The Transcription Factor ATF7 Controls Adipocyte Differentiation and Thermogenic Gene Programming.

Adipocytes function as major players in the regulation of metabolic homeostasis, and factors contributing to adipocyte differentiation and function are promising targets for combatting obesity and associated metabolic disorders. Activating transcription factor 7 (ATF7), a stress-responsive chromatin regulator, is involved in energy metabolism, but the underlying mechanisms remain unknown. Herein, we showed that ATF7 is required for adipocyte differentiation and interacts with histone dimethyltransferase G9a in adipocytes to repress the expression of interferon-stimulated genes, which in turn suppress adipogenesis. Ablation of ATF7 promotes beige fat biogenesis in inguinal white adipose tissue. ATF7 binds to transcriptional regulatory regions of the gene encoding uncoupling protein 1, silencing it by controlling histone H3K9 dimethylation. Our findings demonstrate that ATF7 is a multifunctional adipocyte protein involved in the epigenetic control of development and function in adipose tissues.

Introduction

Obesity is a major contributor to numerous metabolic diseases including type 2 diabetes, hypertension, and atherosclerosis. Adipose tissue plays an essential role in regulating the whole-body energy balance. White adipose tissue (WAT) stores excess energy as lipids, whereas brown adipose tissue (BAT) expends energy as heat. The thermogenic properties of BAT are dependent on high oxidative capacity and mitochondrial density and high levels of uncoupling protein 1 (UCP1). Classical brown adipocytes are mainly located around interscapular BAT, whereas beige adipocytes, inducible brown-fat-like cells, sporadically reside in subcutaneous WAT depots (Bartelt and Heeren, 2014). The emergence of beige adipocytes in WAT is induced by environmental stimuli, including cold exposure, exercise, long-term peroxisome proliferator-activated receptor γ (PPARγ) activation, and β-adrenergic receptor stimulation (Harms and Seale, 2013). The browning of WAT can prevent diet-induced obesity and improve metabolism (Kajimura and Saito, 2014). Notably, recent studies also suggest that these beige adipocytes can significantly contribute to the energy balance in humans (Sidossis and Kajimura, 2015).

Obesity-associated macrophage infiltration in adipose tissues is coupled with the recruitment of proinflammatory M1 macrophages, leading to increased expression of proinflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor alpha (TNF-α) (McLaughlin et al., 2017). These proinflammatory cytokines induce an inflammatory phenotype in adipocytes and prevent adipogenesis of 3T3-L1 cells (Gustafson and Smith, 2006). Elevated circulating lipopolysaccharide (LPS) levels in obesity also trigger inflammation in macrophages and preadipocytes of adipose tissues, and LPS treatment inhibits differentiation of adipocyte precursors in vitro (Zhao and Chen, 2015). In addition to LPS, interferon-α (IFN-α), a key stimulator of the innate immune response, inhibits adipogenesis of 3T3-L1 cells (Lee et al., 2016). Thus, accumulating evidence indicates that activation of innate immune responses negatively regulates adipocyte differentiation. However, the molecular players that repress innate immune responses in adipocytes remain elusive.

Activating transcription factor 7 (ATF7) belongs to the vertebrate ATF2 subfamily of transcription factors, which has three members: ATF2 (originally named CRE-BP1) (Maekawa et al., 1989, Hai et al., 1989), CRE-BPa (Nomura et al., 1993), and ATF7 (originally named ATFa) (Gaire et al., 1990). ATF2 proteins reportedly contribute to the regulation of adipocyte differentiation and function. Atf2Cre-bpa double-heterozygous mice exhibit reduced WAT mass, and bone morphogenetic protein 2 can induce the p38-dependent phosphorylation of ATF2, which binds to the promoter region of Pparγ2 to stimulate adipocyte differentiation (Maekawa et al., 2010a). The p38-dependent activation of ATF2 is also required for the induction of thermogenic genes, including Ucp1 (Cao et al., 2004), Pgc1α (Bordicchia et al., 2012, Yao et al., 2017), and Zfp516 (Dempersmier et al., 2015), in BAT in response to various stimuli. Although ATF7 shares a relatively high amino acid sequence identity with ATF2, it represses rather than activating, gene expression in the absence of stress (Seong et al., 2012). ATF7 recruits histone methyltransferases to silence the transcription of target genes. Histone H3K9 trimethyltransferase ESET/SETDB1 is recruited by ATF7 to promote the formation of heterochromatin-like structure on the regulatory region of the Htr5b gene, which encodes serotonin receptor 5b, in the dorsal raphe nuclei of the brain (Maekawa et al., 2010b). Social isolation stress induces the phosphorylation of ATF7 via p38 and leads to the release of ATF7 and ESET/SETDB1 from target genes, resulting in transcriptional activation. ATF7 represses a group of innate immunity-related genes in macrophages by associating with H3K9 dimethyltransferase G9a. Pathogen-infection-induced phosphorylation of ATF7 stimulates the release of ATF7-G9a from target genes, accompanied by a decrease in repressive histone H3K9me2 levels, leading to elevated gene expression (Yoshida et al., 2015). In mouse embryonic fibroblast cells, ATF7 regulates H3K9me3 levels on pericentromeric heterochromatin and telomeres by interacting with Suv39h1 (Maekawa et al., 2018).

