High-Fat Diet–Induced DeSUMOylation of E4BP4 Promotes Lipid Droplet Biogenesis and Liver Steatosis in Mice

As the most common chronic liver disease, nonalcoholic fatty liver disease (NAFLD) affects approximately two-thirds of obese individuals and is strongly associated with type 2 diabetes (1,2). The early stage of NAFLD is simple liver steatosis, defined as accumulation of hepatic triglycerides (TGs) >5% of total liver weight (3,4). Although simple liver steatosis is considered largely benign, excessive lipid accumulation may sensitize hepatocytes to stress-induced injury and promote the progression of NAFLD toward nonalcoholic steatohepatitis and cirrhosis (5,6). The excess of lipids found during liver steatosis are primarily neutral lipids stored in organelles called lipid droplets (LDs) in hepatocytes (7,8). The biogenesis of LDs is very dynamic and regulated at both transcriptional and posttranslational levels (7,9). It has been shown that abnormalities in LD biology contribute to fatty liver disease (7,10–12). However, detailed molecular mechanisms underlying the impairment of LD metabolism are largely unclear.

Several proteins, particularly cell death–inducing DNA fragmentation factor α-like effector (CIDE) proteins, were found to regulate LD biogenesis (11,13,14). Both CIDEA and CIDEC (also called FSP27 in mice) are upregulated in obese mice and associated with increased TG storage in the liver (7,14,15). Murine models of hepatic steatosis demonstrated a significant increase in levels of FSP27 in the liver (16–18). The liver of obese humans was revealed to express high levels of FSP27, which tended to decrease after significant weight loss with bariatric surgery (19). Functionally, forced expression of FSP27 in hepatocytes in vitro or in vivo led to increased LD and TG accumulation (18,20). Depletion of hepatic Fsp27 with antisense oligo against Fsp27 ameliorated diet-induced steatohepatitis (21–23). Hepatocyte-specific deletion of Fsp27 was reported to reduce liver TGs and injury in mice with alcoholic steatohepatitis (18). Collectively, these studies have indicated that elevated levels of FSP27 in the liver could drive liver steatosis in response to overnutrition or obesity. Thus, suppression of FSP27 expression and function might be a viable approach to treating NAFLD.

Transcriptional regulation has been implicated as the major regulatory pathway for controlling the expression of FSP27 in livers (13,14). So far, several transcription factors have been identified to directly regulate the transcription of Fsp27 within hepatocytes in response to fasting or diets, including peroxisome proliferator–activated receptor-α (PPAR-α), PPAR-γ, liver X receptor-α, hepatocyte nuclear factor-α, and CREB3L3 (also known as CREBH) (16,20,22,24,25). PPAR-γ was found to promote liver Fsp27 expression in ob/ob mice (17), whereas CREBH, a liver-enriched and stress-inducible transcription factor, directly binds to the promoter of Fsp27 and activates its gene expression (20). Enforced expression of CREBH elevates FSP27 levels and promotes the enlargement of LDs and TG accumulation in mouse liver. However, much less is known about transcriptional mechanisms that account for the induction of Fsp27 in liver of NAFLD/nonalcoholic steatohepatitis. Moreover, the signaling pathways that suppress Fsp27 expression in response to overnutrition remain largely unknown.

E4 promoter–binding protein 4 (E4BP4) is a basic leucine zipper transcription factor that has been known for its role in immunomodulation, circadian rhythms, and de novo lipogenesis (26–28). Overexpression of hepatic E4BP4 has been shown to increase gluconeogenesis and induce insulin resistance in both the liver and muscle (29). We previously reported that hepatic E4BP4 is potently induced by endoplasmic reticulum (ER) stress signals. E4bp4 deficiency reduces accumulation of LDs in hepatocytes following ER stress stimulation (30). Hepatocyte E4bp4 deficiency also reduces liver steatosis in mice fed a high-fat, low–l-methionine, and choline-deficient diet. Although E4BP4 may impact lipid metabolism in multiple ways within hepatocytes, it remains unclear whether E4BP4 directly regulates LD biogenesis by controlling the expression of LD binding proteins. Intriguingly, E4BP4 was found to interact with CREBH and contribute to CREBH-mediated gene regulation in mouse liver (31). Yet, no study has been reported to link E4BP4 and the expression of Fsp27 during the pathogenesis of high-fat diet (HFD)–induced liver steatosis.

