Communication between the Regulatory and the Catalytic Region of the cAMP-responsive Guanine Nucleotide Exchange Factor Epac*

Epac1 is a guanine nucleotide exchange factor (GEF) for the small GTPase Rap1 that is directly activated by cAMP. This protein consists of a regulatory region with a cAMP-binding domain and a catalytic region that mediates the GEF activity. Epac is inhibited by an intramolecular interaction between the cAMP-binding domain and the catalytic region in the absence of cAMP. cAMP binding is proposed to induce a conformational change, which allows a LID, an α-helix at the C-terminal end of the cAMP-binding site, to cover the cAMP-binding site (Rehmann, H., Prakash, B., Wolf, E., Rueppel, A., de Rooij, J., Bos, J. L., and Wittinghofer, A. (2003) Nat. Struct. Biol. 10, 26–32). Here we show that mutations of conserved residues in the LID region affect cAMP binding only marginally but have a drastic effect on cAMP-induced GEF activity. Surprisingly, some of the mutants have an increased maximal GEF activity compared with wild type. Furthermore, mutation of the conserved VLVLE sequence at the C-terminal end of the LID into five alanine residues makes Epac constitutively active. From these results we conclude that the LID region plays a pivotal role in the communication between the regulatory and catalytic part of Epac.

A multitude of cellular stimuli, including hormones, growth factors, and neurotransmitters, induce activation of the G␣ s subunit of heterotrimeric G-proteins, which, in turn, activates adenylyl cyclase to synthesize cyclic AMP (cAMP) from ATP. cAMP acts as a second messenger that binds to and regulates PKA 1 and cyclic nucleotide-dependent ion channels (1). More recently, cAMP-responsive guanine nucleotide exchange factors (GEFs) for Rap proteins were identified, i.e. Epac (exchange protein directly activated by cAMP) (2) or cAMP-GEF (3). Rap proteins are members of the Ras family of small GTPbinding proteins. They act as molecular switches that cycle between a GDP-bound inactive and a GTP-bound active state.
Nucleotides are bound tightly to the protein and are released with a very slow dissociation rate. GEFs act by accelerating the slow intrinsic nucleotide dissociation rate by several orders of magnitude. Because the GTP concentration in the cell is much higher then the GDP concentration, Rap is then loaded with GTP. In the GTP-bound form, Rap interacts specifically with proteins to activate downstream targets. Rap1 plays, among other things, a pivotal role in integrin-mediated cell adhesion (4 -9). Rap-GTP is recycled into its inactive state by hydrolyzing GTP to GDP, a process that is catalyzed by Rap-specific GTPase-activating proteins (RapGAPs).
Two isoforms of Epac, Epac1 and Epac2, were found in mammalian cells, both consisting of a regulatory and a catalytic region (2,3). The catalytic region contains a CDC25 homology domain and a REM domain, both of which are characteristic for GEF proteins of Ras-like small GTP-binding proteins such as Ras, Ral, and Rap (Fig. 1). Whereas the CDC25 homology domain contains the active site (10) and forms the interface with its substrate (11), the REM (12) domain stabilizes the CDC25 homology domain without being directly involved in catalysis (11). The regulatory region of Epac contains a cAMPbinding domain and a Dishevelled/Egl-10/pleckstrin (DEP) domain (13). The cAMP-binding domain is responsible for cAMPmediated GEF activation (2,3,14), whereas the DEP domain is responsible for the membrane localization of Epac (15). Whether the DEP domain binds to proteins or lipids is still elusive. Epac2 contains an additional N-terminal cyclic nucleotide monophosphate-binding domain of still unknown function.
