Entropy-driven cAMP-dependent Allosteric Control of Inhibitory Interactions in Exchange Proteins Directly Activated by cAMP*

Exchange proteins directly activated by cAMP (EPACs) are guanine nucleotide-exchange factors for the small GTPases Rap1 and Rap2 and represent a key receptor for the ubiquitous cAMP second messenger in eukaryotes. The cAMP-dependent activation of apoEPAC is typically rationalized in terms of a preexisting equilibrium between inactive and active states. Structural and mutagenesis analyses have shown that one of the critical determinants of the EPAC activation equilibrium is a cluster of salt bridges formed between the catalytic core and helices α1 and α2 at the N terminus of the cAMP binding domain and commonly referred to as ionic latch (IL). The IL stabilizes the inactive states in a closed topology in which access to the catalytic domain is sterically occluded by the regulatory moiety. However, it is currently not fully understood how the IL is allosterically controlled by cAMP. Chemical shift mapping studies consistently indicate that cAMP does not significantly perturb the structure of the IL spanning sites within the regulatory region, pointing to cAMP-dependent dynamic modulations as a key allosteric carrier of the cAMP-signal to the IL sites. Here, we have therefore investigated the dynamic profiles of the EPAC1 cAMP binding domain in its apo, cAMP-bound, and Rp-cAMPS phosphorothioate antagonist-bound forms using several 15N relaxation experiments. Based on the comparative analysis of dynamics in these three states, we have proposed a model of EPAC activation that incorporates the dynamic features allosterically modulated by cAMP and shows that cAMP binding weakens the IL by increasing its entropic penalty due to dynamic enhancements.

The exchange protein directly activated by cAMP (EPAC) 3 is one of the key receptors for the ancient and ubiquitous cAMP second messenger in mammals (1)(2)(3). The interaction of cAMP with EPAC results in the activation of the guanine-nucleotide exchange in the small GTPases Rap1 and Rap2 (1,2), leading to the cAMP-dependent control of a wide array of critical signaling pathways underlying diverse cellular functions, ranging from insulin secretion to memory enhancement and cell adhesion (4 -10). Two cAMP-dependent EPAC isoforms are currently known (Fig. 1a). Both EPAC1 and -2 are multidomain proteins with an N-terminal regulatory region (RR), including the cAMP binding domains (CBDs) and a C-terminal catalytic region (CR), containing a CDC25-homology module (CDC25HD) that functions as a guanine-nucleotide-exchange factor (GEF) (Fig. 1a). In both EPAC isoforms the cAMP dependence of the GEF function is implemented through the CBD at the C terminus of the RR (Fig. 1, a and b) irrespective of the DEP domain, which serves the primary purpose of controlling the membrane localization of EPAC (4,9).
The cAMP-dependent structural changes underlying the regulatory function of the EPAC CBD have been previously mapped by the crystal structures of several structurally homologous CBDs solved in the apo and cAMP-bound states (11)(12)(13)(14)(15)(16)(17)(18). These CBD structures consistently show that the main conformational change caused by cAMP is a hinge-like rotation of the helix C-terminal to the ␤-barrel typical of CBDs (17,18), which is commonly referred to as hinge helix. When cAMP docks into the binding pocket formed by the phosphate binding cassette (PBC) and the base binding ␤4-␤5 region (BBR), the hinge helix approaches the ␤-barrel core, bringing the region C-terminal to it (i.e. the lid) in proximity of the PBC (18). Based on structural and mutagenesis data, it has been proposed that this hinge-like cAMP-dependent motion contributes to the displacement of the RR away from the CR, thus removing steric impediments for the access of substrates to the catalytic domain and explaining the cAMP-dependent activation of the GEF function of EPAC (18). This mechanism of EPAC activation has been formalized in terms of an equilibrium between active and inactive states in which the RR and CR are in an open and closed relative orientation, respectively. The cAMP-dependent hinge rotation then promotes a shift of this equilibrium toward the active (open) state (18,19).
The recently solved structure of apo-EPAC2 (16) has also revealed the involvement of the CBD in another fundamental inhibitory mechanism distinct from the hinge rotation. Specifically, the ␣2 helix in the N-terminal helical bundle (NTHB) of this CBD and the preceding helix (i.e. ␣1) form a cluster of four salt bridges with the catalytic region. Such a CR/RR salt-bridge cluster is called the ionic latch (IL) (Fig. 1b) and serves two main inhibitory functions. First, it secures the RR and CR in close proximity to each other, sterically blocking the access of effector Rap proteins to the catalytic core (16). Second, the IL sequesters and shields a critical CR arginine (i.e. Arg-886 in EPAC2 m ), which has been suggested to be required for the recognition of Rap by EPAC, making it unavailable for a crucial interaction with Rap (16).
A recent site-directed mutagenesis study has shown that the equilibrium between active and inactive states of EPAC is not controlled only by the hinge motion of the CBD C-terminal region but also by the IL (16). For instance, a deletion mutant (i.e. EPAC2⌬306) in which the IL is weakened through the removal of one of the CR/RR salt bridges displays a striking 5-fold increase in the maximum exchange activity (k max ) (16), indicating that the integral IL contributes to shifting the EPAC equilibrium toward the inactive state. However, it is currently not clear how the IL sites located in the N-terminal helical bundle are controlled by cAMP, which docks in the distal PBC and BBR, embedded within the ␤-subdomain. Considering that previous MS investigations (20) support the notion that the region spanning the IL is not subject to major conformational perturbations upon cAMP binding, we propose the hypothesis that the IL is allosterically controlled by cAMP through cAMPdependent changes in dynamics rather than in structure. To test this hypothesis, we have investigated primarily by classical NMR 15 N relaxation experiments as well as by multi-offset NMR dispersion measurements the EPAC1 h -(149 -318) construct in its apo-, cAMP-bound (holo), and Rp-cAMPS-bound states.
The EPAC1 h -(149 -318) segment binds cAMP with an affinity similar to that of full-length EPAC (K D ϳ 4 M) (21) and includes not only the PBC and BBR, where cAMP binds, but also the NTHB and the C-terminal helical hinge region (Fig. 1b) that is allosterically affected by cAMP. Furthermore, EPAC1 h -(149 -318) includes in full the ␣1-helix and the three key sites of the ionic latch (i.e. Gln-168, Asp-172, and Glu-197). A similar N-terminal truncation in EPAC2 (i.e. EPAC2⌬280) did not affect the observed k max (16), indicating that such an N-terminal deletion mutation does not impair the ability of the IL sites to form inhibitory interactions with the catalytic core. In addition, we have noticed that the ␣1-helix with its highly polar N-terminal region (Fig. 1b) ensures excellent solubility in both the presence and absence of ligands resulting in high assignment coverage (i.e. Ͼ82%) and in an unprecedented picture of allostery and dynamics unavailable for other eukaryotic CBDs due to inherent instabilities that have hampered direct apo versus holo comparisons (19,(22)(23)(24)(25).