Protein SUMOylation is an important posttranslational modification that regulates protein turnover, protein-protein interaction, signal transduction, chromatin modifications, and transcription repression (32,33). Recent studies highlighted that SUMOylation of transcriptional factors plays a crucial role in maintaining metabolic homeostasis. It was reported that a SUMOylation-defective mutant of liver receptor homolog 1 enhances SREBP-1 processing and promotes de novo lipogenesis, leading to fatty liver disease (34). In a global SUMOylation analysis, E4BP4 was identified as a putative SUMOylation target (35). Subsequently, E4BP4 was shown to be modified by SUMOylation and identified five specific lysine residues targeted for SUMOylation (36). However, almost nothing is known about the biological actions of E4BP4 SUMOylation in LD formation and lipid accumulation in hepatocytes and during the pathogenesis of fatty liver disease. In this study, we present both in vivo and in vitro evidence supporting the role of hepatic E4BP4 and its SUMOylation during LD expansion in hepatocytes and liver steatosis in response to HFD feeding.

All animal experiments were approved by and performed in accordance with guidelines of the institutional animal care and use committee at the University of Michigan Medical School. Liver-specific E4bp4 knockout (E4bp4-LKO) mice were generated as described previously (30). Eight-week-old E4bp4^flox/flox and E4bp4-LKO mice were challenged with an HFD (45% kcal from fat) for 12 weeks; body weight was recorded weekly. For the FSP27 overexpression experiment, 8-week-old male mice were fed an HFD for 2 weeks. Ad-GFP or Ad-Fsp27 adenovirus was injected via tail vein at a dose of 1.5 × 10¹¹ viral particles/mouse. HFD continued for an additional 10 days prior to liver harvesting for metabolic assays.

Hepa-1c1c cells and Huh-7 cells were purchased from ATCC and maintained according to manufacturer instructions. Isolation of primary mouse hepatocytes (PMHs) was performed by the collagenase perfusion method via inferior vena cava. The reagents used for cell treatment, including norepinephrine forskolin, rapamycin, and AICAR, were all from Cayman Chemical.

E4bp4-LKO PMHs were seeded at the density of 8 × 10⁴ cells/well in a 12-well plate. Ad-LacZ control, Ad-E4bp4, or Ad-Fsp27 adenovirus was applied 2 h after seeding. After being switched to serum-free DMEM (Gibco) the next day, the cells were incubated with 300 μmol/L oleic acid overnight prior to BODIPY (D3922, 493/503, 2 μmol/L; Invitrogen) staining following the protocol described by Qiu and Simon (37). For Huh-7 cells, cells were first transduced with Ad-E4bp4-wild type (WT) or Ad-E4bp4-5KR. Twenty-four hours later, cell culture medium was changed to serum-free minimum essential medium (Gibco), and cells were treated with BSA or palmitic acid (300 μmol/L, dissolved in 0.1 mol/L NaOH at 70°C and conjugated with 10% free fat acid–free BSA in PBS) plus oleic acid (600 μmol/L) for 6 h. Cells were washed with serum-free minimum essential medium once and treated with forskolin at 2.5 μmol/L and 5 μmol/L for 16 h before BODIPY staining.

After fixation in 10% formalin at room temperature overnight, mouse liver samples were sent to the Rogel Cancer Center Tissue and Molecular Pathology Shared Resource Core at the University of Michigan Medical School for paraffin embedding and hematoxylin-eosin (H-E) staining.

For whole-cell lysate preparation, PMHs were washed once in 1× PBS buffer and lysed in radioimmunoprecipitation assay (RIPA) buffer supplemented with 1× protease inhibitor (Roche Applied Science). Liver tissues were weighed and homogenized in RIPA buffer (8 μL/mg tissue weight).

For cytosolic and nuclear protein preparation, liver tissues were weighted, and hypotonic buffer (8 μL/mg tissue weight) was added. Cytosolic protein was collected in the supernatant by using a Dounce homogenizer followed by full-speed centrifugation. The pellet was washed once with hypotonic buffer and resuspended in RIPA buffer (4 μL/mg tissue weight) to obtain the nuclear fraction. After protein lysates were precleared at maximal speed at 4°C in a microfuge, the protein concentration of each supernatant was measured using Bio-Rad Protein Assay Dye. Equal amounts of protein samples were separated using 9% SDS-PAGE and transferred to nitrocellulose membrane (Bio-Rad). The membranes were incubated with primary antibodies at 4°C overnight. Horseradish peroxidase–conjugated secondary antibodies against mouse, rabbit, or goat IgG and homemade enhanced chemiluminescence substrate were used for detecting chemiluminescence on an AlphaImager HD2 (Cell Biosciences).

