CREB3L3 controls fatty acid oxidation and ketogenesis in synergy with PPARα

CREB3L3 is involved in fatty acid oxidation and ketogenesis in a mutual manner with PPARα. To evaluate relative contribution, a combination of knockout and transgenic mice was investigated. On a ketogenic-diet (KD) that highlights capability of hepatic ketogenesis, Creb3l3−/− mice exhibited reduction of expression of genes for fatty oxidation and ketogenesis comparable to Ppara−/− mice. Most of the genes were further suppressed in double knockout mice indicating independent contribution of hepatic CREB3L3. During fasting, dependency of ketogenesis on CREB3L3 is lesser extents than Ppara−/− mice suggesting importance of adipose PPARα for supply of FFA and hyperlipidemia in Creb3l3−/− mice. In conclusion CREB3L3 plays a crucial role in hepatic adaptation to energy starvation via two pathways: direct related gene regulation and an auto-loop activation of PPARα. Furthermore, as KD-fed Creb3l3−/− mice exhibited severe fatty liver, activating inflammation, CREB3L3 could be a therapeutic target for NAFLD.


Effect of nutrient condition on the expression of Creb3l3. First, we examined the expression levels of
Creb3l3 in the liver of wild type (WT) mice that had been fasted for 24 h (Fasted), fed ad libitum (Fed), or fed a KD for 3 days. Both the Fasted and KD groups had significantly higher expression levels of Creb3l3 than did the Fed group (Fig. 1a), which led to higher protein levels of both mature and active forms of CREB3L3 in these groups (Fig. 1b, Supplementary Fig. 1). Ppara and Fgf21 were also upregulated in the Fasted and KD groups (Fig. 1a). Thus, there was a positive correlation between the expression of Creb3l3 and that of Ppara and Fgf21 in response to the feeding conditions. KD-induced hepatic ketogenesis. Mice that are fed on KD, a low-carbohydrate high-fat diet, have enhanced lipid oxidation and ketogenesis in the liver 25 . WT mice that had been fed KD for a short term (3 days) exhibited significantly increased levels of plasma ketone bodies. In both Creb3l3 −/− and Ppara −/− mice, these KD-induced plasma ketone body levels were suppressed by 50% compared with WT mice (Fig. 2a). Plasma glucose levels were identical in all three types of mice when fed a moderate-fat (MF) diet. In contrast, when they were fed KD, Ppara −/− mice showed significantly low levels compared with WT mice, while Creb3l3 −/− mice showed a trend of increase (Fig. 2a). Plasma TG, total cholesterol (TC), and FFA levels were markedly higher in Creb3l3 −/− KD-fed mice than in any other group, and thus increased in Creb3l3 −/− Ppara −/− mice (Fig. 2a). These findings indicate that the deficiency of Creb3l3 mainly contributes to these changes. In contrast, liver TG and TC levels increased in Creb3l3 −/− , Ppara −/− , and Creb3l3 −/− Ppara −/− mice (Fig. 2a). KD-fed induction emphasized the contribution of CREB3L3 along with PPARα and discriminated the effects of these factors on plasma glucose.

Metabolic parameters in WT,
Creb3l3 −/− , but not Ppara −/− mice, exhibited a marked increase in plasma TG levels in a fasted state, but there were no differences in the plasma TC and FFA levels between the groups of mice (Fig. 3b). There were also no differences in liver glycogen contents between WT, Creb3l3 −/− , and Ppara −/− mice; however, these were strongly higher in Creb3l3 −/− Ppara −/− mice (Fig. 3c). The KO mice had higher liver lipid contents (including TG and TC) than WT mice, but there was no difference between the KO mouse genotypes (Fig. 3c).
