GADD45α drives brown adipose tissue formation through upregulating PPARγ in mice

Stress can lead to obesity and metabolic dysfunction, but the underlying mechanisms are unclear. Here we identify GADD45α, a stress-inducible histone folding protein, as a potential regulator for brown adipose tissue biogenesis. Unbiased transcriptomics data indicate a positive correlation between adipose Gadd45a mRNA level and obesity. At the cellular level, Gadd45a knockdown promoted proliferation and lipolysis of brown adipocytes, while Gadd45a overexpression had the opposite effects. Consistently, using a knockout (Gadd45a−/−) mouse line, we found that GADD45α deficiency inhibited lipid accumulation and promoted expression of thermogenic genes in brown adipocytes, leading to improvements in insulin sensitivity, glucose uptake, energy expenditure. At the molecular level, GADD45α deficiency increased proliferation through upregulating expression of cell cycle related genes. GADD45α promoted brown adipogenesis via interacting with PPARγ and upregulating its transcriptional activity. Our new data suggest that GADD45α may be targeted to promote non-shivering thermogenesis and metabolism while counteracting obesity.


Introduction
Stress can lead to metabolic dysfunction and obesity 1 . Obesity has become a global epidemic and is a major risk factor associated with several metabolic syndromes, such as type 2 diabetes, insulin resistance, heart disease, stroke, hyperglycemia, hypertension, and cancer 2,3 . Adipocytes play critical roles in systemic metabolism and energy homeostasis. In mammals, three types of adipocytes, white, brown, and beige or brite adipocytes, have been identified 4 . Among them, white adipocytes store excess energy in lipid droplets 5 , while beige and brown adipocytes burn lipids to produce heat, thus counteracting obesity 6,7 . Unlike white adipocytes, beige and classical brown adipocytes are characterized by their unique ability to transform mitochondrial energy into heat via uncoupling protein 1 (UCP1) 8,9 . In mammals therefore, non-shivering thermogenesis in BAT helps to maintain body temperature in cold environments while spending energy during times of high caloric intake 9 . Thus, better understanding of adipogenesis and its molecular regulation, especially of the brown and beige fat cells, may give rise to efficient and novel strategies for combating obesity and related metabolic disorders.
The growth arrest and DNA damage 45 (GADD45) protein family, consists of three members including GADD45α, GADD45β and GADD45γ 10,11 . GADD45α is a small (18.4 kDa) p53-regulated histone-fold protein known to be induced by varieties of genotoxic stress agents, such as hypoxia, UV radiation, ionizing radiation, oxidants, and alkylating agents 12,13 . GADD45α plays an important role in DNA repair 14 , cell cycle 15 , apoptosis 16,17 , angiogenesis 18 , senescence 19,20 , and DNA demethylation 21 . Gadd45a −/− mice exhibit increased genome instability, reduced nucleotide excision repair and a higher rate of mutations 22 . Recently, Schäfer et al. found that GADD45α and ING1 (inhibitor of growth family member 1) are required for the differentiation of mouse embryonic fibroblasts 23 . The GADD45α/ING1 double-knockout mice display segmental progeria, lipodystrophy and metabolic defects 23 . During white adipogenesis, GADD45α promotes white adipocyte differentiation through epigenetic regulation 24,25 . These results suggest that GADD45α may play an important role in white adipocytes and energy metabolism. However, the role of GADD45α in brown adipocytes was unclear and the molecular mechanisms underlying the functional regulation of GADD45α in BAT remained to be determined. Furthermore, whether the expression of Gadd45a is correlated to obesity remains unclear.
In this study, we used unbiased transcriptomics data analysis and a GADD45α knockout (Gadd45a −/− ) mouse model to determine the regulatory role of GADD45α in brown adipocytes and energy metabolism. We found that Gadd45a mRNA expression is positively correlated with fat deposition. Deficiency of GADD45α affects brown adipocytes proliferation, lipolysis, and mitochondrial biogenesis, and results in obvious metabolic phenotypes. We further revealed the molecular mechanisms underlying the roles of GADD45α in brown adipocytes. Our results demonstrate that GADD45a is a critical regulator of BAT growth and function, and suggest that may be a potential therapeutic target to combat obesity and other metabolic diseases.