Overall, our combined comparative NMR analysis of dynamic features in the EPAC1 CBD shows that cAMP binding leads to significant modulations of the CBD dynamic profile in several time scales and at multiple allosteric sites, pointing to a critical link between flexibility and function, in general, and to a pivotal entropic determinant for the allosteric propagation of the cAMP signal from its binding pocket to the distal IL region, in particular.

EXPERIMENTAL PROCEDURES
Sample Preparation-The human EPAC1 CBD construct EPAC1 h -(149 -318) was expressed and purified according to previously published protocols (26). Further details are available in the supplemental materials.
NMR Measurements-All spectra were acquired at 34°C using a 700-MHz spectrometer equipped with a TCI cryoprobe. The temperature was calibrated using both a thermocouple and an ethylene glycol sample. The heteronuclear single-quantum coherence (HSQC) spectra were acquired with 128 ( 15 N) and 1024 ( 1 H) complex points and spectral widths of 31.8 and 14.2 ppm for the 15 N and 1 H dimensions, respectively. 1 H chemical shifts were calibrated using 2,2-dimethyl-2-silapentanesulfonic acid followed by indirect calibration of the 15 N ppm values using the nitrogen to proton gyro-magnetic ratio (27). The carrier frequencies of the proton and nitrogen channels were centered on the H 2 O resonance and on the middle of the backbone amide region, respectively. 15 N was decoupled using the GARP4 pulse train with a 1.32-kHz radio frequency pulse (RF) strength. All two-dimensional spectra were processed using Xwinnmr (Bruker, Inc.) and/or NMRPipe (28). Cross-peak fit heights were measured using Sparky 3.111 (29) and Gaussian line fitting unless otherwise specified. The resonance assignments were obtained through standard triple resonance experiments (i.e. HNCO, HNCA, HN(CO)CA, CBCA-(CO)NH, HBHA(CO)NH, HNCACB, and HNHAHB) (30), whereas the 1 H, 1 H NOEs were measured using NOESY-HSQC spectra ( mix ϭ 150 ms). All three-dimensional spectra were processed using NMRPipe (28) and analyzed through Sparky 3.111 (29). The secondary structure probabilities were determined using the secondary chemical shifts via the PECAN software (31). Compounded chemical shifts were computed via the following formula: ⌬␦ compound ϭ ((⌬␦ 1 H) 2 ϩ (⌬␦ 15 N/6.5) 2 ) 1/2 , where ⌬␦ 1 H and ⌬␦ 15 N are the differences between the proton and nitrogen chemical shifts, respectively. 15 N NMR Relaxation Measurements-The 15 N longitudinal and transverse relaxation rates R 1 and R 2 as well as the { 1 H} 15 N steady state NOEs were measured for 0.1 mM EPAC1 h -(149 -318) in its apo and Rp-cAMPS-and cAMP-bound forms, with pulse sequences containing water flip back pulses and sensitivity enhancement (32)(33)(34)(35)(36)(37)(38). Before the relaxation experiments the NMR samples were filtered through a 100-kDa cut off filter. For the measurements of R 1 relaxation rates the following relaxation delays were employed: 100 (ϫ2), 200, 300, 400 (ϫ2), 500, 600, 800, and 1000 ms (where ϫ2 indicates duplicate spectra). The offset and duty cycle-compensated 15 N R2 CPMG experiments (39) were measured with a v CPMG of 472 Hz (40,41) and CPMG relaxation delays of 8.48, 16.96, 25.44, 33.92, 42.4, 50.88, 59.36, 76.32, and 93.28 ms. For the NOE experiments, a 10-s recycle delay was used that included a 5-s proton saturation period, whereas a recycle delay of 1.5 s was used for the R 1 and R 2 experiments. The stability of the protein sample was monitored before and after each relaxation experiment via HSQC spectra. To average potential long term instabilities, the entire series of R 1 and R 2 relaxation rates were run in duplicate or triplicate (42), and the { 1 H}-15 N steady state NOEs were collected in 9 or 10 sets of saturated and unsaturated spectra. All spectra were co-added before processing using NMRPipe, resulting in a total of 48 or 72 scans per serial file at each T 2 CPMG length, 32 or 48 scans per serial file at each T 1 inversion recovery delay, and 36 or 40 scans per serial file for each NOE experiment (with and without 1 H saturation). The number of dummy scans was always 96 or higher. The 15 N and 1 H frequencies in the relaxation experiments were digitized with the same number of points as in the HSQC experiments described above. The errors for R 1 and R 2 were evaluated using Gaussian distributed random noise. The steady state NOE values were computed as the ratio of the intensities in saturated to unsaturated spectra. The STDV of the distribution of the differences in fit heights between duplicate spectra was used to estimate the errors of the steady state NOE values (32). All errors were conservatively treated as previously explained (43). Selected crosspeaks were not included in the relaxation analyses due to line broadening and/or overlap.
Reduced Spectral Density Mapping-The 15 N relaxation data were mapped into reduced spectral densities under the assumption that their high frequency values do not vary: J( N ϩ H ) ϭ J( H ) ϭ J( N Ϫ H ), as previously explained (44 -47). Assuming these equalities, it is possible to compute the J( N ϩ H ) and J( N ) values only from the measured 15 N R 1 and { 1 H} 15 N NOEs, whereas the J(0) values are derived from the measured 15 N R 2 rates as well (47). The value of J(0) calculated in this way also includes contributions from chemical exchange. Error propagation was employed to estimate the errors in the reduced spectral densities starting from the uncertainties in the measured 15 N relaxation rates and NOEs.