Our previous study found that deletion of Atf7 reduced adipose tissue mass and increased energy expenditure in mice fed a high-fat diet (Liu et al., 2016), implying that ATF7 may participate in the regulation of energy balance. In the present work, we have more precisely analyzed the role of ATF7 in adipocyte differentiation and showed that ATF7 facilitates adipogenesis by repressing innate immune responses, whereas it suppresses beige adipocyte biogenesis via dimethylation of H3K9 on thermogenic gene enhancers. Thus ATF7 has a dual role for the regulation of white adipocyte differentiation and beige fat biogenesis in inguinal white adipose tissue (iWAT).

Results

ATF7 Deficiency Impairs Adipocyte Differentiation

Atf7-deficient (Atf7) mice exhibited reduced adipose tissue mass, suggesting that ATF7 may contribute to adipocyte differentiation (Liu et al., 2016). To investigate the function of ATF7 in the regulation of adipocytes, we examined whether ATF7 was expressed in fat tissues. ATF7 was detected in BAT, iWAT, and epididymal WAT (Figure 1A). The stroma-vascular fraction (SVF) of adipose tissues provides a rich reservoir of adipocyte precursors. We measured the expression levels of Atf7, Wnt10b, and Prdm16 genes in isolated mature adipocytes and the SVF from BAT and iWAT. Wnt10b expression decreases during adipocyte differentiation (Cawthorn et al., 2012), and Prdm16 is mainly expressed in mature adipocytes (Seale et al., 2007). In line with previous reports, we also observed that Wnt10b was mainly expressed in the SVF, whereas Prdm16 exhibited higher expression levels in adipocytes than in the SVF (Figures S1A and S1B). However, there was no difference in Atf7 gene expression in adipocytes from BAT and iWAT. ATF7 was also expressed at comparable levels in the SVF and mature adipocytes (Figure 1B). To explore the expression of Atf7 during adipocyte differentiation, we used inguinal primary preadipocytes and measured gene expression by quantitative PCR (qPCR). The results indicated that Atf7 expression was reduced by 40% (day 2) after cells were cultured in induction medium, and expression then increased to day 0 levels, whereas expression of the adipogenesis marker Fabp4 gradually increased during adipocyte differentiation (Figure 1C).

Figure 1

Loss of ATF7 Impairs Adipogenesis

(A) ATF7 protein expression levels in BAT, iWAT, and epididymal WAT measured by immunoblotting.

(B) Atf7 mRNA expression levels in mature adipocytes and stromal vascular cells isolated from BAT and iWAT (n = 3, three biological replicates). Data are presented as mean ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests.

(C) Atf7 (left) and Fabp4 (right) gene expression during the differentiation time course of primary iWAT preadipocytes (n = 3, three biological replicates).

(D) Oil red O staining of wild-type (WT) and Atf7−/− primary iWAT preadipocytes at 7 days after induction of differentiation in the absence of rosiglitazone (Rosi).

(E) Expression of the common adipogenic genes Pparγ, C/ebpα, and C/ebpβ, detected by real-time PCR during the differentiation time course of WT and Atf7 primary iWAT preadipocytes (n = 3, three biological replicates).

(F) Oil red O staining of WT and Atf7 primary iWAT preadipocytes at 7 days after induction of differentiation in the presence of Rosi.

(G) Fabp4 gene expression levels in differentiated WT and Atf7 primary iWAT preadipocytes in the presence or absence of Rosi.

Data are presented as means ± SD. Statistical analysis was performed using two-way ANOVA followed by Holm-Sidak multiple comparison tests. *p < 0.05, **p < 0.01, ***p < 0.001 versus WT (without Rosi treatment). See also Figures S1 and S2.

To determine the function of ATF7 in adipocyte differentiation, we used primary inguinal preadipocytes isolated from wild-type (WT) and Atf7 mice to examine whether loss of ATF7 affects adipogenesis. ATF7 ablation reduced the number of oil red O-stained mature adipocytes (Figure 1D), and the expression level of adipogenic markers Pparγ, C/ebpα, and Fabp4 (Figures 1E and 1G). By contrast, C/ebpβ expression levels were not affected by ATF7 deficiency (Figure 1E). When cells were induced to undergo adipocyte differentiation in the presence of the PPARγ agonist rosiglitazone (Rosi), which promotes adipogenesis and activates brown-selective genes (Schoonjans et al., 1996), Atf7 expression exhibited similar changes to those observed in the absence of Rosi, indicating that this agent did not affect Atf7 expression (Figures S2A and S2B). Notably, the presence of Rosi rescued ATF7-deficiency-induced impairment of adipogenesis. Oil red O staining revealed no difference in lipid accumulation between WT and Atf7 cells cultured with Rosi (Figure 1F). Rosi dramatically up-regulated Fabp4 expression in both WT and ATF7 knockout (KO) cells, and importantly, Fabp4 expression in ATF7 KO cells was comparable to that in WT cells in the presence of Rosi (Figure 1G). Compared with WT cells, the common adipogenic markers Pparγ, C/ebpα, and C/ebpβ also exhibited similar expression levels in Atf7 cells when treated with Rosi (Figure S2C). These findings indicate that ATF7 negatively regulates adipogenesis, directly or indirectly, by repressing Pparγ expression.