GADD45α expression is positively correlated with lipid metabolism and obesity
To explore the novel genes associated with obesity, we performed a transcriptome analysis with several published datasets on adipose tissues and disease models. Notably, we found that Gadd45a as well as several genes related to lipid metabolism including Lep, were highly expressed in white adipose tissues of obese mice (Fig. 1a, GSE4692) 26,27 , rats (Fig. 1b, GSE8700) 28 and children (Fig. 1c, GSE9624) 29 . Likewise, in human livers, high levels of Gadd45a were found in both the obese nondiabetic model and the obese diabetic model compared to lean groups (Fig. 1d, GSE121344). When comparing the expression of Gadd45a in white and brown adipocytes, we found that Gadd45a was highly expressed in white adipocytes in mice 27 (Fig. 1e). These results indicate that Gadd45a expression is positively correlated with obesity and may represent a potential regulator of lipid metabolism and brown adipogenesis.
GADD45α deficiency promotes brown adipocyte proliferation through upregulating cell cycle related genes To study the potential role of GADD45α in regulating brown adipocyte function, we first examined whether deletion of Gadd45a affects brown adipocyte proliferation in culture. We designed three independent lentiviral shRNA plasmids to knockdown Gadd45a in brown adipocytes. Infection with the shRNA1 lentivirus led to a 70% reduction in the level of Gadd45a, compared to cells treated with control shRNA. Thus, shRNA1 lentivirus (G45a-sh1) was used to establish a stable Gadd45a knockdown cell line, which was used in the following experiments. Notably, we found a higher percentage of Ki67+ cells in the Gadd45a knockdown cells compared to control cells (Fig. 2a, b). Analysis of colony formation (Fig. 2c) further confirmed that GADD45α deficiency increased the proliferation and colonization of brown adipocytes. Moreover, compared to control cells, mRNA levels of Ki67 and cell cycle markers including Cdkn1a, Cdkn1c, Ccnd1, Ccnd3, and Cdk5r1 were significantly upregulated in the Gadd45a knockdown cells (Fig. 2d, e). By contrast, Gadd45a overexpression inhibited brown adipocyte proliferation in vitro ( Supplementary Fig. 1a-d). In addition, the mRNA level of Ki67 was down-regulated in Gadd45a overexpressing (G45a-oe) cells (Fig. 2f).
To further confirm the effects of GADD45α on brown adipocyte proliferation, we applied RNA-seq to map transcriptional changes upon Gadd45a overexpression. We found a total of 3678 differentially expressed genes, out of which 2096 were increased and 1582 were decreased (Fig. 2g). Gene ontology (GO) enrichment analysis revealed pronounced changes in genes involved in cell cycle and growth (Fig. 2h), particularly the expressions of cell cycle related genes were downregulated ( Fig. 2i) in Gadd45a overexpressing cells. These results suggest that deletion of Gadd45a promotes brown adipocyte proliferation through upregulating cell cyclerelated genes.

GADD45α deficiency inhibits brown adipocyte lipogenesis but promotes lipolysis in vitro
To determine the role of GADD45α for the differentiation of brown adipocytes, we isolated SVF cells from BAT and examined adipogenic differentiation. Oil Red O and bodipy staining results revealed increasing lipid accumulation after adding a differentiation medium into the culture with brown preadipocytes for four days (Fig.  3a). The mRNA levels of lipogenic genes, including Ppara, Pparg, Cebpa, Cebpb, Adipoq, Fabp4, and Fasn were significantly upregulated (Fig. 3b-i). Consistently, mRNA and protein levels of GADD45 family proteins were also observed to be increased ( Fig. 3j-m). Thus, we speculate that GADD45α may be generally related to adipogenic differentiation in brown adipocytes.
We further employed a loss-of-function study in cell culture. We performed adipogenic differentiation in G45a-sh1 treated BAT SVF cells. Our results revealed that Gadd45a knockdown robustly inhibited brown adipocyte lipogenesis and TG accumulation (Fig. 4a-c). Similarly, G45a-sh1 adipocytes expressed lower levels of perilipin protein, and lower mRNA levels of Fabp4 and Adipoq than the control group ( Fig. 4d-f). Lipolysis of triglycerides (TGs) ultimately results in the liberation of glycerol and free fatty acids within the fat cells 30 . We observed higher levels of glycerol release from G45a-sh1 adipocytes as compared with the controls ( Supplementary Fig. 2a), suggesting G45a-sh1 indeed increased lipolysis. We also performed a gain-of-function experiment by using an adenovirus-mediated overexpression of Gadd45a (G45aoe) in cultured SVF cells isolated from BAT. As a consequence, G45a-oe reduced glycerol release (Supplementary Fig. 2b) and promoted adipocyte lipogenesis by increasing mRNA expression of adipogenic related genes including Fabp4, Perilipin, Lep, and Adipoq ( Fig. 4g-l). These results indicate that Gadd45a knockdown suppressed the process of lipogenesis but promoted lipolysis in brown adipocytes in vitro, while Gadd45a overexpression had the opposite effect.
We further used the Mito-tracker staining to confirm our results. When examining mitochondrial abundance by using Mito-tracker, we found that Mito-Tracker Red CMXRos was significantly higher in the G45a-sh1 cells (Fig. 5c). Besides, our data from the transmission electron microscope (TEM) indicated morphological and structural changes between G45a-sh1 cells compared to controls (Fig. 5d). Immunofluorescence results showed higher levels of UCP1 expression in Gadd45a knockdown cells (Fig. 5e). Collectively, these results suggest that GADD45α deficiency increased the expression of mitochondria related genes and enhanced mitochondrial biogenesis, suggesting GADD45α deletion may have beneficial effects on insulin sensitivity and whole-body energy metabolism.

