Selective Activation of Mitogen-Activated Protein (MAP) Kinase Kinase 3 and p38α MAP Kinase Is Essential for Cyclic AMP-Dependent UCP1 Expression in Adipocytes

The sympathetic nervous system regulates the activity and expression of uncoupling protein 1 (UCP1) through the three β-adrenergic receptor subtypes and their ability to raise intracellular cyclic AMP (cAMP) levels. Unexpectedly, we recently discovered that the cAMP-dependent regulation of multiple genes in brown adipocytes, including Ucp1, occurred through the p38 mitogen-activated protein kinases (MAPK) (W. Cao, K. W. Daniel, J. Robidoux, P. Puigserver, A. V. Medvedev, X. Bai, L. M. Floering, B. M. Spiegelman, and S. Collins, Mol. Cell. Biol. 24:3057-3067, 2004). However, no well-defined pathway linking cAMP accumulation or cAMP-dependent protein kinase (PKA) to p38 MAPK has been described. Therefore, in the present study using both in vivo and in vitro models, we have initiated a retrograde approach to define the required components, beginning with the p38 MAPK isoforms themselves and the MAP kinase kinase(s) that regulates them. Our strategy included ectopic expression of wild-type and mutant kinases as well as targeted inhibition of gene expression using small interfering RNA. The results indicate that the β-adrenergic receptors and PKA lead to a highly selective activation of the p38α isoform of MAPK, which in turn promotes Ucp1 gene transcription. In addition, this specific activation of p38α relies solely on the presence of MAP kinase kinase 3, despite the expression in brown fat of MKK3, -4, and -6. Finally, of the three scaffold proteins of the JIP family expressed in brown adipocytes, only JIP2 coimmunoprecipitates p38α MAPK and MKK3. Therefore, in the brown adipocyte the recently described scaffold protein JIP2 assembles the required factors MKK3 and p38α MAPK linking PKA to the control of thermogenic gene expression.

Uncoupling protein 1 (UCP1) is essential for rodents and other small mammals to maintain their body temperatures, since it is the sole mediator of cold-induced nonshivering thermogenesis (4,6,48); UCP1 is also a key contributor to the regulation of diet-induced thermogenesis (6,58). The UCP1 protein resides within the inner membrane of mitochondria, where it serves as a portal for dissipation of the proton gradient such that respiration is uncoupled from ATP production and generates heat (35,49,54). The UCP1 mRNA and protein are found in "brown" and to a lesser extent in "white" adipose tissue; however, its expression is confined to brown adipocytes (53). Similar brown adipocytes exist scattered within white adipose depots in adult humans (22,37), but their contribution to thermogenesis is admittedly modest. Nevertheless, studies in animals or humans exposed to high catecholamine levels or treated with sympathomimetics show that brown adipocytes expressing UCP1 can be recruited within white adipose depots (10,12,13,16,29).
Brown adipose tissue (BAT) and white adipose tissue are innervated by sympathetic noradrenergic nerves (2,3,42,50,63). In response to cold exposure or diet, sympathetic nervous system activation leads to the release of norepinephrine to interact with adrenergic receptors (AR); in particular the family of ␤ARs (39,49,55,72). Catecholamine stimulation of the three ␤ARs present in adipocytes promotes a series of events initiated by the production of cyclic AMP (cAMP) and the activation of cAMP-dependent protein kinase (PKA) (20,56,64). These events result in lipolysis and liberation of free fatty acids (FFA) from triglyceride stores (39). These FFA serve not only as substrates for oxidative respiration but also as allosteric activators of UCP1 function (24,25,60). ␤AR-mediated increases in cAMP also stimulate Ucp1 gene transcription. The cAMP response of the Ucp1 gene is achieved predominantly through an enhancer region (9,15,38). This enhancer, which is well conserved among species (11), confers specificity of expression to brown adipocytes as well as the cAMP response and contains at least two key elements: a peroxisome proliferator response element (PPRE) and a cAMP response element (CRE).
We have recently shown that the cAMP-dependent transcription of the Ucp1 gene is regulated through these two elements by p38 mitogen-activated protein kinase (MAPK) (7). The effect of p38 MAPK on these elements occurs in a coordinated fashion. First, p38 MAPK phosphorylates a protein called PGC-1␣ (7), which is a transcriptional coactivator and mediator of mitochondriogenesis (68), among other functions. This modification of PGC-1␣ enhances its activity as a nuclear coactivator of gene transcription in coordination with peroxi-some proliferator-activated receptor ␥ (PPAR␥); PPAR␥ in turn binds to the UCP1 PPRE (7). Second, p38 directly stimulates expression of the Ucp1 gene through phosphorylation of the transcription factor ATF-2; ATF-2 binds to the CRE2 (7). Finally, the PGC-1␣ gene itself also possesses a CRE (28) but in the brown adipocyte is a target of p38-activated ATF-2 and not CREB (7). By increasing the overall amount of PGC-1␣ protein over time, p38 MAPK primes the cell for a sustained enhancement of UCP1 expression. Despite this new understanding of the role of p38 MAPK in the regulation of the Ucp1 and PGC-1␣ genes in brown fat, the cascade of signaling events downstream of PKA by which p38 MAPK becomes activated is completely unknown.