ATF7 Represses Innate Immune-Related Genes in Adipocytes

To investigate the mechanism by which ATF7 suppresses adipogenesis, we performed RNA sequencing (RNA-seq) on adipocytes (D7) derived from inguinal preadipocytes in the presence and absence of Rosi. In the absence of Rosi, 673 genes were up-regulated and 237 genes were down-regulated by the loss of ATF7, and Gene Ontology analysis revealed that up-regulated genes were related to immune responses (Figure 2A). The most enriched pathways associated with these up-regulated genes included innate immune responses, defense responses to viruses, and cellular responses to IFN-β, whereas the down-regulated genes were mainly related to lipid metabolism and storage (Figure S3A). We also identified 200 genes that were up-regulated and 174 genes that were down-regulated in Atf7 cells compared with WT cells in the presence of Rosi, and down-regulated genes were related to positive regulation of angiogenesis, negative regulation of insulin secretion, and cellular response to IL-1 and TNF-α (Figure S3B). Interestingly, pathway analysis for up-regulated genes gave similar results to those obtained from adipocytes cultured without Rosi (Figure 2B). These up-regulated genes were also mainly associated with immune response pathways, implying that ATF7 may inhibit immune-related gene expression in adipocytes.

Figure 2

ATF7 Represses Innate Immune Gene Expression in Adipocytes

(A and B) Gene Ontology (GO) pathways (Biological Process category) enriched in up-regulated genes in differentiated Atf7 primary iWAT preadipocytes compared with WT controls in the absence (A) or presence (B) of rosiglitazone (Rosi).

(C) The interferon-stimulated response element (ISRE)-binding motif is the most enriched DNA motif within the promoters of up-regulated genes induced by loss of ATF7 in differentiated adipocytes.

(D) Venn diagram showing the number of overlapping up-regulated genes in Atf7 cells in the presence and absence of Rosi.

(E) Heatmap depicting the expression levels of overlapping up-regulated genes in WT and Atf7 cells (n = 2, two biological replicates).

(F) Gene expression levels of IFN-stimulated genes (ISGs) in WT and Atf7 primary preadipocytes and differentiated adipocytes at day 7 (D7, n = 3, three biological replicates).

Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. *p < 0.05, **p < 0.01. See also Figures S3–S5.

To clarify the factors potentially modulating these up-regulated genes in Atf7 adipocytes, we performed motif analysis of promoter regions and found that interferon-stimulated response element (ISRE) was the most enriched motif in genes up-regulated in Atf7 cells in both the presence and absence of Rosi. Without Rosi, 6.53% of the 673 gene promotors contain an ISRE, and in the presence of Rosi, this ratio increases to 22.22% (Figure 2C). Furthermore, 57 genes overlap between the two up-regulated gene sets, including many interferon-stimulated genes (ISGs) such as Irf7, Ifit1, Ifit3, Rsad2, and Oasl2 (Figures 2D and 2E). In accordance with the RNA-seq data, the results of qPCR demonstrated that ISGs were more highly expressed in Atf7 adipocytes (D7), and they were also increased in Atf7 preadipocytes compared with WT controls (Figure 2F). To further validate the role of ATF7 in the regulation of ISG expression, we overexpressed ATF7 in the C3H10T1/2 murine mesenchymal stem cell line (Figure S4A). Expression of ISGs was stimulated by treatment with LPS (0.5 μg/mL) for 12 h. However, overexpression of ATF7 significantly repressed ISG expression in the absence and presence of LPS (Figure S4B). Taken together, these results suggest that ATF7 suppresses the induction of ISGs, and loss of ATF7 stimulates ISG expression during adipocyte differentiation.

In contrast to in vitro data, the results of qPCR and immunofluorescence staining demonstrated that loss of ATF7 did not affect the inflammation level of iWAT (Figures S5A and S5B). Interestingly, we found that loss of ATF7 led to a dramatic reduction in the Retn gene expression in adipocytes (Figure S5C). This finding was consistent with the lower circulating resistin observed in the Atf7 mice (Liu et al., 2016). Given the pro-inflammatory properties of resistin in adipose tissue (Qatanani et al., 2009), the reduction of resistin may counter the adipose tissue inflammation induced by the activation of innate immune response in Atf7 adipocytes. These results suggest that ISGs are up-regulated in Atf7 adipocytes, but it does not cause the inflammation at the tissue level because of negative feedback regulation by resistin.

