cAMP Analog Mapping of Epac1 and cAMP Kinase DISCRIMINATING ANALOGS DEMONSTRATE THAT Epac AND cAMP KINASE ACT SYNERGISTICALLY TO PROMOTE PC-12 CELL NEURITE EXTENSION*

Little is known about the relative role of cAMP-de-pendent protein kinase (cAPK) and guanine exchange factor directly activated by cAMP (Epac) as mediators of cAMP action. We tested cAMP analogs for ability to selectively activate Epac1 or cAPK and discriminate between the binding sites of Epac and of cAPKI and cAPKII. We found that commonly used cAMP analogs, like 8-Br-cAMP and 8-pCPT-cAMP, activate Epac and cAPK equally as well as cAMP, i.e. were full agonists. In contrast, 6-modified cAMP analogs, like N 6 -benzoyl-cAMP, were inefficient Epac activators and full cAPK activators. Analogs modified in the 2 (cid:1) -position of the ribose induced stronger Epac1 activation than cAMP but were only partial agonists for cAPK. 2 (cid:1) - O -Alkyl substitution of SWISS-MODEL

With the exception of cyclic nucleotide gated channels in specialized cells like olfactory neurons (1), the only known direct cAMP effector in mammalian cells was, until recently, the cAMPdependent protein kinase (cAPK), 1 whose mechanism of activa-tion and structure has been studied in detail (2)(3)(4).
It has been questioned whether cAPK is the sole mediator of cAMP action (5,6). The discovery (7,8) of Rap guanine nucleotide exchange factors directly activated by cAMP (Epac1 and Epac2) raised the possibility that effects hitherto attributed to activation of cAPK were in fact mediated by Epac. Epac1 has an N-terminal DEP (dishevelled, Egl-10, pleckstrin) domain, involved in membrane docking (9) and cell adhesion (10), a cAMP binding domain (CNBD), a Ras exchange motif, and a guanine nucleotide exchange factor homology domain (see Fig.  1A). Epac2 has been crystallized recently (11). It has an additional CNBD (8), which is dispensable for cAMP-induced Rap activation (9,12).
There is increasing but still modest knowledge about the biological consequences of cAMP activation of endogenous Epac in intact cells. Epac appears not to mediate modulation of extracellular signal-regulated kinase activity by cAMP (13) but may mediate the stimulation by cAMP of exocytosis (14,15) and the calcitonin-induced H,K-ATPase activation in kidney cells (16) and may modulate integrin-mediated cell adhesion (17). Less is known about how cAMP signaling through cAPK and Epac might be integrated, but there is evidence that overexpressed Epac1 can activate Akt/protein kinase B, whereas stimulation of cAPK inhibits Akt/protein kinase B (18), suggesting that Epac and cAPK serve opposite functions. The aim of the present study was to develop and identify cyclic nucleotide analogs that could help distinguish the roles of cAPK and Epac1 in intact cells. We obtained first a detailed map of the cAMP binding site of Epac1 using more than 50 analogs, many of which are novel. We studied next the ability of selected analogs to activate Epac and cAPK in vitro. We found that 2Ј-O-alkyl-modified cAMP analogs were only partial agonists of cAPK activation, while being stronger than cAMP itself as activators of Epac. On the other hand, several 6-modified cAMP analogs were full cAPK agonists and poor agonists of Epac activation, even if they bound well to Epac.
Intact cell studies confirmed the cAPK specificity of 6-modified analogs and the Epac specificity of 2Ј-O-Me-cAMP analogs like 8-pCPT-2Ј-O-Me-cAMP. Using these analogs as tools we showed that Epac acted synergistically with NGF to promote neurite extension in PC-12 rat pheochromocytoma cells, a model for neuronal differentiation (19,20). Activation of Epac sensitized the cells toward cAPK. This is the first demonstration that Epac and cAPK act in synergy to mediate a cAMP effect.