Hydrodynamic Simulations-The contributions to the relaxation rates and to the reduced spectral densities arising from the overall tumbling and from the effect of diffusion anisotropy were evaluated through hydrodynamic simulations based on the bead method using the HYDRONMR program (48,49). For this purpose the coordinates of several CBD-spanning fragments of the Protein Data Bank code 2BYV structure of EPAC2 m (16) were utilized (supplemental Table S1). In all simulations hydrogen atoms were added through the program Molmol (50), and the atomic element radius was 3.3 Å, which has been previously shown to best fit multiple hydrodynamic properties (i.e. translational diffusion, sedimentation coefficients, rotational diffusion, and intrinsic viscosity) in a set of model proteins (49). All HYDRONMR computations were carried out at a temperature of 307 K and the water viscosity in centipoises (cP) was calculated as: ϭ 1.7753 Ϫ 0.0565t ϩ 1.0751 ϫ 10 Ϫ3 t 2 Ϫ 9.2222 ϫ 10 Ϫ6 t 3 , where t is the temperature in Celsius (49). The HYDRONMR-computed 15 N relaxation rates at a static field of 16.44 tesla assume an N-H distance of 1.02 Å and a chemical shift anisotropy of Ϫ160 ppm (40,41). The D par /D per ratios in supplemental Table S1 were calculated as 2D z /(D x ϩ D y ), in which D x and D y are the two eigenvalues of the rotational diffusion matrix that are closest to each other (48,49).
Relaxation Dispersion NMR-The 15 N relaxation dispersions for backbone amides in the cAMP-and the Rp-cAMPS-bound and apo states of EPAC1 h -(149 -318) were measured using a constant-time relaxation-compensated CPMG pulse sequence (40,41,51). During the total CPMG length (T CP ) of 93.3 ms, either 8 or 88 180°1 5 N pulses were implemented, resulting in CPMG field strengths ( CPMG ) of 43 and 472 Hz, respectively (40,41,51). Six interleaved replicate data sets were recorded and co-added for each CPMG field strength. 128 dummy scans and 16 scans were accumulated per serial file with an interscan delay of 2.2 s. Because of the presence of significant offset effects for the 15 N pulses at 700 MHz and at the CPMG RF (radio frequency pulse) strength employed (3.1 kHz), all constant-time relaxation-compensated CPMG experiments were acquired with three different 15 N carrier frequencies (110, 119, and 127 ppm) to cover the 15 N spectral width through three narrow frequency bands. The NMR relaxation dispersion (⌬R 2 eff ) was computed using the equation ⌬R 2 eff ϭ (1/T CP ) ln(I 472Hz /I 43Hz ), where I CPMG is the cross-peak intensity with a CPMG strength CPMG . NMR dispersion could not be measured for cross-peaks affected by overlap and/or relaxing too quickly to be detected in the relaxation-compensated constanttime CPMG spectra (e.g. several residues in the ␣3 region).
Sequence and Structure Analyses-Sequence alignments were performed with ClustalW (52), whereas Pymol (53) and/or Molmol (50) were used for the structural analyses. shows that no significant cAMP-dependent 1 H 15 N-compounded chemical shift changes are observed for helices ␣1 and ␣2, which span the IL region (i.e. 168 -197). For most ␣1 and ␣2 residues the observed cAMP-induced 1 H 15 N-compounded chemical shift variations are below average (Fig.  1d) and are comparable with or smaller than those observed for residues in other structurally invariant regions in all known CBDs, such as ␤8 (Fig. 1, c and d) (15). In agreement with these observations, the secondary chemical shifts of apo and cAMP-bound EPAC1 h -(149 -318) (Fig. 2, a and b) further support the absence of significant variations in the 2°s tructure for the IL region. Therefore, both the HN and the secondary chemical shifts consistently indicate that the region spanning the IL (i.e. 168 -197) is not affected by major structural variations as a result of cAMP binding.

Evaluation of cAMP-dependent Structural Changes beyond the IL Region Based on a Quantitative Analysis of the cAMPinduced Chemical Shift
Variations-In the other regions of EPAC1 h -(149 -318) C-terminal to ␣2, the observed chemical shift variations (Fig. 1d) are consistent with a model (26,17) Entropy-driven Allosteric Control of EPAC JULY 11, 2008 • VOLUME 283 • NUMBER 28 according to which cAMP docks to the PBC and BBR and relays its signal through the ␤2-␤3 loop and the ␣6-lid regions. According to this model, cAMP binding causes a hinge-like rotation of ␣6 (17) perturbing the environment of adjacent residues in the ␣3 and ␣4 (Fig. 3) and explaining the cAMP-dependent chemical shift variations observed for these helices. This model is also supported by the NMR-based ␣/␤-probabilities of apo and holo EPAC1 h -(149 -318) (Fig. 2, a and b) showing that the most significant 2°s tructure variations occur at the level of the ␣6 helix C terminus (i.e. residues ϳ305-310), which becomes less helical upon cAMP binding, consistently with a similar helicity profile reported for the cAMP-bound state of CBD-A of PKA (23).
Interestingly, the Rp-cAMPS antagonist, unlike cAMP, does not significantly destabilize the ␣6 helix C terminus (Fig. 2c), in full agreement with the absence of significant Rp-cAMPS-dependent HN-ppm variations for ␣6 (Fig. 1d). These observations suggest that Rp-cAMPS does not cause the hinge rotation of ␣6 and explain also why, upon Rp-cAMPS binding, no appreciable HN-ppm changes are detected for residues in ␣3 and ␣4, confirming our interpretation of the cAMP-dependent HN-ppm changes in this region in terms of the ␣6 hinge motion. Another clear difference between the HN-ppm variations caused by cAMP and by Rp-cAMPS is observed for Gly-238 ( Fig. 1d) in the ␤2-␤3 loop, which is very sensitive to cAMP but is only marginally affected by the phosphorothioate antagonist. The HN-ppm changes observed for Gly-238 reflect, therefore, mainly allosteric perturbations.
General Approach to the Characterization of the Dynamics of the EPAC1 CBD-The ps-ns and s-ms dynamic features of EPAC1 h -(149 -318) were probed in the apo and cAMP-and Rp-cAMPS-bound states through the measurement of 15 N R 2 and R 1 relaxation rates and of { 1 H}-15 N NOEs, which are reported in Fig. 4a, b, and d, respectively. One of the primary considerations in the 15 N R 1 , R 2 , and NOE data analysis is the pronounced flexibility of most residues in the long N-terminal helix, as suggested by the fast hydrogen exchange rates (26) and the secondary chemical shifts (Fig. 2). It is well known that this type of conformational heterogeneity in the CBD ␣-subdomain impinges on the diffusion model describing the overall tumbling of the protein in solution (23). In addition, supplemental Table S1 shows that the D par /D per ratios vary considerably as a function of the degree of N-terminal truncation. The significant conformational plasticity of the N-terminal tail, therefore, hampers the accurate description of the overall motion for the EPAC1 CBD in terms of a single well defined diffusion tensor. Hence, we opted to rely for the 15 N relaxation data analysis on the reduced spectral density mapping approach, which does not require assumptions on the overall tumbling of the protein or on the exact form of the spectral density functions (44 -47). The relaxation data were then translated into reduced spectral density maps, which are provided in Fig. 5. Potential contributions arising from the diffusion anisotropy of the overall tumbling were assessed through the evaluation of the R 1 R 2 product (Figs. 4c), which is essentially insensitive to the anisotropy of the diffusion tensor (54), and/or through hydrodynamic simulations of both the 15 N relaxation rates (red lines in Fig. 4) and the corresponding spectral densities (red lines in Fig. 5).
