AMPK-TBC1D1 axis by stabilizing PPARγ

Nonalcoholic fatty liver disease (NAFLD) affects approximately a quarter of the population worldwide, and persistent overnutrition is one of the major causes. However, the underlying molecular basis has not been fully elucidated, and no specific drug has been approved for this disease. Here, we identify a regulatory mechanism that reveals a novel function of Rab2A in the progression of NAFLD based on energy status and PPARγ. The mechanistic analysis shows that nutrition repletion suppresses the phosphorylation of AMPK-TBC1D1 signaling, augments the level of GTP-bound Rab2A, and then increases the protein stability of PPARγ, which ultimately promotes the hepatic accumulation of lipids in vitro and in vivo. Furthermore, we found that blocking the AMPK-TBC1D1 pathway in TBC1D1S231A-knock-in (KI) mice led to a markedly increased GTP-bound Rab2A and subsequent fatty liver in aged mice. Our studies also showed that inhibition of Rab2A expression alleviated hepatic lipid deposition in western diet-induced obesity (DIO) mice by reducing the protein level of PPARγ and the expression of PPARγ target genes. Our findings not only reveal a new molecular mechanism regulating the progression of NAFLD during persistent overnutrition but also have potential implications for drug discovery to combat this disease.


Introduction
Nonalcoholic fatty liver disease (NAFLD) is the main cause of chronic liver disease worldwide and has a global prevalence of 25.2% [1]. Clinically, patients with NAFLD often somehow suffer from other metabolic syndromes, such as obesity, type 2 diabetes, and insulin resistance [2]. Many studies have shown that genetic mutations can regulate the progression of NAFLD, including patatin-like phospholipase domain-containing 3 (PNPLA3), transmembrane 6 superfamily member 2 (TM6SF2), and membrane-bound O-acyltransferase domain- further explore the underlying role of AMPK-TBC1D1 signaling in hepatic steatosis, we detected NAFLD-related phenotypes in TBC1D1-KI mice at 4 to 6 months, 12 months, and 18 months of age. The observations of 4-to 6-month-old KI mice revealed normal lipid droplets in liver sections (S1A Fig), consistent with normal levels of TGs (S1B Fig). However, 12-and 18-month-old KI mice showed markedly increased accumulation of larger lipid droplets in liver sections (Fig 1A and 1B, S1C Fig) accompanied by severely increased TG levels (Fig 1C,  S1D Fig). These abovedescribed data indicate that blocking the AMPK-TBC1D1 axis leads to hepatic lipid accumulation in aged mice.
To investigate the underlying molecular mechanisms that mediate the storage of lipids in the liver of aged TBC1D1-KI mice, we first detected the mRNA levels of general fatty acid metabolism-related genes, such as genes involved in lipolysis and fatty acid oxidation (S1E The results showed no significant differences between the 2 genotypes. Our previous studies revealed increased activities of IGF1-AKT/PKB-mTOR-SREBP-1c signaling and lipogenesis in the adipose tissue of TBC1D1-KI mice at 4 to 6 months [9], and we thus also detected the protein levels of core lipogenic proteins such as fatty acid synthase (FASN) and cytoplasmic acetyl-CoA synthase (ACS1). All of these proteins showed normal expression levels in the livers of 18-month-old (S1I and S1J Fig) and 12-month-old mice (S1K and S1L Fig).
We then performed RNA sequencing of the livers of 18-month-old mice, and the KEGG analysis showed the marked enrichment of genes related to the PPARγ signaling pathway (S2A Fig). Furthermore, a heatmap analysis revealed substantial increases in lipid storage genes, such as Cidea, Cidec, and Plin4, which are targets of PPARγ (Fig 1D), and the Q-PCR results confirmed these findings in the livers of 18-month-old ( Fig 1E) and 12-month-old mice (S2B Fig). Then, we found significantly increased PPARγ protein levels with no obvious differences in the PPARγ mRNA levels in the 12-month-old (S2C- S2E Fig) and 18-month-old TBC1D1-KI mice (Fig 1F-1H), and these effects were followed by increased PPARγ activity. Together, the abovementioned data suggest that blocking the AMPK-TBC1D1 axis leads to NAFLD in aged mice, possibly through the activation of PPARγ signaling.