To begin to unravel this new pathway, we realized that it was necessary to tackle this problem in a "bottom-up" approach. Therefore, we reasoned that a strategy that would best serve this effort should first identify the actual p38 MAPK isoform(s) involved and proceed in a retrograde manner. The p38 MAPK group is composed of four isoforms: p38␣ (26,41), p38␤ (32), p38␥ (43), and p38␦ (66). Among them, p38␣ and -␤ are sensitive to the pyrimidyl imidazoles SB202190 and SB205380 (14,23). These two isoforms are expressed in adipocytes (36). Depending on cell type and stimulus, p38 MAPK can be activated by MKK3 (17) or MKK6 (27,46,52,61) or by both of them. In some cell types MKK4 can activate p38 MAPK (17,44). However, depending upon the stimulus or physiological state, there are circumstances in which these MKKs can clearly display substrate preferences or noninterchangeable roles (62,69). For example, MKK3 tends to prefer p38␣ while MKK6 is equally efficient at both p38␣ and p38␤ (18). We also embarked on the current series of studies because defining the exact p38 isoform(s) and its immediate activator(s) in the control of UCP1 transcription may provide clear targets to modulate thermogenesis. Using a variety of experimental approaches, we show that p38␣ and its activator MKK3 are the sole players in the control of Ucp1 gene transcription.
Cell culture and transfection. The HIB-1B brown preadipocytes (57) were maintained in Dulbecco's modified Eagle medium (DMEM) with 10% fetal bovine serum. Cells in 6-well plates were transfected with a total amount of plasmid DNA up to 2.2 g/well and 5 to 10 l Lipofectamine. As needed, these DNA mixtures included pGL2-␤ 3 AR (0.5 g), pCDNA3-kinases (0.2 to 1 g), UCP1 enhancer-TK-CAT (1.0 g), and ␤-actin-luc (0.2 g) or cytomegalovirus-␤-galactosidase (CMV-␤-GAL) (0.125 g). Transfection with siRNAs (20 to 60 nM) were performed using siPORT lipid (4 to 8 l). In cotransfection experiments involving siRNAs, the siRNA was added to the well at the time of the seeding the cells, and the plasmid transfection was performed 12 h later. In all cases, the PPAR␥ agonist rosiglitazone (1 M) was added at the same time as the serum following the serum-free period of the transfection protocols.
GGUGGCUGUAAAGAAGCUGTT CAGCUUCUUUACAGCCACCTT a Note that for some of the targeted genes, two (p38␣ MAPK) or three (MKK3) siRNA duplexes were used because they achieved Ͼ80% specific gene knockdown.
VOL. 25,2005 PKA ACTIVATION OF MKK3 AND p38␣ MAPK 5467 were then incubated for 5 min with either CL (10 M) or Forsk (10 M). Cells were then washed once with phosphate-buffered saline (PBS) containing 5 mM ␤-glycerophosphate and 1 mM sodium orthovanadate, followed by a lysis buffer (25 mM Tris-HEPES, 150 mM NaCl, 10 mM ␤-mercaptoethanol, 5 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, 0.5 mM EDTA, 0.5 mM EGTA, 1% triton X-100, 0.1% IGEPAL, and 1 CPIC tablet per 10 ml) for 15 min. Five microliters of this cell lysate was incubated for 30 min at 30°C with 5 l of the 5ϫ PepTag PKA reaction buffer, 2 g of PepTag A1 peptide (fluorescent kemptide), and 1 ml of the peptide protection solution (all parts of the PKA assay kit were from Promega) in a 25-l total volume. For the positive control, the sample has been replaced by 10 ng of PKA catalytic subunit, and for the negative only lysing buffer is added to the reaction mixture. The reaction was stopped by boiling the sample for 10 min, a final concentration of 3.2% glycerol was added, and 3.5 l sample was loaded and resolved on a 0.8% agarose gel. Image acquisition was performed on a typhoon 9410 variable modes imager and analyzed using Image-Quant TL v2003.03 software. Rap1 activation assay using RalGDS-RBD. Rap1 pull-down assays were performed essentially as described previously (19) using the reagents from Upstate (Charlottesville, VA). HIB-1B cells were seeded in 10-cm-diameter dishes. Cells were washed twice in cold PBS and lysed on ice in 1 ml of lysis buffer (50 mM Tris-HCl [pH 7.5], 500 mM NaCl, 2.5 mM MgCl 2 , 1% NP-40, 10% glycerol, and 1 Complete Mini Antiprotease tablet per 10 ml of lysis buffer). Cell debris was removed by centrifugation at 13,200 ϫ g for 10 min at 4°C. Fifty microliters of the GST-RalGDS-RBD-agarose slurry was added to each supernatant, and the mixture was incubated at 4°C for 45 min on a rotating wheel. Beads were washed three times with lysis buffer. Samples were denatured for 3 min at 95°C and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Western blotting was performed using the anti-Rap1 antibody included in the kit and an alkaline phosphatase-conjugated anti-rabbit secondary antibody and ECF detection kit from Amersham Biosciences. Image acquisition was performed on a typhoon 9410 variable modes imager and analyzed using image-Quant TL v2003.03 software, both from GE healthcare (Piscataway, NJ).