Hepatic gene expression in WT, Creb3l3 −/− , Ppara −/− , and Creb3l3 −/− Ppara −/− mice. We determined the effect of CREB3L3 on the expression of genes involved in gluconeogenesis, β -oxidation, and ketogenesis in the liver by comparing gene expression in WT, Creb3l3 −/− , Ppara −/− , and Creb3l3 −/− Ppara −/− mice in a fasted state (Fig. 4a). We found that Creb3l3 mRNA decreased by approximately 50% in Ppara −/− mice, and Ppara mRNA decreased by approximately 50% in Creb3l3 −/− mice presumably due to the mutual auto-loop system of the two factors 23,24 . Furthermore, the PPARα coactivator PPAR gamma coactivator 1α (Ppargc1a) decreased by 40% in both single KO and further in double KO mice, and Fgf21 expression was completely blunted in all KO mice compared with WT mice, indicating that both factors are crucial for Fgf21 expression.
Cpt1a and Bdh1 demonstrated stronger and comparable suppression in Creb3l3 −/− , respectively, with a further decrease in Creb3l3 −/− Ppara −/− mice (Fig. 4a), suggesting that Cpt1a and Bdh1 are direct targets for CREB3L3 (Fig. 4b). Most of the genes related to peroxisomal β -oxidation, microsomal β -oxidation, and microsomal ω -oxidation were significantly downregulated in Creb3l3 −/− mice, but more markedly in Ppara −/− mice with further suppression in both KO mice (Fig. 4a), suggesting that both factors are involved, but Ppara is more important in these gene expressions. Cyp4a10 was not regulated by CREB3L3 but by PPARα. The gluconeogenic genes, including Pck1 and G6pc, were downregulated in Creb3l3 −/− mice, which is consistent with previous reports 8,11 as being a direct target, whereas only G6pc was downregulated in Ppara −/− mice. Fatty acid synthase (Fasn) was significantly higher in Creb3l3 −/− mice.
In terms of ketogenic enzymes, the expression of Hmgcs2 was blunted in Ppara −/− and Creb3l3 −/− Ppara −/− mice and not in Creb3l3 −/− mice, indicating that PPARα governs Hmgcs2 expression. Hmgcl wass similarly   in the liver of these mice, as determined using qPCR. All data are represented as mean ± SEM. Significant differences were determined by one-way ANOVA followed by Tukey-Kramer test and shown by *P < 0.05, **P < 0.01. n = 5-16 per group.
Scientific RepoRts | 6:39182 | DOI: 10.1038/srep39182 decreased in all KO mice. There is the possibility that both factors co-operate its expression. The expression of Bdh1 decreased in Creb3l3 −/− and Ppara −/− mice and further in Creb3l3 −/− Ppara −/− mice, suggesting that this is a target for CREB3L3 as well as PPARα .
These findings indicated that CREB3L3 may play a role in regulating ketogenesis, although the contribution was less as compared to PPARα . Therefore, to confirm this, we directly estimated ketogenesis in mice by fasting (c) Liver contents of glycogen, TG, and TC. Data are represented as mean ± SEM. Significant differences were determined by one-way ANOVA followed by Tukey-Kramer test and shown by *P < 0.05, **P < 0.01. n = 4-11 per group. the mice overnight and then intraperitoneally injecting them with sodium octanoate, a mitochondrial-permeable ketogenic substrate 27 . Creb3l3 −/− mice had significantly low β -OH butyrate levels compared with WT mice during this test (Fig. 4b), showing the impairment of hepatic ketogenesis activity in Creb3l3 −/− mice. Ppara −/− and Creb3l3 −/− Ppara −/− mice showed a further decrease in β -OH butyrate levels during its test (Fig. 4b), indicating PPARα is more effective for ketogenesis.
In vivo expression of Bdh1 and Cpt1a. The Creb3l3 knockout mouse studies outlined above identified Cpt1a, Hmgcl, and Bdh1 as candidates of target genes for CREB3L3. To determine whether CREB3L3 directly regulates the expression of these genes, Ad-CREB3L3 was introduced into the mouse hepatoma cell line AML12.2 and quantitative polymerase chain reaction (qPCR) was performed. Ad-CREB3L3 significantly activated the expression of Bdh1 and Cpt1a but did not affect the expression of any other β -oxidation and ketogenesis genes, such as Acox1, Hmgcs2, and Hmgcl (Fig. 5a).