The AMPK-TBC1D1 axis regulates the protein stability and functions of PPARγ
To clearly elucidate the relationship between the AMPK-TBC1D1 axis and PPARγ, we constructed a relevant system in human hepatoma cell line (HepG2), human embryonic kidney cell line (HEK293T), and primary hepatocytes. Primarily, we found that transient overexpression of TBC1D1 in HepG2 cells markedly increased the endogenous protein level of PPARγ To uncover the underlying mechanism in detail, we selected PPARγ2, which is the longer form, as the protein of interest in the following studies. TBC1D1 S237A , which refers to TBC1D1 with an alanine substitution at the S-237 site, is not phosphorylatable by AMPK at this site.  . (D) Heatmap shows the genes with significant differences in expression between the livers of WT and TBC1D1-KI male mice at 18 months of age based on RNA sequencing data (random feed, n = 3 per group). (E) mRNA expression levels of genes in the heatmap were confirmed by Q-PCR (random feed, n = 4 per group). (F) Increased protein level of PPARγ in the liver of TBC1D1-KI mice aged 18 months. The data were obtained by immunoblotting (random feed, n = 4 per group). The red asterisk indicates nonspecific signal. (G) Statistical analysis of protein levels shown in F. (H) mRNA expression levels of PPARγ-related genes were confirmed by Q-PCR (random feed, n = 4 per group). The data were PPARγ protein, we utilized an AMPK activator A769662 to treat cells expressing an empty vector, or WT TBC1D1, or TBC1D1 S237A mutant. A769662 treatment expectedly increased the phosphorylation of AMPK and its bona fide substrate acetyl-CoA carboxylase (ACC) in a dose-dependent manner, and the activation states of AMPK were comparable among the cells expressing the empty vector, or WT TBC1D1, or TBC1D1 S237A mutant (Fig 2A and 2B). Notably, activation of AMPK by A769662 caused a gradual decrease of exogenous PPARγ2 (Fig 2A) and endogenous PPARγ (Fig 2B), which correlates with the dosage of A769662. Expression of WT TBC1D1 elevated both exogenous PPARγ2 and endogenous PPARγ, which were still decreased upon AMPK activation by A769662 treatment (Fig 2A and 2B). Importantly, the A769662-induced diminution of exogenous PPARγ2 and endogenous PPARγ was prevented when the AMPK-insensitive TBC1D1 S237A mutant was expressed in cells (Fig 2A and 2B).
Taken together, these data firmly establish a causal role of the AMPK-TBC1D1 axis in the regulation of PPARγ protein, and disruption of the AMPK-TBC1D1 nexus increases PPARγ protein to promote TG accumulation by elevating the expression of lipid storage genes in hepatocytes.

Rab2A, as a downstream protein of the AMPK-TBC1D1 axis, regulates the protein level of PPARγ
We next sought to find out how TBC1D1 regulates PPARγ protein levels. Interestingly, we found that Flag-TBC1D1 could interact with exogenous PPARγ2-MYC protein (S3K Fig) as well as endogenous PPARγ protein (S3L Fig). Moreover, the in vitro binding assay also demonstrated a direct interaction between TBC1D1 and PPARγ2 ( Fig 3A). Notably, overexpression of TBC1D1 attenuated the degradation of PPARγ2 when cells were treated with cycloheximide (CHX) (S3M and S3N Fig), suggesting that TBC1D1 may regulate the protein stability of PPARγ.
The in vitro binding assay showed that TBC1D1 S237A mutant protein still possessed the ability to interact with PPARγ in a manner similar to WT TBC1D1 (Fig 3A), suggesting that TBC1D1-S237 phosphorylation does not affect the TBC1D1 interaction with PPARγ. TBC1D1 is a GTPase-activating protein (GAP) and exhibits significant GAP activity toward Rab2A, Rab8A, Rab8B, Rab10, and Rab14 in an in vitro assay [23]. Overexpression of a GAP-inactive TBC1D1 R854K mutant increased PPARγ2 protein to an extent similar to TBC1D1 S237A mutant protein ( Fig 3B). It is currently not clear whether the TBC1D1 R854K mutant exerts its effect on PPARγ2 through a GAP-independent mechanism or via a dominant-negative mechanism. To further study the possible mechanisms by which TBC1D1 regulates the PPARγ protein, we analyzed with unpaired 2-tailed Student t test and are presented as the mean ± s.e.m.s. " � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001. Raw data are given in S1 Excel spreadsheet with raw data from all figures. AAU : AbbreviationlistshavebeencompiledforthoseusedthroughoutFigs1to7:Pleaseveri MPK, adenosine monophosphate-activated protein kinase; KI, knock-in; TG, triglyceride; WT, wild-type. We then investigated how the AMPK-TBC1D1 axis regulates Rab2A via a pulldown assay utilizing the bait protein GST-GRASP55, which binds to the GTP-bound form of Rab2A [24]. As expected, GST-GRASP55 preferentially interacted with Rab2A-Q65L, the GTP-bound form, over Rab2A-S20N, the GDP-bound form (S6A Fig). In HepG2 cells, overexpression of the TBC1D1 S237A protein but not WT TBC1D1 significantly increased the GTP-bound form of Rab2A ( Fig 3E). Moreover, the GTP-bound form of Rab2A was markedly increased, while total Rab2A levels were unaltered, in the livers of TBC1D1-KI mice (18 months old) (Fig 3F  and 3G). These data show that the TBC1D1-S231A mutation causes activation of Rab2A in cells, although Rab2A might not be a physiological substrate of TBC1D1.