Protein kinase assay for p38 MAPK activity. HIB-1B cells were transfected or not with FLAG-tagged p38 MAPK, and total cell p38 MAPK activity was assessed from cellular lysate while isoform-specific activity was assessed following immunoprecipitation using M2 anti-FLAG agarose antibody. The cells were washed twice with PBS containing 5 mM ␤-glycerophosphate and 1 mM sodium orthovanadate and then lysed for 30 min in a 25 mM Tris-HEPES buffer containing 150 mM NaCl, 5 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, 5 mM EDTA, 5 mM EGTA, 1% Triton X-100, 0.1% IGEPAL, and 1 CPIC tablet per 10 ml. Kinase assays in vitro were performed either on whole-cell lysates or after immunoprecipitation. In the latter case, the lysate was incubated overnight with 40 l of M2 anti-FLAG agarose antibody. Cell lysate or immune complexes were incubated at 30°C for 45 min with 2 g GST-ATF-2(1-109), 250 M ATP (containing or not 10 Ci [␥-32 P]ATP) in 40 l of kinase reaction buffer (43). For the radioactive version of the protocol, an equal amount of 2ϫ Laemmli sample buffer was added to terminate the reactions. In the nonradioactive version, 40 l of glutathione Sepharose 4B was used to pull down the substrate. Proteins were resolved with 4 to 20% acrylamide gradient Tris-glycine gels. Protein phosphorylation was visualized either by autoradiography for the radioisotopic protocol or with the Pro-Q Diamond phosphor-protein stain for the nonradioactive method. In both cases image acquisition was performed on a Typhoon 9410 variablemode imager and analyzed using ImageQuant TL v2003.03 software.
Western blot for MAPK phosphorylation. MAPK phosphorylation was evaluated by Western blot using specific anti-phospho-MAPK and total MAPK antibodies (1:1,000 dilution) and secondary antibodies (1:10,000 for the Amersham antibody and 1:25,000 for the Sigma antibodies). The alkaline phosphatase activity was determined using the ECF detection kit from Amersham Biosciences. Image acquisition was performed on a Typhoon 9410 variable modes imager and analyzed using ImageQuant TL v2003.03 software.
CAT and luciferase assays. Cells were harvested to assay UCP1 enhancer activities 48 h after transfection. CL316,243 or Forsk was added for the last 6 to 8 h of the transfection to stimulate cAMP production, after which cell extracts were prepared in lysis buffer from a CAT enzyme-linked immunosorbent assay (ELISA) kit (Roche Molecular Biochemicals). CAT and luciferase assays were performed as previously described (8).
RNA isolation, reverse transcription-PCR, and real-time PCR. Total mRNA was extracted from cultured cells using RNAaquous4PCR and from tissue using the RNAaquous MIDI RNA purification kits. For tissues, the samples were submerged in "RNAlater" prior to the extraction. These RNA reagents were from Ambion. cDNA was generated using the High Capacity cDNA Archive kit from Applied Biosystems exactly as described in the kit (although scaled down to a 50-l total volume). Real-time PCR was performed using TaqMan probes from Applied Biosystems (Foster City, CA) on an ABI PRISM 7700 Sequence Detector from Perkin Elmer (Boston, MA) exactly as indicated by Applied Biosystems. For the quantification of the p38␣ and p38␤ MAPK mRNA, standard curves (0 to 4,000 amol) were generated using plasmids containing the cDNAs of the mouse genes. GAPDH was used as internal standard.