To test our hypothesis that if Cpt1a and Bdh1 are direct targets for CREB3L3, the reduction of them in Ppara −/− mice might be dependent on CREB3L3. To determine the effects of CREB3L3 on the expression of both Cpt1a and Bdh1, we crossed Ppara −/− mice with mice overexpressing the active form of human CREB3L3 in the liver (CREB3L3 Tg) to generate CREB3L3 Tg × Ppara −/− mice, and then performed qPCR. The expression levels of the PPARα target genes Acox1, Hmgcs2, Hmgcl, and Bdh1 were significantly downregulated in the liver of Ppara −/− mice compared with WT mice (Fig. 5b). The expression of Bdh1 and Cpt1a was significantly increased in CREB3L3 Tg × Ppara −/− mice compared with that in Ppara −/− mice (Fig. 5b). In contrast, there was no difference in the expression levels of the other genes, such as Hmgcs2, and Acox1, between Ppara −/− mice and CREB3L3 Tg × Ppara −/− mice (Fig. 5b). These results functionally confirmed that CREB3L3 could directly upregulate Bdh1 and Cpt1a expression in the liver.
Confirmation of Cpt1 and Bdh1 as CREB3L3-target genes using promoter analysis. To investigate whether CREB3L3 directly increases the expression of Cpt1 and Bdh1, we performed a luciferase analysis using the Cpt1a and Bdh1 promoters. CREB3L3 can bind to the cAMP response element (CRE) consensus sequence, which is a specific binding site of the CREB/ATF family of proteins 8 . We identified the CRE site at approximately − 3 kb from the transcriptional initiation site of the mouse Cpt1a promoter. When the Cpt1a promoter luciferase vector containing this region was used in the luciferase assay, we found that it was CREB3L3 rather than PPARα that significantly activated the Cpt1a luciferase activity in the AML12.2 cells (Fig. 6a). We were unable to find the CRE sequence in the Bdh1 promoter; however, we successfully cloned the − 750-bp region of its promoter. As in the Cpt1 promoter analysis, we found that CREB3L3, not PPARα , significantly activated Bdh1 luciferase activity in the AML12.2 cells (Fig. 6a). These results indicate that CREB3L3 activates these promoters by regulating these regions independent of PPARα .
The CRE-deleted and mutated Cpt1a promoter luciferase vectors were completely blunted to activation by CREB3L3 (Fig. 6b). The electrophoretic mobility shift assay (EMSA) showed that CREB3L3 could bind to the CRE sequence in the Cpt1a promoter (Fig. 6b). We also constructed a series of deletion mutants of the Bdh1 luciferase vectors, which showed that CREBL3 activated − 200 bp of the Bdh1 luciferase vector, but the − 100-bp region was blunted to activation by CREB3L3. The EMSA assay showed that CREB3L3 bound to the region between − 126 bp and − 98 bp in the Bdh1 promoter, but no other regions (Fig. 6c). CREB3L3 activated the luciferase vector containing 4x the region between − 126 bp and − 98 bp in the Bdh1 promoter (Fig. 6c). The chromatin immunoprecipitation (ChIP) assay using the liver of mice showed that CREB3L3 directly bound to the Bdh1 and Cpt1a promoter region more efficiently in mice that were in a fasted state than those that were in a fed state (Fig. 6d), which is consistent with the finding that the active form of the CREB3L3 protein is upregulated in fasted mice. Moreover, the  PPARα on the Cpt1a and Bdh1 promoter activity in mouse AML12.2 cells, as estimated using the luciferase promoter assay. Data are represented as mean ± SEM. Significant differences were determined by one-way ANOVA followed by Tukey-Kramer test and shown by *P < 0.05, **P < 0.01. n = 3 per group. (b) The effect of CREB3L3 on the Cpt1a promoter, as estimated using cAMP response element (CRE)-deleted and mutated vectors using the luciferase assay; the electrophoretic mobility shift assay (EMSA) showed that CREB3L3 directly bound to the CRE sequence. Data are represented as mean ± SEM. Significant differences were determined by unpaired Student's t test and shown by **P < 0.01. n = 3 per group. (c) The effects of CREB3L3 on the Bdh1 promoter, as estimated with a luciferase analysis