GTP-bound form of Rab2A binds and inhibits the proteasomal degradation of PPARγ
We next sought to delineate the mechanism how Rab2A regulates PPARγ protein. In a coimmunoprecipitation assay, endogenous PPARγ was found in the immunoprecipitates of . (E, F) Knockdown of PPARγ rescues the effects of lipid storage in TBC1D1-S231A primary hepatocytes. Primary hepatocytes were isolated, cultured, and transfected with the indicated lentivirus-expressing plasmids. Then the cells were chosen and harvested for Q-PCR (E) and TG testing (F). The TG level of control cells was normalized to 1, and all the above statistical data were analyzed with unpaired 2-tailed Student t test (n = 3 per group) and are presented as the means ± s.e.m.s. shNC versus shPPARγ in Flag-TBC1D1 WT -overexpressing HepG2 cells or WT primary hepatocytes (" � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001). shNC versus shPPARγ in Flag-TBC1D1 S237A -overexpressing HepG2 cells or TBC1D1-S231A primary hepatocytes ("z" indicates p < 0.05, "zz" indicates p < 0.01, and "zzz" indicates p < 0.001). Flag-TBC1D1 WT versus Flag-TBC1D1 S237A in HepG2 cells or WT versus TBC1D1-S231A primary hepatocytes ("#" indicates p < 0.05, "##" indicates p < 0.01, and "###" indicates p < 0.001). Raw data are given in S1 Excel spreadsheet with raw data from all figures. All experiments were performed at least 3 times with similar results. AMPK, adenosine monophosphate-activated protein kinase; WT, wild-type.
As a small G protein, Rab2A can switch between the GTP-bound active form and the GDPbound inactive form [25]. The GTP-bound form, but not GDP-bound form, of Rab2A markedly augmented protein levels of exogenous PPARγ2-MYC (S7A Fig) and endogenous PPARγ ( Fig 4D). Moreover, the GTP-bound form, but not GDP-bound form, of Rab2A interacted with PPARγ2-MYC in the co-immunoprecipitation assay ( Fig 4E). PPARγ contains several domains, namely N terminal region (NAU : PleasenotethatNTRhasbeenchangedtoNTDinthesentenc TR), DNA-binding domain (DBD), ligand-binding domain (LBD), and a special region at the carboxyl terminus known as activation function 2 (AF-2) [15]. Besides these domains commonly found in both PPARγ isoforms, PPARγ2 has an additional 30 amino acids at its N-terminus as compared to PPARγ1. Interestingly, analyses with internal deletion mutants revealed that PPARγ2 with deletion of the AF-2 domain, but not other domains, lost its ability to respond to Rab2A (Fig 4F, S7B and S7C Fig). Notably, deficiency of the AF-2 domain prevented the binding of PPARγ2 to Rab2A in the co-immunoprecipitation assay ( Fig 4F). These data suggest that the GTP-bound active form of Rab2A might interact with the AF-2 domain of PPARγ to increase the protein level of PPARγ.
We next examined whether Rab2A might regulate PPARγ stability to control its protein level. To this end, we first determined the effect of Rab2A overexpression on the degradation rate of PPARγ2-MYC. Addition of CHX triggered a rapid decrease of PPARγ2-MYC protein, and overexpression of Rab2A markedly slowed down the CHX-induced decrease of PPARγ2-MYC (S7D and S7E Fig). These data demonstrate that Rab2A indeed regulates PPARγ stability. We next examined whether Rab2A regulates PPARγ stability through the lysosome or proteasome pathway. Treatments with the inhibitors of proteasome, MG132 and ALLN (Ac-LLnL-CHO), both caused accumulation of PPARγ2-MYC in a concentration-and timedependent manner (Fig 4G-4N). Notably, overexpression of Rab2A markedly attenuated the effects of these proteasome inhibitors on the accumulation rate of PPARγ2-MYC (Fig 4G-4N). By contrast, overexpression of Rab2A did not alleviate the accumulation of PPARγ2-MYC induced by the lysosomal inhibitors, NH 4 Cl and bafilomycin A1 (S7F- S7M Fig).