Epac and cAPK Purification-Recombinant GST-hEpac1fl, GST-hEpac12-329, GST-hEpac1149 -318, GST-hEpac1149 -881, GST-hRI␣, and His-hRII␣ were expressed in E. coli BL21 cells. The GST fusion proteins were adsorbed to GST-agarose and either eluted with glutathione, or the GST cleaved off in situ with thrombin (7,24 . Panel D shows the saturation of GFP-Epacfl (Ⅺ) and GFP-Epac149 -881 (‚) as a function of free [ 3 H]cAMP, as assayed by scintillation proximity assay. The proteins were immobilized to GSH-coated wells of a microplate with solid scintillant (flashplate), and the radioactivity close to the wall (bound to immobilized Epac) was determined. The data points represent the mean Ϯ S.E. of at least four separate experiments for Epac149 -318 and for Epac149 -881 and the mean of two experiments for Epacfl. The concentration of [ 3 H]cAMP at which half-maximal saturation was observed (apparent K D ) is indicated in panels C and D. 318 was a gift from Drs. A. Krä mer and A. Wittinghofer (25). The proteins were further purified by size exclusion fast protein liquid chromatography as described (24). The catalytic subunit of cAPK was prepared as described (24 Determination of the cAMP Analog Binding Affinity for Sites A and B of RI and RII and for Epac1-The equilibrium inhibition constant of binding (K i ) of cAMP analog was determined by competitive displacement of [ 3 H]cAMP binding to R or Epac. The analog affinity relative to cAMP (KЈ i analog) is K i cAMP/K i analog. The determination of KЈ i analog for Epac was, in principle, as described previously (28) and validated for RI and RII.

In Vitro Assay of Epac-induced Rap Activation and cAMP-dependent
Protein Kinase Activity-Cyclic AMP analogs were tested for in vitro Rap1 or Rap2 activation by determination of their effect on the rate of Epac-induced fluorescent GDP analog (3Ј-O-(N-methylantraniloyl)-GDP) exchange from Rap (9). The determination of cAMP analog effects on protein kinase activity was routinely by incubation for 40 min at 37°C in Buffer A with 1 mM [␥-32 P]ATP, using kemptide as substrate (24).
Assay of cAMP Analog Actions in Intact Cells-The activation of Rap was determined by GST pull-down assay as described (29), except that NaF was omitted, and 5 mM Mg(CH 3 COO) 2 added in lysis and wash buffers. The amount of immobilized Ral-GDS-RBD-GST protein (where GDS is GDP dissociation stimulator and RBD is Ras binding domain) used to capture Rap1-GTP was in excess to ensure that variation in pipetting error of the slurry would not affect the pull down of Rap-GTP. Phospho-CREB was determined by Western blotting using the Ab5322 (a kind gift from Dr. M. Montminy, Salk Institute, La Jolla, CA) and compared with the amount of total CREB using a non-discriminating CREB antibody (number 9192; www.cellsignal.com). Primary dog thyroid cells were cultured and assessed for rounding as described earlier (5,30). Rat pheochromocytoma PC-12 cells were cultured in RPMI with 10% horse serum and 5% fetal calf serum. The cells were kept for 20 h in intact medium after seeding and 1 h with low (5% horse and 2.5% fetal calf) serum, before being stimulated with cAMP analog or NGF for 3 to 72 h, and fixed for evaluation by phase and differential interference contrast microscopy (31).
Molecular Modeling-The structures of the CNBDs A and B of RI␣ (Protein Data Bank accession code 1RGS) (4) were used as templates to construct a structural model of the CNBD of Epac1. The model obtained using the Homology module of InsightII 2000 (MSI found at www.msi.

TABLE I Affinity of purine base-modified cAMP analogs the for Epac1 and site A and B of RI␣ and RII␣
The table shows the affinity of cAMP analogs relative to cAMP (i.e. KЈ i , which is K i cAMP/K i analog) for Epac1 and site A and B of RI␣ (AI and BI) and RII␣ (AII and BII). The data were obtained by competitive isotope displacement, as detailed under "Experimental Procedures." The results represent the mean of three determinations, ranging Ϯ 15% from the average.

