MAP Kinase Stimulation by cAMP Does Not Require RAP1 but SRC Family Kinases*

The small G protein RAP1 and the kinase B-RAF have been proposed to link elevations of cAMP to activation of ERK/mitogen-activated protein (MAP) kinase. In order to delineate signaling pathways that link receptor-generated cAMP to the activation of MAP kinase, the human A2A-adenosine receptor, a prototypical Gs-coupled receptor, was heterologously expressed in Chinese hamster ovary cells (referred as CHO-A2A cells). In CHO-A2A cells, the stimulation of the A2A-receptor resulted in an activation of RAP1 and formation of RAP1-B-RAF complexes. However, overexpression of a RAP1 GTPase-activating protein (RAP1GAP), which efficiently clamped cellular RAP1 in the inactive GDP-bound form, did not affect A2A-agonist-mediated MAP kinase stimulation. In contrast, the inhibitor of protein kinase A H89 efficiently suppressed A2A-agonist-mediated MAP kinase stimulation. Neither dynamin-dependent receptor internalization nor receptor-promoted shedding of matrix-bound growth factors accounted for A2A-receptor-dependent MAP kinase activation. PP1, an inhibitor of SRC family kinases, blunted both the A2A-receptor- and the forskolin-induced MAP kinase stimulation (IC50 = 50 nm); this was also seen in PC12 cells, which express the A2A-receptor endogenously, and in NIH3T3 fibroblasts, in which cAMP causes MAP kinase stimulation. In the corresponding murine fibroblast cell line SYF, which lacks the ubiquitously expressed SRC family kinases SRC, YES, and FYN, forskolin barely stimulated MAP kinase; this reduction was reversed in cells in which c-SRC had been reintroduced. These findings show that activation of MAP kinase by cAMP requires a SRC family kinase that lies downstream of protein kinase A. A role for RAP1, as documented for the β2-adrenergic receptor, is apparently contingent on receptor endocytosis.

the interaction between p21 ras and c-RAF; this results in cAMP-mediated suppression of mitogen-activated protein kinase pathway (1)(2)(3)(4). However, in some cells, agonist occupancy of G s -coupled receptors is associated with both increases in cellular cAMP and with stimulation of MAP kinase. It has been argued that stimulation of MAP kinase is dependent on cAMP and that effectors other than adenylyl cyclase generate the signal that link G s -coupled receptor to MAP kinase activation. Furthermore, in each of the proposed models, the signal to MAP kinase diverges at a different level from the signaling cascade composed of receptor, G s , and adenylyl cyclase/cAMP. (i) The cAMP-dependent signaling mechanism is mediated by the p21 ras -related, monomeric G protein RAP1 that preferentially activates B-RAF (5,6). Loading of RAP1A with GTP requires guanine nucleotide exchange factors, a class of which (Epac) is directly activated by cAMP, i.e. in a manner independent of PKA (7,8). (ii) Alternatively, a role has been invoked for the non-receptor tyrosine kinase SRC, because G␣ s directly binds to and activates SRC in vitro (9). In this case, stimulation of MAP kinase by G s -coupled receptors is independent of cAMP but sensitive to inhibition or genetic ablation of SRC. (iii) A large set of experiments support a third model of MAP kinase activation in which the receptor is endocytosed in a dynamindependent fashion. Here ␤-arrestin functions as adapter protein for the recruitment of SRC and for the assembly of a large signaling complex. This model has been primarily developed with the ␤ 2 -adrenergic receptor (10) and predicts that stimulation of MAP kinase by the receptor is blocked by abrogating dynamin and SRC. (iv) Finally, G protein-coupled receptors may promote transactivation of tyrosine kinase receptors by causing the shedding of matrix-bound growth factors; this effect depends on the activation of matrix metalloproteases (11).
Although G protein-coupled receptors may recruit multiple and redundant pathways to stimulate MAP kinase (12), it is evident that some of the proposed mechanisms are mutually exclusive. It is also difficult to understand why protein kinase A is required for cAMP-dependent stimulation of MAP kinase (13) if GTP-liganded RAP1 is formed by the action of Epac (7,8). In the present study, we have therefore tested the predictions of the four models; we compared the action of receptorindependent elevations of cAMP (by membrane-permeable analogues and forskolin, the direct activator of adenylyl cyclase) with the effect of the A 2A -adenosine receptor. This G s -coupled receptor activates MAP kinase in CHO cells in a manner dependent on cAMP (14). Our experiments show that RAP1 and dynamin-dependent receptor endocytosis are dispensable for receptor and cAMP-dependent activation of MAP kinase; similarly, MAP kinase stimulation cannot be accounted for by transactivation of a receptor tyrosine kinase due to release of a matrix-bound release growth factor. In contrast, SRC (or an SRC-like kinase) plays an essential role, but it is downstream of protein kinase A.