Total RNA was extracted with TRIzol (Invitrogen) and chloroform. cDNA was synthesized with M-MLV Reverse Transcriptase and random hexamers (Thermo Fisher Scientific, Surrey, U.K.) and subjected to quantitative PCR (qPCR) using 2× Fast qPCR Mix (ABclonal Science) on an ABI QuantStudio 5 Real-Time PCR System (Applied Biosystems, Foster City, CA). The value of each cDNA was calculated using the ΔΔCt method and normalized to the value of the housekeeping gene controls, including β-actin, Gapdh, or 18S rRNA. The data were plotted as fold change.

Serum cholesterol, TGs, and ALT were measured using commercial kits (A7510, A7525, and T7532; Pointe Scientific) according to the manufacturer’s instructions. For liver cholesterol and TG content, liver tissues were homogenized for total lipid extraction according to Bligh and Dyer (38). Total lipids were resuspended in 400 μL ethanol and incubated at 55°C for 10 min.

A 1.7-kb Fsp27 promoter containing the CREBH responsive element was amplified with primers and ligated with the pGl4 plasmid, which was double digested by KpnI and XhoI, to generate the Fsp27 luciferase reporter construct. The mouse Fsp27 coding sequence was first cloned into pDONR/Zeo vector via Gateway BP recombination and then transferred into pAdCMV-DEST plasmid via Gateway LR recombination. All the adenoviruses were generated after transfection of 293AD cells and concentrated by ultracentrifuge in CsCl gradient and purified by Vivaspin 20 concentrators. Ad-shCrebh with the mouse Crebh-targeting sequence of GAGCAGAAAGTTCTACTAAAT was a gift of K.Z.

A total of 2 × 10⁵ Hepa1 cells/well were seeded in 24-well plates. Transfection was performed with Opti-MEM (2021431; Gibco) and polyethylenimine. The cells were cotransfected with active Crebh plasmid, E4bp4 plasmid, and Fsp27 luciferase reporter construct along with β-galactosidase (β-gal) plasmid. Twenty-four hours later, luciferase activity was measured with luciferin (103404-75-7; Gold Biotechnology) on a BioTek Synergy 2 microplate reader and normalized by β-gal assay.

Total RNA was extracted from the liver of HFD-fed E4bp4^flox/flox (n = 4) and E4bp4-LKO mice (n = 4). Equal amounts of RNA from the livers of mice of the same genotype were pooled and analyzed at the University of Michigan Medical School Advanced Genomic Core for Affymetrix microarray analysis. Gene expression levels were shown as expression values of log₂-transformed data generated via robust multiarray averages (39). The heat map was generated using the pheatmap R package to compare gene expression levels between HFD-fed E4bp4^flox/flox and E4bp4-LKO mice. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed by using the Database for Annotation, Visualization and Integrated Discovery and visualized using the ggplot2 R package.

A total of 1 × 10⁸ 293AD cells was transduced with Ad-GFP or Ad-Flag-E4bp4 for 16 h prior to whole-cell lysate preparation. Protein samples were separated using 9% SDS-PAGE. The gel was stained with Coomassie blue dye for 30 min and destained overnight. The bands showed up only in the Ad-Flag-E4bp4 group and were cut out for mass spectrometry at the Proteomics Resource Facility, Department of Pathology, University of Michigan Medical School.

Statistical analysis was performed using GraphPad Prism 9.3 software. For comparison between two groups, statistical significance was determined by the unpaired two-tailed Student t test. To compare multiple groups, one-way ANOVA with post hoc Tukey test was used in studies with one independent variable, whereas two-way ANOVA with post hoc Tukey test was used for studies with two independent variables. All results are presented as mean ± SEM. Differences were considered statistically significant at P < 0.05.

All the data and critical resources supporting our findings, as well as methods and conclusions, are available from the corresponding authors upon request. The original Affymetrix microarray data are available in the National Center for Biotechnology Information Gene Expression Omnibus database (GSE 200528).

Given the role of E4BP4 in promoting lipogenesis and enhancing lipid accumulation upon ER stress (26,28,30), we hypothesized that hepatic E4bp4 deficiency may suppress the development of liver steatosis following HFD feeding. To test this hypothesis, both E4bp4^flox/flox and E4BP4-LKO mice were subjected to 12-week HFD feeding. Through the whole feeding period, no differences were observed between E4BP4-LKO and E4bp4^flox/flox mice in body weight, weight of white adipose tissue, and weight of liver (Fig. 1A–C). Insulin tolerance test after 5 weeks of HFD and glucose tolerance test after 8 weeks of HFD showed no differences between E4BP4-LKO and E4bp4^flox/flox mice (Supplementary Fig. 1A and B). In addition, overnight fasting glucose was only slightly lower in E4BP4-LKO mice, without statistical significance. The expression levels of the key gluconeogenic genes were comparable between the two groups (Supplementary Fig. 1C and D). These findings indicate that hepatocyte E4BP4 is largely dispensable for body weight gain and glucose metabolism in mice following HFD feeding.