Immunoprecipitation. For immunoprecipitation experiments, the tissue or the cells were lysed with the same buffer as for the kinase assays. The lysate (1 to 2 mg of total protein) was precleared by a preincubation of 2 h with 40 l protein G-agarose. The cleared lysate was then incubated for 3 h with the antibody (kinase assay) or overnight (coimmunoprecipitation) with antibodies. The Gprotein agarose was added, and the mixture was incubated for an additional 3 h and washed once with the lysing buffer and five times with the washing buffer composed of 50 mM Tris-HCl, 150 mM NaCl, and antiproteases. The proteins are eluted from the resin in an ultrafree-mc 5-m centrifugal filtration device by exposing the resin to sample buffer without reducing agent at room temperature for 10 min.

RESULTS
In previous studies we demonstrated that p38 MAPK activity is involved in the ␤ARand cAMP-dependent induction of the Ucp1 gene in brown adipocytes (7,8). However, neither the identity of the p38 MAPK nor the molecular intermediaries linking PKA to the activation of p38 MAPK are known. Therefore, we used a series of experiments designed to define the p38 MAPK isoforms and immediate upstream activators involved. In our earlier studies, we presumed that the elevated cAMP levels generated in response to ␤-agonist stimulation are activating PKA, concluding that this kinase is solely responsible for conveying the cAMP signal that leads to p38 MAPK activation and Ucp1 gene expression. This conclusion was based on the ability of two mechanistically different "inhibitors" of cAMP, the competitive antagonist Rp-cAMPS and the catalytic inhibitor H89, to suppress both p38 MAPK activation and transcription of the Ucp1 gene. This "signature" typically indicates involvement of PKA. However, in a variety of cell types, cAMP has been shown to activate the small G-protein Rap1 through its interaction with a family of guanine nucleotide exchange factors (GEFs) that include Epac (exchange protein directly activated by cAMP), cAMP-GEF-I, and cAMP-GEF-II (1,33,65). The activities of these molecules are blocked by Rp-cAMPS but are unaffected by H89. Importantly, Rap1 has been shown to be an activator of p38 MAPK (30,59). Therefore, as we began this series of studies it was necessary to unequivocally determine whether PKA or a GEF (or some combination of both) leads to stimulation of p38 MAPK and Ucp1 gene expression. We treated HIB-1B brown adipocytes with the ␤ 3 AR agonist, CL316,243 (CL), or the adenylyl cyclase stimulator, forskolin (Forsk), in the presence of H89 or Rp-cAMPS or the p38 MAPK inhibitor SB202190 (SB). As shown in Fig. 1A, PKA was activated by either CL or Forsk. The response to both activators was blocked by H89 or Rp-cAMPS but not by SB. Therefore, these results indicate activation of PKA and additionally show that inhibition of p38 MAPK does not affect PKA. As shown in Fig. 1B, neither CL nor Forsk was able to elicit GTP loading to Rap1. Together these results strongly support the conclusion that p38 MAPK activation by cAMP does not depend upon a cAMP-GEF and Rap1 activation but, rather, solely requires PKA.
To confirm the role of both PKA and p38 MAPK in UCP1 induction, HIB-1B cells were pretreated with H89, Rp-cAMPS, or SB, followed by stimulation with either CL or Forsk. Activation of p38 MAPK was measured using glutathione-S-trans-ferase (GST)-tagged ATF-2 as a substrate. As shown in Fig.  1C, p38 MAPK enzyme activity stimulated by CL or Forsk was abrogated by H89, Rp-cAMPS, and SB. These inhibitors similarly blocked the transactivation of the UPC1 enhancer in re-sponse to CL and Forsk (Fig. 1D). These results indicate that, irrespective of the stimulus, PKA is necessary for p38 activation and that, in turn, induction of UCP1 enhancer activity requires p38 MAPK activity. Since all three ␤ARs are expressed in brown adipocytes and can stimulate cAMP production (56), we proposed that all of them can activate p38 MAPK and UCP1 transcription. To test this hypothesis, we treated HIB-1B cells with specific agonists of these receptors. As shown in Fig. 1E, the ␤ 1 AR-selective agonist, dobutamine, and the ␤ 2 AR-selective agonist, salbutamol, both activated p38 MAPK in a PKA-dependent manner. The nonselective ␤AR activator isoproterenol and the natural adrenergic agonist norepinephrine also activated p38 MAPK in a PKA-dependent fashion (Fig. 1E). Furthermore, these four ␤AR agonists also induced UCP1 enhancer activation, which was blocked by p38 MAPK inhibition (Fig. 1F). These results show that all three ␤AR subtypes can stimulate p38 MAPK activity and subsequently Ucp1 gene transcription.
