Signal-dependent Control of Gluconeogenic Key Enzyme Genes through Coactivator-associated Arginine Methyltransferase 1*

Together with impaired glucose uptake in skeletal muscle, elevated hepatic gluconeogenesis is largely responsible for the hyperglycemic phenotype in type II diabetic patients. Intracellular glucocorticoid and cyclic adenosine monophosphate (cAMP)/protein kinase A-dependent signaling pathways contribute to aberrant hepatic glucose production through the induction of gluconeogenic enzyme gene expression. Here we show that the coactivator-associated arginine methyltransferase 1 (CARM1) is required for cAMP-mediated activation of rate-limiting gluconeogenic phosphoenolpyruvate carboxykinase (PEPCK; EC 4.1.1.32) and glucose-6-phosphatase genes. Mutational analysis showed that CARM1 mediates its effect via the cAMP-responsive element within the PEPCK promoter, which is identified here as a CARM1 target in vivo. In hepatocytes, endogenous CARM1 physically interacts with cAMP-responsive element binding factor CREB and is recruited to the PEPCK and glucose-6-phosphatase promoters in a cAMP-dependent manner associated with increased promoter methylation. CARM1 might, therefore, represent a critical component of cAMP-dependent glucose metabolism in the liver.

According to recent estimates, 200 -300 million people worldwide will be diagnosed with type II diabetes in 2010 (1,2). Chronic hyperglycemia as observed in type II diabetic patients represents the major cause for vascular complications and results from the development of a peripheral resistance against the action of the pancreatic ␤-cell hormone insulin (3). A defective insulin response in liver has been shown to importantly contribute to the development of peripheral insulin resistance. Mice bearing a targeted disruption of the insulin receptor gene in liver display hyperglycemia and impaired glucose tolerance (4).
The coactivator-associated arginine methyltransferase CARM1 has originally been identified as a binding partner for p160 nuclear receptor co-factor family members (10). CARM1 is considered to represent a transcriptional activator as associated with specific methylation of histone H3 (11).
Attempts to elucidate the biological function of CARM1 have been constrained by the perinatal mortality of CARM1 knock-out mice (12). To elucidate the role of CARM1 on a tissue-specific basis, we explored the consequences of acute CARM1 activity modulation for hepatic gene expression. We have identified CARM1 as a critical component of the cAMP-dependent regulatory complex involved in the control of hepatic gluconeogenic enzyme gene expression.

MATERIALS AND METHODS
Plasmids, Cell Culture, and Transfections-Plasmids pSG5-CARM1 containing the CARM1 cDNA cloned into pSG5 (Stratagene, Heidelberg, Germany) was kindly provided by M. T. Bedford (M. D. Anderson Cancer Center, Smithville, TX). Mutant PKA (mutPKA), wild-type PKA (wtPKA), G6Pase-1220, PEPCK-490, PEPCK-355, and PEPCK⌬CRE have been described previously (6,13,14). Human HepG2 hepatocytes and human embryonic kidney (HEK) cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Rat H4IIE cells were maintained in minimum essential medium including 10% fetal bovine serum and 1% non-essential amino acids. For transfection, the calcium phosphate precipitation method was used as described previously (100 ng of reporter plasmid/well) (15). Where indicated, expression plasmids encoding CARM1 or mutant or wild-type PKA were co-transfected (75 or 100 ng of plasmid/well, respectively). For H4IIE transfections, 1800 ng of indicator plasmid and 500 ng of CARM1 expression vector were delivered by the calcium phosphate method using 6-well cell culture plates. The cells were treated with either Me 2 SO or forskolin (10 M) (Sigma) for 18 h and were harvested for luciferase assays. The luciferase activity was normalized by ␤-galactosidase activity.
RNA Interference (RNAi)-Oligonucleotides targeting human CARM1 (5Ј-GTAACCTCCTGGATCTGA A-3Ј) were annealed, cloned into pSuper RNAi vector (OligoEngine, Seattle, WA), and transfected into human HepG2 cells (250 ng of plasmid/well). The effect of RNAi on PEPCK or G6Pase promoter activity was measured after 48 h. Oligonucleotides containing two point mutations within the wild-type CARM1 sequence (5Ј-GTAGCCTCCTGGATCGGAA-3Ј) or unspecific oligonucleotides (5Ј-CATTACAGTATCGATCA GA-3Ј) with no * This work was supported by grants from the Deutsche Forschungsgemeinschaft (Sonder Forschungsbereich 635) (to J. C. B.) and (He3260/2-1) (to S. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. 1 significant homology to any mammalian gene sequence were used as non-silencing controls in all experiments. RNAi efficiency was determined by Western analysis of human HEK cells transfected with either CARM1-specific or -unspecific RNAi vectors using CARM1-or CREB-specific antibody (provided by M. T. Bedford and Upstate Biotechnology, Lake Placid, NY, respectively). HEK cells were transfected (1000 ng of plasmid/dish) using Lipofectamine according to the manufacturer's instructions (Invitrogen). In contrast to HepG2 cells, HEK cells were transfected with Ͼ80% efficiency (not shown) and are, therefore, useful in establishing RNAi efficiency and specificity.