However, we observed a >40% reduction in serum ALT, a serum marker for liver injury (Fig. 1D), 20% reduction in serum cholesterol (Fig. 1E), and no change in serum TGs in HFD-fed E4bp4-LKO mice (Fig. 1F). Liver tissues from E4bp4-LKO mice contained significantly less TG and cholesterol (Fig. 1G and H). Further H-E staining also showed greatly reduced lipid accumulation within E4bp4-LKO liver (Fig. 1I). Liver steatosis at the late stage of NAFLD is associated with increased immune cell infiltration and hepatic stellate cell activation (1,5). By RT-qPCR, a downward trend of fibrosis markers was observed in the liver of E4bp4-LKO mice (Col1a1 and Col1a2) (Supplementary Fig. 2A). The expression of inflammatory markers was not significantly altered (Supplementary Fig. 2B). In summary, hepatic E4bp4 deficiency protects mice against HFD-induced liver steatosis and liver injury without affecting adiposity, glucose metabolism, and insulin sensitivity. Altogether, these findings are consistent with our previous report on high-fat, low–l-methionine, and choline-deficient–fed E4bp4-LKO mice (30).

To identify downstream pathways regulated by E4BP4 during HFD feeding, we performed a microarray analysis with liver samples from both E4bp4^flox/flox and E4BP4-LKO mice on a 12-week HFD. The microarray heat map highlighted the top 10 upregulated and downregulated genes (>1.5 fold change) (Fig. 2A). Among these downregulated genes, Cidea and Cidec are two classical LD binding proteins (11), whereas Gpnmb and Mmp12 are shown to be involved in liver fibrosis (40,41). As for the upregulated genes, Cyp2c54 was found to be primarily expressed in the liver, and its function is related to fatty acid oxidation (42). As a solute carrier anion transporter responsible for transporting bile acids into hepatocytes, Slco1a1 (also called Oatp1) expression level was markedly suppressed in the mouse liver following a methionine-choline–deficient diet (43). Serpina1e was recently found to be involved in hepatic gluconeogenesis (44). The KEGG pathway analysis showed that metabolic pathways are the most significantly affected by the loss of E4bp4 in the liver of E4bp4-LKO mice (Fig. 2B), consistent with the profile of genes observed in microarray analysis.

Next, RT-qPCR analysis was performed to verify the microarray data. Several major LD binding proteins, including Cidea, Fsp27, and Plin 2, 3, 4, and 5 were indeed potently downregulated in the E4bp4-LKO liver (Fig. 2C). In contrast, no major changes in lipogenic pathway (Fasn, Scd1, and Acc1), fatty acid oxidation (Cpt1a, Acox1, and Ehhadh), and lipolysis (Pnpla2 and Lipe) pathways were detected (Fig. 2D–F). CD36, a fatty acid translocase, was significantly downregulated in the liver of E4BP4-LKO mice (Fig. 2G). This is interesting since hepatocyte-specific disruption of CD36 attenuates liver steatosis upon HFD feeding in mice (28,45). The RT-qPCR data showed no differences in genes involved in hepatic cholesterol metabolism between E4bp4^flox/flox and E4bp4-LKO mice (Supplementary Fig. 3). Meanwhile, both E4BP4 and FSP27 protein levels in the liver were markedly reduced in E4bp4-LKO mice (Fig. 2H). Taken together, our results indicate that hepatic E4bp4 deficiency reduces HFD-induced liver steatosis and liver injury with a concomitant suppression of LD binding and lipid uptake genes.

So far, whether E4BP4 can regulate the expression of Fsp27 in a cell-autonomous manner remains unclear. To address this question, we manipulated E4BP4 expression in E4bp4-LKO PMHs and examined its effect on LD formation and Fsp27 expression. Loss of E4bp4 led to a significant decrease of LD numbers in Ad-LacZ–transduced E4bp4-LKO PMHs compared with E4bp4^flox/flox PMHs. In contrast, adenovirus-mediated overexpression of E4BP4 indeed increased the number of LDs and the expression of the liver-specific Fsp27β isoform and Cidea in E4bp4-LKO PMHs (Fig. 3A and B). In contrast, acute depletion of E4bp4 abrogated not only the basal expression of Fsp27β but also its induction after stimulation with the saturated fatty acid (palmitate) plus the inflammatory cytokine tumor necrosis factor-α (TNF-α) (Fig. 3C and D), an in vitro condition mimicking diet-induced nonalcoholic steatohepatitis (46). These results indicate that E4BP4 is both necessary and sufficient to promote the transcription of Fsp27β in hepatocytes.