Considering that the ␣1 N-terminal helix is predominantly unstructured, the hydrodynamic modeling of the relaxation rates and of the reduced spectral densities were repeated at three different levels of N-terminal truncation (i.e. starting from residues 149, 161, and 173) (Figs. 4 and 5; supplemental Table S1). Although the best match with the average values of the relaxation rates and of the spectral densities was obtained for the simulations starting at residue 161 (Fig. 4), the data calculated for the constructs truncated at residues 149 and 173 were still employed to provide a conservative estimate of the FIGURE 1. a, EPAC1 and EPAC2 domain organization. The black circle indicates the cAMP ligand. The dotted contour and the empty circle (cAMP) for the CBD:A of EPAC2 mean that this module is not strictly necessary for the regulation of EPAC2 (18). The question mark denotes a domain with currently unknown function. The domain investigated here for EPAC1 is highlighted in orange. b, sequence alignment of CBDs in human EPAC 1 (EPAC1 h ), human EPAC2 (EPAC2 h ), and mouse EPAC2 (EPAC2 h ). The secondary structure is shown in red, and it was obtained from the crystal structure of apoEPAC2 m (Protein Data Bank (PDB) code 1O7F) (17). Strictly conserved residues in all three sequences are highlighted in green, whereas residues that are identical only in two of the three sequences are marked in yellow. Residues conserved only in terms of side-chain type are highlighted in cyan. Residues that play pivotal allosteric roles based on mutational analyses (17) or on the apo-EPAC2 m structure (16) are marked by blue rectangles. c, representative expansion of the 1 H, 15 N HSQC spectra of EPAC1 h -(149 -318) in its apo (red) and cAMP-bound (blue) states. All data were acquired at 700 MHz, 34°C, and in 50 mM Tris (pH 7.6 with Mes), 50 mM NaCl, 1 mM Tris(2-carboxyethyl)phosphine, 0.02% w/v NaN 3 using a Bruker TCI cryo-probe. d, HN-compounded chemical shift changes for EPAC1 h -(149 -318) occurring upon binding to cAMP (black) or to the antagonist Rp-cAMPS (green). The value reported for Ala-280 is downscaled by a factor of two. The dashed horizontal line indicates the average cAMP-dependent chemical shift change. The gray background highlights the region spanning the ionic latch. The dotted lines indicate the secondary structure expected based on the coordinates of apoEPAC2 m (PDB code 1O7F) (20), which is highly homologous to EPAC1 h as shown in panel b.  range of rate and spectral density variability associated with the overall tumbling anisotropy (Figs. 4 and 5). In addition, dynamics in the ϳms time scale were probed in a diffusion anisotropyindependent manner through NMR dispersion (NMRD) measurements reported in Fig. 6.
In conclusion, our analysis of the dynamic profile of the EPAC1 CBD is largely structure-independent, as it relies on a combination of HN-NOE, R 1 R 2 , reduced spectral densities, and NMR dispersion measurements. HN-NOEs and J( H ϩ N ) are used to probe fast local ps-ns motions, whereas J(0), the R 1 R 2 product, and the NMRD rates report on dynamics in the ms-s range. This approach was previously successfully applied to the characterization of the PKA CBD-A dynamics (23).
Dynamics of ApoEPAC1 h -(149 -318)-The dynamic profile of apoEPAC1 h -(149 -318) was investigated in the ps-ns and ms-s time scales. For the ps-ns dynamics of apoEPAC1 h -(149 -318) the most significant feature is the flexibility of the N-terminal moiety of ␣1, as indicated by the relatively low NOE (Fig. 4d) and high J( H ϩ N ) (Fig. 5c) values observed for the corresponding residues. This marked flexibility at the N termi-nus explains why the overall tumbling is best modeled by hydrodynamic simulations of the N-terminal-truncated EPAC1 constructs (supplemental Table S1; Figs. 4 and 5). Other sites of apoEPAC1 h -(149 -318) that appear significantly affected by local ps-ns motions as conservatively indicated by NOE values Յϳ0.5 are Asp-236 in the ␤2-␤3 loop and several C-terminal lid residues (Figs. 4d and 5c). The flexibility of Asp-236 is also confirmed by its fast H/H exchange in the apo state (26) despite the fact that its amide hydrogen is involved in a backbone-to-backbone hydrogen bond (17). The dynamics of the C-terminal tail is further supported by the secondary chemical shifts pointing to a predominantly random-coil preference in the 310 -318 region (Fig. 2).
ApoEPAC1 h -(149 -318) displays also multiple sites affected by ms dynamics as revealed by the NMRD data (Fig. 6). Residues with greater than average ⌬R 2,eff. in the apo state cluster at several loci distributed throughout the domain, including the ␣1 C terminus, the ␣4-␤1 and ␤2-␤3 regions, the PBC (i.e. Ala-277), and the ␣6 (hinge) helix (Fig. 6). In addition, significant line-broadening is observed for several residues in the PBC and  Fig. 1c. a, spin-spin relaxation rate R 2 . b, spin-lattice relaxation rate R 1 . c, product of the R 1 and R 2 relaxation rates. In a-c residues displaying the most significant enhancements in their R 1 R 2 values upon cAMP binding are highlighted in red. d, 15 N{ 1 H} NOE calculated as I sat /I nonsat . In this panel residues for which the HN NOE increases upon cAMP binding are highlighted in blue, whereas those for which the HN NOE decreases upon cAMP binding are highlighted in red. In panels a, c, and d, residues for which the ps-ns or ms-s dynamics is enhanced or quenched in the cAMP-bound state relative to both the apo and the Rp-cAMPS states are labeled by a star. The red lines in panels a-d indicate relaxation rates calculated based on a hydrodynamic bead model for the overall tumbling motion. The hydrodynamic computations were carried out for three different deletion constructs of the highly homologous EPAC2 m , 284 -444, 296 -444, and 308 -444, corresponding to EPAC1 h 149 -309, 161-309, and 173-309, respectively. Residues for which no relaxation data are available are prolines or are overlapped and/or broadened beyond detection.
in the BBR (supplemental Fig. S1) making them undetectable in the relaxation experiments but also suggesting the presence of ms-s dynamics at these sites, consistent with their lack of protection from hydrogen/deuterium exchange in the absence of cAMP (19,26). Other sites affected by ms-s dynamics in apoEPAC1 h -(149 -318) are mostly located in the NTHB region spanning helix ␣3 (Figs. 4c and 5a). For instance, residues Glu-197, His-200, Ile-201, and Ala-203 are highly dynamic in the ms-s time-scale as indicated by their R 2 and J(0) values, which are significantly higher than the values computed based on hydrodynamic modeling of the overall tumbling (Figs. 4c and 5a). These R 2 and J(0) enhancements, therefore, cannot be accounted for by diffusion anisotropy effects, as also independently confirmed by the corresponding R 1 R 2 products (Fig. 4c).