To determine whether ␤ 3 AR stimulation leads to p38 MAPK activation and to a p38 MAPK-dependent Ucp1 gene expression in brown fat in vivo, SB (12.5 mg/kg of body weight) and CL (1 mg/kg) were administered to mice. Phosphorylation and activation of p38 MAPK and JNK was assessed by Western blotting and kinase assays and Ucp1 gene expression by realtime PCR. As shown in Fig. 2A, CL treatment induced phosphorylation of p38 MAPK by 2.5-Ϯ 0.2-fold and p38 MAPK enzyme activity by 2.4-Ϯ 0.1-fold. In contrast, following CL treatment, phosphorylation of JNK could not be detected (Fig.  2B). The ability of antibody to recognize phospho-JNK was confirmed by treating HIB-1B cells with 5 g/ml anisomycin for 15 min (Fig. 2B). Under these same treatment conditions, CL injection stimulated Ucp1 gene expression, and this stimulation was largely prevented (70%) by prior p38 MAPK inhibition (Fig. 2C). Consistent with what we have previously reported (7), these results clearly show that selective ␤ 3 AR agonist stimulation in vivo triggers p38 MAPK activity to regulation of Ucp1 gene transcription.
To identify the p38 MAPK isoforms(s) responsible for UCP1 enhancer activation, we first assessed which SB-sensitive isoforms were expressed in BAT and in the brown adipocyte cell line used to dissect the molecular pathway between PKA and Ucp1 gene expression. As shown in Fig. 3A, both p38␣ and -␤ mRNAs were expressed in BAT as well as in HIB-1B cells. Consistent with this finding, both proteins were detected by Western blot (Fig. 3B). We next overexpressed the p38␣ or -␤ isoforms in HIB-1B cells and measured UCP1 promoter activity. As shown in Fig. 3C, both isoforms could stimulate UCP1 enhancer activation equally (with a slight preference for the ␣ isoform). Next, we coexpressed MKK6E (a constitutively active form of this kinase that can phosphorylate and activate p38 MAPK) with either p38␣ or -␤ in HIB-1B cells, followed by measurements of UCP1 enhancer activity. As shown in Fig.  3D, MKK6E could activate either of the p38 isoforms as measured by significant amplification of UCP1 enhancer activity, but there was a greater preference for p38␣ MAPK. Together, these results indicate that both p38␣ and -␤ isoforms are capable of stimulating UCP1 transcription and that under conditions of maximal stimulation p38␣ MAPK might couple more efficiently to UCP1 induction. However, these data do not indicate whether either or both isoforms play a role under adrenergic stimulation. To address this issue, we introduced FLAG-tagged p38␣ or -␤ MAPK into HIB-1B cells and subsequently treated the cells with CL or Forsk. As clearly shown in Fig. 4A and B, p38␣ MAPK but not p38␤ was activated by CL or Forsk. We also performed immunoprecipitation experiments of the endogenous p38 MAPK isoforms and confirmed that Forsk-induced p38 MAPK activity could be recovered only from the p38␣ MAPK immunoprecipitate (Fig. 4C). Interestingly, using our brown adipocyte model, transactivation of the UCP1 enhancer by CL or Forsk was potentiated only by p38␣ but not by the p38␤ isoform (Fig. 4D). In order to validate this selectivity in vivo, mice were either injected with 1 mg/kg CL or exposed to a 4°C environment. As shown in Fig.  4E and F, both manipulations led to the sole activation of the p38␣ MAPK isoform in interscapular BAT. Altogether, these data establish that p38␣ MAPK but not p38␤ is activated during sympathetic nervous system stimulation of the thermogenic program and following exposure of BAT and brown adipocytes to sympathomimetic drugs. Establishing that the ␣ isoform of p38 MAPK is the one that is activated, we used siRNA gene silencing to demonstrate that p38␣ MAPK and not p38␤ MAPK was responsible for the induction of UCP1 expression. In these studies it was first necessary to demonstrate the efficacy of the siRNAs direct-ed against either the p38␣ or -␤ isoforms. This was examined in HIB-1B cells. As shown in Fig. 5A, the siRNA against either form of p38 MAPK reduced the targeted protein level by more than 80% without affecting the other isoform. More importantly, as shown in Fig. 5B, using this approach we found that essentially all CL-and Forsk-promoted p38 activity can be attributed to the p38␣ MAPK isoform. Figure 5C further shows that the siRNA against p38␣ MAPK completely inhibited UCP1 enhancer activation, while the siRNA against