EXPERIMENTAL PROCEDURES
Materials-Adenosine deaminase, basic fibroblast growth factor, 12CA5 anti-hemagglutinin mouse monoclonal antibody, and enzymes for DNA manipulation were from Roche Molecular Biochemicals. CGS21680 was from Tocris Cookson Ltd. (Bristol, UK). Hepes was from Biomol (Munich, Germany). The materials required for SDS-PAGE were from Bio-Rad. Fetal calf serum was from PAA Laboratories (Linz, Austria); Dulbecco's modified Eagle's medium, Opti-MEM medium, horse serum, non-essential amino acids, ␤-mercaptoethanol, and G418 (geneticin) were obtained from Invitrogen. Ham's F-12 medium was from BioConcept (Allschwil, Switzerland). Collagen was from Biomedical Technologies Inc. (Stoughton, MA), Centrifuge tubes and tissue culture plates were from Greiner (Vienna, Austria) and from Corning Costar (Acton, MA). Forskolin, 8-Br-cAMP, L-glutamine, penicillin G, streptomycin, Triton X-100, PMSF, leupeptin, and thrombin were purchased from Sigma. Aprotinin, PP3, and PDBu were from Calbiochem. PP1 was from Alexis Biochemicals (San Diego, CA). The inhibitor of protein kinase A H89 was from Alexis Corp. (Laeufelfingen, Switzerland). The Micro BCA protein assay reagent kit was from Pierce. Buffers and salts were from Merck. Glutathione-Sepharose and protein G-Sepharose was from Amersham Biosciences. Polyclonal rabbit antisera recognizing the diphosphorylated sequence of ERK1 and ERK2 was from New England Biolabs (Beverly, MA). The monoclonal mouse antibody directed against RAP1 was from Transduction Laboratories (Lexington, KY). Polyclonal rabbit antisera recognizing the carboxyl terminus of ERK1/ERK2 and B-RAF were purchased from and from Santa Cruz Biotechnology (Santa Cruz, CA). The 16B12 anti-hemagglutinin mouse monoclonal antibody was generous gift of E. Ogris (Vienna Biocenter). Horseradish peroxidase-conjugated anti-mouse and anti-rabbit immunoglobulin antibodies were from Amersham Biosciences. The immunoreactive bands on nitrocellulose blots were detected by chemiluminescence using SuperSignal chemiluminescence substrate from Pierce. SuperFect polycationic transfection reagent and plasmid preparation kits were from Qiagen (Hilden, Germany). PC12 were from ECACC (Salisbury, UK). SYF and SYF cells, in which c-SRC had been reintroduced via retroviral infection (referred as SYF ϩ c-SRC), were purchased from ATCC (Manassas, VA). The plasmid driving the mammalian expression of RAP1GAP was a gift from P. Polakis and J. L. Bos, which encoding mutated c-SRC was generously provided by R. J. Resnick and D. Shalloway, those encoding wild type and mutated versions of dynamin were kindly provided by C. van Koppen. In order to generate an alternative reporter MAP kinase, we amplified the cDNA encoding a hemagglutinin-tagged version of ERK1/p44 MAP kinase by PCR using the following primer (where underlines indicate the restriction sites for the enzymes shown in parentheses): ERK-5Ј-GCA GGA GCT CCG ATG TAC CCA TAC GAT GTT CCA (SacI); ERK-3Ј GCA TGA ATT CTC GGG GCC TCT GGT GCC CCT GG (EcoRI). The purified PCR fragment was ligated via the SacI-EcoRI restriction site into pGFP-N1 (CLONTECH, BD PharMingen) linking the HA-ERK in-frame to the amino terminus of the GFP protein. The sources for the other plasmids employed has been listed previously (14).
For co-culture of CHO or CHO-A 2A with reporter CHO cells, expressing reporter MAP kinase, CHO cells were transiently transfected with a plasmid expressing HA-tagged reporter MAP kinase. After 24 h, the transfected reporter CHO cells were seeded with CHO-A 2A or control CHO cells (ratio 1:1) and allowed to adhere for 12 h. Thereafter, the cells were rendered quiescent by withdrawing serum for 12 h, and MAP kinase assays were subsequently performed as described below.