Binding of [ 3 H]cAMP and cAMP
Analogs to Full-length Epac1, the CNBD Fragment of Epac, and Sites AI, BI, AII, and BII of cAPK-The isolated CNBD (Epac149 -318) of Epac is well expressed and does not aggregate in the presence of cAMP, unlike Epacfl and Epac149 -881 (25). To know whether the binding properties of CNBD are relevant for full-length Epac, we compared their affinity for cAMP. The isolated CNBD (Epac149 -318) of Epac1 bound [ 3 H]cAMP with an apparent K D of 2.9 M (Fig. 1, B and C) at close to physiological pH (7.2) and ionic strength, as determined by the time-honored size exclusion gel chromatography method (27). To prevent aggregation, we anchored GST-Epacfl and GST-Epac149 -881 to GSHcoated plates with intrinsic scintillant and determined [ 3 H]cAMP binding by scintillation proximity assay. GST-Epacfl and GST-Epac149 -881 bound [ 3 H]cAMP with an apparent K D of 2.8 M (Fig. 1D).
We conclude that the CNBD of Epac has the same affinity for cAMP whether in a 170-residue peptide (Epac149 -318) or included in the full-length Epac molecule. Cyclic AMP analog mapping of the binding sites of Epac1, RI and RII, was undertaken to probe for differences between the Epac and R subunit binding sites. We showed first that the estimated KЈ i of cAMP analogs for binding to Epac149 -318 was similar whether assayed by the routine ammonium sulfate precipitation method or the equilibrium binding chromatography assay (Fig. 2).
Mapping data for analogs modified only in the adenine moiety are shown in Table I. We noted that the commonly used cAPK activators 8-Br-cAMP, 8-AHA-cAMP, and particularly 8-pCPT-cAMP had higher affinity than cAMP itself for Epac. No analog was severely restricted from binding to Epac, whereas several had very low KЈ i for binding to one or more of the binding sites of RI␣ (AI, BI) or RII␣ (AII, BII). All the R subunit binding sites discriminated better than Epac against cGMP (Table I). The modestly (12-fold) decreased KЈ i of cGMP for Epac was because of a combined effect of the introduction of 2-NH 2 and 6ϭO into the cAMP molecule (Table I). The Sp-and Rp-diastereoisomers of cAMPS bound to Epac with a relative affinity similar to that for the cAPK sites (Table II). The presumed cAPK-specific agonists Sp-8-pCPT-cAMPS and Sp-5,6-DCl-cBIMPs (33,34) bound to Epac with an affinity similar to that of cAMP itself (Table II).
The loss of the 2Ј-OH of cAMP was more detrimental for binding to cAPK than to Epac, because both 2Ј-deoxy-and 2Ј-O-alkyl-cAMP had higher KЈ i for Epac than for RI or RII (Table II). A more detailed survey of 8-substituted analogs of 2Ј-O-Me-cAMP indicated that the highest affinity was achieved when the S-phenyl ring was substituted in the para-position with a polar group, like Cl Ϫ , HO Ϫ , O-methyl (Table II), or F Ϫ (not shown). All cAMP analogs modified in the 6-position of the adenine ring, including cPuMP, 6-Cl-cPuMP, cIMP, cGMP, 6-MB-cAMP, 6-Bnz-cAMP, and to a lesser extent 6-Phe-cAMP, were only partial Epac agonists (data for 6-Bnz-cAMP are shown in Fig. 6A). Furthermore, the partial stimulation was observed at a higher concentration than expected from the analog affinity for free Epac (not shown).