Taken together, these data demonstrate that the GTP-bound form of Rab2A binds to the AF-2 domain of PPARγ and inhibits the proteasomal degradation of PPARγ. level of GTP-bound form of Rab2A. HepG2 cell lines stably expressing TBC1D1-WT or TBC1D1-S237A proteins were harvested, and the data were obtained by GST-GRASP55 pulldown and immunoblotting assays. (F, G) Significantly increased levels of GTP-bound form of Rab2A in the livers of TBC1D1-KI male mice aged 18 months. The data are shown by GST-GRASP55 pulldown and immunoblotting assays (F), and the percentage of GTP-Rab2A was quantified and is shown (G) (n = 5 per group, " ��� " indicates p < 0.001). (H) Knockdown of Rab2A attenuates the function of TBC1D1 in the regulation of exogenous PPARγ2 stability. HEK293T cells were cultured and transfected with the indicated plasmids and siRNAs for 2 days, and the cells were then harvested and analyzed by immunoblotting. The level of MYC was quantified and normalized to lane 1. (I, J) Knockdown of Rab2A rescues the effects of lipid storage in HepG2 cells with TBC1D1-S237A protein overexpression. HepG2 cells were cultured and transfected with the indicated lentivirus-expressing plasmids, and then the positive cells were chosen and harvested for Q-PCR (I) and TG testing (J). (K, L) Knockdown of Rab2A rescues the effects of lipid storage in TBC1D1-S231A primary hepatocytes. Primary hepatocytes were isolated, cultured, and transfected with the indicated lentivirus-expressing plasmids. Then, the cells were chosen and harvested for Q-PCR (K) and TG testing (L). The TG level of control cells was normalized to 1, and all the above statistical data were analyzed with unpaired 2-tailed Student t test (n = 3 per group) and are presented as the means ± s.e.m.s. shNC versus shRab2A in Flag-TBC1D1 WToverexpressing HepG2 cells or WT primary hepatocytes (" � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001). shNC versus shRab2A in Flag-TBC1D1 S237A -overexpressing HepG2 cells or TBC1D1-S231A primary hepatocytes ("z" indicates p < 0.05, "zz" indicates p < 0.01, and "zzz" indicates p < 0.001). Flag-TBC1D1 WT versus Flag-TBC1D1 S237A in HepG2 cells or WT versus TBC1D1-S231A primary hepatocytes ("#" indicates p < 0.05, "##" indicates p < 0.01, and "###" indicates p < 0.001). Raw data are given in S1 Excel spreadsheet with raw data from all figures. All the above cellular experiments were performed at least twice with similar results. AMPK, adenosine monophosphate-activated protein kinase; KI, knock-in; siRNA, small interfering RNA; TG, triglyceride; WT, wild-type. https://doi.org/10.1371/journal.pbio.3001522.g003 Nutrition status regulates the activity of the AMPK-TBC1D1-Rab2A axis and subsequently the protein level of PPARγ AMPK is a well-known kinase that senses the energy/nutrient status including cellular glucose levels [26]. Therefore, we investigated how the AMPK-TBC1D1-Rab2A axis regulates PPARγ protein in response to nutrition status. We first utilized a mouse model of western dietinduced obesity (DIO), which develops fatty liver due to overnutrition. Overnutrition in the DIO mice expectedly inactivated AMPK as evidenced by decreased phosphorylation of AMPK, which consequently resulted in lower TBC1D1 phosphorylation in the liver (Fig 5A  and 5B). Notably, Rab2A became GTP-loaded active form in the liver of DIO mice (Fig 5A  and 5B). Both Pparγ1 and Pparγ2 mRNA levels did not change in the liver of DIO mice ( Fig  5C), while their protein levels were significantly increased (Fig 5A and 5B). Moreover, PPARγ target genes were also significantly increased in the liver of DIO mice ( Fig 5C). These data show that the AMPK-TBC1D1-Rab2A axis responds to overnutrition to increase PPARγ protein in mouse livers.
We next employed cell models of glucose starvation to further investigate how the AMPK-TBC1D1-Rab2A axis regulates PPARγ protein in response to nutrition status. As expected, glucose starvation activated the AMPK pathway in both HEK293T and HepG2 cells as evidenced by the increased phosphorylation of AMPK and ACC, which further phosphorylated TBC1D1 at serine 237 site (Fig 5D-5G). Notably, glucose starvation decreased the protein levels of exogenous PPARγ2-MYC in HEK293T cells (Fig 5D and 5F) and also endogenous PPARγ in HepG2 cells (Fig 5E and 5G). We then used an AMPK inhibitor, compound C, to treat cells that were subjected to glucose starvation. Compound C dose dependently inhibited the AMPK activation induced by glucose starvation, which further led to inhibition of glucose starvation-induced TBC1D1 phosphorylation in both cell types (Fig 5H  and 5I). Importantly, Compound C dose dependently restored the protein levels of exogenous PPARγ2-MYC in glucose-starved HEK293T cells (Fig 5H) as well as endogenous PPARγ in glucose-starved HepG2 cells (Fig 5I). Furthermore, glucose depletion lowered the GTP-bound form of Rab2A in HepG2 cells, and compound C reversed this effect (Fig 5J).
Taken together, these data demonstrate that the AMPK-TCB1D1-Rab2A axis regulates PPARγ protein level in response to nutrition status.