Activation of Epac as well as PKA and PKG (cGMP-regulated kinase) is mediated by the binding of cyclic nucleotide to the cyclic nucleoside monophosphate (cNMP)-binding domains. Whereas Epac and PKG consist of only one polypeptide chain, PKA is composed of catalytic (C) and regulatory (R) subunits. The R subunit contains two cAMP-binding domains and forms a homodimer that, in the absence of cAMP, binds to and inhibits the protein kinase activity of the two C subunits. Binding of cAMP leads to the release of the C subunits from the complex and its activation. PKG also contains a dimerization sequence in its regulatory part. Inactive PKG thus resembles inactive PKA, although it contains a different number of polypeptide chains. A chimera of PKG and PKA can be created that maintains the nucleotide specificity of the cyclic nucleoside monophosphate-binding domains (16). These findings suggest that, in PKA and PKG, the inhibition is realized in the same way by the binding of a regulatory domain to the kinase domain either in an intra or an intermolecular manner. For Epac, the regulatory region also inhibits GEF activity in the absence of cAMP, and cAMP relieves this inhibition (2). Indeed, in vitro, an isolated catalytic domain, which is constitutively active, is inhibited by an isolated regulatory domain, suggesting a direct * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ physical interaction between the regulatory and the catalytic regions of Epac. This inhibition is released by cAMP (14).
The kinase domain of the C subunit of PKA was structurally analyzed in great detail and was, in fact, the first protein kinase structure to be solved (17,18). Structures of the different R subunits of PKA have been analyzed in the presence of bound cAMP (19,20). However a structure of the holoenzyme is, to date, not available, and the interaction of the R and the C subunits is incompletely understood in molecular terms. Likewise, it is not understood in any detail how the binding of cAMP induces dissociation of the holoenzyme. Recently, we have solved the structure of the regulatory region of Epac2 in the cAMP-free conformation (21). When compared with the cAMPbound regulatory subunits of PKA, the most obvious difference between these structures is a C-terminal helix, also called LID, which, in PKA, covers cAMP against the solvent and is unstructured in the determined Epac structure, most likely because it is flexible due to the absence of cAMP. It was argued that the HINGE region regulates the orientation of the LID. In turn, the conformation of the HINGE is directly regulated by the presence or absence of cAMP (21). Here we present results indicating that the region of Epac, which would correspondent to the LID region in PKA, is involved in the direct communication between the regulatory domain and the catalytic region. A mutation has been found that completely relieves the autoinhibition of Epac.

EXPERIMENTAL PROCEDURES
Preparation of Proteins-All Epac constructs were expressed as GST fusion proteins from the pGEX-4T2 vector in the Escherichia coli strain CK600K as described previously (22). These constructs are from human Epac1 and comprise the following amino acid residues: Epac, 149 -881; Epac-RD 1-328 , 1-328; Epac-RD 149 -317 , 149 -317; Epac-RD 149 -328 , 149 -328; and Epac-CAT, 324 -881. Single/double amino acid and Ala 5 mutations were made in the Epac-(149 -881) background. Mutations were introduced according to the QuikChange procedure (Stratagene). Because of low expression levels and poor solubility, only low and impure amounts of Epac-CAT and Epac(Ala) 5 were available. The concentration of these proteins was therefore determined after SDS-PAGE gel electrophoresis. The C-terminal truncated version of Rap1B was expressed as described (10). We refer to this construct as Rap1 in the text.
Isothermal Titration Calorimetry-To analyze the binding of cAMP to the isolated cAMP-binding domain, a microcalorimeter (MicroCal Inc.) was used. The protein was brought to 25°C in the 1.38-ml cell of the apparatus, and cAMP was injected in 40 steps of 6 l every 4 min up to a 2-3-fold molar excess. The concentration of the protein was typically 50 M, and the concentration of the nucleotide solution was 10 -20 times higher. The cAMP solution was in the same buffer as the purified protein. The data were analyzed using the manufacturer's software.