Activation of ISGs Impairs Adipogenesis

To examine the effect of the induction of ISGs on adipogenesis, we treated preadipocytes with LPS (0.5 μg/mL) for 24 h and found that LPS treatment dramatically stimulated the expression of ISGs (Figure S6A). By contrast, Pparγ expression levels in LPS-treated cells were reduced by approximately half compared with those in phosphate-buffered saline (PBS)-treated control cells (Figure S6B), implying that LPS treatment may impair adipogenesis by repressing Pparγ expression. A previous study demonstrated that LPS treatment for 24 h before the induction of adipocyte differentiation results in the profound impairment of adipogenesis (Zhao and Chen, 2015). In line with these previous observations, we also found that adipocyte differentiation was severely blunted by LPS exposure for 24 h before addition of induction medium. In the absence of Rosi, LPS treatment led to decreased lipid accumulation and lower expression levels of adipogenic markers (Figures S6C and S6D). Interestingly, the presence of Rosi during differentiation clearly rescued the impairment of adipogenesis induced by LPS treatment. The results of oil red O staining demonstrated that Rosi mitigates the effect of LPS treatment on lipid accumulation (Figure S6C). Moreover, there was no difference in expression levels of Pparγ between LPS- and PBS-treated cells in the presence of Rosi, although Fabp4 and C/ebpα expression remained slightly reduced following LPS treatment (Figure S6D).

IFN-α can stimulate the expression of ISGs and inhibit adipogenesis in 3T3-L1 cells (Lee et al., 2016). However, the inhibitory effect of IFN-α on adipogenesis in brown adipocyte precursors was not observed (Kissig et al., 2017). To examine the effect of IFN-α on adipogenesis in iWAT preadipocytes, we measured the expression of ISGs in preadipocytes following exposure to IFN-α (500 U/mL) for 24 h and found that expression levels of ISGs Stat1 and Stat2 were significantly elevated by IFN-α (Figure 3A). By contrast, Pparγ expression levels in IFN-α-treated cells were reduced by 80% compared with those in control cells (Figure 3B). Exposure to IFN-α during adipocyte differentiation severely inhibited lipid accumulation and expression of adipogenic markers (Figure 3C and 3D). Taken together, these results indicate that induction of ISGs by LPS and IFN-α is accompanied by a reduction in Pparγ expression, leading to inhibition of adipogenesis.

Figure 3

IFN-α Inhibits Adipocyte Differentiation of Primary iWAT Preadipocytes

(A) Induction of ISG expression in preadipocytes by IFN-α treatment (n = 3, three biological replicates).

(B) Relative mRNA levels are reduced by IFN-α treatment in preadipocytes (n = 3, three biological replicates).

(C) Oil red O staining of differentiated preadipocytes in the presence or absence of IFN-α.

(D) Expression levels of adipogenic genes Fabp4, Pparγ, and C/ebpα in differentiated pre-adipocytes with or without IFN-α treatment (n = 3, three biological replicates).

Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. *p < 0.05, **p < 0.01, ***p < 0.001. See also Figures S6.

ATF7 Recruits G9a to Repress Stat1 Expression in Preadipocytes

Given that ATF7 associates with histone dimethyltransferase G9a to suppress the expression of innate immune-related genes in macrophages (Yoshida et al., 2015), we speculated that inhibition of ISG expression in adipocytes by ATF7 may occur in the same manner. To test this hypothesis, we first compared genes up-regulated by loss of ATF7 in macrophages and adipocytes. Among the top 250 genes up-regulated in Atf7 macrophages, 58 also exhibited higher expression levels in Atf7 adipocytes than in WT controls (Figure S7A). Pathway analysis of these 58 overlapping genes revealed their association with innate immune responses (Figure S7B). To examine whether ATF7 inhibits the expression of ISGs via recruitment of G9a, we performed a co-immunoprecipitation (coIP) experiment in preadipocytes using anti-ATF7 antibody. The results demonstrated that ATF7 does indeed form a complex with endogenous G9a in preadipocytes (Figure 4A).

Figure 4

ATF7 Recruits G9a to Repress Stat1 Expression in Preadipocytes

(A) Cell extracts of preadipocytes were immunoprecipitated with antibody (2F10)-recognizing ATF7 and immunoblotted with antibody against G9a.

(B) Chromatin immunoprecipitation-qPCR analysis showing the binding of ATF7 to the Stat1 gene promoter (n = 3, three biological replicates).

(C) H3K9me2 enrichment on the Stat1 promoter in WT and Atf7 preadipocytes (n = 3, three biological replicates).

(D) Relative Stat1 mRNA levels in WT and Atf7 preadipocytes (n = 3, three biological replicates).

Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. **p < 0.01. See also Figures S7.

STAT1 functions as a master regulator of the induction of ISGs by binding to ISREs. In macrophages, ATF7 directly binds the promoter region of Stat1 to repress its expression (Yoshida et al., 2015). Thus we speculated that Stat1 may also be a target gene of ATF7 in preadipocytes. Chromatin immunoprecipitation-qPCR results confirmed that ATF7 binds directly to the promoter of the Stat1 gene in preadipocytes (Figure 4B). Importantly, the loss of ATF7 led to lower H3K9me2 levels at the promoter region of the Stat1 gene, and increased expression of Stat1 in Atf7 preadipocytes (Figures 4C and 4D). These findings imply that ATF7 may recruit G9a to the promoter region of Stat1 and repress its expression through dimethylation of H3K9 in preadipocytes. Elevated STAT1 induced by loss of ATF7 in turn stimulates the expression of ISGs and inhibits adipogenesis.