Rab2A regulates cellular lipid accumulation
We next investigated whether Rab2A regulates cellular TG storage through PPARγ. Stable overexpression of Rab2A resulted in accumulation of lipid droplets in HepG2 cells as revealed by Oil Red O staining (Fig 6A and 6B). In agreement, cellular TG and total cholesterol (TC)   (Fig 6C and 6D). To define the functions of Rab2A in an unbiased manner, we then performed an RNA sequencing analysis of HepG2 cells stably overexpressing Rab2A. The Gene Ontology (GAU : PleasenotethatGOhasbeendefinedasGeneOntologyinthesentenceTheGeneOntolo O) enrichment analysis revealed that the differentially expressed genes were mainly involved in TG homeostasis and lipoprotein particle remodeling and transport ( Fig  6E). Moreover, we found markedly increased transcription levels of PPARγ target genes in HepG2 cells stably overexpressing Rab2A, such as PLIN4, CIDEA, and FSP27/CIDEC (Fig 6F), as similarly seen in the livers of TBC1D1-KI mice (Fig 1E, S2B Fig). The effects of Rab2A on the expression of PPARγ target genes and the cellular TG accumulation were dependent on its guanine nucleotide binding states. The GTP-bound Rab2A Q65L , but not the GDP-bound Rab2A S20N , promoted the expression of PPARγ target genes ( Fig 6G) and caused cellular TG accumulation (Fig 6H). Importantly, knockdown of PPARγ suppressed Rab2A Q65L -induced expression of PPARγ target genes ( Fig 6G) and prevented Rab2A Q65L -elicited cellular TG accumulation (Fig 6H). In contrast to Rab2A overexpression, stable knockdown of Rab2A decreased the amounts of cellular lipid droplets (Fig 6I and 6J) and suppressed mRNA expression of PPARγ target genes (Fig 6K) in HepG2 cells. Together, these data demonstrate that Rab2A regulates lipid storage gene expression and cellular TG accumulation through PPARγ.

Suppression of Rab2A alleviates diet-induced hepatic lipid accumulation
We next sought to find out whether suppression of Rab2A might help to alleviate diet-induced hepatic lipid accumulation in vivo. To address this question, we down-regulated Rab2A expression in the livers of DIO mice through adeno-associated virus serotype 8 (AAV8) mediated expression of shRNA (shRab2A). After the delivery of shRNA-expressing AAV8 for 2 months, the mice were subjected to molecular and physiological analyses. The molecular analysis of mouse livers confirmed a pronounced decrease of Rab2A at both mRNA and protein levels (Fig 7A and 7B, S8A Fig). Notably, endogenous PPARγ2 protein was significantly decreased in the livers of shRab2A mice, and a less pronounced reduction was also observed for endogenous PPARγ1 protein (p = 0.057) in these mice, although Pparγ mRNA remained normal (Fig 7A and 7B, S8A Fig). These data indicated that the knockdown of Rab2A attenuated the protein stability of PPARγ, which is consistent with the cellular results (Fig 3D, S6C  Fig). Besides the decrease of PPARγ protein, we also observed lower expression of genes involved in lipid metabolism, such as SREBP-1c, FASN, ACL, and ACC, at the protein level (Fig 7A and 7B) and mRNA level (S8A and S8B Fig). In addition, the genes involved in the cholesterol synthesis pathway were also inhibited (Fig 7A and 7B, S8C Fig) and the results showed that the mRNA levels of most of these genes were normal. Surprisingly, unpaired 2-tailed Student t test and are presented as the means ± s.e.m.s. " � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001. Raw data are given in S1 Excel spreadsheet with raw data from all figures. (D-G) Glucose starvation attenuates the protein level of PPARγ. HEK293T cells (D and F) and HepG2 (E and G) cells were cultured and handled with various hours of glucose depletion (D and E) or various concentrations of glucose (F and G) as methods described. The results were analyzed via immunoblotting. (H, I) Inhibition of AMPK signaling with compound C increases the protein level of PPARγ. As described in the methods, HepG2 cells (H) and HEK293T cells (I), were pretreated with various concentrations of inhibitor for 30 minutes and then incubated with the indicated culture medium for 8 hours. The results were analyzed by immunoblotting. (J) Glucose starvation decreases the GTPbound form of Rab2A via the AMPK pathway. HepG2 cells were cultured, stimulated, and harvested for the GST-GRASP55 pulldown assay. The above cellular experiments were performed at least twice with similar results. AMPK, adenosine monophosphate-activated protein kinase.
We subsequently studied the phenotypes of shRab2A mice, particularly those related to lipid metabolism. The liver weight of the shRab2A mice was significantly lower than that of the control mice (shNC) ( Given the impacts of Rab2A on PPARγ and lipid metabolism, the change in liver weight might owe to decreases in hepatic TG and cholesterol storage. Indeed, the TG contents were significantly decreased in the livers of shRab2A mice (Fig 7C), which was in agreement with the reduced lipid droplets detected in liver sections (S8I Fig). We also observed a decreased TC level in the livers of shRab2A mice (Fig 7D). Interestingly, both serum TG and TC levels were significantly decreased in the shRab2A mice (Fig 7E and 7F Taken together, these data demonstrate that suppression of Rab2A lowers PPARγ protein and thereby alleviates diet-induced hepatic lipid accumulation in vivo.