Stimulation of MAP Kinase Phosphorylation, Immunoprecipitation, and Immunoblots-Confluent cell layers (in 6-cm dishes) were rendered quiescent by serum starvation for 12-24 h. The starving medium was supplemented with 1 unit/ml adenosine deaminase to remove any endogenously produced adenosine; 30 min prior to MAP kinase assays the medium was again changed against prewarmed (37°C) medium in order to minimize basal activity. If not otherwise indicated, cells were subsequently stimulated by addition of medium containing or lacking agonists and maintained at 37°C for 5 min. Control incubations were carried out in order to verify that the carry-over of dimethyl sulfoxide, which resulted in final concentrations of Յ0.1%, neither affected the basal levels of MAP kinase phosphorylation nor the response to agonists. The exposure to agonists or vehicle was terminated by rapidly rinsing with ice-cold phosphate-buffered saline; thereafter, the dish was immediately immersed in liquid nitrogen; after rapid thawing, cells were lysed by addition of 80 l of lysis buffer (in mM: 50 Tris, 40 ␤-glycerophosphate, 100 NaCl, 10 EDTA, 10 p-nitrophenol phosphate, 1 PMSF, 1 Na 3 VO 4 , 10 NaF, pH adjusted to 7.4 with HCl), 1% Nonidet P-40, 0.1% SDS, 250 units/ml aprotinin, 40 g/l leupeptin. The cellular debris was removed by centrifugation at 10,000 ϫ g for 10 min, and the total protein content was measured photometrically using bicinchoninic acid (Micro-BCA kit, Pierce). Aliquots corresponding to 2.5-5⅐10 4 cells (10 -30 g of protein) were dissolved in Laemmli sample buffer containing 30 mM dithiothreitol and applied to SDS-polyacrylamide gels (monomer concentration 10 -15% acrylamide, 0.26 -0.4% bisacrylamide). MAP kinase phosphorylation was assayed by incubating nitrocellulose blots with an antiserum that recognizes only the dually phosphorylated forms of p42 and p44 MAP kinase; in order to rule out that the differences observed were due to the application of unequal amounts of lysates, control blots were also probed with an antiserum recognizing both the unphosphorylated (inactive) and phosphorylated (active form). The immunoreactive bands were visualized by enhanced chemiluminescence using horseradish peroxidase-linked secondary antibodies. Immunodetection of the other proteins was performed in an analogous manner, using the appropriate antibodies or antisera. In several instances the reporter MAP kinase was first immunoprecipitated using the monoclonal 12CA5 antibody directed against the HA epitope and protein G-Sepharose that had been pre-equilibrated in lysis buffer; HA-tagged or HA-GFP-tagged ERK1 were used interchangeably with equivalent results. Co-immunoprecipitations of HA-tagged RAP1 with B-RAF were done in a similar manner.
Pull-down Assays for the Determination of RAP1 Activation-GST fusion protein of the minimal RAP1-binding domain of RalGDS (ral-RBD, see Ref. 15) were expressed in Escherichia coli (strain BL21DE3); following induction by isopropyl-1-thio-␤-D-galactopyranoside, bacterial lysates were prepared as described. GST fusion proteins were immobilized by incubating bacterial lysates for 1 h at 4°C with GSH-Sepharose pre-equilibrated in RIPA buffer (50 mM Tris⅐HCl, pH 7.5, 150 mM NaCl, 0.5% deoxycholate, 1% Nonidet P-40, 0.1% SDS) supplemented with 2 g/ml aprotinin, 1 g/ml leupeptin, and 1 M PMSF. The Sepharose beads were washed 3 times in order to remove excess GST fusion protein. Cells were prepared for the assay in a similar way as outlined above for MAP kinase assays if not otherwise stated; incubation with agonists was carried out for 5 min followed by rapidly rinsing with ice-cold phosphate-buffered saline and addition of RIPA buffer to achieve cell lysis. Cell lysates were cleared by centrifugation (20,000 ϫ g for 1 min). The resulting supernatants were incubated together with the GSH-Sepharose beads (50 l of a 1:1 slurry containing about 10 g of immobilized GST fusion protein) for 1 h to allow for the association of activated RAP1 with the RalGDS-GST fusion protein. Samples were washed 3 times in RIPA buffer, resuspended in Laemmli sample buffer, and applied to SDS-polyacrylamide gels; RAP1 was visualized using specific antibodies in a dilution of 1:250. If HA-tagged RAP1 was cotransfected, the assay was carried out in a similar way as described except that the transfection procedure preceded the assay and that the immunoblot was done with the 16B12 monoclonal antibody directed against HA tag sequence.