Comparison of cAMP Analogs as Modulators of cAPK and Epac Activity in Vitro and in Intact Cells
In fibroblasts with enforced expression of Rap1 and Epac1, the analogs 6-MB-cAMP and 6-Bnz-cAMP stimulated strongly the phosphorylation of the cAPK substrate CREB but failed to activate Rap1 (not shown). Preferential stimulation by 6-modified cAMP analog of CREB phosphorylation was observed also in primary dog thyrocytes. In these cells rounding is a specific reaction to cAPK (5). Cell rounding was induced by 6-MB-cAMP (Fig. 7A) and 6-Bnz-cAMP (not shown) without Rap1 activation (Fig. 7A). In contrast, 8-pCPT-cAMP and the adenylate cyclase activator forskolin induced Rap activation, CREB phosphorylation, and cell rounding, whereas 8-pCPT-2Ј-O-Me-cAMP only induced Rap activation (Fig. 7A). We conclude that 6-modified cAMP analogs, notably 6-Bnz-cAMP and 6-MB-cAMP, may be useful to preferentially activate cAPK in intact cells.
It was of interest to know whether Rp-cAMPS analogs could be used to inhibit Epac, and we used Rp-8-p-CPT-cAMPS to test this hypothesis, because this compound had a high affinity for Epac (Table II). Rp-8-pCPT-cAMPS is an inhibitor of both cAPKI and cAPKII in intact cells (33,35). It was a very weak Epac agonist in vitro (Fig. 6B) and did not activate Rap in intact thyrocytes (Fig. 7A). Rp-8-pCPT-cAMPS had a weak ability to counteract 8-Br-cAMP-induced Rap activation in vitro (Fig. 6B) although it could inhibit 8-pCPT-cAMP-induced Rap activation in thyrocytes when present at high concentration. The compound was a stronger inhibitor of CREB phosphorylation and rounding than of Rap activation in thyrocytes (Fig.  7, B and C). This suggests that Rp-8-pCPT-cAMPS would have to be modified to act as a specific Epac antagonist in intact cells. Rp-8-Br-cAMPS, which inhibits cAPKI in intact cells (35), was a weak partial Epac agonist in vitro (Fig. 6B) and failed to inhibit either basal (Fig. 7A) or 8-pCPT-cAMP-induced (not shown) Rap activation in intact thyroid cells or fibroblasts. We conclude that presently available Rp-cAMPS analogs are unable to preferentially inhibit Epac and may be more potent antagonists of cAPK than Epac.

Use of cAMP Analogs to Elucidate the Roles of cAPK and Epac in Mediating cAMP-induced PC-12 Cell Neurite Extensions and Rap1
Activation-It is uncertain whether Epac or cAPK is responsible for the cAMP-induced Rap1 activation in PC-12 cells (20). The Epac activator 8-pCPT-2Ј-O-Me-cAMP activated Rap1 without stimulating CREB phosphorylation. The cAPK activator 6-Bnz-cAMP stimulated CREB phosphorylation but did not significantly activate Rap1 (Fig. 7D). We conclude that Epac stimulation can activate Rap1 in the absence of cAPK activation.
To ensure that 6-Bnz-cAMP and not 8-pCPT-2Ј-O-Me-cAMP acted via cAPK also with respect to neurite extensions, we tested them in the presence of the ATP site-directed cAPK inhibitor H-89 (36,37) and a combination of Rp-cAMPS and Rp-8-Br-cAMPS. These Rp analogs have lower affinity for Epac than Rp-8-pCPT-cAMPS (Table II) and should therefore be more cAPK selective. Furthermore, when combined, they inhibit both cAPKI and cAPKII, because Rp-8-Br-cAMPS is an inhibitor of cAPKI and Rp-cAMPS of cAPKII in intact cells (35). The cAPK antagonists inhibited the action of 6-Bnz-cAMP much more strongly than the action of 8-pCPT-2Ј-O-Me-cAMP and failed to affect the NGF action (Fig. 9, A and C). The cAMP analog 8-pCPT-cAMP is expected to activate both cAPK and Epac. It had an effect comparable with that of 8-pCPT-2Ј-O-Me-cAMP and 6-Bnz-cAMP combined and was inhibited by Rp-cAMPS analogs to a similar extent as 8-pCPT-2Ј-O-Me-cAMP and 6-Bnz-cAMP in combination (Fig. 9C). Similar results were obtained with 8-Br-cAMP (data not shown).