How E4BP4 enhances the transcription of Fsp27β intrigued us, since the canonical function of E4BP4 acts as a transcriptional repressor (26,47,48). Furthermore, a genome-wide chromatin immunoprecipitation assay with anti-E4BP4 did not reveal direct binding of E4BP4 to either Fsp27 or Cidea promoter (49). Previously, we reported that E4BP4 stabilizes nuclear SREBP-1c and subsequently enhances SREBP-1c–driven lipogenic gene expression via protein-protein interaction (28). We postulated that E4BP4 might promote the transcription of Fsp27 indirectly via collaboration with other transcriptional activators of Fsp27. We then analyzed the protein levels of several FSP27 regulators (16,20,50) in PMHs transduced with Ad-E4BP4. The immunoblotting results showed that E4BP4 overexpression elevated the abundance of CREBH and SREBP-1c (to a lesser extent) while showing no effects on PPAR-γ (Fig. 3E). In contrast, a marked reduction of the nuclear CREBH was observed in the liver of E4BP4-LKO mice following HFD feeding (Fig. 3F). Consistent with these observations, knockdown of Crebh by Ad-shCrebh blocked the induction of Fsp27β by E4BP4 overexpression in hepatocytes (Fig. 3G). Furthermore, we tested the effects of E4BP4 on the activity of the mouse Fsp27 promoter, which was reported to contain a binding site for CREBH (20). Compared with active CREBH overexpression, E4BP4 alone had minimal induction of the Fsp27 promoter–driven luciferase reporter in Hepa1 cells (Fig. 3H). However, its overexpression significantly boosted the CREBH-induced transaction of Fsp27, supporting that E4BP4 most likely promotes the Fsp27 transcription via CREBH.

We also examined whether SREBP-1c could be involved in E4BP4-stimulated expression of Fsp27. Acute depletion of Srebp-1c also tended to reduce the ability of E4BP4 to induce the Fsp27 expression without reaching statistical significance (Supplementary Fig. 4). In summary, these data suggest a link between E4BP4 and CREBH, pointing to E4BP4 as a critical regulator of FSP27 expression mainly via CREBH in hepatocytes.

Next, we investigated to what extent FSP27 mediates the effects of E4BP4 on lipid content and the number of LDs in the liver. For this purpose, an adenovirus expressing the Fsp27α isoform (Ad-Fsp27) was generated and used to transduce PMHs along with Ad-LacZ control. Of note, the functional motifs of FSP27α and FSP27β are almost indistinguishable despite tissue-specific expression patterns (18,20). Indeed, compared with Ad-LacZ–transduced E4bp4-LKO PMHs, Ad-Fsp27 markedly increased the number of LDs in E4bp4-LKO PMHs (Fig. 4A). Next, we injected E4bp4^flox/flox mice with Ad-GFP and E4bp4-LKO mice with Ad-GFP control or Ad-Fsp27 after 2 weeks of HFD feeding and then examined liver lipid metabolism 10 days later. By the end of the experiment, we confirmed overexpression of Fsp27 at both mRNA and protein levels (Fig. 4B and C). The total liver TG content was increased significantly in the Ad-Fsp27–injected E4bp4-LKO mice versus the Ad-GFP–injected E4bp4-LKO mice (∼2.8-fold, P < 0.01) (Fig. 4D). The total cholesterol content showed no significant difference among the three groups (Fig. 4E). Liver H-E staining revealed that both the size and number of LDs were markedly increased by Ad-Fsp27 in E4bp4-LKO mice (Fig. 4F). However, no differences were found in terms of serum levels of ALT, TGs, and cholesterol among the three groups of mice (Supplementary Fig. 5A–C). Thus, both in vitro and in vivo data support that FSP27 is a downstream effector that links E4BP4 and liver steatosis in response to HFD feeding.