In Vitro Activation of Rap1-In vitro activation of Rap1 was performed as described previously (10). Briefly, 200 nM of Rap1B loaded with the fluorescent GDP analogue 2Ј(3Ј)-O-(N-methylanthraniloyl) guanosine diphosphate (mantGDP) were incubated in the presence of 20 M GDP (Sigma) and either 100 nM (mutant) Epac or isolated domains as indicated. cAMP (Sigma) was added as indicated for the individual experiments. The nucleotide exchange was measured in real time as decay in fluorescence using a Spex1 spectrofluorometer (Spex Industries). The decay is caused by the release of protein-bound man-tGDP, which shows higher fluorescence intensity in the hydrophobic environment of the protein than in the buffer solution. The obtained data were fitted to a single exponential decay, and the rate constants (k obs ) calculated were plotted against the cAMP concentration. The concentration dependence of the rate constants was treated as a normal titration experiment. All data analysis, fitting, and plotting were done with the Grafit 3.0 program (Erithacus Software).

Regulation of GEF Activity via the C-terminal Region of the cAMP Domain-
The isolated recombinant catalytic domain of Epac1 (Epac-CAT) is active as a GEF and accelerates release of GDP bound to Rap1. The activity is independent of cAMP but can be inhibited by the addition of recombinant regulatory domain (Epac-RD 1-328 ) (14). The heterodimeric Epac-RD 1-328 -Epac-CAT complex can be reactivated by cAMP ( Fig. 2A). To investigate whether an isolated cAMP-binding domain was sufficient for trans-inhibition, we tested Epac-RD 149 -328 . As expected, this cAMP-binding domain was sufficient to inhibit Epac-CAT. Interestingly however, a cAMP-binding domain that is shorter by only eleven amino acids at the C terminus (Epac-RD 149 -317 ) did not inhibit Epac-CAT, not even at much higher concentrations (Figs. 1 and 2B).
To exclude the possibility that Epac-RD 149 -317 is not properly folded, we determined the affinity for cAMP, because an unfolded protein should not be able to bind cAMP. As determined by isothermal titration calorimetry, the cAMP affinities for Epac-RD 149 -328 (4 M) and for Epac-RD 149 -317 (3 M) are similar (Fig. 3). From these results we conclude that the additional eleven amino acids are required for the trans-inhibition of Epac-CAT and, thus, for the communication of the cAMPbinding domain to Epac-CAT.
The VLVLE Sequence Is Required for Auto-inhibition-Sequence comparison identifies a VLVLE motif within the additional eleven residues, which is conserved between different Epac proteins from man to nematode and fly (Fig. 1). To inves-FIG. 1. Schematic representation of Epac1. Indicated is the domain structure of Epac1 with a DEP domain, a cAMP-binding domain (cAMP), a REM, and a CDC25 homology domain (CDC25-HD). The enlarged region is the C-terminal region of the cAMP-binding domain and is aligned with the homologous regions in Epac2, Drosophila Epac (dEpac), and Caenorhabditis elegans Epac (ceEpac). The region comprises part of the HINGE region and the LID region as defined from the structure of the regulatory region of Epac2 (21). Indicated by arrows are the conserved residues that have been mutated to Ala and analyzed in this study. The single nucleotide polymorphism G332S in Epac1, identified by Vanvooren et al. (23), is highlighted by an asterisk. Also indicated are the deletion mutants of Epac1 used in this study. tigate the function of these residues in the regulation of Epac1, we mutated the VLVLE motif in Epac to AAAAA creating Epac(Ala) 5 . Epac is inactive in the absence of cAMP but becomes active after the addition of cAMP (Fig. 4A). In contrast, Epac(Ala) 5 is active in the absence of cAMP in a concentrationdependent manner (Fig. 4, B and C). Importantly, cAMP does not stimulate this activity any further (Fig. 4B). From these results we conclude that the VLVLE sequence is required for the inhibition of Epac by the regulatory domain.