We considered next whether 8-pCPT-2Ј-O-Me-cAMP could act by binding to an as yet uncharacterized cAMP receptor. We tested therefore whether cAMP analogs induced neurite extension with potency expected from their affinity for Epac1. The analog 8-pCPT-2Ј-deoxy-cAMP had 27-fold lower affinity than 8-pCPT-2Ј-O-Me-cAMP for Epac and failed to induce neurite extensions. 8-Br-2Ј-O-Me-cAMP had 5-fold lower affinity and promoted neurite extensions only weakly and at high (Ͼ0.5 mM) concentrations (not shown). 8-pHPT-and 8-pMeOPT-2Ј-O-Me-cAMP bound to Epac1 with 50% higher affinity than 8-pCPT-2Ј-O-Me-cAMP (Table II) and were slightly more potent than 8-pCPT-2Ј-O-Me-cAMP in inducing neurite exten- FIG. 7. Effect of cAMP analogs on intact cell activation of Rap1, phosphorylation of CREB, and rounding. Panels A-C show total Rap1, active Rap1 (Rap1-GTP), phospho-CREB (P-CREB), and total CREB in primary dog thyrocytes. Rp-8-pCPT-cAMPS was present for 60 min before addition of 8-pCPT-cAMP (panels B and C). Otherwise, analogs were present for 10, 30, or 60 min for the determination of Rap activation, rounding, and P-CREB, respectively. The concentration of forskolin was always 10 M. For panels A and D all cAMP analogs were at 0.5 mM. Thyrocyte rounding was determined by phase contrast microscopy. 3ϩ, Ͼ90% of cells rounded; 2ϩ, 30-60% cell rounding; 0, less than 4% rounded cells. For PC-12 cells (panel D) analogs were present for 30 min before cell lysis. sions (Fig. 9D). We conclude that 8-pCPT-2Ј-O-Me-cAMP acted via Epac1 or a cAMP receptor with binding properties nearly identical to Epac1.
It could be argued that 8-pCPT-2Ј-O-Me-cAMP had an extracellular action, e.g. by binding to a surface receptor. To test this possibility we prepared the benzyl ester pro-drug of 8-pCPT-2Ј-O-Me-cAMP. This inactive compound penetrates cells better than the parent compound and is converted to the parent compound within the cell (38). It was more potent than the parent compound as inducer of neurite extensions (Fig. 9E). We conclude that Epac agonists like 8-pCPT-2Ј-O-Me-cAMP most likely acted intracellularly.
It was finally considered whether 8-pCPT-2Ј-O-Me-cAMP acted through inhibition of cyclic nucleotide phosphodiesterase leading to increased cGMP and activation of cGMP targets. We found no inhibition of the action of 8-pCPT-2Ј-O-Me-cAMP by the cGMP-dependent protein kinase inhibitor Rp-8-pCPT-cGMPS, and we failed to promote neurite outgrowths with the cGMP-dependent protein kinase activators 8-Br-cGMP and Sp-8-Br-PET-cGMPS (not shown). We conclude that cAMP analogs can be used to dissect separately the roles of Epac and cAPK as mediators of cAMP action and that Epac and cAPK act in synergy to promote PC-12 cell neurite extensions.

DISCUSSION
The cAMP analog mapping revealed significant differences between the binding sites of Epac and the R subunits of cAPK. In general, Epac had less stringent requirements for binding of either adenine-or 2Ј-ribose-modified analogs than any of the cAPK binding sites (AI, BI, AII, BII). Molecular docking suggests that Epac could accommodate bulky adenine substituents because of little steric hindrance in the region facing the adenine. In addition, perfect stacking between the adenine of analogs and an aromatic residue is not required for binding to Epac, which has no aromatic residue in a position corresponding to Tyr-371 in bRI␣ (Fig. 10A). Introduction of an 8-thiophenyl substituent with a polar group like Cl-, HO-or MeO-in para-position enhanced the affinity for Epac more than 50-fold. Docking of 8-pCPT-2Ј-O-Me-cAMP into the CNBD of Epac1 (Fig. 10B) suggested that the high affinity of 8-thiophenyl substituted cAMP related to the ability of His-317 to provide stacking interactions with the phenyl ring. The further affinity enhancement from a polar group in the para-position of the thiophenyl substituent could be explained by the predicted solvent exposure of the apical part of the phenyl ring (Fig. 10B). Modification of the 2Ј-position of the ribose was less detrimental for binding to Epac than to cAPK. Comparative docking of 8-pCPT-2Ј-O-Me-cAMP (Fig. 10B) and cAMP (Fig. 10A)  activate cAPK above the normal resting level (20 -30% of full activity).