Co-immunoprecipitation-Rat hepatoma H4IIE cells were grown to 90% confluence, treated with forskolin (10 M) when indicated, and the interaction assay and Western analysis were performed as described elsewhere (16). Antibodies against CREB were obtained from Upstate Biotechnology.
Statistical Analysis-Statistical analyses were performed using a twoway analysis of variance (ANOVA) with interaction and logarithmic data. In the ANOVA, Bonferroni-adjusted comparisons between individual experimental groups were performed as indicated. The significance level was p ϭ 0.05.

CARM1 Controls Gluconeogenic Gene Expression-To test whether
CARM1 is able to regulate the gluconeogenic gene program, we transfected two different lines of cultured hepatocytes with reporter gene constructs carrying PEPCK and G6Pase proximal promoter regions together with CARM1 or control expression plasmids. In both (human HepG2 as well as rat H4IIE hepatocytes), CARM1 overexpression had . Two-way ANOVA revealed p Ͻ 0.05 for the interaction term (forskolin ϫ CARM1), respectively. B, identical to A, except that rat H4IIE hepatocytes were used. no significant effect on PEPCK and G6Pase promoter activities under basal conditions (Fig. 1). However, CARM1 overexpression potentiated cAMP-induced reporter gene activities (Fig. 1, p Ͻ 0.05). In contrast, CARM1 had no effect on glucocorticoid-induced PEPCK and G6Pase transcription (data not shown), suggesting that CARM1 is specifically involved in the cAMP/PKA-dependent control of gluconeogenic gene expression. To further confirm this notion, we employed CARM1-specific RNAi constructs to knock down expression of endogenous CARM1 protein in transcriptional activation studies. To this end, Western analysis of human HEK cells using CARM1-or CREB-specific antibodies demonstrated the efficient and specific knockdown of endogenous CARM1 protein levels by CARM1-specific RNAi vectors as compared with two unspecific control vectors ( Fig. 2A and supplemental Fig. 1A). Because of the extremely low transfection efficiency of H4IIE cells (data not shown) (17,18) and the observed qualitatively identical impact of CARM1 on gluconeogenic gene transcription in H4IIE and HepG2 cells (Fig. 1), human HepG2 cells were used for all subsequent promoter studies. As shown in Fig. 2B, RNAi-mediated CARM1 gene ablation strongly inhibited PKA-induced PEPCK and G6Pase promoter activities in human HepG2 hepatocytes (left and right panels; p Ͻ 0.001 and p Ͻ 0.005 for wtPKA, plus CARM1 RNAi versus wtPKA, plus control RNAi, respectively). Supporting the specificity of the RNAi effects, a point-mutated CARM1 RNAi oligonucleotide had no influence on PKA-stimulated PEPCK activity (supplemental Fig. 1B). In line with a largely signal-dependent function of CARM1, RNAi-mediated gene knockdown had only a relatively mild effect on basal PEPCK promoter activity (Fig. 2B, left panel; p ϭ 0.0468 for mutPKA, plus CARM1 RNAi versus mutPKA, plus control RNAi). In contrast, CARM1 gene ablation also strongly reduced non-stimulated G6Pase promoter activity (Fig. 2B, right panel; p Ͻ 0.001 for mutPKA, plus CARM1 RNAi versus mutPKA, plus control RNAi), suggesting that endogenous CARM1 might also play a role in the maintenance of basal G6Pase gene expression.
CARM1 Is Recruited to Gluconeogenic Gene Promoters in Vivo-The ability of CARM1 to regulate gluconeogenic gene promoter activity prompted us to examine CARM1 recruitment to endogenous PEPCK and G6Pase chromatin templates. To this end, only weak CARM1 association with the PEPCK promoter could be detected by ChIP analysis in untreated rat hepatocytes (Fig. 3A, upper left panel, ␣CARM1, lane 1).