Each experiment was at least carried out three times. Addition of the A 2A -selective agonist CGS21680 resulted in biphasic stimulation of MAP kinase phosphorylation. An early increment in the phosphorylation of the endogenous p42 and p44 (ERK2 and ERK1) isoforms of MAP kinase was seen at 5-10 min and was followed by a sustained phase of activation after 60 -90 min (Fig. 1A). It is evident from Fig. 1B (left panel) that GTP-liganded active RAP1 was also formed with biphasic kinetics. We also verified by co-immunoprecipitation that receptor-dependent activation of RAP1 (by CGS21680 or by thrombin as a positive control) resulted in formation of a complex between of RAP1 with its putative effector B-RAF (Fig. 1B, right panel). RAP1 is under the control of cAMP-dependent exchange factors (7,8). Thus, protein kinase A should be dis-pensable for both activation of RAP1 and stimulation of MAP kinase. This was clearly not the case. Formation of GTP-bound RAP1 was insensitive to the PKA-inhibitor H89 (Fig. 1C, lower panel); in contrast, activation of MAP kinase was suppressed by H89 (Fig. 1C, upper panel). In order to directly activate cAMP-dependent effectors, we also performed a similar set of experiments with the membrane-permeable cAMP analogue 8-Br-cAMP; the results were equivalent to those shown in Fig.  1, A-C, for the A 2A -agonist CGS21680 (data not shown).

Time Course of RAP1 and MAP Kinase Activation following Stimulation by an
Inhibitory Effect of RAP1GAP on A 2A -agonist-stimulated RAP1 Activation but Not on MAP Kinase Activation-The time course of RAP1 activation and the kinetics of MAP kinase phosphorylation were compatible with a cause and effect relationship. Similarly, the fact that following stimulation by an A 2A -agonist, RAP1 associated with B-RAF, which is a MAP kinase kinase kinase and thus, by definition, upstream of ERK1/2. These observations clearly argued for a role of RAP1 in mediating the stimulatory effect of the A 2A -receptor on MAP kinase. However, the protein kinase A inhibitor H89 discriminated between stimulation of MAP kinase phosphorylation and GTP loading of RAP1. This indicated a requirement for protein kinase A rather than for an exchange factor of the Epac family. It has been argued that protein kinase A acted upstream of RAP1 activation, possibly on the RAP1 exchange factor C3G (16). In order to address this discrepancy, we employed RAP1GAP, a member of the GAP family that regulates RAP1 (17). The intrinsic GTPase activity of small G proteins is very low, and it is, in most cases, accelerated by a GTPase-activating protein (GAP) that provides one or more residues required for catalysis. Thus, GAPs switch off the active, GTP-bound state. We transiently expressed an HA-tagged version of RAP1GAP together with epitope versions of reporter RAP1 ( Fig. 2A) and ERK1 (Fig. 2B). Overexpression of HA-tagged RAP1GAP efficiently prevented the accumulation of GTP-bound RAP1 after stimulation of CHO-A 2A cells by the A 2A -agonist ( Fig. 2A, top  row).
Cells that expressed HA-RAP1GAP (middle row in Fig. 2A) synthesized less HA-RAP1 ( Fig. 2A, bottom row). However, GTP-bound RAP1 was below the detection limit, even if HA-RAP1GAP-transfected cells had been stimulated by the A 2Aagonist. In contrast, GTP-loaded RAP1A was detected under basal conditions in vector-transfected control cells; it is evident from a comparison of the level of total (bottom row in Fig. 2A) and GTP-bound RAP1 (top row, Fig. 2A) that the lower level of reporter HA-RAP1 in HA-RAP1GAP-transfected cells clearly still would have sufficed to allow for the detection of receptordependent activation. Although the accumulation of GTPbound RAP1 was completely abrogated by the overexpression of RAP1GAP, it did not have any appreciable effect on the phosphorylation of a co-transfected reporter MAP kinase in response to stimulation by CGS21680 (Fig. 2B, cf. lanes C in vector and HA-RAP1GAP-transfected cells). Similar observations were also made if cells were stimulated with the membrane-permeable cAMP analogue 8-Br-cAMP (Fig. 2B, lanes labeled 8 Br). These results were difficult to reconcile with published data, namely that RAP1GAP attenuated MAP kinase activation by the ␤ 2 -adrenergic receptor (16). We therefore also transiently co-expressed the ␤ 2 -adrenergic receptor in CHO-A 2A cells together with the reporter MAP kinase in the absence and presence of HA-RAP1-GAP. By contrast with MAP kinase stimulation with the A 2A -agonist (and 8-Br-cAMP), RAP1GAP did diminish isoproterenol-stimulated phosphorylation of the reporter MAP kinase construct (Fig. 2C, cf. lanes labeled I and C for isoproterenol and CGS21680, respectively). Finally, we ruled out that any differences observed are due to a variation in the amount of reporter MAP kinase in the immunoprecipitates because comparable levels were detected with an antiserum that recognizes holo-ERK1/2 (bottom rows in Fig. 2, B and C).