p38␤ failed to do so. Similar results were obtained in experiments FIG. 3. p38␣ or p38␤ MAPK can promote UCP1 enhancer activity. The expression of p38␣ and p38␤ MAPKs was measured in BAT and HIB-1B cells by quantitative real-time PCR (A) and Western blotting with selective antibodies for each isoform (B). (C) HIB-1B cells were transfected with increasing concentration of FLAG-tagged p38␣ MAPK. Two days later, the kinases were immunoprecipitated and their expression was evaluated by Western blot using anti-p38 MAPK antibody (second and fourth blots); we measured their activity using GST-ATF2 as a substrate (first and third blots), and we evaluated UCP1 enhancer activity using a CAT assay (graph). (D) HIB-1B cells were transfected with FLAG-tagged p38 MAPK and/or FLAG-tagged MKK6E. Two days later, theirs kinases were immunoprecipitated and their expression was measured by Western blot using anti-p38 MAPK and anti-MKK6 antibodies (second, third, and fourth blots), p38 MAPK activity was measured using GST-ATF2 as substrate (first blot), and UCP1 enhancer activity was evaluated by CAT assay (graph). The results shown are means Ϯ standard deviations of three independent experiments, each performed in triplicate, while the blots are from one of three experiments.  Fig. 1A. Cells were lysed, the kinase was immunoprecipitated (Western blot, bottom blots in A and B), and activity was measured using GST-ATF2 as a substrate (top blots in A and B). The results shown are from one of three independent experiments. (C) HIB-1B cells were treated as shown, and p38␣ and p38␤ MAPKs were immunoprecipitated and kinase activity measured using GST-ATF2 as a substrate. (D) HIB-1B cells were transfected with the ␤ 3 AR and with FLAG-tagged p38␣ MAPK or FLAG-tagged p38␤ MAPK. Cells were treated as follows: Basal (white bars), CL (gray bars), and Forsk (black bar). For measurement of UCP1 enhancer activity cells were harvested after 6 h. The results shown are means Ϯ standard deviations of three independent experiments, each performed in duplicate. For measurement of kinase activity, 20 min posttreatment the kinases were immunoprecipitated and activity was measured using GST-ATF2 as a substrate (upper blot). Relative amounts of the kinases in the assay are shown by Western blot using anti-FLAG antibody (lower blot). (E and F) Mice were either treated with CL (1 mg/kg intraperitoneally) for 30 min (E) or placed at 4°C for 1 h (F), and BAT was excised and processed for immunoprecipitation of p38␣ MAPK or p38␤ MAPK. Kinase activity and the protein levels of the p38 MAPK isoforms was measured as above.
employing dominant-negative constructs of p38␣ and p38␤ MAPK (not shown). Finally, it was rather remarkable to find that even the 60-to 80-fold induction of the endogenous Ucp1 gene in HIB-1B cells was totally eliminated by the p38␣ MAPK siRNA (Fig. 5C). Altogether these data leave little doubt about the highly specific activation of p38␣ MAPK by ␤AR agonists and PKA and its essential role in the activation of the Ucp1 gene. The next objective was to address the origin of p38␣ MAPK selectivity. For this purpose we explored which of the MKKs were activated by CL and Forsk. In Fig. 6A (middle panel), MKK3 and/or MKK6 was phosphorylated in a PKA-dependent manner following CL or Forsk. However, neither CL nor Forsk was able to promote the phosphorylation of MKK4 or MKK7 (Fig. 6B). These results are consistent with the fact that the JNK pathway is not activated in brown adipose tissue in vivo (Fig. 3B) or in brown adipocytes in vitro (Fig. 6B). However, the nonisoform selective nature of the MKK3/6 phospho-antibody required additional experimentation in order to address whether one (or both) of these two MKK isoforms was acti-vated. We performed selective immunoprecipitation of MKK3 and MKK6 under basal and cAMP-stimulated conditions. As shown in Fig. 6C, Forsk-induced phospho-MKK3/6 immunoreactivity was detected only upon immunoprecipitation of MKK3. Confirmation of this exclusive activation of MKK3 versus MKK6 in vivo was obtained from BAT samples of mice treated with CL or placed in a 4°C environment. Collectively, these findings demonstrate a selective activation at the level of the direct upstream kinase within the p38 MAPK module, which might be an underlying mechanism of the above-described specific p38␣ MAPK activation.