GADD45α deficiency inhibits brown adipogenesis and upregulates expression of BAT-selective genes in vivo
To precisely explore the function of GADD45α in BAT, we used the Gadd45a −/− mice (KO) model to verify our results. Genotyping and real-time polymerase chain reaction (PCR) analyses all confirmed efficient deletion of GADD45α in BAT and WAT ( Supplementary Fig. 3a, b), as well as in non-adipose tissues ( Supplementary Fig. 3c). The Gadd45a −/− mice were born at expected Mendelian ratios and were morphologically indistinguishable from their wildtype (WT) littermates ( Supplementary Fig. 3d). On the normal chow diet, the Gadd45a −/− male mice showed similar body weights but higher food intakes compared to WT mice at 8 weeks of age ( Supplementary Fig. 3e, f). The BAT masses were similar, while the masses of iWAT from the KO mice were lower than in the WT mice (Supplementary Fig. 3g-i). All other non-adipose tissue masses were not affected by GADD45α deficiency (Supplementary Fig.  3j, k). Our results revealed that GADD45α deficiency did not seem to affect BAT development.
Interestingly, hematoxylin-eosin (H&E) staining revealed an obvious decrease in adipocyte size in the Gadd45a −/− BAT compared with WT BAT (Fig. 6a). Nuclear densities (number of nuclei per unit area) were also higher in the KO mice than in the WT mice, confirming smaller adipocyte size in the KO mice (Fig. 6b). In addition, genomic DNA content per BAT depot was higher in the KO mice than in the WT mice (Fig. 6c), suggesting that Gadd45a −/ Fig. 2 GADD45α deletion promotes proliferation of brown adipocytes. a, b Brown adipocytes stained with proliferating marker Ki67 (red) and DAPI (blue) in control and G45a-sh1 cells (a), and the percentage of Ki67 + cells (b). c Colony assay comparing growth of control and G45a-sh1 cells. d, e mRNA levels of Ki67 and cell cycle related genes in control and G45a-sh1 cells. n = 4. f mRNA levels of Ki67 in control and G45a-oe cells. n = 4. g Scatterplot of gene expression differences between control and G45a-oe BAT cells as determined by RNA sequencing. Shades of blue correspond to genes downregulated in the G45a-oe cells, and shades of red indicate upregulation in the G45a-oe cells. n = 4. |log2(Fold Change)| > 2, p < 0.001. h Gene Ontology (GO) enrichment analysis was performed with −log10 (p value) plotted (x-axis) as a function of classification meeting a p value of <0.001. i Heatmap of expressions of selected cell cycle related genes from the RNA-seq dataset. Error bars represent SEM, *P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student's t test. Scale bars: 200 µm. − BAT contained more cells per depot than the WT BAT deports. Expansion of fat mass can result from increased intracellular lipids and greater adipocyte size (hypertrophy) as well as increased numbers of adipocytes (hyperplasia) 31 . Our in vivo and in vitro data suggest that GADD45α deficiency may promote proliferation and lipolysis in brown adipocytes.
We further analyzed the expression of adipogenic related genes and found that the mRNA levels of BATselective and mitochondrial marker genes, such as Ucp1, Pgc1a, Prdm16, Uqcr10, Cox5b, Cox7a, and Ppara were significantly higher in the KO BAT compared with WT BAT (Fig. 6d). By contrast, the mRNA levels of panadipocyte genes Pparg, Adipoq and Lep were significantly lower in the KO BAT than WT BAT (Fig. 6d). The expression of the WAT-specific genes Agt, Retn and Trim14 was similar between the two genotypes ( Fig. 6d). In addition, the KO BAT expressed higher levels of UCP1 and PGC1α protein than the WT BAT (Fig. 6e, f). We also determined the mitochondrial proteins and found that the protein levels of complex CII (SDHB) and CV (ATP5A), key enzymes in oxidative phosphorylation and responsible for energy production 32 , were dramatically elevated in the KO BAT ( Supplementary Fig. 4). Overall, our results indicate that GADD45α deficiency affected brown BAT development and upregulated expression of BATselective and mitochondrial marker genes.