In conclusion, the combined analysis of the NMRD profile (Fig.  5) and of the R 2 , J(0), and R 1 R 2 values (Fig. 4, a and c, and Fig. 5a) reveals that in the apo state of the EPAC1 h CBD, the sites affected by motions in the ms-s time scale are distributed in both the ␣ and ␤ subdomains, as shown also by Fig. 7 in which the dynamic hot spots of the apoEAPC CBD have been mapped onto the three-dimensional structure.
In a previous NMR investigation of a shorter EPAC1 construct (i.e. EPAC1 h (169 -318)), a minor (i.e. 20%) set of HSQC cross-peaks was detected for 9 residues in the apo state (19). This minor set of apo peaks appears at chemical shifts similar to those observed in the cAMP-bound spectrum for the corresponding residues (19). Even though these minor apo peaks were not reported for other regions with significant cAMP-dependent chemical shift changes such as the ␣6-helix, they were interpreted as supporting evidence of a dynamic equilibrium between active and inactive states in apoEPAC. For the longer construct EPAC1 h -(149 -318) and under our experimental conditions, we could not observe this minor set of HSQC peaks. However, minimally populated states that escape direct detection in the HSQC experiment are readily probed through NMRD measurements (Fig. 6) when they exchange with the main set of peaks in the ϳms time scale (40,41,51).
cAMP-dependent Dynamics of EPAC1 h -(149 -318)-Despite the fact that the dynamic hot spots of the apoEPAC1 h CBD are present in both ␣ and ␤ subdomains, the flexibility of these two subdomains is affected by cAMP in remarkably different ways. Although cAMP binding results in an overall quenching of ms dynamics in several ␤-subdomain loci as indicated by the corresponding cAMP-dependent decreases in the NMRD ⌬R 2,eff. values (for instance, Phe-232, Ile-243, and Ile-244 in the ␤2-␤3 regions and Ala-277 in the PBC; Fig. 6), a significant enhance-  Fig. 4. In all three panels, the red lines indicate the reduced spectral densities computed starting from the relaxation rates predicted based on the hydrodynamic bead models, as shown in Fig. 4. JULY 11, 2008 • VOLUME 283 • NUMBER 28 JOURNAL OF BIOLOGICAL CHEMISTRY 19697 ment of ms motions is observed for multiple sites in the ␣-subdomain. For instance, residues Val-177, Leu-179, Lys-181, and Ile-194 in the ␣1,2 region and Glu308 at the C terminus of ␣6 all display a significant increase in their NMRD ⌬R 2,eff. values upon cAMP binding (Fig. 6). A similar pattern of ␤-subdomain ms-s dynamics quenching and concurrent ␣-subdomain ms-s dynamic enhancement upon cAMP binding is also supported by the reduction in line-broadening observed for the BBR and the PBC regions in the holo relative to the apo state (supplemental Fig. S1) and by the cAMP-dependent increases in the R 1 R 2 and J(0) values observed for Glu-197, the ionic latch residue in ␣2, and for Ile-201 and Ala-203 in ␣3 (Figs. 4c and 5a).

Entropy-driven Allosteric Control of EPAC
The cAMP-dependent ␣/␤-dynamics enhancement/quenching overall trend observed for the ms-s time-scale extends also to the fast ps-ns motions. The only ␤-subdomain site that in the apo state is highly flexible in the ps-ns time scale, i.e. Asp-236 in the ␤2-␤3 loop, is rigidified upon cAMP binding as indicated by its high NOE and low J( H ϩ N ) values, which in the cAMPbound state are within error from the values expected based only on overall tumbling (Figs. 4d and 5c). The quenching of dynamics for Asp-236 is also consistent with a 3 order of magnitude increase in hydrogen exchange protection factor observed at this locus upon cAMP binding (26). In sharp contrast to Asp-236, several NTHB sites spanning ␣1 and ␣3 become more flexible in the ps-ns time scale as shown by the cAMP-dependent decreases in NOE and corresponding increases in the J( H ϩ N ) values observed for residues Val-150, Arg-169, Asp-172, Ala-173, and Ile-201 (Figs. 4d and 5c). Interestingly, enhanced ps-ns dynamics upon cAMP binding is also observed for Lys-305 in ␣6 and Arg-313 in the adjacent lid (Figs. 4d and 5c), consistent with the destabilization of the C-terminal half of ␣6 (i.e. residues 305 onward) occurring upon cAMP binding as revealed by the secondary chemical shifts (Fig. 2, a  and b).