Involvement of Endocytosis and of Transactivation-Activation of MAP kinase by ␤ 2 -adrenoceptors is dependent on receptor internalization (18). In order to test if this was also true for the A 2A -receptor, CHO-A 2A cells were transiently co-transfected with dynamin K44A, a dominant negative suppressor of dynamin-dependent formation of endocytotic vesicles (19), the reporter MAP kinase, and the ␤ 2 -adrenergic receptor. As a control, we employed the plasmid encoding wild type dynamin (which did not alter the response to the agonists CGS21680 and isoproterenol, not shown). Fig. 3A shows that the response to the ␤ 2 -agonist isoproterenol was blunted in cells that expressed dynamin K44A; in contrast, dynamin K44A did not have any appreciable effect on the stimulation of reporter MAP kinase by the A 2A -adenosine receptor (Fig. 3A, lanes C). Two additional manipulations highlighted the fundamental difference between activation of ERK1/2 by the ␤ 2 -adrenergic receptor and A 2A -adenosine receptor; neither overexpression of the carboxyl terminus of the ␤-adrenergic receptor kinase nor of phosducin, which both act as scavengers for free ␤␥, impaired A 2A -receptor-stimulated MAP kinase activation (data not shown). In contrast, receptor-generated G␤␥ is important for MAP kinase activation by the ␤ 2 -adrenergic receptor (20).
Taken together the observations indicated that there was a fundamental difference in the mechanism by which the A 2Aadenosine receptor and the ␤ 2 -adrenergic receptor impinged on MAP kinase. We have therefore explored if the A 2A -adenosine receptor signaled to MAP kinase via transactivation. Originally, receptors with tyrosine kinase activity were proposed to act as scaffolds for the assembly of signaling complexes and to thereby support MAP kinase activation by G protein-coupled receptors, and this was referred to as transactivation; more recently, the emphasis has shifted to the ability of G proteincoupled receptors to promote the release of cell surface-bound growth factors, e.g. heparin-binding EGF, via activation of a matrix metalloprotease (11). By definition, this type of stimulation is paracrine stimulation; we have therefore used a coculture system that was analogous to the one originally employed by Prenzel et al. (11; see also ref. 21); CHO-A 2A or control CHO cells were mixed at 1:1 ratio with reporter CHO cells that harbored an epitope-tagged MAP kinase and seeded at high density. Growth factor release and the ensuing para-

FIG. 2. Effect of RAP1GAP overexpression on activation of RAP1 (A) and of MAP kinase phosphorylation by CGS21680 and 8-Br-cAMP (B) and by isoproterenol (C) in CHO-A 2A cells. A,
CHO-A 2A were transiently transfected with plasmids encoding HAtagged RAP1GAP or empty vector with an HA-tagged RAP1 (1.5 g/ 6-cm dish each). After the cells had reached confluency (24 h), they were maintained in medium containing adenosine deaminase in the absence of serum for 24 h. Subsequently cell were incubated with 1 unit/ml adenosine deaminase in the absence (lanes A) and presence of 0.5 M CGS21680 (lanes C) for 5 min. The GST pull-down was done as indicated under "Experimental Procedures"; aliquots (30%) of each sample (top row) and 20 g of the corresponding cellular lysate (middle and bottom rows) were applied to SDS-polyacrylamide gels. Immunoblotting (IB) was performed with monoclonal 16B12 anti-HA antibody. Data are from a representative experiment that was reproduced twice. B and C, CHO-A 2A cells were transiently transfected with plasmids encoding HA-tagged RAP1GAP or empty vector with an HA-tagged p44 MAP kinase (1.5 g/6-cm dish each) and ␤ 2 -adrenoreceptor (or the appropriate empty vector; also at 1.5 g/dish; C) and subsequently maintained as in A. After serum starvation, the cells were incubated with adenosine 1 unit/ml deaminase (lanes A), 0.5 M CGS21680 (lanes C), 100 M 8-Br-cAMP (lanes 8 Br), or 1 M isoproterenol (lanes I in C) for 5 min. The HA-tagged reporter MAP kinase was immunoprecipitated (IP) from cellular lysates as outlined under "Material and Methods." Aliquots (30%) of the immunoprecipitates were applied onto SDS-polyacrylamide gels. Immunoblotting was done with antisera directed against the bisphosphorylated MAP kinase (IB, anti P-ERK) and the carboxyl terminus of MAP kinase (IB, anti-ERK). crine stimulation ought to be detected as A 2A -receptor-dependent stimulation of MAP kinase in the reporter cells. This was clearly not observed; on the contrary, the reporter MAP kinase was phosphorylated to a negligible extent irrespective of whether the co-culture contained cells endowed with or lacking the A 2A -adenosine receptor (Fig. 3B, top row). In contrast, we readily detected the phosphorylation of endogenous MAP kinase isoforms in the cellular lysates if these were prepared from co-cultures containing CHO-A 2A cells; as expected, MAP kinase was irresponsive to the A 2A -agonist in the control cocultures that lacked A 2A -receptor-bearing cells (Fig. 3B, bottom  row). We therefore conclude that the A 2A -adenosine receptor relies on components that are intrinsically present within the stimulated cell to activate MAP kinase and that there is no evidence for the involvement of an additional, extracellular signal.