Loss of ATF7 Promotes the Thermogenic Programming and Browning of iWAT

The PPARγ agonist Rosi induces thermogenic gene programming of beige and brown adipocytes (Ohno et al., 2012). RNA-seq data showed that the majority of BAT-selective genes, including Ucp1, Cidea, Pparα, and Otop1, were evidently stimulated by Rosi treatment in both WT and Atf7 cells (Figure 5A). Interestingly, Ucp1 exhibited higher expression in Atf7 cells than in WT cells in the presence of Rosi (Figure 5B). Other thermogenic genes, such as Cidea, Cox8b, and Pgc1α, were also increased by the loss of ATF7, whereas Prdm16 expression was not significantly different between WT and Atf7 cells (Figure 5C). The acute stimulation of thermogenic genes in response to the activation of β-adrenergic receptors is one of the main functional characteristics of beige/brown adipocytes. To examine whether ATF7 is essential for this, differentiated WT and Atf7 cells were treated with isoproterenol (Iso), a synthetic pan β-adrenergic receptor agonist. Ucp1 expression levels were significantly increased in both WT and Atf7 cells after Iso exposure for 4 h (Figure 5D). Of note, the increase in Ucp1 expression stimulated by Iso was comparable in WT and Atf7 cells (∼4-fold), implying that ATF7 may be dispensable for activation of beige cells via β-adrenergic receptors. These findings indicate that loss of ATF7 promotes thermogenic programming in a cell-autonomous manner.

Figure 5

Loss of ATF7 Promotes Thermogenic Gene Programming

(A) Heatmap showing the expression levels of BAT-selective genes in WT and Atf7 cells in the presence or absence of rosiglitazone (Rosi; n = 2, two biological replicates).

(B) Ucp1 gene expression levels in WT and Atf7 differentiated primary iWAT preadipocytes in the presence or absence of Rosi (n = 3, three biological replicates). Data are presented as means ± SD. Statistical analysis was performed using two-way ANOVA followed by Holm-Sidak multiple comparison tests. **p < 0.01, ***p < 0.001.

(C) Thermogenic gene expression levels in differentiated WT and Atf7 primary iWAT preadipocytes treated with Rosi (n = 3, three biological replicates). Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. *p < 0.05.

(D) Relative mRNA levels of Ucp1 in differentiated WT and Atf7 primary iWAT preadipocytes before and after treatment with 2 μM isoproterenol (Iso) for 4 h (n = 3, three biological replicates). Data are presented as means ± SD. Statistical analysis was performed using two-way ANOVA followed by Holm-Sidak multiple comparison tests. *p < 0.05, ***p < 0.001 versus WT (no Iso treatment).

To determine whether ATF7 contributes to thermogenic programming of beige or brown adipocyte in vivo, we examined UCP1 protein levels in BAT and iWAT in WT and Atf7 mice. We found that loss of ATF7 did not affect UCP1 abundance in BAT, but UCP1 expression was increased in iWAT in Atf7 mice compared with WT controls (Figure 6A). To assess the morphology of adipocytes in iWAT, we performed hematoxylin and eosin staining of iWAT sections, and more multilocular adipocytes were observed in iWAT from Atf7 mice than from WT controls (Figure 6B). Importantly, immunohistochemical staining showed that the number of UCP1-positive cells was markedly increased by ablation of ATF7, consistent with evaluated UCP1 protein levels in iWAT from ATF7 KO cells (Figures 6A and 6B). To explore gene expression profiles in iWAT induced by the loss of ATF7, we performed microarray analysis using RNA isolated from iWAT in Atf7 and WT control cells. The results revealed that 115 genes, including Ucp1, Cidea, Pgc1α, and Pparα, were up-regulated in Atf7 iWAT compared with control tissue (Figure 6C). Pathway analysis revealed that these genes are associated with adaptive thermogenesis and fatty acid metabolism (Figure 6D). Subsequent qPCR results further confirmed that several brown-fat-selective genes including Ucp1, Cidea, Pparα, and Ntrk3, and mitochondrial genes including Cox7a1 and Cox8b, were increased in iWAT following loss of ATF7, whereas there was no change in the expression of common adipocyte genes including C/ebpα, Fabp4, and Pparγ (Figure 6E). However, except for Ntrk3, expression of brown-fat-selective genes did not differ between WT and Atf7 BAT. In addition, there were no changes in the expression of mitochondrial and common adipocyte genes in Atf7 BAT compared with WT controls (Figure S8A). In line with the results of thermogenic gene expression in BAT, there is no significant difference in the tolerance to acute cold exposure between Atf7 and WT littermates (Figure S8B). Taken together, the in vivo and in vitro results suggest that ATF7 contributes to the regulation of beige adipocyte biogenesis by suppressing thermogenic gene programming.