Rp-cAMPS-dependent Dynamics of EPAC1 h -(149 -318)-For the purpose of separating cAMPdependent variations in dynamics merely due to binding from those due to allosteric effects, the 15 N R 1 , R 2 , and HN NOE together with the corresponding reduced spectral densities (Figs. 4 and 5, green circles) and the NMRD ⌬R 2,eff. rates (Fig. 6, green circles) were also measured for the antagonist Rp-cAMPS-bound state of EPAC1 h -(149 -318). Figs. 4 -6 show that overall most of the cAMP-dependent changes in dynamics result from allosteric rather than from binding effects. For instance, the ms dynamics at Leu-179, Lys-181, Val-218, Glu-222, and Glu-308 is reduced in both the apo and Rp-cAMPS-bound states relative to the cAMP-bound form, as indicated by the consistently higher ⌬R 2,eff. values observed for these residues in the cAMP-saturated state compared with both apo and Rp-cAMPS-bound forms (Fig. 6). If these variations in ms dynamics were just the result of binding, the dynamics in the antagonist-bound state should have instead resembled the cAMP-bound rather than the apo state. The NMRD data in Fig. 6 indicate, therefore, that the large majority of the cAMP-dependent enhancements in ms dynamics observed for EPAC1 h -(149 -318) result from allosteric effects. Similarly, most of the sites affected by a cAMP-dependent quenching in ms dynamics (Figs. 6) are also the result of allosteric perturbations as indicated by the ⌬R 2,eff. dispersions in the cAMP-bound state being lower than those of both apo and Rp-cAMPS-bound forms (Fig. 6). Furthermore, the R 1 R 2 analysis shows that allosteric effects account also for the up-regulation of ms-s dynamics of Arg-169, Glu-197, and Ala-203 (Fig. 4c). Glu-197 mediates critical IL salt bridges, and Arg-169 is adjacent to Gln-168, another IL residue.  Fig. 1c. The effective R 2 relaxation rate dispersion was measured using multi-offset relaxation-compensated constant time CPMG measurements and computed as ⌬R 2,eff. ϭ R 2,eff. (43 Hz) Ϫ R 2,eff. (472 Hz), where the numbers in parenthesis denote the CPMG field strengths. If ⌬R 2,eff. Յ 2 s Ϫ1 , only an upper limit is reported and is denoted by a down-pointing triangle set at the maximum of 2 s Ϫ1 and (⌬R 2,eff. ϩ ⑀ ⌬R2,eff. ), where ⑀ ⌬R2,eff. is the experimental error on ⌬R 2,eff. determined through constant-time NMRD measurements. Residues for which no data are reported correspond to cross-peaks that are prolines or are overlapped and/or broadened beyond detection.

Entropy-driven Allosteric Control of EPAC
The allosteric propagation of the cAMP signal affects the dynamic profile of EPAC1 h -(149 -318) not only in the ms-s but also in the ps-ns time scale. This modulation of ps-ns flexibility involves several functionally critical residues such as Asp-172 in the ionic latch, Asp-236 in the ␤2-␤3 loop, and Lys-305 in the hinge ␣6 helix (Fig. 4d). Specifically, the ps-ns dynamics at Asp-172 and Lys-305 is reduced in both the apo and Rp-cAMPS-bound states relative to the cAMP-bound form (Figs. 4d and 5c). The HN NOE and J( H ϩ N ) values in Figs. 4d and 5c, respectively, indicate therefore that the cAMP-dependent enhancement of ps-ns dynamics at Asp-172 and Lys-305 is an allosteric effect. Similarly, the cAMP-dependent quenching of ps-ns dynamics at Asp-236 (Figs. 4d and 5c) is also the result of allosteric coupling between the ␤2-␤3 loop and the PBC because for this residue the J( H ϩ N ) spectral densities of the apo and Rp-cAMPS-bound states are both significantly higher than that of the cAMP-saturated form.

Functional Relevance of the ApoEPAC1 h -(149 -318)
Dynamics-The ps-ns and ms-s dynamic "hot spots" of apoEPAC1 h -(149 -318) are summarized in the three-dimen-sional map of Fig. 7. To the best of our knowledge this is the first time the full ps-ns and ms-s dynamic profile of a ligand-free eukaryotic CBD could be investigated at atomic resolution by 15 N relaxation experiments, because previous attempt to analyze the apoCBD of PKA were unsuccessful due to aggregation (24). Fig. 7 shows that these dynamic hot spots cluster into a limited set of patches which match well the previously identified functional sites of the EPAC1 CBD. For instance, not only the N and C termini of the PBC appear dynamic in the ms-s time scale (Fig. 7b) but also the adjacent ␤2-␤3 loop is affected by both ms-s and ps-ns motions (Fig. 7a), suggesting that in the absence of cAMP this turn is unstable. Furthermore, two other ms-s dynamic patches involve the hinge ␣6 helix and the IL spanning ␣1-2 helices, respectively (Fig. 7). The patch including ␣6 also affects the adjacent ␣4 and that, including ␣1, extends to the proximal ␤1 strand as well (Fig. 7). Overall, the high degree of co-localization between the functionally critical regions of the EPAC1 CBD (i.e. the PBC, the ␤2-␤3 loop, the hinge, and the ␣1-2 helices) and the dynamic patches is fully consistent with the apo state of this domain existing in an equilibrium of active and inactive conformations, as previously hypothesized based on cAMP-dependent bioassays on fulllength EPAC (18).
Inspection of Fig. 7 reveals also two major clusters of residues affected by ps-ns dynamics and localized in the vicinity of the N and C termini. Although the flexibility in the C-terminal tail is likely to be to a large extent the result of the C-terminal truncation, the dynamics observed in the N-terminal region is consistent with the elevated B-factors observed for this region in the context of the full-length EPAC structure (16). As shown in Fig. S2, the B-factors for residues 149 -165 are significantly higher than the average value observed for the 149 -318 segment, suggesting that the flexibility observed for the N terminus of EPAC1 h -(149 -318) may reflect at least in part an intrinsic property of the EPAC1 CBD rather than just a truncation artifact. In addition, the apo state dynamic profile serves as a key reference to evaluate the changes in flexibility caused by cAMP.
Functional Relevance of the Allosteric cAMP-dependent Modulations of Dynamics in EPAC1 h -(149 -318)-The cAMPdependent changes in ps-ns and ms-s dynamics that result from allosteric rather than simple binding effects are summarized in the three-dimensional map shown in Fig. 8, which reveals the presence of multiple clusters of residues for which dynamics is either quenched (referred to as "cold patches") or enhanced (defined as "hot patches") by cAMP. Several of these cAMP-dependent dynamic patches match well to the known functional sites of this CBD. For instance, a first striking feature that emerges from Fig. 8 is that the stabilizing effect of cAMP extends well beyond the PBC to the ␤2-␤3 loop. Specifically, the ␤-turn involving the hydrogen bond between the Asp-236 HN and the Ser-233 CO is stabilized only in the presence of cAMP, which also quenches the ms dynamics of Phe-232 located at the C terminus of the highly distorted ␤2-strand ( L231 ϭ Ϫ59°; L231 ϭ Ϫ47°). Interestingly, the PBC and ␤2-␤3 loop residues for which dynamics is detectably quenched by cAMP (i.e. Ala-277, Asp-236, and Phe-232) cluster around the two highly conserved and FIGURE 7. a, three-dimensional map of the dynamic hot spots for EPAC1 h -(149 -318) in its apo state. Unless otherwise specified residues affected by ps-ns dynamics (i.e. HN NOE Ͻ0.5) are marked in yellow, whereas residues affected by ms-s dynamics (i.e. ⌬R 2,eff. Ͼ 2 s Ϫ1 or R 1 R 2 Ͼ 17.6 s Ϫ2 , which is the maximum value conservatively predicted through hydrodynamic simulations) are highlighted in red. Residues Ala-277 and Glu-308 are dynamic in both the ps-ns and ms-s time scales. The dashed lines indicate clusters of dynamic residues involved in the functional regions of this domain. Selected 2°structure elements and functional sites are labeled. PBC stands for phosphate binding cassette. b, opposite side view of a.