using deletion constructs of the Bdh1 promoter; the EMSA assay indicated that CREB3L3 directly bound to the region from − 128 to − 98 bp of the Bdh1 promoter. The effect of CREB3L3 on the region from − 128 to − 98 bp of the Bdh1 promoter, as estimated using the luciferase assay with 4-tandem region from − 128 to − 98 bp of the Bdh1 promoter vector Data are represented as mean ± SEM. Significant differences were determined by one-way ANOVA followed by Tukey-Kramer test and shown by *P < 0.05, **P < 0.01. n = 3 per group. (d) Results of the chromatin immunoprecipitation (ChIP) assay demonstrating that CREB3L3 directly binds to the Cpt1a and Bdh1 promoters in vivo in fasted and fed mice (n = 4 per group), and in vitro in AML12.2 cells infected with adenoviral GFP or the active form of CREB3L3 (n = 3 per group). Data are represented as mean ± SEM. Significant differences were determined by unpaired Student's t test and shown by **P < 0.01.

Effect of Creb3l3 deficiency on KD-induced hepatosteatosis. It has previously been shown that
KD-fed mice exhibit suppressed lipogenic gene expression and increased expression of genes involved in fatty acid oxidation, ketogenesis, and inflammation of the liver 28,29 . Furthermore, long-term KD causes an injury pattern in WT mice similar to nonalcoholic fatty liver disease (NAFLD) phenotypes 28,30 . To clarify the effects of CREB3L3 on ketogenesis-related hepatosteatosis, WT and Creb3l3 −/− mice were fed MF or KD for 4 weeks, when WT mice does not show steatohepatitis. Histological analysis showed severe lipid accumulation and inflammation in the liver of Creb3l3 −/− mice compared with WT mice (Fig. 7a). Consistently, liver TG and TC levels were also significantly higher in Creb3l3 −/− mice than in WT mice (Fig. 7a). Immunohistochemistry using the anti-CD11b antibody as the macrophage marker showed that the liver sections from Creb3l3 −/− mice exhibited increased infiltration by inflammatory cells compared with WT mice (Fig. 7a), which resulted in increased levels of both plasma aspartate aminotransferase (AST) and alanine aminotransferase (ALT), which are markers of liver injury (Fig. 7b).
The hepatic expression of inflammatory marker genes, such as chemokine (C-C motif) ligand 2 (Ccl2), interleukin 6 (Il6), and tumour necrosis factor (Tnf), and fibrosis marker genes, such as alpha 2, smooth muscle, mice that had been fed KD for 4 weeks, using haematoxylin and eosin staining and immunohistochemistry using the anti-CD11b antibody. Liver contents of triglycerides (TG) and total cholesterol (TC). Data are represented as mean ± SEM. Significant differences were determined by unpaired Student's t test and shown by *P < 0.05. n = 5 per group. (b) Plasma aspartate aminotransferase (AST), alanine aminotransferase (ALT), and FGF21 levels, and (c) hepatic gene expression profiles of WT and Creb3l3 −/− mice that had been fed KD for 4 weeks. Data are represented as mean ± SEM. Significant differences were determined by one-way ANOVA followed by Tukey-Kramer test (b,c) and shown by *P < 0.05, **P < 0.01. n = 5-10 per group. aorta (Acta2), and collagen, type I, alpha 1 (Col1a1), was significantly upregulated in KD-fed Creb3l3 −/− mice compared with that in MF diet-fed WT mice and KD-fed WT mice (Fig. 7c). These results confirmed that inflammation was enhanced in Creb3l3 −/− KD-fed mice. The expression of fatty acid oxidation genes Ppara and Cpt1a, and the ketogenesis genes Hmgcs2, Bdh1, Fgf21, and Creb3l3 were also significantly upregulated in KD-fed WT mice compared with that in MF-fed WT mice (Fig. 7c) but significantly downregulated in KD-fed Creb3l3 −/− mice compared with that in KD-fed WT mice. Thus, the Creb3l3 deficiency suppressed fatty acid oxidation-and ketogenesis-related gene expression in the liver, even in mice that had been fed KD for a long time. Collectively, these data indicate that the dysfunction of fatty acid oxidation and ketogenesis in the liver of Creb3l3 −/− mice exacerbated the development of KD-induced hepatosteatosis.