GADD45α deficiency enhances insulin sensitivity and energy expenditure
Adipose tissue depots regulate systemic glucose metabolism and insulin sensitivity 33,34 . To determine whether the Gadd45a −/− mice may have beneficial metabolic health effects, we conducted glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs). Compared to the WT littermates, KO mice had lower blood glucose levels after glucose injection (Fig. 7a, b) and a faster rate of insulin-stimulated glucose clearance (Fig. 7c). To gain further insight into the effect of GADD45α deficiency on whole-body metabolism, metabolic cages were used for the simultaneous measurement of food intake, energy expenditure, heat production and physical activity in the mice. We observed no significant change in body weights between WT and KO mice (Fig. 7d). However, food intake The mRNA level of BAT-specific and mitochondrial related genes were measured by qPCR. n = 6. c The mitochondrial staining (Mito Tracker Red CMXRos) of control and G45a-sh1 BAT cells. Scale bar, 50 µm. d The structure of each mitochondrion was detected by transmission electron microscope (TEM) in control and G45a-sh1 cells. Scale bar, 0.5 µm. e UCP1 staining in control and G45a-sh1 treated BAT cells. Scale bar, 50 µm. Error bars represent SEM, *P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student's t test.
was significantly higher in the KO mice compared to WT mice (Fig. 7e). In addition, the Gadd45a −/− mice had an increased general activity, higher rates of O 2 consumption and CO 2 production ( Fig. 7f-h). However, no significant differences in heat production were observed between WT and KO mice (Fig. 7i). Our results demonstrate that GADD45α deficiency improved systemic insulin sensitivity and glucose tolerance, and ameliorated the metabolic profile of mice.
GADD45α promotes the differentiation of brown adipocytes through interacting with PPARγ and enhancing its transcriptional activity To closer investigate the cellular and molecular mechanisms through which GADD45α leads to brown adipocyte differentiation, we measured gene expression of the genes involved in regulating adipogenesis based on knockdown or overexpressing of GADD45α (Fig. 8a, b). In brown adipocytes, Gadd45a knockdown dramatically decreased the mRNA expression of Pparg (Fig. 8a), a key transcription factor that regulates adipogenesis 35,36 , while Gadd45a overexpression had the opposite effects (Fig.  8b). Similar results were obtained using western blotting (Fig. 8c, d). GADD45α may thus be associated with PPARγ to regulate BAT adipogenic differentiation. Moreover, rosiglitazone (a PPARγ agonist) treatment rescued the differentiation of the Gadd45a-deficient brown adipocytes ( Fig. 8e; Supplementary Fig. 5a). By contrast, PPARγ inhibitor GW9662 treatment partly attenuated the lipid droplet formation of Gadd45a-oe brown adipocytes (Fig. 8f; Supplementary Fig. 5b, c). From the gene expression and pharmacological rescue data above, we demonstrate that GADD45α positively regulated PPARγ expression at both mRNA and protein level.
We next performed Co-IP experiments to examine whether GADD45α interacts with PPARγ. After GADD45α overexpression in BAT, we found that PPARγ can be pulled down by FLAG-G45α and vice versa (Fig.  8g). Likewise, endogenous interactions between GADD45α and PPARγ were also found in differentiated BAT cells (Fig. 8h). Moreover, immunofluorescence microscopy verified that GADD45α and PPARγ protein had similar subcellular localization in differentiated brown adipocytes ( Supplementary Fig. 6a), suggesting GADD45α interacts with PPARγ. To examine whether the interaction regulates the transcriptional activities of PPARγ, we performed a chromatin immunoprecipitation and luciferase reporter assay of FABP4, which is a downstream target of PPARγ and directly regulated by PPARγ [37][38][39] . The chromatin immunoprecipitation (ChIP) and luciferase reporter assays indicated that PPARγ directly bound to the promoter of FABP4 to enhance FABP4 gene transcription ( Supplementary Fig. 6b-d).