Blockage of SRC Family Kinases or Ablation of SRC Blunts MAP Kinase Stimulation by the A 2A -adenosine Receptor and cAMP-
The observations presented so far suggested that cAMPdependent activation of PKA played an essential role in linking MAP kinase to the A 2A -adenosine receptor. However, G␣ s may have effectors other than adenylyl cyclase isoforms; the nonreceptor tyrosine kinase SRC has, in particular, been reported as direct target of G␣ s (9). If the signaling pathway A 2A -adenosine receptor/G s /SRC was upstream of MAP kinase activation, inhibition of SRC (or related kinases) ought to discriminate between receptor-and cAMP-dependent MAP kinase stimulation. We therefore tested incubated CHO-A 2A cell in the presence of 1 M PP1, an inhibitor of SRC family kinases. As can be seen in Fig. 4A, preincubation with PP1 specifically abrogated the ability of the A 2A -agonist CGS21680 (lanes C) and of forskolin (lanes F) to stimulate MAP kinase phosphorylation. In control, PP1 did not blunt the effect of basic fibroblast growth factor (lanes FGF in Fig. 4A) and of thrombin (lanes T in Fig.  4A). The following control experiments were carried out to document the physiological relevance and the specificity of these findings. (i) Instead of CHO-A 2A cells, in which the adenosine receptor was heterologously expressed, we used PC12 cells, in which the A 2A -adenosine receptor is endogenously expressed and signals to MAP kinase (22). PP1 blunted A 2Aagonist-and forskolin-induced phosphorylation of MAP kinase (Fig. 5B). (ii) Cells were also incubated in the presence of PP3, an analog that is inactive against SRC family kinases (but inhibits the activity of the EGF receptor kinase with an IC 50 of 2.7 M; Ref. 23). PP3 neither blocked MAP kinase stimulation by forskolin and CGS21680 in PC12 cells (Fig. 4B) nor in CHO-A 2A cells (Fig. 5B). (iii) When assayed in vitro in a kinase assay, PP1 inhibits enzymatic activity of SRC family kinases with an IC 50 value in the low nanomolar range (6 and 5 nM for FYN and LCK, respectively). However, the spectrum of kinases inhibited by PP1 is not strictly limited to SRC family kinases. PP1 also inhibits the activity of EGF receptor kinases, ZAP-70 and JAK2, but the concentrations required are substantially higher (i.e. in the submicromolar to micromolar range; Ref. 23). We generated concentration-response curves to test if the effect of PP1 on MAP kinase phosphorylation reflects a specific inhibition of SRC family kinases. A representative blot is depicted in Fig. 5A (top); we estimated from densitometric quantitation that the IC 50 was about 6 nM (see Fig. 6E). As mentioned earlier, over a similar concentration range, the inactive analog PP3 did not affect MAP kinase stimulation (Fig. 5B). (iv) As further proof of specificity, we exploited the fact that, in HEK293 cells, the A 2A -adenosine stimulates MAP kinase in a manner independent of cAMP (14). Accordingly, in stably transfected HEK A 2A cells (i.e. HEK293 cells endowed with the A 2A -adenosine receptor), PP1 did not blunt the response to the A 2A -agonist CGS21680 (Fig. 5A, bottom).
SRC family kinases comprise a large group of non-receptor kinases. They can be divided into two classes as follows: those that are widely expressed such as SRC, FYN, YES; and those with more restricted patterns of expression, such as LCK, HCK, FGR, BLK, and LYN (24). Experiments with NIH3T3 cells verified that these murine fibroblasts responded to elevations of cAMP by forskolin with an increase in MAP kinase phosphorylation (Fig. 6A, top), a response seen with the membrane-permeable cAMP analog 8-Br-cAMP and upon transient expression of the human A 2A -adenosine receptor (not shown). Furthermore, in NIH3T3 fibroblasts, PP1 inhibited MAP kinase stimulation by forskolin (Fig. 6A) or 8-Br-cAMP (not shown). We therefore employed murine embryonic fibroblasts that were devoid of SRC, YES, and FYN (SYF cells) to evaluate the importance of SRC family kinases in the cAMP-dependent activation of MAP kinase. In SYF cells, we observed a pronounced reduction in the capacity of forskolin to activate MAP kinase (Fig. 6B). In cells, in which c-SRC was reintroduced via retroviral infection (SYF ϩ c-SRC), the response to forskolin was restored. The stimulation by the phorbol ester PDBu (lanes P in Fig. 6) was used as an internal control; it is evident that the response to PDBu was reasonably similar in NIH3T3 cells, SYF cells, and SYF ϩ c-SRC cells.