So far, we have demonstrated that E4BP4 overexpression stimulates the expression of Fsp27 in hepatocytes. We also showed that such action primarily depends on CREBH. These findings raised an important question of how exactly E4BP4, a transcriptional repressor, activates the expression of Fsp27. To test the possibility that E4BP4 could change its transcriptional activity via posttranslational modifications, we transduced PMHs with Ad-Flag-E4bp4 and performed immunoprecipitation with anti-FLAG for proteomic analysis to identify novel interaction proteins and posttranslational modifications of E4BP4. Two distinctive bands were cut out for proteomic analysis of the immunocomplex of E4BP4 (Fig. 5A). The top band had a higher molecular weight than that of E4BP4 and was shown to be an E4BP4-dominant protein complex that contained metastasis associated 1 family member 2 (MTA2), small ubiquitin-like modifier 2 (SUMO2), SUMO3, GATA binding protein 2A (GATA2A), and GATA2B (Fig. 5B). The lower band matched the molecular weight of E4BP4 that was identified as the major constituent along with a panel of other transcriptional regulators. Transducin β-1 X-linked (TBL1) and its receptor TBLR1 were reported to function as a transcription cofactor in controlling lipid mobilization in white adipose tissue (51). WD-repeat domain 1 (WDR1) is a major cofactor of the actin depolymerizing factor and accelerates actin disassembly. Mice with Wdr1 deficiency in cardiomyocytes exhibit cardiac hypertrophy and myocardial fibrosis (52). Histone deacetylase 2 (HDAC2), one of the class I HDAC enzymes, has been shown to have diverse biological function in cellular metabolism and liver diseases (53). Finally, lysing acetyltransferase 8 (KAT), is an important epigenetic regulator for embryonic development and normal chromatin architecture (54). Recently, KAT8 was also found to play a role in liver metabolism and liver injury (55,56). Until now, the interactions between E4BP4 and these epigenetic regulators are novel and unexpected. We suspect that E4BP4 might function differently via interacting with different transcription cofactors or epigenetic regulators.

The proteomic finding that E4BP4 coexists with SUMO2 and SUMO3 peptides in hepatocytes prompted us to postulate that E4BP4 might be modified by SUMOylation in hepatocytes. In fact, E4BP4 was found to be modified by SUMOylation in 293T cells via an unbiased screen (35). The five lysine residues within E4BP4 for SUMOylation were identified in a recent report (36) (Fig. 5C). To verify our proteomic finding, we performed an in vitro SUMOylation assay using purified glutathione S-transferase (GST)-E4BP4 and recombinant SUMO enzymes with or without Mg²⁺ plus ATP (Mg-ATP). Indeed, compared with GST alone control, GST-E4BP4 showed distinct bands of SUMO2/3 conjugations in the presence of SUMO enzymes in an Mg-ATP–dependent manner (Fig. 5D). Taken together, our data not only confirm that E4BP4 is a direct target of SUMOylation but also demonstrate for the first time that E4BP4 SUMOylation occurs in hepatocytes.

Protein SUMOylation is dynamically regulated by hormonal and nutritional signaling (32). For example, SUMOylation of hepatic bile acid sensor farnesoid X receptor was found to be reduced upon HFD feeding, while its acetylation was elevated in the liver (57). We therefore asked whether hepatic E4BP4 SUMOylation is also regulated by nutritional signaling. We first examined E4BP4 SUMOylation in the liver after HFD feeding. E4BP4 SUMOylation was detected in the liver of regular chow–fed mice but became largely undetectable after 10 weeks of HFD feeding despite elevated hepatic E4BP4 protein (Fig. 5E). Next, we addressed whether specific nutritional signals could promote deSUMOylation of E4BP4 within hepatocytes. We chose to test palmitate, the saturated fatty acid known to be associated with liver steatosis, insulin resistance, and ER stress (6). After overnight treatment with palmitate at 400 μmol/L, E4BP4 SUMOylation was markedly reduced, whereas the E4BP4 total protein increased, reminiscent of the pattern in mouse liver after HFD feeding (Fig. 5F). Taken together, our data suggest that hepatic E4BP4 exists in both SUMOylated and deSUMOylated states. More importantly, nutritional stress, including HFD feeding and palmitate treatment, can lead to deSUMOylation of E4BP4 in the liver and hepatocytes.

To further investigate the biological significance of E4BP4 SUMOylation in regulating hepatic lipid metabolism, we generated adenoviruses to express E4BP4-WT and E4BP4-5KR respectively. Since protein SUMOylation affects both protein stability and activity, we compared the protein stability of E4BP4-WT and E4BP4-5KR mutant in a cycloheximide chase experiment. As shown in Supplementary Fig. 6A, both E4BP4-WT and E4BP4-5KR exhibited similar rates of degradation, consistent with a previous report (36). Functionally, both E4BP4-WT and E4BP4-5KR significantly upregulated Fsp27β and enhanced the protein abundance of CREBH in PMHs (Supplementary Fig. 6B and C). Furthermore, we found that both E4BP4-WT and E4BP4-5KR interacted with CREBH in a coimmunoprecipitation assay (Supplementary Fig. 6D). Taken together, nonSUMOylatable E4BP4-5KR mutant behaves similarly to E4BP4-WT at basal and nonstimulated conditions.