Discussion
Our study reveals a novel role of GADD45α in regulating the brown adipogenesis and function. We have provided functional physiological, histological and cellular evidence to demonstrate that GADD45α deficiency improves energy metabolism and mitochondrial biogenesis in mice. We found out that GADD45α deficiency promotes BAT adipocyte proliferation and decreases lipid accumulation. As the GADD45 protein family is considered highly conserved in evolution, we anticipate that our results from mice may be well applicable to humans, although future studies dissecting the role of GADD45α signaling in human adipose tissues will be necessary.
GADD45α is a p53-targeted protein whose expression is induced by several stress agents 12,13 . Previous study demonstrated that stress can lead to metabolic dysfunction and obesity 1 . Using unbiased transcriptomics data analysis, we found that Gadd45a mRNA expression was indeed correlated with obesity and may regulate lipid . n = 6. c, d Protein levels of PPARγ in G45a-sh1 (c) and G45a-oe treated brown adipocytes (d) during adipogenic differentiation. e Protein levels of PPARγ and Perilipin treated with or without rosiglitazone (Ros). f Protein levels of PPARγ and Perilipin treated with or without GW9662. g GADD45α interacts with PPARγ. Brown adipocytes were transfected with pCDNA-FLAG-GADD45α and the lysates were immunoprecipitated (IP) with FLAG and PPARγ antibodies and blotted with FLAG and PPARγ antibodies. h Endogenously expressed GADD45α interacts with PPARγ. Cell lysates from differentiated brown adipocytes were IP with GADD45α or PPARγ antibodies, and blotted with these antibodies. i Brown adipocytes were infected with adenovirus expressing Gadd45a and differentiated. Chromatin immunoprecipitation using a PPARγ specific antibody followed by qPCR amplified by primers flanking FABP4 specific (SP) DNA sequence or nonspecific (NS) sequences in the promoter region of FABP4 gene. n = 3. j Luciferase assay of 293T cells after co-transfected with the plasmids shown. n = 3. k Schematic summary illustrating GADD45α regulates brown adipocyte differentiation. Error bars represent SEM, *P < 0.05, **P < 0.01, two-tailed Student's t test.
metabolism and brown adipogenesis. Our data indicated that Gadd45a knockdown dramatically upregulated the expression of Ki67 and other cell cycle markers and promoted brown adipocyte proliferation. Vice versa, Gadd45a overexpression markedly inhibited the process of brown adipocyte proliferation through inhibiting growth and cell cycle genes. Consistent with our results, GADD45α blocks cell proliferation in hepatocellular carcinoma cells through cell cycle arrest in the G2/M 40,41 . Likewise, Gadd45a deletion increased cell proliferation of mouse embryo fibroblasts 22 . In addition, GADD45α acted as an autoimmune suppressor in vivo by inhibiting T cell proliferation in response to TCR activation 42 . In our study presented here, GADD45α may be required for adipogenic differentiation in brown pre-adipocytes. Gadd45a overexpression promoted brown adipogenesis and lipid accumulation, accompanied by an increased expression of adipogenic genes and decreased glycerol release into the medium. Instead, knockdown of GADD45α inhibited brown pre-adipocytes lipogenesis and facilitated the lipolysis of triglycerides in vitro. These observations are suggesting that GADD45α deficiency may regulate the balance between proliferation and differentiation of the precursor cells in BAT.
We found that the Gadd45a −/− mice had lower iWAT mass and identical BAT masses than the WT mice. The number of brown adipocytes was increased but the cell size was decreased in Gadd45a −/− mice. This observation suggests that GADD45α deletion may promote BAT proliferation while suppressing differentiation and accumulation of lipids in brown adipocytes in vivo. The unchanged BAT mass in Gadd45a −/− mice may be due to the combinatory effects of i) the increase in cell number and ii) the decrease in cell size. These results are consistent with our above described phenotypes in vitro. Several obesity genes driving food intake and energy expenditure were previously characterized and revealed a homeostatic system for energy metabolism 43,44 . It is interesting that the Gadd45a −/− mice exhibited improved insulin sensitivity and food intake compared to the WT mice. The beneficial effects on insulin sensitivity and food intake may have been caused due to the upregulated BAT-selective gene