Discussion
Our findings reveal a previously unrecognized role of Rab2A in regulation of hepatic TG storage, which functions genetically at the downstream of the AMPK-TBC1D1 axis to regulate PPARγ protein stability and thereby the expression of PPARγ target genes. Our results are consistent with a composite model in which overnutrition attenuates the phosphorylation levels of AMPK-TBC1D1 signaling and augments the GTP-bound active Rab2A, which consequently promotes hepatic steatosis through increasing the protein stability of PPARγ and the expression of PPARγ target genes (Fig 7G).
AMPK has been implicated in whole-body energy/nutrient metabolism via various downstream targets [4,6,27]. Many studies have indicated that the activation of AMPK in the liver partially attenuates high-fat diet-induced fatty liver, but it is not clear whether, and if so, how a reduction in AMPK activity contributes to the development of this disease [28][29][30]. Here, we show that phosphorylation of AMPK and its downstream target, TBC1D1, is attenuated in the liver of western diet-induced obese mice, which promotes the GTP-bound form of Rab2A and PPARγ protein levels. Our data reveal a previously unknown molecular link between AMPK and PPARγ and strongly demonstrate that genetically blocking the AMPK-TBC1D1 axis test and are presented as the means ± s.e.m.s. " ��� " indicates p < 0.001. (E) GO enrichment analysis based on RNA sequencing data show the major roles of Rab2A in lipid metabolism. Free-Flag versus Flag-Rab2A (n = 3 per group). (F) Marked increases in PPARγ-targeted genes were observed in Rab2Aoverexpressing HepG2 cells. Cell lines with stable Rab2A overexpression were harvested and analyzed by Q-PCR (n = 3 per group). The data were analyzed with unpaired 2-tailed Student t test and are presented as the means ± s.e.m.s. " � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001. (G, H) Knockdown of PPARγ rescues the effects of lipogenesis in HepG2 cells overexpressing Rab2A. HepG2 cells were cultured and transfected with the indicated lentivirus-expressing plasmids, and then the positive cells were chosen and harvested for Q-PCR (G) and TG testing (H). The TG level of control cells was normalized to 1, and the statistical data were analyzed with unpaired 2-tailed Student t test (n = 3 per group) and are presented as the means ± s.e.m.s. shNC versus shPPARγ in Flag-Rab2A WT overexpressing HepG2 cells (" � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001). shNC versus shPPARγ in Flag-Rab2A Q65L -overexpressing HepG2 cells ("z" indicates p < 0.05, "zz" indicates p < 0.01, and "zzz" indicates p < 0.001). shNC versus shPPARγ in Flag-Rab2A S20N -overexpressing HepG2 cells ("#" indicates p < 0.05, "##" indicates p < 0.01, and "###" indicates p < 0.001). (I) The BODIPY staining of Rab2A-knockdown and control HepG2 cells. The cells were incubated with BODIPY for 30 minutes and then fixed for imaging. Representative images are shown. (J) Quantification of the IF intensity in I. The data were analyzed with unpaired 2-tailed Student t test and are presented as the means ± s.e.m.s. " � " indicates p < 0.05. (K) Marked decreases in PPARγ-targeted genes were found in Rab2A-knockdown HepG2 cells. Cell lines with stable Rab2A knockdown were harvested and analyzed by Q-PCR (n = 3 per group). The data were analyzed with unpaired 2-tailed Student t test and are presented as the means ± s.e.m. s. " � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001. Raw data are given in S1 Excel spreadsheet with raw data from all figures. All the above experiments were performed at least 3 times with similar results except for assay in E (once). GO, Gene Ontology; HDL, high-density lipoprotein; TC, total cholesterol; TG, triglyceride; VLDL, very low-density lipoprotein.
https://doi.org/10.1371/journal.pbio.3001522.g006 aggravates the progression of NAFLD in aged mice. A previous study shows that a cholinedeficient high-fat diet (CD-HFD) can also reduce the phosphorylation level of AMPK and demonstrates the important role of the AMPK-Caspase 6 axis in nonalcoholic steatohepatitis [31]. Therefore, AMPK functions as an energy/nutrient sensor, whose inhibition in response to overnutrition contributes to the development of NAFLD at multiple disease stages through distinct downstream targets.
Genetic studies have implicated that TBC1D1 is an important candidate gene of obesity. For example, a human epidemiologic study shows that an R125W coding variant of TBC1D1 confers a risk for familial obesity; however, the underlying molecular mechanism is not clear [32,33]. In mice, whole-body deletion of TBC1D1 confers leanness and protects against obesity induced by a high-fat diet [34,35]. Deficiency of TBC1D1 impairs glucose uptake, but enhances fatty acid uptake and oxidation in skeletal muscle [34,35]. The TBC1D1-Ser231 KI mutation in mice gives rise to obesity and type 2 diabetes through promoting IGF1 secretion and thereby increasing lipogenesis in the adipose tissue [9]. These studies show that TBC1D1 may regulate lipid metabolism in different tissues via distinct mechanisms. Here, we reveal a previously unrecognized role of TBC1D1 in the control of hepatic lipid storage through composite actions involving both direct binding and Rab2A-dependent regulation of PPARγ ( Fig  7G), which not only further demonstrates the importance of TBC1D1 in lipid metabolism but also links this protein to another metabolic disease NAFLD.