In contrast, CARM1 occupancy over the PEPCK promoter was readily detected after a 1-h treatment of the cells with the cAMP agonist forskolin (4.2-fold induction as determined by quantitative PCR analysis, Table 1) (Fig. 3A, upper left panel, ␣CARM1, lane 2), supporting the idea that CARM1 is indeed involved in the cAMP response of the PEPCK  gene promoter. Promoter-bound CARM1 levels returned to basal conditions after 4 h of cAMP stimulus (Fig. 3A, upper left panel, ␣CARM1, lane 3; Table 1). Consistent with the kinetics of CARM1 recruitment, PEPCK promoter histone methylation was increased after 1 h of cAMP stimulus and declined after 4 h of forskolin treatment (Fig. 3A, upper left  panel, ␣H3-Methyl, compare lanes 2 and 3). Similar dynamics of PEPCK promoter histone modification was also observed for histone H4 acetylation (Fig. 3A, upper left panel, ␣H4-Acetyl, lanes 1-3), further strengthening the assumption that CARM1-dependent PEPCK promoter methylation is indicative of a transcriptionally active chromatin status. Interestingly, analysis of CARM1 occupancy of the G6Pase promoter showed strong CARM1 binding already under basal conditions, confirming the notion that CARM1 might be required not only for signal-dependent but also basal G6Pase activity (Fig. 3A, upper right panel, ␣CARM1, lane 1; compare with Fig. 2B, right). In analogy to PEPCK, CARM1 binding to the G6Pase promoter also increased (3.4fold as determined by quantitative PCR, Table 1) after 1 h of forskolin stimulation and declined after 4 h of treatment (Fig. 3A, upper right  panel, ␣CARM1, lanes 2 and 3; Table 1). In correlation with these effects, histone H3 methylation as well as H4 acetylation were also induced after 1 h of forskolin treatment and declined after a 4-h exposure time (Fig. 3A, upper right panel, ␣H3-Methyl and ␣H4-Acetyl). Together, these ChIP experiments confirmed the importance of CARM1 for the regulation of multiple steps within the gluconeogenic pathway.
CARM1 Activates the PEPCK Promoter through the CREB-binding Site-The observed kinetics of CARM1 occupancy over the PEPCK and G6Pase promoters was tightly correlated with the phosphorylation status of cAMP-responsive element-binding protein CREB in hepatic cells exposed to cAMP stimulus for the indicated time points (Fig. 3B). Following a typical burst (1 h) attenuation (4 h) pattern, PKA-mediated CREB phosphorylation at Ser-133 is closely coupled to the recruitment of histone acetylases CBP/P300 and the subsequent transcriptional activation of cAMP-responsive target genes (5). The PEPCK promoter region harbors a conserved CREB-binding site (CRE) around position Ϫ80 bp from the transcription start site that is required for the cAMP/ PKA-dependent induction of PEPCK gene transcription during fasting (6,9,19). This CRE has been shown to be critical for the synergistic induction of PEPCK transcription through glucagon and glucocorticoids, the latter of which exert their transcriptional impact via the well defined glucocorticoid response unit located between Ϫ355 bp and Ϫ490 bp relative to the transcriptional start site (20). To define the CARM1 target site within the PEPCK promoter, we employed PEPCK promoter deletion constructs in transient assays of cultured human HepG2 hepatocytes. When co-transfected with an expression vector for CARM1, PKA-stimulated activity of reporter constructs containing 490 bp of the PEPCK promoter was robustly induced through CARM1 overexpression as demonstrated above (Fig. 4A, PEPCK-490; compare with Fig. 1; p Ͻ 0.001). Deletion of the glucocorticoid response unit had no effect on the CARM1-mediated promoter stimulation (Fig. 4A, PEPCK-355; p Ͻ 0.05). Interestingly, in the context of the PEPCK-355 construct, CARM1 also induced basal promoter activity, the effect of which did not reach statistical significance (p Ͼ 0.05). In contrast to PEPCK-490 and PEPCK-355, site-specific mutation of the CRE in the context of the PEPCK-490 parent plasmid completely abolished the stimulatory effect of CARM1 overexpression on PEPCK promoter transcription (Fig. 4A, PEPCK⌬CRE, p Ͼ 0.05), demonstrating the requirement of the CRE for CARM1 activity. To this end, we finally performed co-immunoprecipitation studies using cell extracts from cultured hepatocytes. As shown by Western analysis, no CARM1 could be recovered from precipitates using CREB-specific antibodies under basal conditions (Fig. 4B, ␣CREB, lane 1). However, CARM1 was readily detected in CREB precipitates from forskolin-treated hepatocytes (Fig. 4B, ␣CREB, lane 2), demonstrating the signal-dependent in vivo interaction of both endogenous proteins. In contrast, no CARM1 was recovered by precipitation with unspecific IgG, serving as negative control (Fig. 4B, ␣IgG, lanes 1 and 2).