Figure 6

ATF7 Ablation Promotes Browning of iWAT

(A) Relative UCP1 protein levels in BAT and iWAT of WT and Atf7 littermates (n = 2, two biological replicates).

(B) Hematoxylin and eosin staining (upper) and UCP1 immunohistochemical analysis (lower) of iWAT from WT and Atf7 littermates.

(C) Heatmap depicting expression levels of differentially expressed genes induced by loss of ATF7 in WT and Atf7 littermates (n = 3, three biological replicates).

(D) Pathway analysis for up-regulated genes in Atf7 iWAT.

(E) Expression of brown-fat-selective genes, common adipocyte genes, and mitochondrial genes in iWAT from WT and Atf7 littermates (n = 3, three biological replicates).

Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. *p < 0.05. See also Figures S8.

Overexpression of ATF7 Inhibits Thermogenic Gene Programming

To further explore the molecular mechanism of ATF7 functions in beige cell biogenesis, we used C3H10T1/2 multipotent mesenchymal cells, which have the capacity to undergo brown or beige adipogenic differentiation when treated with an adipogenic cocktail. The results revealed no obvious change in ATF7 protein abundance during differentiation, but phosphorylation of ATF7 was induced during the later stages of differentiation (Figure 7A). To examine the role of ATF7 in differentiation, FLAG-tagged ATF7 was expressed in C3H10T1/2 cells via a lentiviral vector, and overexpression of ATF7 did not affect lipid accumulation, as shown by oil red O staining (Figure 7B). However, RNA-seq data indicated that expression of brown-fat-selective genes, including Ucp1, Cidea, Pparα, Kcnk3, and Otop1, was reduced following overexpression of ATF7 (Figure 7C). Notably, these down-regulated genes induced by ATF7 overexpression are related to fatty acid and lipid metabolism (Figure 7D), similar to the pathways associated with genes up-regulated in Atf7 iWAT (Figure 6D). We next compared genes down-regulated by ATF7 overexpression in differentiated C3H10T1/2 cells and up-regulated by ATF7 ablation in iWAT. As shown in the Venn diagram (Figures 7E), 22 genes including Ucp1, Cidea, Pparα, and Cox7a1 overlap between these two gene sets. The results of qPCR experiments further confirmed that ATF7 overexpression led to a reduction in the expression of brown-fat-selective genes Cidea, Elovl3, and Pparα, and mitochondrial genes Cox7a1 and Cox8b (Figure 7F). Although overexpression of ATF7 strongly inhibited Ucp1 expression, there was a 3-fold increase in Ucp1 expression after Iso treatment in ATF7-overexpressing cells, whereas Iso stimulation led to a 3-fold increase in Ucp1 expression in control cells (Figure 7G). These results further suggest that ATF7 does not participate in the acute activation of Ucp1 via β-adrenergic stimulation.

Figure 7

Overexpression of ATF7 Represses Thermogenic Gene Programming

(A) ATF7 protein and phosphorylation levels during C3H10T1/2 cell differentiation measured by immunoblotting using antibodies against ATF7 (1A7) and Phospho-ATF7 (Thr53), respectively.

(B) Oil red O staining of control and ATF7-overexpressing cells 7 days after induction of differentiation.

(C) Heatmap depicting expression levels of differentially expressed genes in ATF7-overexpressing C3H10T1/2 cells compared with control cells (n = 2, two biological replicates).

(D) Pathway analysis for down-regulated genes induced by ATF7 overexpression.

(E) Venn diagram showing the number of overlapping up-regulated genes in Atf7 iWAT and down-regulated genes in ATF7-overexpressing cells.

(F) Relative mRNA levels of thermogenic genes in differentiated control and ATF7-overexpressing cells (n = 3, three biological replicates). Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. **p < 0.01, ***p < 0.001.

(G) Ucp1 gene expression levels before and after treatment with 2 μM isoproterenol (Iso) for 4 h in control and ATF7-overexpressing cells (n = 3, three biological replicates). Data are presented as means ± SD. Statistical analysis was performed using two-way ANOVA followed by Holm-Sidak multiple comparison tests. *p <0.05, **p < 0.01 versus controls (without Iso treatment), ***p < 0.001 versus controls (with Iso treatment).