Entropy-driven Allosteric Control of EPAC
co-evolved residues, i.e. Gly-238 and Arg-279 (Fig. 8a). These two residues have been proposed to be critical elements of the CBD allosteric network, based on recent extensive evolutionary analyses of genomic CBD sequences revealing that Gly-238 and Arg-279 have co-evolved for the purpose of coupling cAMP binding to distal regulatory regions (55). The chemical shift changes of the Gly-238 site could not be probed through previous spectroscopic analyses (19), but this site is clearly detectable in our spectra (Fig. 4d), which indicate that Gly-238 is highly sensitive to cAMP, although not to the antagonist Rp-cAMPS (Fig. 1d), in full agreement with the allosteric role of Gly-238 anticipated based on the co-evolutionary genomic analyses (55). Furthermore, sequence alignments show that Gly-238 in EPAC1h corresponds to Gly-169 in PKA RI␣ (26), which plays a pivotal role in the cAMP-mediated activation of PKA as revealed by genetic screening (56). Overall, these observations corroborate the functional relevance of the allosteric change in dynamics in this region and suggest that the CHinteraction between the guanidinium of Arg-279 and the C␣ of the Gly-238 (55) may account for the observed dynamic allosteric coupling between the PBC and the ␤2-␤3 loop.
Another cluster of allosteric cAMP-dependent dynamic changes involves the hinge (␣6) region and the adjacent ␣4 helix. Unlike the PBC/␤2-␤3 loop cold patch, the ␣6,␣4 patch involves both hot and cold sites. This mixed pattern is fully consistent with a rearrangement of the packing contacts between the ␣6 and ␣4 helices occurring as a result of the cAMPdependent hinge rotation of ␣6. As shown in Fig. 8b, according to the proposed hinge model of cAMP activation, Asn-301 and Arg-302 in the central moiety of ␣6 approach Val-211 at the N terminus of ␣4 upon cAMP binding, explaining why for these three residues the ms dynamics is quenched by cAMP. However, the cAMP-dependent ␣6 rotation also perturbs the ␣4/6 contacts involving the C termini of these helices, accounting for their enhanced dynamics upon cAMP binding. For instance, the cAMP-dependent hinge rotation of ␣6 brings the N terminus of ␣6 away from Val-218 at the C terminus of ␣4 and simultaneously displaces Lys-305 and Glu-308 located in the C-ter- minal region of ␣6, away from ␣4 (Fig. 8b). Consistent with these observations, for all these three residues (Val-218, Lys-305, and Glu-308) cAMP binding results in increased dynamics (Fig. 8).
Notably, Lys-305, which is the only ␣6 residue for which dynamics is significantly enhanced in the ps-ns time scale (Fig.  8a), marks the beginning of the ␣6 region that is destabilized by cAMP (Fig. 2, a and b) and includes also Glu-308, which is involved in a backbone-to-backbone hydrogen bond with Lys-305. The reduction of ␣-helix probability observed for the 305-309 segment upon cAMP binding (Fig. 2, a and b) reflects, therefore, an overall increase in flexibility at this site that connects the EPAC1 CBD to the EPAC1 catalytic region and is critical to control the relative orientation of the regulatory and catalytic regions (i.e. RR and CR in Fig. 1a), as required for the modulation of the GEF activity of EPAC through steric occlusion of its catalytic site.
Another site playing a pivotal role in the activation of EPAC is the IL, which provides additional RR/CR contact points. Interestingly, two of the three IL residues (i.e. Glu-197 and Asp-172) are part of a third major cluster of residues for which dynamics is enhanced by cAMP. This cluster involves part of the ␣2 helix and mainly the inner side of the ␣1 helix (Fig. 8c). Considering that the chemical shift maps (Figs. 1 and 2) indicate that cAMP alone does not cause major structural rearrangements in the IL spanning region (i.e. helices ␣1 and ␣2), the cAMP-dependent dynamic enhancements observed for Glu-197 and Asp-172 and the adjacent residues suggest that cAMP weakens the inhibitory CR/RR IL interactions primarily by increasing the entropic penalty associated with the formation of the IL salt bridges. It is also possible that this entropic control exerted by cAMP may become even more relevant in full-length EPAC, where the IL dynamics of the apo state is likely to be further quenched by the presence of the catalytic region forming multiple stabilizing salt bridges with the IL (16). Therefore, in full-length EPAC1 the increase in the configuration entropy of the IL region caused by cAMP binding is expected to be even more significant than in the single EPAC1 CBD. In other words, the recognition of cAMP and the formation of the CR/RR IL salt bridges are negatively cooperative events, and dynamics is a key carrier of the allosteric free energy for this negative cAMP/IL cooperativity. Similar allosteric roles of ligand-dependent entropic modulations in the absence of significant structural variations have been previously reported for other systems either on theoretical grounds (57) or based on 15 N NMR relaxation measurements (58,59).
These conclusions imply that the mechanism of EPAC GEF activation by cAMP should be modified to include multiple allosteric pathways that involve not only conformational and dynamic changes in the region C-terminal to the CBD ␤-barrel, as previously proposed (18), but also a cAMP-mediated entropy-driven control of the IL interactions. Fig. 9 shows schematically such a modified mechanism of EPAC activation taking into account the allosteric cAMP-dependent changes in dynamics and their impact on the coupled active/inactive and apo/bound equilibria. Upon cAMP binding, the EPAC equilibrium is shifted toward the active state by the combined action of the hinge motion and of the increased entropic cost for the IL between the CR and the NTHB (Fig. 9) as well as by other possible currently uncharacterized cAMP-dependent perturbations in the lid/Ras exchange motif region. Our data show that the Rp-cAMPS ligand does not activate either the ␣6 hinge motion or the IL entropic weakening, explaining why it functions as an antagonist.