TABLE 1 Occupancy of the PEPCK and G6Pase gene promoters by CARM1 upon treatment with forskolin
The numbers shown are derived from quantitative real-time PCR analysis of PEPCK and G6Pase promoter fragments co-immunoprecipitated with a CARM1-specific antibody after forskolin treatment of H4IIE hepatocytes, as indicated. The data are expressed as the fold difference relative to the control (0 h forskolin) samples from three independent assays as shown in Fig. 3.

DISCUSSION
De novo glucose production, e.g. gluconeogenesis, represents one of the key features of hepatic metabolism under fasting conditions to maintain blood glucose levels and energy substrates for brain function (7). In this report, we have defined the arginine methyltransferase CARM1 as a novel regulatory checkpoint in the cAMP-dependent control of gluconeogenic gene expression in the liver. The PEPCK and G6Pase genes have been identified as rate-limiting steps in the gluconeogenic pathway. Their activity is largely controlled on the transcriptional level and predominantly exerted by counter-regulatory hormonal cues. Although insulin efficiently down-regulates expression in the postprandial phase, glucagon acting via cAMP strongly induces promoter activity during fasting (14,19,21,22). To this end, Roesler and colleagues (23,24) proposed a model for the cAMP-dependent control of PEPCK gene transcription. In this context, CREB binding to the CRE seems to be a prerequisite for maximum cAMP responsivity of the PEPCK promoter. Given its association with CREB and the dependence of its activity on the CRE, CARM1 might, therefore, represent a novel component of the PEPCK cAMP-response unit contributing to the maximum induction of hepatic PEPCK gene transcription in response to hormonal signaling. CARM1 was initially identified as a co-activator for the estrogen receptor, working in combination with members of the p160 family of nuclear receptor co-factors, e.g. SRC-1/TIF2/GRIP1, as well as CREBbinding protein CBP/P300 (25,26). Recent studies have extended the spectrum of factors co-activated by CARM1, including p53, ␤-catenin, NF-B, and MEF2C (27)(28)(29)(30). Our current studies have identified CREB as a novel, signal-dependent CARM1 binding partner conferring cAMPdependent recruitment of CARM1 to gluconeogenic gene promoters. However, the observed induction of the PEPCK-355 reporter construct under non-stimulated conditions as well as the significant effect of CARM1 depletion upon basal G6Pase activity suggests that CARM1 might also interact with additional factors in the absence of hormonal stimulation. Both the PEPCK as well as the G6Pase promoter harbor binding sites for members of the C/EBP transcription factor family (24,31). C/EBP proteins are critically involved in the cAMP-mediated as well as basal transcription of both enzyme genes and are recruited to the G6Pase promoter in a constitutive manner (31)(32)(33)(34)(35)(36). C/EBP family members, therefore, represent attractive candidates for additional CARM1 interaction partners that might explain the impact of CARM1 on basal gene activities under certain metabolic conditions or promoter arrangements.
Previous reports suggested an inhibitory effect of CARM1 on CREBdependent transcription mediated by methylation of its co-factor CBP within the CREB interaction domain, the so-called KIX domain (37). However, subsequent studies identified more prominent methylation sites within CBP that did not inhibit CREB-dependent transcription but seem to contribute to factor/promoter-specific transcriptional activity of CBP (38). The observed stimulatory effect of CARM1 in hepatocytes might, therefore, represent a cell-specific CBP methylation pattern that would allow efficient co-activation by CARM1 of a CREB-CBP transcriptional complex in the context of gluconeogenic gene promoters.
PEPCK and G6Pase are overexpressed under diabetic conditions, thereby promoting diabetic hyperglycemia (39,40). The critical involvement of CARM1 in the hormone-dependent control of these key regulatory steps in hepatic glucose production raises the intriguing possibility that CARM1 is intrinsically contributing to the aberrant activity of the gluconeogenic pathway under diabetic conditions.