ATF7 Associates with C/EBPβ and Regulates Histone Modification of Ucp1 Enhancers

To explore the molecular mechanism underlying the repressive role of ATF7 in thermogenic gene programming, we examined whether ATF7 associates with PRDM16, which is the master regulator of the thermogenic gene programming. However, coIP experiment was performed using 293T cells, and the results indicated that ATF7 interacts with C/EBPβ, but not with PRDM16 (Figure 8A), implying that ATF7-induced suppression of thermogenic gene programming is likely not mediated by PRDM16. To further confirm whether ATF7 associated with C/EBPβ in the adipocytes, coIP experiment was performed using ATF7-overexpressing C3H10T1/2 cells after differentiation in the presence of Rosi for 4 days. We found that the interaction between ATF7 and endogenous C/EBPβ could be detected in the adipocytes (Figure 8B). Recent study demonstrated that C/EBPβ binds to the enhancer regions of thermogenic genes and controls their expression in response to IL-10 treatment (Lai et al., 2017, Rajbhandari et al., 2018). Although overexpression of ATF7 did not affect the binding of C/EBPβ to Ucp1 enhancers located 12, 5, and 2.5 kb upstream of transcription start sites, the elevated H3K9me2 levels on these regions were detected in the ATF7-overexpressing C3H10T1/2 cells after differentiation for 5 days (Figure 8C). As these enhancer regions contain the CRE (cAMP response element) sequence, we speculated that ATF7 may bind to these regions (Rajbhandari et al., 2018). The results of quantitative chromatin immunoprecipitation (qChIP) analysis confirmed that ATF7 directly binds to these enhancers in C3H10T1/2 cells after differentiation for 5 days (Figure 8D). We also investigated the effect of ATF7 deficiency on H3K9me2 levels on those enhancers, and levels on the enhancer located 5 kb upstream from the transcription start site in Atf7 primary iWAT preadipocytes were lower than in WT cells after differentiation in the presence of Rosi for 7 days (Figure 8E), indicating that ATF7 silences Ucp1 gene expression by increasing H3K9me2 levels. Thus these findings demonstrate that ATF7 associates with C/EBPβ, binds enhancers of the Ucp1 gene, and represses gene expression via H3K9me2.

Figure 8

ATF7 Interacts with C/EBPβ and Regulates H3K9me2 Levels of Ucp1 Enhancers

(A) 293T cells transfected with C/EBPβ-His, HA-Prdm16, and FLAG-ATF7. Lysates were immunoprecipitated with anti-hemagglutinin (HA) antibody and immunoblotted with antibodies against Prdm16, ATF7, and C/EBPβ, respectively.

(B) Co-immunoprecipitation of FLAG-ATF7 and C/EBPβ in ATF7-overexpressing C3H10T1/2 cells after differentiation in the presence of Rosi for 4 days.

(C) Chromatin immunoprecipitation (ChIP)-PCR analyses showing binding of C/EBPβ and enrichment of H3K9me2 on Ucp1 enhancers in control and ATF7-overexpressing C3H10T1/2 cells after differentiation in the presence of Rosi for 5 days (n = 3, three biological replicates).

(D) ChIP-PCR analyses showing binding of ATF7 to Ucp1 enhancers in differentiated C3H10T1/2 cells after differentiation in the presence of Rosi for 5 days (n = 3, three biological replicates).

(E) H3K9me2 enrichment on enhancers of Ucp1 genes in WT and Atf7 primary iWAT preadipocytes after differentiation in the presence of Rosi for 7 days (n = 3, three biological replicates).

Data are presented as means ± SD. Statistical analysis was performed using two-tailed unpaired Student's t tests. *p < 0.05, **p < 0.01, ***p < 0.001.

Discussion

In this study, we found that loss of ATF7 impairs adipogenesis and induces expression of innate immune-related genes. A previous study demonstrated that ATF7 represses innate immune gene expression in macrophages via modulation of H3K9 dimethylation levels at promoter regions (Yoshida et al., 2015). Consistent with this function in macrophages, our results demonstrated that ATF7 also functions as a repressive regulator of innate immune genes in preadipocytes. Interestingly, preadipocytes can act as macrophage-like immune cells and express the same cytokines as macrophages (Cousin et al., 1999). Under certain conditions, preadipocytes can be converted to macrophages, further implying that they may share the same regulatory mechanisms regarding regulation of innate immune responses (Charrière et al., 2003). We therefore predicted that the inhibitory function of ATF7 in innate immune-related gene expression in preadipocytes is mediated by association with G9a. In line with our hypothesis, deletion of G9a led to the activation of innate immune genes in brown preadipocytes, although G9a blocked adipogenesis by suppressing Pparγ expression (Wang et al., 2013).

Our results demonstrated that the promoter regions of up-regulated genes contain ISRE motifs that are recognized by STAT1. ATF7 directly binds to the promoter of Stat1 and represses its transcription in preadipocytes via epigenetic regulation, implying that the inhibitory effect of ATF7 on innate immune gene expression is mainly mediated by STAT1. Of note, the JAK/STAT1 signaling pathway reportedly mediates the inhibitory effect of IFN-α on adipocyte differentiation (Lee et al., 2016). Moreover, in vitro studies suggest that STAT1 binds to the Pparγ2 promoter in 3T3-L1 adipocytes and represses its transcription (Hogan and Stephens, 2001). Thus ATF7 may facilitate adipogenesis by indirectly regulating Pparγ expression via STAT1. Consistent with this possibility, we observed that the Pparγ agonist rosiglitazone could rescue impaired adipogenesis in ATF7-deficient primary cells. It was previously demonstrated that LPS treatment results in the defective adipogenic potential of preadipocytes independently of nuclear factor-κB (Zhao and Chen, 2015), and phosphorylation of ATF7 in response to LPS treatment contributes to the activation of innate immune genes (Yoshida et al., 2015). Our results showed that overexpression of ATF7 blocks LPS-induced activation of innate immune genes, further suggesting that ATF7 mediates the inhibitory effect of LPS on adipogenesis.