Although it is difficult to compare the levels of signaling molecules across different cell types, it is worth pointing out that the levels of SRC in SYF ϩ c-SRC cells were not vastly different from those seen in PC12 cells (Fig. 6C), i.e. in a cell line in which MAP kinase stimulation via cAMP is a physiological response controlled by the endogenously expressed A 2Aadenosine receptor. Accordingly, in SYF ϩ c-SRC inhibition by PP1 of cAMP-dependent stimulation of MAP kinase was observed over a concentration range (Fig. 6D) that was similar to that observed in PC12 cells (Fig. 5A, middle). In fact, a comparison of all concentration-response curves showed that PP1 inhibited cAMP-dependent formation of phosphorylated MAP kinase with essentially identical IC 50 values (range 5-10 nM) in CHO-A 2A , PC12, and SYF ϩ c-SRC irrespective of whether cAMP accumulation had been induced by activation of the A 2A -adenosine receptor or by direct activation of adenylyl cyclase (Fig. 6E).
Protein kinase A phosphorylates SRC on Ser 17 (25). Accordingly, we employed c-SRC(S17A), a mutated version of SRC in which Ser 17 was replaced by Ala, to test whether PKA-dependent phosphorylation of SRC was required for MAP kinase stimulation. SYF cells were transiently co-transfected with a plasmid encoding c-SRC or c-SRC(S17A) and a vector driving the expression of the reporter MAP kinase. As expected, addition of forskolin resulted in increased levels of phosphorylated MAP kinase in cells that expressed c-SRC (Fig. 6F, top); in contrast, in cells that contained c-SRC(S17A) forskolin failed to stimulate MAP kinase phosphorylation over the unstimulated con- for immunodetection of phosphorylated p44 and p42 MAP kinase (ERK1 and ERK2) as described in the legend to Fig. 1. E, concentrationresponse curves for the inhibitory effect of PP1 on MAP kinase/ERK trol (cf. lanes F and U in Fig. 6F, top). The observed differences were not accounted for by different levels of reporter MAP kinase or SRC expression (Fig. 6F, middle and bottom). DISCUSSION It has long been appreciated that intracellular cAMP levels impinge on cell growth and cell survival (26); the effects range from inhibition of proliferation and apoptosis to trophic stimulation and differentiation. Because of the many targets of cAMPdependent protein kinase, it has been notoriously difficult to pinpoint the underlying molecular mechanisms. More recently, a plausible mechanism was proposed that suggested that MAP kinase stimulation by cAMP was elicited via activation of RAP1 which, in its activated GTP-loaded form, combined with B-RAF (6). This model is attractive because RAS and RAP1 are closely related (with RAP standing for RAS proximate) and because in vitro B-RAF can be activated by RAP1 (5); thus, RAP1-dependent regulation of B-RAF is congruent with RAS-dependent activation of c-RAF (ϭ RAF-1). Finally, the discovery of Epacexchange factors that are direct targets of cAMP seemingly closed the gap between adenylyl cyclase and RAP1 (7,8) such that the cascade upstream of Mek1 can be delineated as consisting of receptor G s adenylyl cyclase/cAMP-Epac-RAP1-B-RAF. However, our observations are inconsistent with this model; the experiments relied on the use of both the A 2Aadenosine receptor, a prototypical G s -coupled receptor (27), which stimulates MAP kinase in CHO cells via cAMP and G s (14,28), and on stimuli that acted downstream of the receptor and G␣ s (by employing forskolin as a direct activator of adenylyl cyclase isoforms and the membrane-permeable cAMP analog 8-Br-cAMP). The rationale for using these two approaches is the evidence that receptor-generated cAMP may be compartmentalized and elicit effects distinct from those produced by downstream stimuli (29). Our data unequivocally show that protein kinase A and an SRC family kinase are essential components of the signaling cascade that links cAMP to MAP kinase. In contrast, RAP1 is dispensable for cAMP-dependent stimulation of MAP kinase irrespective of the source of cAMP because the deactivation of RAP1 by RAP1GAP did not affect stimulation of ERK1/2.
Earlier work (30) also found that the extent of RAP1 activation did not show any correlation to the level of MAP kinase phosphorylation. We therefore conclude that cAMP generated in response to a prototypical G s -coupled receptor (the A 2Aadenosine receptor) or to forskolin (a direct activator of adenylyl cyclases) does lead to complex formation between RAP1 and B-RAF but that these complexes are irrelevant to activation of ERK1/2 by cAMP, possibly because complexes formed between RAP1 and B-RAF are segregated from Mek, the kinase upstream of MAP kinase. It is not clear which biological response is elicited by this reaction, but it is evident that RAF kinases have targets other than Mek1 (31,32).