expression and the improved mitochondrial biogenesis in the Gadd45a −/− mice. The increase of food intake (hyperphagia) triggered by fasting is a simple but compelling example of food intake regulation 45 . The higher food intake and physical activity may be due to an alteration in the central nervous system (CNS), critical for normal energy balance 45 . The activation of sympathetic nerves can increase lipolysis and increase thermogenesis in brown adipose tissue, as well as other central and peripheral pathways increasing energy expenditure 46 . Besides, beige adipocyte homeostasis can be bi-directionally converted from and to white adipocytes under the control of environmental cues or innervation 47 .
Previous studies have revealed that sympathetic activation induces heat production by stimulating the lipolysis of cytosolic lipid droplets (LDs) through the β3-adrenergic signaling in BAT. The released fatty acids from glycerol serve as fuel for thermogenesis during cold exposure 48 . Thermogenic respiration is initiated by lipolysis through the cyclic AMP-protein kinase A signaling pathway, and activation of thermogenesis in BAT increases energy expenditure 49,50 . The lipolysis and mobilization of lipid droplets may explain the observed BAT phenotypes in the Gadd45a −/− mice. It is well known that the thermogenic capacity of BAT depends on UCP1, as well as on the tissue's high mitochondrial density and oxidative capacity 36,51 . When activated, UCP1 catalyzes the mitochondrial proton gradient, thereby using oxidative respiration to generate heat instead of ATP 35,52 . In addition, both mitochondrial biogenesis and respiration are highly dependent on PGC1α 53 , and ablation of PGC1α leads to reduced mitochondrial content and impaired capacity for cold-induced adaptive, nonshivering thermogenesis 54 . Mice lacking GADD45γ have an impaired UCP1 function and thermogenic response in the cold 55 . GADD45γ overexpression in BAT adipocytes instead enhanced ERRγ-dependent transcription and thermogenesis as well 55 . Consistently, we found that deletion of GADD45α increased both mRNA and protein levels of PGC1α and UCP1 in vivo and in vitro, suggesting GADD45α deficiency promoted mitochondria biogenesis through upregulating the expression of PGC1α and UCP1. However, the exact mechanism needs being explored in full detail in the near future.
PPARγ is a master regulator of adipocytes differentiation, playing a critical role in systemic lipid and glucose metabolism. We found that GADD45α activates PPARγ expression during brown adipogenesis. GADD45α is a vital mediator in gene-specific activated DNA demethylation during adult stem cell differentiation and white adipogenesis 24,56 . Newly emerging evidence indicates that DNA demethylation plays an important role in regulating PPARγ expression and adipogenesis in intramuscular preadipocytes and 3T3-L1 cells 57,58 . GADD45α could recruit demethylation proteins to CpG island promoters 59 , and the CpG demethylation of the PPARγ promoter may contribute to its expression 58 . Here, we demonstrated that GADD45α interacts with PPARγ and upregulates its transcriptional activity. The activated form of PPARγ is well-accepted to be upstream of FABP4, which is also known as adipocyte protein 2 (aP2) and is involved in the intracellular fatty acid transport and glucose and lipid homeostasis in mature adipocytes [37][38][39] . The promoter of FABP4 has been widely used in adipocytespecific recombination in mice 60,61 . In both brown and white adipose tissue, FABP4 marks a distinct population of adipocyte progenitors 62 . Our findings show that PPARγ directly binds to the FABP4 promoter to enhance its transcription. Importantly, we discovered that GADD45α interacts with PPARγ and significantly upregulated the transcriptional regulation of PPARγ on FABP4 expression, thus demonstrating the functional significance of the interaction between GADD45α and PPARγ. This observation is consistent with our result that GADD45α promotes FABP4 expression at both protein and mRNA level in differentiated brown adipocytes.
In summary, our results reveal important regulatory roles of GADD45α in brown adipocytes. We highlight the function of GADD45α in BAT adipogenesis and demonstrate that GADD45α interacts with PPARγ by enhancing its transcriptional activities in brown adipocytes. Our results provide novel insights into the mechanistic role of GADD45α in counteracting obesity and other metabolic diseases.