MKK6 is a universal p38 MAPK activator, but the ability of MKK3 to activate p38␤ MAPK is modest at best (18). It was tempting to speculate that the specific involvement of p38␣ MAPK would be recapitulated at the level of MKK3. We therefore tested the functional impact of overexpressing MKK3, MKK4, MKK6, and MKK7 individually on UCP1 enhancer activity and p38 MAPK activity. The results clearly show that MKK3, and to a much lower extent MKK6, can potentiate the effects of CL and Forsk on both UCP1 expres- VOL. 25,2005 PKA ACTIVATION OF MKK3 AND p38␣ MAPK 5473 sion (Fig. 7A) and p38 MAPK activity (Fig. 7B). The reciprocal experiments, using siRNAs that specifically downregulated MKK3 and MKK6 by more than 80% (Fig. 7C), show that only the selective siRNA directed against MKK3 could block CL and Forsk induction of p38 MAPK activity (Fig. 7D) as well as UCP1 expression (Fig. 7E). Finally, the siRNA that specifically targets MKK3 was the only one capable of completely interfering with the expression of the endogenous Ucp1 gene (Fig.   FIG. 6. ␤-Adrenergic agonists and cAMP promotes activation of MKK3. (A and B) HIB-1B cells were treated as described in Fig. 1A and then lysed for measurement of the activity and amounts of the indicated kinases by Western blotting as detailed in Materials and Methods, and MKK3/6 phosphorylation was evaluated by Western blot (middle blot). Each blot shown is one of three experiments. (C) HIB-1B cells were treated or not with H89 and forskolin as described in Fig. 1A. The cells were lysed, MKK3 and MKK6 were immunoprecipitated, and their phosphorylation state was measured by Western blotting. (D and E) Mice were either treated with CL (1 mg/kg intraperitoneally) for 30 min (E) or placed at 4°C for 1 h (F), and BAT was excised and processed for immunoprecipitation of MKK3 or MKK6. The ability to phosphorylate GST-tagged p38␣ MAPK and the protein levels of the MKKs were measured by Western blotting.
7F). All together, these results unequivocally define the proximal steps in the cAMP-and PKA-dependent activation of the Ucp1 gene in brown fat as being mediated solely by MKK3 and p38␣ MAPK.
The MAP kinases are usually assembled together with their upstream MKKs into a signaling module that is coordinated by large scaffolding molecules such as the JIPs (JNK-interacting proteins) (21,47). As a result, these scaffold proteins can concentrate interacting signaling partners in the vicinity of an upstream activator in order to favor a particular pathway.
Since there is no established link between PKA and p38 MAPK, we attempted to determine which, if any, of the known JIP family members in brown adipocytes might serve as the scaffold to specifically bind p38␣ MAPK and MKK3. The first objective was to determine the relative expression levels of the four known JIPs: JIP1, JIP2, JIP3, and JLP. As shown in Fig.  8A, there was no detectable expression of JIP1 in either BAT or the HIB-1B cell line (RT-PCR cycle, Ͼ45). However, JIP2, JIP3, and JLP were all found in both samples, with JIP2 being the least abundant at the mRNA level (Fig. 8A). Based on these results, each of these three JIPs was analyzed for its ability to interact specifically with MKK3 and p38␣ MAPK. His-tagged or S-protein-tagged constructs of JIP2, JIP3, and JLP were transfected into HIB-1B cells, followed by their immunoprecipitation in order to assess the identity of any interacting kinases. Figure 8B clearly shows that only JIP2 was able to specifically recover both MKK3 and p38␣ MAPK. It is noteworthy that neither MKK6 nor p38␤ MAPK was found under any circumstance, although all these molecules are clearly present as shown in the lysate.

DISCUSSION
Catecholamine regulation of brown fat thermogenesis has been clearly shown to involve increased transcription of the Ucp1 gene as a result of stimulation of the ␤ARs and cAMP production (see references 48 and 51 for reviews). Despite this, the details of the signaling cascade beyond this point have been ambiguous. It has been generally assumed that the elevated levels of cAMP lead to activation of PKA, which in turn phosphorylates the nuclear factor CREB, resulting in transcription of the various target genes in brown fat, including UCP1. In the last few years we have reported results from studies in white and brown adipocyte cell models and in vivo manipulations to show that p38 MAPK is activated in response to ␤-adrenergic stimuli (7,8). This was a novel and unexpected link between cAMP and p38 MAPK, but one that relied heavily on the use of chemical inhibitors of p38 MAPK and PKA. Since there is no clearly delineated series of steps linking PKA to p38 MAPK in the literature, unequivocal identification of the individual molecules in this cascade requires the use of more stringent approaches.