Adipocyte differentiation and thermogenic gene programming are controlled by distinct transcriptional regulatory mechanisms, even though several common transcriptional regulators, such as TLE3 and ZFP423, are shared by these two processes. Both TLE3 and ZFP423 stimulate adipogenesis through activation of PPARγ (Villanueva et al., 2011, Gupta et al., 2010); TLE3 suppresses thermogenic gene programming by disrupting interactions between PRDM16 and PPARγ (Villanueva et al., 2013), whereas ZFP423 inhibits thermogenic gene programming by suppressing Ebf2 (Shao et al., 2016). However, ATF2 acts as a positive regulator during adipogenesis and stimulates thermogenic genes (Maekawa et al., 2010a, Cao et al., 2004). Our results reveal that, similar to TLE3 and ZFP423, ATF7 promotes adipogenesis while inhibiting adipocyte thermogenic gene programming. Interestingly, unlike ATF2, ATF7 was found to be dispensable for acute induction of Ucp1 by activation of β-adrenergic receptors. Although the browning of iWAT was observed in Atf7 mice, a difference in thermogenic gene expression in BAT was not detected. These findings suggest that ATF7 mainly contributes to the regulation of beige cell biogenesis, rather than the stimuli-induced acute enhancement of Ucp1 expression.

Beige adipocyte biogenesis is regulated by a number of epigenetic regulators (Inagaki et al., 2016), and accumulating evidence suggests that the dynamic enhancer epigenome is tightly correlated with thermogenic gene programming in adipocytes (Roh et al., 2018, Lee et al., 2017). Notably, the amount of the repressive chromatin marker H3K9me2 on enhancers of thermogenic genes, including Ucp1 and Cidea, is reduced, whereas levels of the active histone marker H3K27ac are increased during brown or beige cell biogenesis (Lai et al., 2017, Brunmeir et al., 2016). Deletion of HDAC3 in adipose tissue leads to increased H3K27 acetylation of Ucp1 and Pparα enhancers, resulting in the browning of WAT (Ferrari et al., 2017). Our results showed that ATF7 associates with C/EBPβ and represses thermogenic gene programming via H3K9 dimethylation on enhancers of Ucp1. A recent study demonstrated that the recruitment of C/EBPβ and ATF2 is accompanied by altered histone modification on enhancer regions, which mediates the inhibitory effect of IL-10 signaling on thermogenic gene programming (Rajbhandari et al., 2018). C/EBPβ is essential for brown and beige adipocyte development through formation of a transcriptional complex with PRDM16 (Kajimura et al., 2009).

We identified that ATF7 as the transcriptional repressor of IFN signaling in adipocytes. Recent study shows that the activation of ISGs induced by IFN-α impairs the thermogenic gene programming and disrupts the mitochondrial structure in the brown adipocytes (Kissig et al., 2017). Importantly, repression of the IFN signaling by blocking of JAK-STAT pathway promotes the browning of human adipocytes (Moisan et al., 2015), highlighting the repressive role of IFN-JAK-STAT signaling in thermogenic gene expression. However, we found that loss of ATF7 promotes thermogenic gene programming, although ISGs exhibited higher level in Atf7 adipocytes, suggesting that ATF7 controls the expression of thermogenic genes mainly through repressive epigenetic modification on the enhancers rather than the repression of JAK-STAT pathway in adipocytes. Enhancement of beige adipocyte biogenesis by a loss of ATF7 by regulating the dimethylation of H3K9 on thermogenic gene enhancers might be more dominant compared with suppression of the same by the JAK-STAT pathway.

Expression of thermogenic genes is closely linked to environmental factors such as low temperature. Repeated cold exposure leads to increased activity in brown and beige cells, which confers beneficial effects to metabolic homeostasis (Bartelt and Heeren, 2014). A recent study indicates that paternal cold exposure before conception can enhance thermogenic gene expression in offspring by reprogramming the sperm epigenome (Sun et al., 2018). Notably, dATF2, the homolog of mammalian ATF7 in Drosophila, acts as a transcriptional repressor and mediates stress-induced epigenetic inheritance (Seong et al., 2011). Combined with our current results, these findings suggest that ATF7 may mediate epigenetic changes in sperm induced by cold exposure and contribute to enhanced brown or beige cell activity in the offspring of a father that experienced cold before conception.

Limitations of Study

Our results demonstrate that ATF7 can associate with C/EBPβ in adipocytes, although ATF7 may not disrupt the interaction between Prdm16 and C/EBPβ. However, at present, we cannot rule out the possibility that ATF7 blocks the association of C/EBPβ with other partners, such as the SWI/SNF complex (Kowenz-Leutz and Leutz, 1999), resulting in a reduction in C/EBPβ transcriptional activity. Furthermore, whether binding of ATF7 to enhancers of Ucp1 is dependent on C/EBPβ also needs further investigation.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.