Our work also confirmed that MAP kinase stimulation by the ␤ 2 -adrenergic receptor was abrogated if RAP1 was clamped in the inactive conformation by overexpression of RAP1GAP (16). These findings are in marked contrast to the observations obtained in parallel by directly raising cAMP or by activating the A 2A -adenosine receptor. This seeming contradiction can be resolved as follows: the ␤ 2 -adrenergic receptor is rapidly (i.e. within minutes) internalized, and dynamin-dependent internalization is important for stimulation of MAP kinase (18). In contrast, the A 2A -adenosine receptor is not internalized to an appreciable extent over the time frame required to induce MAP kinase phosphorylation 2 (see also Ref. 34); accordingly, the interfering mutant dynamin-K44A did not affect the response to A 2A -agonist stimulation. It is also worth noting that, when visualized in living cells, active GTP-bound RAP1 mainly resides on intracellular vesicles (35). We therefore propose that the role of RAP1 in MAP kinase activation is indirect, possibly by controlling the assembly of signaling complexes on intracellular vesicles, and contingent on receptor internalization.
In the three cell lines investigated (CHO-cells, PC12 cells, and NIH3T3/murine embryonic fibroblasts) in which elevations of cAMP caused MAP kinase phosphorylation, the response was inhibited by the inhibitor of SRC family kinase PP1 (but not by its inactive analog PP3); obviously, these findings argue for a role of SRC or SRC-like kinases upstream of MAP kinase. This argument is further supported by three sets of observations: (i) in SYF cells that are deficient in the three ubiquitous SRC-like kinases, the MAP kinase response to elevated cAMP was strongly impaired; (ii) expression of SRC restored the ability of forskolin to stimulate MAP kinase phosphorylation; (iii) the IC 50 values of PP1 were similar in the three cell lines investigated in which cAMP promoted MAP kinase phosphorylation regardless of whether cAMP was raised by the endogenous receptor (PC12 cells) or the heterologously expressed receptor (CHO-A 2A ) or by forskolin (SYF ϩ c-SRC). Several distinct sites of action have been proposed for SRC (or SRC-like kinase) in the downstream cascade controlled by G s -coupled receptors. SRC binds directly to the ␤ 3 -adrenergic but not the ␤ 2 -adrenergic receptor, and this interaction supports activation of MAP kinase (36). However, in adipocytes, i.e. the cellular background in which the ␤ 3 -adrenergic receptor occurs physiologically, the ability of the ␤ 3 -receptor to stimulate MAP kinase is fully accounted for by its ability to raise cAMP (37). It is worth pointing out that, for the A 2A -adenosine receptor, this was also true not only in CHO cells (where the receptor had been introduced by stable transfection) but in PC12 cells, which endogenously express the receptor. With the ␤ 2 -adrenergic receptor, ␤-arrestin serves as a docking site for SRC (10), and SRC-dependent phosphorylation of dynamin is essential for MAP kinase activation (38). It is evident that these mechanisms of recruiting are restricted to receptors that can directly bind SRC-like kinases or that require dynamin-dependent internalization. SRC (9) and the SRC-like kinase LCK (39) have been reported to be directly activated by G␣ s . This provides for a more general mechanism of activation. In endothelial cells (38) and HEK293 cells (14), stimulation of MAP kinase is independent of G␣ s . Accordingly, the SRC inhibitor PP1 did not block MAP kinase activation. However, direct activation of SRC (or SRC-like kinase) by G␣ s fails to explain (i) the ability of forskolin and 8-Br-cAMP to stimulate MAP kinase in an SRC-dependent manner and (ii) the ability of the protein kinase inhibitor H89 to block stimulation of MAP kinase. It has long been known that the major site of serine phosphorylation is on Ser 17 of SRC (25); this is also the site targeted by protein kinase A (41), and PKA-dependent phosphorylation of SRC on Ser 17 promotes its release from the plasma membrane (33,42). This may facilitate signaling by redirecting SRC to a subset of specific substrates (40). Thus, the action of PKA may be accounted for, at least in part, by a direct effect on SRC (or an SRC-like kinase). There are, however, examples of specialized cells in which MAP kinase stimulation by cAMP does not rely on PKA (and is independent of RAP1) (43). We also do not intend to claim that the link between protein kinase A and SRC (or a SRC-like kinase) is the only mechanism that accounts for MAP kinase stimulation; there was, for instance, still a slight stimulation of MAP kinase by forskolin in SYF cells. However, our data provide firm evidence for an important role of protein kinase A upstream of SRC (or SRC-like kinases) and that this is an essential component in the signaling cascade that links G s -coupled receptors to stimulation of MAP kinase.