The CNBD of free Epac showed only partial discrimination (about 10-fold) between cAMP and cGMP (Table I). This would not be sufficient to prevent cGMP binding to Epac in cells where the cGMP level exceeds that of cAMP. We found that 6-modified cAMP analogs, including cGMP, were only weak partial agonists of Epac activation, indicating that Epac is only weakly activated by cGMP under physiological conditions. Basically, guanine nucleotide exchange factors like Epac act by stabilizing the free, relative to the GDP-complexed form, of G-proteins (40). An intact amino group in the 6-position appears therefore to be crucial for cAMP analogs to induce the active conformation of Epac with enhanced affinity for free Rap relative to Rap-GDP. The low efficiency of the 6-modified analogs was not because of introduction of bulk causing steric hindrance, because cPuMP and cIMP/cGMP have less bulk than cAMP at the 6-position but still were only weak agonists.
The intact cell studies showed that 6-Bnz-cAMP and 6-MB-cAMP failed to activate Rap1. In contrast, 8-pCPT-2Ј-O-Me-cAMP, 8-pCPT-cAMP, 8-Br-cAMP, and the adenylate cyclase stimulator forskolin activated Rap1 in the same cells (fibroblasts, dog thyrocytes, PC-12 cells) under comparable conditions. This suggests that 6-modified analogs are indeed poor Epac activators in the intact cell. It also suggests that Rap1 activation by cAMP is more likely to be mediated by Epac than cAPK in the cells under the conditions of the present study.
Armed with discriminatory cAMP analogs it was possible to dissect the relative role of Epac and cAPK in cAMPinduced neurite extension in rat pheochromocytoma PC-12 cells. The Epac activator 8-pCPT-2Ј-O-Me-cAMP appeared able to induce neurite extensions on its own. Only analogs with a high affinity for Epac could substitute for 8-pCPT-2Ј-O-Me-cAMP, indicating that Epac and not another cAMP receptor was the target. This suggests Epac to have an important, hitherto unrecognized, role in cAMP-induced neurite extension. Rp-cAMPS analogs and H-89 counteracted the action of 6-Bnz-cAMP much more strongly than the action of 8-pCPT-2Ј-O-Me-cAMP, further underscoring the Epac specificity of 8-pCPT-2Ј-O-Me-cAMP.
Activation of Epac had a strongly enhancing effect on cAPK. One obvious consequence of the synergy between Epac and cAPK is to provide positive cooperativity of cAMP action, because activation of Epac enhanced the action of cAPK activator, which in turn enhanced the action of Epac activator (Fig. 9).
NGF sensitized the cells to both cAPK and Epac activators. A challenging observation was that the Epac stimulator 8-pCPT-2Ј-O-Me-cAMP acted more rapidly than NGF to induce neurite extensions, although NGF gave rise to a higher early activation of Rap1. 2 This suggests that activation of Epac does not merely act to mimic and replace the Rap1 activation by NGF but probably has a distinct action, possibly by activating Rap1 in a different compartment than NGF. The Epac activator can also have hitherto unrecognized effects not mediated via activation of Rap (41).
In conclusion, we have mapped fine structural differences between the binding sites of Epac1 and cAPK and elucidated differences in the mechanism of activation of Epac and cAPK using cAMP analogs. Selective analogs have been demonstrated to discriminately activate Epac or cAPK and thereby point out a hitherto unknown synergism between Epac and cAPK in PC-12 cell differentiation.