Animals
All procedures involving mice were approved by the Zhejiang University Animal Care and Use Committee. Gadd45a −/− mice 22 were directly contributed from Professor Albert J. Fornace Jr. (Gene Response Section, DBS, National Cancer Institute, USA) and were maintained on a C57BL/6 background. All mice used in this study, the Gadd45a −/− mice and their WT littermate controls, were produced from intercrossing of Gadd45a +/− mice obtained from Hangzhou Normal University. Male mice were single housed in the animal facility with free access to water and standard rodent chow food. The age of the mice was between 8 and 10 weeks in the experiments. PCR genotyping was carried out as described by the supplier. Food intakes were measured by weighing total individual food consumption once per week.

Blood glucose measurements
For GTT, mice were given an i.p. injection of 100 mg ml −1 D-glucose (2 g kg −1 body weight) after overnight fasting 3 , and tail blood glucose concentrations were measured by a glucometer (Accu-Check Active, Roche). For ITT, mice were fasted for 4 h before the i.p. administration of human insulin (Santa Cruz) (0.75 U per kg body weight) 3 , and tail blood glucose concentrations were monitored. For both GTT and ITT, each mouse was singly caged with blinded cage number and random orders.

Indirect calorimetry study
Oxygen consumption (VO 2 ), carbon dioxide production (VCO 2 ), respiratory exchange ratios and heat production were measured using an indirect calorimetry system (Oxymax, Columbus Instruments), installed under a constant environmental temperature (22 or 30°C) and a 12-h light (06:00-18:00 h), 12-h dark cycle (18:00-06:00 h). All mice had free unlimited access to food and drinking water. The raw data were normalized to the lean mass of the mice.

Cell culture and adipogenic differentiation
Primary BAT stromal vascular fraction (SVF) cells were isolated using collagenase digestion followed by density separation. Briefly, the interscapular brown adipose (BAT) was minced and digested in 1.5 mg/ml collagenase at 37°C for 0.5 and 1 h, respectively. The digestions were terminated with Dulbecco's modification of Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS) (Gibco, CA, USA), and filtered through 100 µm filters to remove connective tissues and undigested trunks of tissues. Cells were then centrifuged at 450 g for 5 min to separate the SVF cells. The freshly isolated SVF cells were seeded and cultured in growth medium containing DMEM, 20% FBS, 1% penicillin/streptomycin (Invitrogen) at 37°C with 5% CO 2 for 3 days, followed by feeding with fresh medium every 2 days. The BAT cell lines, were cultured under the same conditions as SVF cells. For BAT SVF cells adipogenic differentiation, the cells were induced with induction medium (IM) contains DMEM, 10% FBS, 2.85 mM insulin, 0.3 mM dexamethasone (DEXA) and 0.63 mM 3-isobutylmethylxanthine (IBMX) for 4 days on confluence and then differentiated in differentiation medium (DM) contains DMEM, 10% FBS, 200 nM insulin and 10 nM T3 for 2 days until adipocytes mature. To avoid a cell density effect on adipogenic differentiation, cells were induced to differentiate when they reach 90% confluence.

Oil red O staining
Cultured cells were washed with PBS and fixed with 4% formaldehyde for 15 min at room temperature. Then the cells were stained using the Oil red O working solutions containing 6 ml Oil red O stock solution (5 g l −1 in isopropanol) and 4 ml ddH 2 O for 30 min. After staining, the cells were washed with 60% isopropanol in PBS and pictured. Oil red O dye were extracted from stained adipocytes with 100% isopropanol, and the Oil red signal were quantified by measuring the optical density at 490 nm (OD 490).

Glycerol release measurements
Glycerol release was assessed using the free glycerol reagens (Sigma-Aldrich, USA). For in vitro lipolysis, glycerol release from differentiated adipocytes was measured as previously described 63 . The results are expressed in µg glycerol released per mg protein.

H&E and immunostaining
Adipose tissues were fixed in 4% formalin for 24 h at room temperature. Then the tissues were embedded into paraffin and cut at 4-µm thick slices. For H&E staining, the sections were deparaffinized, rehydrated and the nuclei stained with haematoxylin for 15 min. Sections were then rinsed in running tap water and stained with eosin for 1 min, dehydrated and mounted. Whole-slide digital images were collected at ×20 magnification with an Aperio Scan Scope slide scanner (Aperio, Vista, CA). Scanned images of H&E staining were analyzed by Photoshop CS3 to calculate numbers of nuclei. For immunostaining, the sections were blocked with blocking buffer containing 5% goat serum, 2% BSA, 0.2% triton X-100 and 0.1% sodium azide in PBS for 1 h after deparaffinized and antigen retrieval. Then the samples were incubated with Ki67 (Abcam, ab16667, 1:500), PPARγ (C26H12, 1:500) and GADD45α (sc-6850, 1:200) primary antibodies diluted in blocking buffer overnight. After washing with PBS, the samples were incubated with secondary antibodies and DAPI for 45 min at room temperature. Fluorescent images were captured as single-channel grayscale images using a Leica DM 6000B fluorescent microscope with a ×20 objective (NA 0.70).

Mito-tracker and bodipy staining
Control and Gadd45a knockdown (G45a-sh1) or Gadd45a overexpressing (G45a-oe) cells were incubated for 15 minutes with 20 nM MitoTracker ® Red CMXRos (Invitrogen). Then cells were washed with PBS for 3 times, then were added fresh DMEM medium and pictures were taken. Intracellular lipids were visualized by staining with 0.5 nM BODIPY FL (Invitrogen) for 10 min. Cells were fixed afterwards with 4% paraformaldehyde and were observed by fluorescence microscopy.