Discussion
CREB3L3 and PPARα have previously been shown to reciprocally activate each other and have overlapping functions such as changes of gene expression in the liver during fasting 23,24 . However, the functional differences between these proteins remained unclear. In this study, we investigated the direct effects of CREB3L3 on ketogenesis under various feeding conditions by comparing the phenotypes of WT, Creb3l3 −/− , Ppara −/− , and Creb3l3 −/− Ppara −/− mice. We found that the deletion of either Creb3l3 or Ppara resulted in considerable defects in the adaptation to energy starvation in mice. It should be noted that in either of the single KO mice, expression of the other gene was also decreased and thus additional effects of double deletion were further severe. Roughly, the deletion of Ppara largely blunted the hepatic gene expression of fatty acid oxidation and ketogenesis. It is conceivable that the synergistic interaction between CREB3L3 and PPARα should contribute to defected ketogenesis in Creb3l3 −/− mice as well as Ppara −/− mice. We also found that CREB3L3 directly transactivates Cpt1a and Bdh1, thereby regulating ketogenesis independent of PPARα . Creb3l3 −/− mice, especially on a KD, accelerated KD-induced hepatosteatosis.
PPARα is ubiquitously expressed in not only the liver but some peripheral tissues. In fasting, PPARα induces lipolysis in WAT by activating ATGL and HSL 31,32 . In turn, fatty acids derived from lipolysis are important activators of PPARα 33,34 and the substrates of fatty acid oxidation and ketogenesis in liver. Hepatic PPARα governs fatty acid oxidation and ketogenesis gene expression. Therefore, it is thought that the deficiency of adipose Ppara aggravates the defects of hepatic fatty acid oxidation and ketogenesis due to insufficiency of energy source supply from peripheral tissues and metabolic activity in the liver. On the other hand, as CREB3L3 is expressed in only the liver and small intestine, lipolysis in WAT in Creb3l3 −/− mice were comparative with that in WT mice. This could explain that reduced ketosis in Creb3l3 −/− mice was limited in fasting.
KD is a high-fat, low-carbohydrate diet containing approximately 80% fat. Although both fasting and a KD feeding induce ketogenesis, a KD feeding supplies energy, the sources of which are mainly from dietary fat, and not lipolysis in adipose tissue. Therefore, a KD feeding study can reflect on the functions in hepatic fatty acid oxidation and ketogenesis. The changes of fatty acid oxidation and ketogenesis in Creb3l3 −/− mice were comparable with those in Ppara −/− mice, indicating both hepatic factors are equally important in this metabolism.
In this study, we identified that CREB3L3 directly regulates Cpt1a, the rate limiting molecule for FA transport into the mitochondria, and Bdh1, a component gene in the ketogenesis pathway. Moreover, CREB3L3 upregulates PPARα target genes via the regulation of both the gene expression and transcriptional activity of PPARα . The expression of Cpt1a was solely regulated by CREB3L3 in a KD feeding (Fig. 2), whereas it was regulated by both CREB3L3 and PPARα during fasting (Fig. 4a). These finding indicates that the PPARα -CREB3L3 interaction in a KD feeding is more important for the PPARα -mediated Cpt1a expression than in fasting.