Our previous study shows that the TBC1D1 S231A mutation inhibits its binding to 14-3-3s and increases its GAP activity toward Rab8A, thus resulting in a decrease of GTP-bound Rab8A [9]. The increased GAP activity by the TBC1D1 S231A mutation causes a partial inhibition of AICAR-stimulated GLUT4 transport and glucose uptake in skeletal muscle presumably through decreasing the GTP-bound form of a yet unknown Rab small G protein [7]. In this study, we find that TBC1D1 S231A mutation and TBC1D1-Ser231 hypophosphorylation through inhibition of AMPK both result in an increase of the GTP-bound active Rab2A, suggesting that Rab2A might not be a direct substrate of TBC1D1 in vivo (Fig 3E-3G). Small G proteins on a pathway may be networked to one another via their GAPs or guanine nucleotide exchange factors (GEFs) to form signaling cascades [36]. It is possible that TBC1D1 might regulate a Rab protein that, in turn, controls some GAPs or GEFs for Rab2A. Alternatively, TBC1D1 might have crosstalk with other pathways to indirectly regulate Rab2A activation. A third possibility is that Ser231 dephosphorylation might relocate TBC1D1 away from Rab2A thereby allowing a Rab-GEF to increase the GTP-loaded form of Rab2A. Nevertheless, our genetic analysis demonstrates that Rab2A functions genetically at the downstream of TBC1D1 to regulate PPARγ protein stability. Both TBC1D1 and Rab2A interact with PPARγ and may regulate PPARγ protein stability in a cooperative manner.
The functions of Rab GTPases in lipid droplet metabolism, such as Rab8A [37] and Rab18 [38,39], have been discovered but have not yet been implicated in the progression of NAFLD. The results from this study reveal the novel functions of Rab2A regulated by the AMPK-TBC1D1 axis in hepatic steatosis. Previous studies have demonstrated that Rab2A is localized at the Golgi apparatus and that this protein is essential for protein transport [24]. It also plays a critical role in regulation of the formation of autophagosomes [40]. Whole-body deletion of Rab2A leads to preweaning lethality in mice, thus preventing the utilization of this bound Rab2A and then increases the protein stability of PPARγ, which ultimately promotes the progression of NAFLD. Meanwhile, TBC1D1 can also mildly promote the protein stability of PPARγ via constitutive binding independent of phosphorylation at the serine-231 site. These data were analyzed with unpaired 2-tailed Student t test and are presented as the means ± s.e.m.s. " � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001. Raw data are given in S1 Excel spreadsheet with raw data from all figures. NAFLD, nonalcoholic fatty liver disease; shRNA, short hairpin RNA; TG, triglyceride.
https://doi.org/10.1371/journal.pbio.3001522.g007 model in studying its functions in the adulthood [22]. Therefore, liver-specific Rab2A knockout models are needed to further delineate the in vivo functions of Rab2A in regulation of hepatic lipid metabolism and in the pathogenesis of NAFLD in the future.
It is intriguing that the protein levels of PPARγ, including PPARγ1 and PPARγ2, in the livers of mice with western DIO are significantly higher than those in the livers of mice with normal chow diet, while the Pparγ mRNA level is comparable between the 2 groups of mice. This is in contrast to previous reports in which the Pparγ transcripts are increased in the liver of high-fat diet-fed mice or leptin-deficient mice [12,13,41]. In our study, the GTP-bound form of Rab2A regulates the stability of PPARγ mainly by interacting with the AF-2 domain to prevent its proteasomal degradation. The AF-2 domain is responsible for ligand binding and has been shown to be critical for ligand-induced degradation [42]. It is possible that Rab2A may play a critical role in the ligand-induced PPARγ degradation. Future studies are required to elucidate the molecular mechanisms through which Rab2A binds to PPARγ, the components involved in this process, and the relationship with ligand binding. Recently, researchers have demonstrated that the serine 273 site of PPARγ is linked to obesity and insulin resistance [11], and other studies have shown that the phosphorylation of S273 can inversely reduce the total protein level of PPARγ in white adipose tissue through a yet unknown mechanism [16,43]. Given that PPARγ is primarily expressed in the adipose tissue, it is intriguing to find out whether the mechanism regulating hepatic PPARγ stability might also play a role in regulating adipose PPARγ protein.
In summary, we show that Rab2A functions genetically at the downstream of the AMPK-TBC1D1 axis to regulate hepatic PPARγ protein stability and thereby the expression of PPARγ target genes for TG accumulation in the liver in response to energy/nutrient status. Our findings may have therapeutic implications for treatment of fatty liver disease.