To understand in more detail the contribution of the individual residues of the VLVLE motif, mutations were generated as indicated (Table II). First, residues were mutated in a pair-wise alanine scan generating Epac(VAALE), Epac(VLAAE), and Epac(VLVAA). The mutant proteins were characterized by determination of the cAMP concentration necessary to obtain half-maximal activity (AC 50 ) as well as the maximal GEF activity measured as the maximal rate constant k max for dissociation of mantGDP from Rap1-mantGDP, which was achieved under standard experimental conditions and saturating concentrations of cAMP. The results are shown in Fig. 5 and summarized in Table I. All of the mutations were inactive in the absence of cAMP and show cAMP-dependent GEF activity. Importantly, AC 50 and k max values of the mutants did not differ more than 2-fold compared with the wild type protein. Whereas the k max of Epac(VLVAA) is comparable with wild type, it shows an AC 50 of 25 M as compared with 45 M for wild type. Epac(VAALE) and Epac(VLAAE) are characterized by a reduced k max but an almost unchanged AC 50 . Apparently, the VLVLE motif tolerates a number of mutational assaults.
In a second series of experiments, more drastic mutations to charged or bulky residues were introduced into the VLVLE  Table II). Replacing leucine with the more bulky hydrophobic residue tryptophane in Epac(VWVLE) is well tolerated, as is the positive charge in Epac(VLRLE). This latter mutant shows a 2.5-fold reduced maximal catalytic activity. Epac(VLDLE), wherein valine is replaced by the small negatively charged aspartate, is not soluble. Apparently, this alteration has a drastic effect on the structure of the protein. Replacement of the negative glutamate by alanine in Epac(VLVAA) showed no gross changes in properties. However, changing the negatively charged glutamate into the positively charged arginine in Epac(VLVLR) resulted in a 2-fold decrease in AC 50 and, very surprisingly, a 2.3-fold increase in k max as compared with wild type. Because the concentration of active protein is crucial for these experiments, we used different preparations of wild type and mutant proteins and obtained similar results, not deviating by more than 15% for both AC50 and k max . From these measurements we conclude that the VLVLE sequence can tolerate extensive mutations without affecting the inhibition by the regulatory region. However, the mutations do modulate the ability of cAMP to activate Epac. Most dramatic is the 2.3-fold increase activity of Epac(VLVLR), indicating that wild type Epac is activated submaximally by cAMP, at least in vitro.
Additional Conserved Amino Acids in the LID Region-From the comparison between the crystal structure of the cAMP-free regulatory domain of Epac2 and the cAMP-bound regulatory domain of PKA (19,20), a model of the molecular mechanism of cAMP-induced activation was developed (21). It predicts that cAMP binding induces a conformational change in the phosphate binding cassette (PBC) and the HINGE of the cAMPbinding domain. This, in turn, induces the C-terminal helix of the cAMP-binding domain, the LID, to move toward the core structure and shield cAMP from the solvent. The VLVLE motif is located at the end of the LID (Figs. 1 and 8). A number of additional conserved residues in the LID (Fig. 1) were analyzed for their possible involvement in binding the adenine base of cAMP and/or mediating the interaction between the regulatory region and the catalytic region. Four mutants were made in Epac-RD 148 -328 and analyzed by measuring the affinity of cAMP to the isolated cAMP-binding domain by isothermal titration calorimetry. Whereas H317A showed wild type affinity, E308A, T311A, and R313A showed a 2-fold reduction in affinity for cAMP, arguing that these residues make a small contribution to binding of cAMP. In addition, the same mutations were made in Epac to determine cAMP-dependent catalytic activity using the fluorescent GEF assay (Fig. 6 and Table II). The AC 50 values of R313A and H317A are similar to wild type, whereas E308A and T311A show a 3-fold lower AC 50 . Remarkably, both E308A and T311A showed a 2-3-fold increase in cAMP-induced GEF activity compared with wild type, whereas H317A has a 5-fold reduced maximal activity.