CPT1a is one of the rate-limiting enzymes involved in mitochondrial β -oxidation, catalysing the esterification of long-chain acyl-CoAs to L-carnitine for transport into the mitochondria for fatty acid oxidation. Therefore, the mitochondrial carnitine system plays a crucial role in β -oxidation. The reduction in Cpt1a expression in the liver of Creb3l3 −/− mice inhibit the transport of fatty acids to the mitochondria and suppress fatty acid oxidation. KD contains approximately 80% fat, which mainly consists of long-chain acyl-CoAs. Thus, Creb3l3 −/− mice cannot catalyse the excess long-chain acyl-CoAs.
Ketogenesis is catalysed by the enzymes HMGCS2, HMGCL, and BDH1, which converts acetyl-CoA into β -OH butyrate. BDH1 is involved in the final step of this process, catalysing Acac to β -OH butyrate. In this study, we demonstrated that CREB3L3 directly regulates Bdh1 expression. To increase the enzymatic activity of HMGCS2, deacetylation by SIRT3 is required 26 . The deficiency of Creb3l3 reduced the expression of Sirt3 in feeding a KD diet, indicating the reduction of Hmgcs2 enzymatic activity in Creb3l3 −/− mice. Taken together, these results demonstrate that CREB3L3 plays a pivotal role in the ketogenic process that converts acetyl-CoA into β -OH butyrate in the liver.
These findings could propose that CREB3L3 contributes to ketogenesis; however, in the ketogenesis activity test, the deficiency of Creb3l3 did not show the predicted loss of its activity. Because its test uses sodium octanoate as a substrate, octanoate can cross the mitochondrial membranes independently of CPT1a 35 . Therefore, the differences of Cpt1a expression between Creb3l3 −/− mice and Ppara −/− mice could not reflected a change in the ketogenesis activity (Fig. 4b). The deficiency of Ppara had the further decrease of other fatty acid oxidation and ketogenesis gene expression compared with the deficiency of Creb3l3, thereby further decreasing the ketogenesis activity.
It has previously been shown that Fgf21 −/− mice exhibit impaired adaptation to ketosis 20 . In the present study, we found that the expression of hepatic Fgf21 was blunted in both Creb3l3 −/− and Ppara −/− mice, which could at least partially explain why the Creb3l3 deficiency led to impaired ketogenesis.
It is well established that hepatic fatty acid oxidation disorders, including a deficiency in PPARα , lead to NAFLD [5][6][7]36 . High fat feeding of mice with insufficiency of ketogenesis results in NAFLD 37  In summary, our findings indicate that CREB3L3 regulates ketogenesis via two pathways: (1) by directly activating Ppara expression to activate PPARα -induced genes related to fatty acid oxidation and ketogenesis and (2) by directly activating Cpt1a and Bdh1 expression. We propose that CREB3L3 co-operates with PPARα by directly and indirectly regulating the expression of genes involved in fatty acid oxidation and ketogenesis. In particular, CREB3L3 plays a role in exogenous (dietary) fatty acid homeostasis, while PPARα plays dual roles in exogenous and endogenous (released from lipolysis in adipose tissues) fatty acid homeostasis. Individuals with NASH have been shown to have lower levels of ketone bodies than individuals with simple steatosis 38 , and studies on humans have revealed that the levels of plasma ketone bodies is negatively correlated with the pathology of NASH, suggesting a reduction in ketone body metabolism in individuals with NASH 38 . Thus, as ketogenesis prevents diet-induced fatty liver injury and hyperglycaemia 37 , CREB3L3 may represent a new therapeutic target for NAFLD.