Materials and methods
All mice were housed in a pathogen-free environment with a 12-hour light/12-hour dark cycle and had free access to water consumption and food intake. All animal breeding, husbandry, care, euthanasia, and use procedures followed the guidelines provided by the Ethics Committees of Nanjing University (Approval number MARC-CS3) and Anhui Medical University (Approval number LLSC20200327). Extended methods and information about reagents, cell culture, TBC1D1 S231A -KI mouse model, DIO mouse model, AAV8-mediated Rab2A gene knockdown in the liver, transfection and plasmids, immunoblotting and antibodies, glucose starvation, immunoprecipitation, RNA isolation and quantitative PCR, RNA sequencing, prokaryotic expression and purification, pull down assay of Rab2A-GTP, cytoplasmic and nuclear extraction, immunofluorescence staining and imaging, Oil Red O staining and imaging, histology and imaging, blood chemistry, measurement of liver and cellular TC and TG levels, oral glucose tolerance test and insulin tolerance test, and statistical analysis are described in S1 Supplemental Materials and Methods. The original RNA sequencing data are listed in S1 RNA Sequencing data. The raw data of statistical results are given in S1 Excel spreadsheet with raw data from all figures and raw western blot images are shown in S1 Raw images. HepG2 (E) and HEK293T (F and G) cells were cultured and transfected with the indicated plasmids for 2 days, and the cells were collected and analyzed by immunoblotting. (H) Gradient overexpression of TBC1D1 S237A plasmids increases the protein stability of exogenous PPARγ2. HEK293T cells were cultured and transfected with the indicated plasmids, and 2 days later, the cells were collected and analyzed by immunoblotting. (I) Blocking the phosphorylation of TBC1D1 at serine 237 increases endogenous protein levels of PPARγ. HepG2 cells were cultured, transfected with the indicated plasmids for 2 days, harvested and analyzed by immunoblotting. (J) Blocking the phosphorylation of TBC1D1 at serine 237 increases the activation of PPARγ. TBC1D1 WT -and TBC1D1 S237A -overexpressing HepG2 cells were harvested and analyzed by Q-PCR (n = 3 per group). The data were analyzed with unpaired 2-tailed Student t test and are presented as the means ± s.e.m.s. Free-Flag versus Flag-TBC1D1 WT (" � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001). Free-Flag versus Flag-TBC1D1 S237A ("z" indicates p < 0.05, "zz" indicates p < 0.01, and "zzz" indicates p < 0.001). Flag-TBC1D1 WT versus Flag-TBC1D1 S237A ("#" indicates p < 0.05, "##" indicates p < 0.01, and "###" indicates p < 0.001). (K) Binding assay between WT TBC1D1 and exogenous PPARγ2. HEK293T cells were transfected with Flag-TBC1D1 and PPARγ2-MYC plasmids, harvested, and immunoprecipitated with anti-Flag beads and analyzed by immunoblotting. (L) Binding assay between WT TBC1D1 and endogenous PPARγ. HepG2 cells stably expressing Flag-TBC1D1 were harvested and immunoprecipitated with anti-Flag beads and analyzed by immunoblotting. (M, N) Overexpression of WT TBC1D1 attenuated the degradation of PPARγ2. HEK293T cells were cultured and transfected with the indicated plasmids, and 2 days later, the cells were stimulated with 200-μM CHX for the indicated hours. The cells were then harvested and analyzed by immunoblotting; the MYC blots were spliced to obtain a similar baseline protein level (M). (N) Quantification of PPARγ2 levels in C. The ratios in lanes 1 and 7 were defined as 1, respectively. Raw data are given in S1 Excel spreadsheet with raw data from all figures. All experiments were performed at least twice with similar results. CHX, cycloheximide; WT, wild-type.  N) ITT of AAV8-Rab2A-shRNA and control mice (n = 5 per group). The data were analyzed with unpaired 2-tailed Student t test and are presented as the means ± s.e.m.s. " � " indicates p < 0.05, " �� " indicates p < 0.01, and " ��� " indicates p < 0.001. Raw data are given in S1 Excel spreadsheet with raw data from all figures. AAV8, adeno-associated virus serotype 8; EDL, extensor digitorum longus; epWAT, epididymis white adipose tissue; ITT, insulin tolerance test; NEFA, nonesterified fatty acid; OGTT, Oral glucose tolerance test; prWAT, perirenal white adipose tissue; scWAT, subcutaneous white adipose tissue; shRNA, short hairpin RNA; TA, tibialis anterior muscle. (TIF) S1 Supplemental Materials and Methods. Extended reagents, materials and methods with detail description. (DOCX) S1 RNA Sequencing data. Original RNA sequencing data. (XLSX) S1 Excel spreadsheet with raw data from all figures. The raw data of statistical results. (XLSX) S1 Raw images. Raw western blot data. (PDF)