The G332S Polymorphism Does Not Affect the Regulation of Epac by cAMP-Interestingly, a G322S polymorphism is localized C-terminal to the cAMP-binding domain in close proximity to the VLVLE sequence (23). To investigate whether this mutation may affect the regulation of Epac by cAMP, this mutation was introduced in Epac, and cAMP-induced activation was analyzed (Fig. 7). However, G332S showed the same AC 50 value as wild type, whereas the maximal activity was slightly increased by 20%. DISCUSSION A previous comparison of the structures of the cAMP-free (open) cAMP-binding domain of Epac and the cAMP-bound (closed) cAMP-binding domain of PKA revealed that the main difference between the open and the closed structure is the orientation of the LID. In the closed conformation of PKA, this LID covers the cAMP-binding site and interacts with the adenine group of cAMP. This suggests that this LID region may be the main determinant of the interaction between the cAMPbinding domain and the catalytic region of Epac (21). This would imply that, in the open conformation, the LID is able to interact with the catalytic region, resulting in an inhibition of the activity (see Fig. 8). Here we show that the LID region of Epac1 indeed plays a pivotal role in the communication between the regulatory and catalytic domain of Epac. Most nota- The obtained individual reaction rates (k obs ) were plotted against the cAMP concentration. wt, wild type.

TABLE I Biochemical properties of mutations in the VLVLE region
The AC 50 values and the relative k max as determined from Fig. 5   This was shown in a trans-inhibition experiment in which an isolated cAMP-binding domain is capable of inhibiting an isolated catalytic region only when the region containing the VLVLE sequence is present. In addition, the Epac(Ala) 5 mutant, wherein the VLVLE sequence is replaced by alanine residues, is constitutively active, and cAMP cannot activate this mutant protein further. Thus, the VLVLE sequence is essential for inhibition of the catalytic region by the open cAMP-binding domain even when the two domains are separated. This strongly suggests that this region directly interacts with the catalytic region. Further analysis of the VLVLE sequence revealed that mutating individual residues into alanine had modifying effects on cAMP-induced activation. This is reflected by either a change in the AC 50 for cAMP and/or a change in the maximal activity. Apparently, the sequence requirement for the VLVLE sequence to inhibit the catalytic domain is not very strict. Nevertheless the sequence is highly conserved. This may indicate that the small changes we observe can nevertheless not be tolerated in vivo or imply that the VLVLE sequence, in addition to inhibiting the regulatory domain, serves an additional function. Indeed, single amino acid changes did affect cAMP-induced regulation. This is reflected by either a change in the AC 50 for cAMP and/or a change in k max . Most surprisingly, mutating glutamate into the positively charged arginine resulted in a 2-3-fold higher k max of Epac. From these results, we conclude that the VLVLE sequence also serves as an element that is responsible for translating cAMP binding in a correct activation response. Most likely, the VLVLE sequence is conserved to serve both in the inhibition of the catalytic region and the establishment of the correct conformational response to cAMP. We have also identified conserved residues in the N-terminal part of the LID region that are involved in the regulation mechanism. From their positioning compared with PKA, we assumed that these residues might be involved in the interaction with the base of cAMP. Indeed, the affinity of cAMP for isolated cAMP domains of the E308A, T311A, and R313A mutants is reduced ϳ2-fold. This difference is relatively minor and suggests that the core structure and the phosphate-binding cassette mainly provide the affinity of cAMP for Epac. As for the single mutations in the VLVLE sequence, mutants with mutations in the N-terminal LID region also affect the AC 50 of cAMP to activate Epac and the maximal activation at saturating cAMP levels. The H317A mutant is still responsive to cAMP with a wild type AC 50 value but a k max of only one-fifth of wild type. This indicates that H317 plays a key role in releasing Epac from auto-inhibition, perhaps by sensing the correct base  Table II. C, 200 nM Rap-mantGDP were incubated in the presence of excess GDP and different cAMP concentrations with 100 nM Epac and the indicated mutants, respectively. The reaction rates (k obs ) were determined by fitting analysis and plotted against the cAMP concentration. wt, wild type. structure of cAMP. A similar interaction is found between cAMP and tryptophane or a tyrosine residue in the LID of PKA (19,20). In contrast, E308A and T311A show a reduced AC 50 value and a higher k max . Apparently, these mutants make Epac more active.