Transmission electron microscopy (TEM)
TEM assay was performed as described 64 . Electron photomicrographs were taken from cell ultrastructures under a transmission electron microscopy (Hitachi, H-7650).

Cell growth rate
Cell growth rates were determined as described previously 65 . BAT cells were seeded in six-well plates (1 × 10 4 cells per well) and cultured under standard adipocyte conditions with or without drug treatment. The cells were harvested and counted using a hemocytometer.
Total RNA extraction and real-time PCR Total RNA was extracted from cells or tissues using Trizol Reagent (Invitrogen, CA, USA) and following the manufacturer's instructions. The purity and concentration of total RNA were measured by a spectrophotometer (Nanodrop 2000; Thermo Fisher Scientific) at 260 and 280 nm. Absorption rates (260/280 nm) of all samples were between 1.8 and 2.0. Then the first-strand cDNA was synthesized using random primers with a reverse transcription kit (Invitrogen, USA). Real-time PCR was carried out with a Roche Lightcycler 480 PCR System using SYBR Green Master Mix and gene-specific primers (Table S1). The 2 −ΔΔCT method was used to analyze the relative changes in gene expression normalized against 18 S rRNA as internal control.

RNA-seq analysis
RNA extraction and RNA-seq analysis were performed by Novogene Bioinformatics Institute (Beijing, China). Sequencing libraries were generated from 1 µg total RNA using NEBNext ® UltraTM RNA Library Prep Kit for Illumina ® (NEB, USA), following manufacturer's recommendations. The libraries were then quantified and pooled. Paired-end sequencing of the library was performed on the HiSeq XTen sequencers (Illumina, San Diego, CA). FastQC (version 0.11.2) was used for evaluating the quality of sequenced data. Gene expression values of the transcripts were computed by StringTie (version 1.3.3b). The TPM eliminated the influence of gene lengths and sequencing discrepancies to enable direct comparison of gene expression between samples. Differential expression analysis of two groups was performed using the DESeq2 R package (1.16.1). Genes were considered as significantly differentially expressed if p value < 0.001 and |foldchange | > 1.5. GO enrichment analysis of differentially expressed genes was implemented by the clusterProfiler R package.

Co-IP assay
Total protein was extracted from differentiated brown adipocytes. The lysate was precleared with protein A/G agarose at 4°C for 1 h. Then 2 mg of primary antibody anti-GADD45α (sc-6850, Santa), anti-PPARγ (C26H12, CST) anti-FLAG (M20008S, Abmart) was added into lysate contains 500 mg total protein and rotating at 4°C overnight. In the next morning the protein A/G agarose was added and rotated for 2 h. The samples were washed with cold PBS for three times and collected for western blotting.

ChIP assay
Brown preadipocytes were seeded on to 10 cm plates and grown to confluence. Cells were harvested 6 days after adipogenic differentiation. Protein-DNA complexes were cross-linked using 1% formaldehyde for 10 min and the cross-linking was then quenched with the addition of 125 mM glycine for 5 min. Samples were washed twice with cold PBS and placed in SDS lysis buffer containing 20 mM Tris, 0.1% SDS, 1% Triton-100, 150 mM NaCl, 1 Mm EDTA and protease inhibitor. The samples were further sonicated and diluted for IP with the indicated antibodies PPARγ (C26H12, 1:100) or rabbit IgG (sc-2027, Santa Cruz) and incubation at 4°C overnight. Then, the immunoprecipitates were eluted and reverse crosslinked overnight at 65°C. DNA was purified using the Cycle Pure kit (Omega Bio-Tek), and qPCR was performed.

Luciferase assay
HEK293T cells were seeded into 24-well plates for 24 h and then transfected with different plasmids using Lipofectamine 2000 (Invitrogen, USA). The pGL3-FABP4 promoter luciferase plasmid was generated. For transfection of each well, 80 ng Renilla plasmid (pRL-TK), 250 ng pGL3-FABP4 and 300 ng pcDNA-GADD45A plasmid (or its blank control plasmid) and/or 300 ng pcDNA-FLAG-PPARγ (or its blank control plasmid) were co-transfected following the manufacturer's instructions. Cells were harvested 36 h after transfection and analyzed with the Dual-Luciferase Reporter Assay System (Promega).

Statistical analysis
Data were presented as means ± SEM from at least three independent experiments. GraphPad (Prism 6) was used for data analyses. Comparisons were made by two-tailed Student's t tests. Differences were considered significant at P < 0.05.