For KD analysis, 8-week-old male mice were fed for 3 days or 4 weeks and sacrificed in a fed state. KD consisted 76% fat, 8.8% protein, and 0.74% carbohydrates (no sucrose) (w/w) 24 . For adenoviral infection, 8-week-old male mice were infected with the indicated adenovirus at 5.0 × 10 8 pfu/g body weight, following which samples were collected 6 days after infection from mice in a fed state. For the fasting and re-feeding experiments, mice were fasted for 24 h and then fed a high-sucrose/fat-free diet for 12 h, as previously described 42 . All animal husbandry procedures and animal experiments were performed in accordance with the Regulation of Animal Experiments of the University of Tsukuba and were approved by the Animal Experiment Committee, University of Tsukuba.
Histological analysis. Mouse livers were fixed, embedded in paraffin, sectioned, and stained with haematoxylin and eosin, and Masson trichrome.
Immunohistochemistry. Mouse livers were fixed in 10% formalin and embedded in paraffin blocks. The liver sections were then deparaffinized, and the antigens were retrieved by heating at 80 °C for 20 min. Following blocking, the sections were incubated with the primary antibodies for CD11b (1:600) (abcam, ab75476). After washing the sections with phosphate buffered saline (PBS), Histostar ™ mouse (Medical & Biological Laboratories Co. Ltd. Japan) was applied at room temperature for 30 min, following which the samples were again washed with PBS. The reaction product was visualized by applying diaminobezidine (Dako Japan) for 5 min and then counterstaining the sections with haematoxylin.
Cell culture. Mouse AML12.2 hepatoma cells were maintained in Dulbecco's Modified Eagle Medium/Ham's F12 media supplemented with ITS Liquid Media Supplement (SIGMA) and 10% fetal calf serum. Cells were infected with adenoviral green fluorescent protein (GFP) or the active form of CREB3L3 at a multiplicity of infection of 100 and then incubated for 24 h. Ketogenesis assay. To measure ketogenic activities in vivo, 8-week-old male mice were administered 0.5 M sodium octanoate at a dose of 6 μ l per gram of body weight after 18 h of fasting 27 . Blood samples were collected 0, 1, 2, and 4 h after the injection, and plasma β -OH butyrate levels were determined using the β -OH butyrate colorimetric assay (Sanwa kagaku).
Plasmids. The active form of human CREB3L3 (1-320 aa) (AB0050902) was cloned using PCR into the pcDNA3 vector. The mouse Cpt1a promoter from − 2200 bp to − 1480 bp was cloned using PCR and subcloned into the pGL3-promoter luciferase vector (Promega). The mouse Bdh1 promoter from − 720 bp to + 130 bp and the mouse Ppara promoter from − 480 bp to + 100 bp were cloned using PCR and subcloned into the pGL3-basic luciferase vector (Promega).
Preparation of recombinant adenovirus. The cDNAs coding the active form (1-320 aa) of human CREB3L3 and GFP were cloned into pShuttle vectors (Clontech). In addition, the recombinant adenovirus vectors were recombined with pAdeno-X vectors (Clontech). The recombinant adenoviruses were produced in 293A cells (Invitrogen) and purified using CsCl gradient centrifugation, as previously described 43 . Metabolic measurements. The levels of glucose, insulin, TG, FFA, TC, AST, ALT, and β -OH butyrate in the plasma, and the levels of glycogen, TG, and TC in the liver were measured as previously described 43 . Plasma FGF21 levels were measured using mouse FGF21 Quantikine ® ELISA kits (R&D systems).
Immunoblotting. Total cell lysates were immunoblotted as previously described 43 , using antibodies to α -tubulin (Millipore) and an anti-mouse CREB3L3 polyclonal antibody, which was generated as previously described 24 . Analysis of gene expression. Total RNA from cells and tissues was prepared using TRIzol ® Reagent (Invitrogen). Real-time PCR analysis was performed using total RNA for cDNA synthesis (Invitrogen) with the