The p38 MAPK group is composed of four isoforms, p38␣ (26,41), p38␤ (32), p38␥ (43), and p38␦ (66). Among them, p38␣ and -␤ are sensitive to the pyrimidyl imidazoles SB202190 and SB205380 (14,23). Our earlier reports indicated that ␤AR and PKA stimulation of UCP1 expression in brown adipocytes was sensitive to SB (7,8) and thus narrowed the scope of inquiry. The pyridinyl imidazole-sensitive p38 MAPK isoforms are often considered to be redundant enzymes, as their substrate specificities overlap significantly. However, in some cases these two isoforms have been shown to be able to discriminate between substrates, at least under conditions of forced overexpression of dominant-negative mutants (67,73).
In the studies that we report here, we used a combination of in vivo and in vitro approaches designed to distinguish between the p38 MAPK isoforms and to identify the upstream MKK enzyme(s) responsible for mediating the PKA signal to activate Ucp1 gene expression. Our results clearly establish that p38␣ MAPK is a central obligatory component of this signaling cascade, with arguably no contribution from p38␤ MAPK, despite its presence. We do not at this point rule out the possibility that p38␤ MAPK might regulate other genes in the brown adipocyte that might be triggered by different stimuli. For example, insulin has been reported to selectively activate p38␤ MAPK but not p38␣ in brown adipocytes, and this effect was associated with increased glucose transport (36).
Our studies also indicate that there is a unique requirement for MKK3 as the immediate upstream kinase for p38␣. The selectivity of these two elements to be activated in succession is not particularly surprising, but nevertheless it does occur in spite of the existence in brown adipocytes of at least four MKKs, three of which have been demonstrated to activate p38␣ in other settings (17,27,44,46,52,61). This tight coupling is most likely indicative of their existence in a multimolecular signaling complex, as is known to exist for the stressactivated kinases from yeast to mammals (21,47). These signaling modules are maintained together by scaffolding proteins, such as KSR for the ERK pathway and the JIP proteins for the JNK pathway. A recent estimate of the size of this "functional family" of scaffold proteins suggests more than 19 members (47). Clearly, this complexity necessitates more extensive investigation in order to assign individual scaffolds to specific kinase members but which will probably also depend upon the stimuli and the cell types in which they exist. Much is known about the role of these JIP scaffolding molecules for the JNK pathway (for which they were named), while in the case of p38 MAPK three members of the family (JIP2, JLP, and JIP4) have been proposed to be able to interact with p38 MAPK and/or MKK3 (5,34,40). When considering adipocytes it must be acknowledged that relatively little is known about these molecules aside from a recent report concerning a role for JIP1 in insulin-induced JNK activation (31). Nevertheless, since we have now established that PKA specifically utilizes MKK3 and p38␣ MAPK to regulate genes in the brown adipocyte, this information directed our quest for the scaffolding protein involved. Our results clearly show that at least three putative p38 MAPK scaffolds are expressed in brown adipose tissue. The absence of JIP1 and the presence of JIP2 was somewhat surprising, since JIP1 is clearly expressed at low levels in white adipose tissue (31) and JIP2 is expressed predominantly in neuronal tissue (71). However, as usual, adipose tissue was absent from the panel of tissue surveyed for JIP2 expression in this latter study. Since JIP3 and JLP are more widely expressed, their existence in adipose tissue was not unexpected. Interestingly, when JIP2 was immunoprecipitated, only MKK3 and p38␣ MAPK were recovered within the coprecipitate. Although not a definitive proof of the necessity of JIP2 for the PKA-dependent activation of p38␣ MAPK, it at least places these components within close proximity to each other. Also, during the revision of the manuscript JIP4 was cloned (34), therefore the role for JIP4 in parallel with JIP2 cannot be excluded.
From a signaling perspective, the adipocyte is unique in that all three known ␤AR subtypes are expressed there and each is coupled to the production of cAMP (56). Here we have also shown that all three ␤ARs can stimulate p38 MAPK in the adipocyte to increase Ucp1 gene expression. Based on the present results and our earlier findings that cAMP-dependent activation of p38 MAPK elicits an orchestrated response to increase the thermogenic capacity of brown adipocytes by increasing the expression of PGC-1␣, a master regulator of mitochondriogenesis (68), an interesting speculation arises. Since human adipocytes also express multiple ␤AR subtypes and the critical region of the Ucp1 gene that responds to cAMP and p38 MAPK is conserved between rodents and humans, there may well be a similar regulation of p38 MAPK and activation of UCP1 and PGC-1␣ expression in human adipocytes. This prospect will require careful examination from human visceral adipose samples, since this depot is the main location of brown adipocytes in adult humans and may potentially be a reservoir of quiescent cells capable of thermogenic activity upon appropriate stimulation.