Dynamics and Allostery in PKA Versus EPAC-The NMR 15 N relaxation rates of the PKA RI␣ CBD-A have been previously measured in the cAMP-and in the Rp-cAMPS-bound states but not in the apo form due to its poor solubility (23). The lack of this key reference state for the PKA CBD, therefore, limits our EPAC versus PKA comparative analysis of dynamic profiles to the cAMP and to the Rp phosphorothioate antagonist-bound forms. In PKA the main effect of the oxygen-tosulfur isolobal substitution at the exocyclic equatorial phosphate position is an increase of ms-s dynamics at the PBC and ␤2-␤3 regions (23). Our data indicate that a similar enhancement of ms-s flexibility occurs also in EPAC, as indicated by the NMRD dispersions measured for residues Ala-272 and -277 in the EPAC1 PBC and Phe-232 and Ile-243 in the EPAC1 ␤2-␤3 site, which are consistently higher in the Rp-cAMPSbound state relative to the cAMP-bound form (Fig. 6). Similarly, the R 1 R 2 product for Leu-273 is higher in the antagonistbound state than when the EPAC1 CBD is bound to cAMP (Fig.  4c), pointing to an Rp-cAMPS-specific ms-s dynamic enhancement at this critical PBC site. Furthermore, in EPAC1 the Rp-cAMPS antagonist causes an increase of the ␤2-␤3  Fig. 1; CDB:␤ refers to the CBD ␤-subdomain, and NTHB refers to the N-terminal helical bundle. For the sake of simplicity, the ␣1-helix connecting the DEP to the CBD is considered to be part of the NTHB, and it is not explicitly shown. ␣6 is the hinge-helix, and IL stands for ionic latch. The Ras exchange motif domain is shown with different shapes in the inactive and active states to indicate that it may be subject to conformational changes upon activation (20). The EPAC1 CBD regions in which the overall dynamics is quenched by cAMP are colored red in the apo states and blue in the cAMP-bound states. A reversed color code is used for the EPAC1 CBD regions in which the overall dynamics is enhanced by cAMP. In the apo state EPAC exists in an equilibrium between closed (autoinhibited or inactive) and open (active) states. Upon cAMP binding the equilibrium is shifted toward the open (active) state due to the combined action of the hinge helix rotation and of the weakening of the IL salt bridges due to the increased entropic cost arising from the enhanced dynamics of the NTHB caused by cAMP. REM, Ras exchange motif; RA, Ras-association domain. dynamics in the ps-ns time scale as well. This is supported by the low HN NOE value observed for the ␤2-␤3 loop Asp-236 residue in the antagonist-bound state (Fig. 4d). The enhanced PBC and ␤2-␤3 flexibility observed in EPAC and in PKA by the replacement of cAMP with the Rp-antagonist suggests that for both systems the dynamics at these sites is a key allosteric determinant, further confirming the pivotal role of the ␤2-␤3 loop in the cAMP-dependent allostery of EPAC.
Another dynamic hot spot common to both the EPAC1 and the PKA CBDs (23) is that observed in the cAMP-bound state for the C terminus of the hinge helix after the ␤-barrel (Fig. 8b). In PKA the conformational heterogeneity in this region of CBD-A is likely pivotal in the control of the relative orientations of the two CBDs of the regulatory region (CBD-A and -B) (14), whereas in EPAC the dynamic nature of this site may contribute to the re-orientation of the RR relative to the CR as required for the activation of the GEF function (Fig. 9).
As to the dynamics in the helical bundle, which is N-terminal to the ␤-barrel, at present it is not possible to know whether the cAMP-dependent flexibility enhancements observed in EPAC1 are common to PKA as well due to the limited sequence homology between PKA and EPAC1 in this region. However, in both EPAC and PKA systems the ␣3 site is highly dynamic mainly in the ms-s time scale, and also in both EPAC and PKA the ␣3 flexibility decreases when cAMP is replaced by the Rp antagonist (Figs. 4 and 5) (23). Considering that ␣3 in the PKA CBD-A is the site of key contacts between the regulatory and catalytic subunits (12), these observations suggest that the cAMP-dependent control of the configuration entropy cost for inhibitory interactions, as proposed here for EPAC1, may represent a more general CBD allosteric mechanism common to both signaling units. This conclusion is also supported by a recent structure of the PKA R:C complex (14), revealing that cAMP binding to the PKA CBD-B nucleates a network of stabilizing interactions in the C-terminal helices while concurrently causing the disruption of a critical salt bridge that stabilizes the N-terminal helical bundle of CBD-B (14). Overall such a mechanism of cAMP activation proposed for PKA (14) points to the existence of a set of mutually exclusive intra-CBD interactions in full agreement with the enhanced dynamics observed here upon cAMP binding for the N-terminal helices of EPAC1.
Conclusions-We have mapped by classical 15 N relaxation and NMRD experiments the dynamic profiles of the EPAC1 CBD in its apo and cAMP-and Rp-cAMPS-bound states. Such a three-state comparative analysis has revealed that cAMP-dependent variations of dynamics in the ms-s and ps-ns time scales are key carriers of allosteric free energy in this domain. Specifically, we observe positive cooperativity between the PBC and the ␤2-␤3 region, whereby cAMP docking in the PBC results in an extended rigidity of the ␤2-␤3 loop. This turn region emerges as a key allosteric hot spot, in full agreement with the recent finding about the co-evolution of conserved residues at the PBC and ␤2-␤3 sites. In sharp contrast to the quenching of dynamics in the ␤2-␤3 locus upon cAMP-binding, several residue clusters were found in the ␣-helical subdomain in which dynamics was allosterically enhanced by cAMP. These include not only the C-terminal region of the hinge helix that plays a critical role in defining the RR/CR relative orientation but also the ionic latch spanning region in helices ␣1-2 that mediates key GEF inhibitory interactions. Considering that cAMP binding alone does not cause any significant structural rearrangement for ␣1-2, the increase of conformational entropy promoted at this site by cAMP emerges as a key mechanism for a cAMP-dependent weakening of the inhibitory salt bridges mediated by the ionic latch. Based on these results we have proposed a mechanism of EPAC activation that incorporates the dynamic features allosterically modulated by cAMP. According to this model, the cAMP-dependent entropic control of the ionic latch represents an additional allosteric pathway that acts in concert with the previously proposed ␣6 hinge motion to shift the EPAC equilibrium toward the active (open) state, removing the steric hindrance exerted by the RR on the CR and exposing the catalytic core to the downstream Rap effector. Although a similar three-state comparative analysis of dynamic profiles is currently unavailable for PKA, the existing data suggest that the proposed model for the entropy-driven allosteric cAMP control of the N-terminal helical bundle may be at least in part generalized to the CBDs of PKA.