During fasting, breakdown of LDs and lipolysis are stimulated by signals, including cAMP/cAMP-dependent protein kinase (PKA), AMPK, and inhibition of mammalian target of rapamycin (58–60). We therefore examined whether E4BP4-WT and E4BP4-5KR respond differently to these signaling pathways. Both E4BP4-WT and E4BP4-5KR behaved similarly in the presence and absence of rapamycin or AICAR treatment (Supplementary Fig. 7A and B). However, in the presence of forskolin, a cAMP/PKA activator, E4BP-WT degraded within 6 h after treatment, significantly faster than E4BP4-5KR (Fig. 6A). Given the potent activation of PKA and lipolysis by norepinephrine in hepatocytes (61), we also treated hepatocytes with norepinephrine and observed a similar pattern where norepinephrine stimulated the degradation of E4BP4-WT but not E4BP4-5KR (Fig. 6B), supporting that deSUMOylation renders E4BP4 more resistant to protein degradation induced by activation of the cAMP pathway.

To further determine whether forskolin affects E4BP4-dependent Fsp27 expression in hepatocytes, we transduced Hepa-1c1c7 cells with Ad-E4bp4-WT versus Ad-E4bp4-5KR and treated cells with forskolin. Forskolin potently downregulated Fsp27β in Ad-E4bp4-WT–transduced Hepa1 cells. However, the suppression of Fsp27β by forskolin was completely lost in Hepa1 cells transduced with Ad-E4bp4-5KR (Fig. 6C). Meanwhile, other forskolin-sensitive genes, such as G6pse and Il11, were not affected by overexpression of E4BP4 (Supplementary Fig. 8A) (62). Interestingly, forskolin also reduced CREBH protein in Hepa1 cells transduced with either Ad-GFP or Ad-E4bp4-WT. However, the suppression of CREBH by forskolin was completely lost in Hepa1 cells overexpressing E4bp4-5KR (Supplementary Fig. 8B), consistent with the unchanged mRNA levels of Fsp27β in Ad-E4bp4-5KR–transduced Hepa1 cells (Fig. 6C). These observations provide further evidence supporting that deSUMOylation of E4BP4 protects CREBH-dependent Fsp27 expression under the cAMP activation condition.

Finally, we examined LD formation in Huh-7 cells following treatment of forskolin in the presence of overexpression of E4BP4-WT versus E4BP4-5KR. After incubation with a mixture of palmitate and oleate, which was shown to induce LD formation in hepatocytes (63), both E4BP4-WT and E4BP4-5KR–expressing cells accumulated similar numbers of LDs. In E4BP4-WT–expressing Huh-7 cells, forskolin treatment at both 2.5 μmol/L and 5 μmol/L significantly reduced the number of LDs. In contrast, forskolin treatment at 2.5 μmol/L failed to block the increase of LDs in E4BP4-5KR–expressing cells (Fig. 6D). However, forskolin at 5 μmol/L still showed a minimal inhibitory effect on LD formation in cells with E4BP4-5KR overexpression, suggesting that there are E4BP4-independent mechanisms for the suppression of LD formation by forskolin. Taken together, our data uncover a novel link between the status of E4BP4 SUMOylation and cAMP activation–mediated suppression of LD biogenesis.

In this study, we demonstrated the essential role of E4BP4 in driving hepatic FSP27 through CREBH to expand LDs in response to HFD feeding. Our data highlight that deSUMOylation of E4BP4 blocks the inhibitory effects of cAMP pathway activation on Fsp27 expression and LD formation in hepatocytes, potentially leading to the development of NAFLD.

Our study is the first to demonstrate that E4BP4 is one of the master regulators of LD binding proteins during NAFLD. There is mounting evidence that LD binding proteins play important roles in regulating LD biogenesis and catabolism (10,12). In our study, we discovered and confirmed the marked downregulation of several major LD binding proteins in the liver of E4bp4-LKO mice, including Cidea, Cidec (Fsp27 in mice), Plin2, Plin3, and Plin4, and to a lesser degree Plin1 and Plin5. We focused on how hepatic E4BP4 promotes Fsp27 expression because of the following reasons: 1) FSP27 is a liver-enriched CIDE-type LD binding protein (20); 2) upregulation of Fsp27 by CREBH, PPAR-γ, and SREBP-1c has been linked to LD growth and hepatic steatosis (16,20,50); and 3) administration of Fsp27 antisense oligos improves diet-induced steatohepatitis in mice (21,23), suggesting that a tight control of Fsp27 expression could be a key step to treating fatty liver diseases. Although the exact mechanisms remain to be further defined, we demonstrated that knockdown of Crebh abrogates the induction of Fsp27β by E4BP4, supporting that E4BP4 promotes the transcription of Fsp27β via CREBH. Hepatic CREBH protein has been shown to depend on E4BP4 (31), prompting us to speculate that E4BP4 might promote Fsp27 transcription by maintaining the CREBH stability.