The observation that certain LID mutants have a higher or lower k max than wild type is very intriguing and reminiscent of our findings that certain cAMP analogues also show either reduced (22) or increased maximal activity. 2 This indicates that the LID region in the presence of cAMP still influences the catalytic region and precludes a simple model as for PKA, wherein the activation of the catalytic region is caused by the dissociation of the regulatory region. For Epac, the interaction between the regulatory region and the catalytic region apparently remains after the binding of cAMP and imposes a restraint on the activity of Epac. One possible reason for this continuing restraint in the presence of cAMP is that Epac has an additional level of regulation that can modulate the effect of cAMP. This additional level of regulation may impinge on Epac through the binding of regulators to, for instance, the DEP domain, affecting the orientation of the LID region.
This explanation for the various levels of maximal activity, which is consistent with the mutational data and thermodynamic considerations, can be put forward by considering the four-state model of Epac activation (Fig. 9) proposed earlier (22). Regulation of Epac is thus described by a system of coupled equilibria between a bound and an unbound state and between an inactive and an active state. Whereas ligand-free Epac exists mostly (but not exclusively) in the inactive conformation, the binding of cAMP to the cAMP-binding domain shifts the equilibrium more (but not totally) to the active conformation. One should note that the AC 50 values measured here reflect the overall equilibrium of cAMP binding and activation, whereas k max reflects the equilibrium between the cAMP-bound inactive and active conformation The VLVLE-(Ala) 5 mutation would thus favor the active conformation even in the absence of ligand, described by K d3 . Other mutations would influence the conformational equilibrium K d2 between cAMP-bound inactive and active conformation. The fact that Epac(VLVLR), Epac(E308A), and Epac(T311A) have a higher maximal activity than wild type protein would indicate, keeping with the same model, that cAMP does not induce maximal activity even in the wild type protein and that Epac can therefore exist in a cAMP-bound but inactive conformation. The fact that maximal activity can be increased 2-3-fold indicates that the equilibrium described by K d2 is close to or even a bit lower than unity and that a low energy barrier for the conformational change (described by K d2 ) might be advantageous for cAMP signaling or will be influenced by other cellular components. Preliminary NMR shows that, in the absence and presence of cAMP, there is indeed a fast dynamic inter-conversion between two conformations, which we assume corresponds to the active and inactive conformations. 3 At this moment, the residues in the catalytic region in contact with the LID are elusive. It could be that the LID, in particular the VLVLE sequence, interacts with the CDC25 homology domain, thereby preventing binding of Rap (Fig. 8). 2 H. Rehmann, unpublished observations. 3 A. Shimada, unpublished observations FIG. 9. Coupled equilibria for Epac activation. Epac exists in four states, i.e. in cAMP-bound and-unbound states as well as active (subscript a) and inactive (subscript i) states. The occupancy of the cAMP-free but -active state (colored in gray) is very low for Epac wild type and all analyzed mutations, except for Epac(Ala) 5 . For wild type Epac, the equilibrium K d2 is not completely on the right side.
FIG. 8. Model of the regulation of Epac by cAMP. Schematic model for the involvement of the LID region and its VLVLE sequence in mediating the communication between the regulatory and catalytic regions of Epac. The LID region, in particular the VLVLE motif, is interacting with the catalytic region of Epac in the absence of cAMP. This prevents the interaction between Epac and Rap. Upon binding of cAMP, a conformational change is induced that allows the C-terminal helix at the hinge to move closer to the core structure of the cAMP-binding domain. This movement is, in turn, responsible for the flip of the LID region, including placement of the VLVLE motif over the cAMP-binding pocket. This conformation is stabilized by the interaction of the LID with cAMP.
Alternatively, the LID region may induce a conformational change in the catalytic region, for instance through an interaction with the REM domain. Indeed, recently structural and biochemical evidences were presented that the REM domain could accelerate the intrinsic activity of the CDC25 homology domain of the Ras GEF Sos (24).