In this study, we examined the role of protein SUMOylation in regulating E4BP4 action on FSP27 expression. Protein SUMOylation has emerged as a critical modification during cellular responses to various stresses, including hypoxic and oxidative stress, and more recently ER stress (64). It has been reported that XBP1, the key transcription factor regulating ER stress response, is modified by SUMOylation, and SUMOylated XBP1 is more stable than the unmodified XBP1 (65). Moreover, ATF6 protein can be modified by SUMOylation, and its transcriptional activity is suppressed by SUMOylation (66). These findings suggest that the outcomes of SUMOylation are transcription factor specific. We also found that E4BP4 is modified by SUMOylation in a SUMO2/3-dependent manner and that SUMOylation of E4BP4 is sensitive to nutritional status in hepatocytes. We speculate that SUMOylation status of E4BP4 depends on the activity of both SUMO E3 ligases and deSUMOylases within hepatocytes. So far, PIASy, one of the SUMO E3 ligases, has been reported to regulate hepatic lipid metabolism upon fasting signaling (67). In contrast, SENP3, one of the SUMO-specific proteases, was found to act as a key regulator of hepatic lipid metabolism in NAFLD (68). We intend to identify the specific enzymes that control the reversible E4BP4 SUMOylation in future work.

Our data show that SUMOylation does not impact E4BP4 protein degradation or its ability to induce the CREBH-Fsp27β axis in hepatocytes under basal conditions. However, our data did reveal that the nonSUMOylatable E4BP4-5KR mutant is resistant to activation of the cAMP pathway, a key signaling pathway in regulating carbohydrate and lipid metabolism in the liver during fasting-feeding cycles and nutritional stress (58,59,69). Accumulated evidence suggests that elevated cAMP signaling confers protection against diet-induced NAFLD in mice, whereas inhibition of cAMP signaling is sufficient to trigger NAFLD (70–72). Our work uncovered that activation of the cAMP pathway by forskolin not only lowers the protein levels of E4BP4-WT but also downregulates the CREBH-Fsp27β axis, subsequently reducing LDs within hepatocytes. However, these effects are largely absent in hepatocytes expressing the SUMOylation-defective E4BP4-5KR mutant, supporting a critical role of E4BP4 SUMOylation in mediating the inhibitory effects of cAMP pathway activation on CREBH-FSP27–dependent LD biogenesis. Given the importance of the cAMP pathway in liver physiology and liver disease, it would be of great interest to identify both upstream signaling and downstream effectors of the cAMP pathway in regulating the SUMOylation and stability of E4BP4 in NAFLD.

Funding. This work was supported by National Institutes of Health grant R01 DK121170 (to X.T.). Part of the work was also supported by Michigan Nutrition Obesity Research Center pilot grant P30 DK089503 (to L.Y.) and Michigan Diabetes Research Training Center pilot grant P60 DK020572 (to X.T.). J.S. and A.W. were supported by the Molecular and Integrative Physiology Short Term Educational Program (R25 DK088752).

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

Author Contributions. S.W., M.Y., J.S., K.R., and D.L. performed the RNA extraction from cells/tissues, RT-qPCR, genotyping PCR, and phenotyping of E4bp4-LKO mice after HFD feeding. M.Y., J.S., and L.Y. performed the BODIPY and biochemical experiments. M.Y. and D.Z. performed the primary hepatocyte isolation. M.Y. and X.T. generated all the expression vectors and recombinant adenoviruses for the in vitro and in vivo experiments. P.L. performed affinity purification of the E4BP4 protein for mass spectrometry. J.S. performed the in vitro SUMOylation assay. A.W. created the Fsp27 promoter–driven luciferase reporter construct. K.Z. provided the Ad-shCrebh adenovirus. L.Y. provided input on the manuscript. X.T. supervised the project and wrote the manuscript. All authors reviewed and commented on the manuscript. X.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 2022 Federation of American Societies for Experimental Biology Meeting on Lipid Droplets, Asheville, NC, 26–30 June 2022.