Identification of a tetrahydroquinoline analog as a pharmacological inhibitor of the cAMP-binding protein Epac

Background : The guanine exchange factor Epac is a cAMP sensor. non-selective β-adrenergic receptor agonist. Our study shows that CE3F4 and related


Background:
The guanine exchange factor Epac is a cAMP sensor.

Results:
We have identified a tetrahydroquinoline analog named CE3F4 that blocks Epac activation in response to cAMP in vitro and in living cultured cells.

Conclusion:
CE3F4 behaves as an uncompetitive antagonist of Epac with respect to cAMP. Significance: CE3F4 may serve as a basis for the development of new therapeutic drugs.

SUMMARY
The cAMP-binding protein Epac is a therapeutical target for the treatment of various diseases such as cardiac hypertrophy and tumor invasion. This points out the importance to develop Epac inhibitors to better understand the involvement of these cAMP sensors in physiology and pathophysiology. Here, we have developed a functional fluorescence-based highthroughput assay with a Z' value around 0.7 for screening Epac specific antagonists. We identified an Epac1 inhibitor compound named CE3F4 that blocked Epac1 guanine nucleotide exchange activity towards its effector Rap1 both in cell-free systems and in intact cells. CE3F4 is a tetrahydroquinoline analog that fails to influence PKA holoenzyme activity. CE3F4 inhibited neither the interaction of Rap1 with Epac1 nor directly the GDP exchange on Rap1. Kinetics of inhibition by CE3F4 indicated that this compound did not compete for binding of agonists to Epac1, and suggested an uncompetitive inhibition mechanism with respect to Epac1 agonists. A structure-activity study showed that the formyl group on position 1 and the bromine atom on position 5 of the tetrahydroquinoline skeleton were important for CE3F4 to exert its inhibitory activity. Finally, CE3F4 inhibited Rap1 activation in living cultured cells, following Epac activation by either 8-(4chlorophenylthio)-2'-O-methyl-cAMP, an Epac selective agonist, or isoprenaline, a nonselective β-adrenergic receptor agonist. Our study shows that CE3F4 and related Epac in cAMP signaling and to better understand its pathophysiological roles, it is crucial to further develop Epac inhibitors and to test their functional effects in living cultured cells. To this end, we established a high throughput screening (HTS) assay to isolate Epac inhibitor compounds from chemical libraries. We identified an Epac inhibitor compound named CE3F4 that blocked Epac guanine nucleotide exchange activity towards its effector Rap1 in vitro. Although the Epac1 regulatory domain was necessary for CE3F4 to inhibit Epac1 activity, the compound did not competitively inhibit agonist activation of Epac1. Importantly, CE3F4 inhibited Epac downstream effector, Rap1 following Epac activation by either an Epac selective agonist, 007, or isoprenaline (Iso), a non-selective βadrenergic receptor agonist in living cultured cells. Our study shows that CE3F4 and related compounds may serve as a basis for the development of new therapeutic drugs.

EXPERIMENTAL PROCEDURES
Compound library -The screening library is a collection of 640 chemically diverse compounds called "Chimiothèque Essentielle" (CE), which was developed as a representative part of the French National Chemical Library (http://chimiotheque-nationale.enscm.fr), the latter containing more than 50,000 compounds. The CE library was provided on the basis of 80 compounds per 96-well plate, at a concentration of 1 mg/mL in 100% DMSO. The compounds were then diluted to 0.1 µg/mL with H 2 O, and the "daughter" plates were kept at -20°C until use.
Protein expression and purification -Part of the coding sequence of human Epac1 (residues 149-881, i.e. Epac1 deleted of its DEP (Dishevelled, Egl-10, Pleckstrin) domain, was PCR-amplified using the plasmid pSC-A-HA-hEpac1 as a template, and inserted in pET41a (Novagen). The coding sequence of the catalytic domain of Epac1 (residues 321-881) was PCR-amplified from the same plasmid. The full-length coding sequence of human Rap1A was excised from the plasmid pRK5-HA-hRap1A and inserted in pET41a. All constructs were verified by sequencing and expressed in Escherichia coli Rosetta 2(DE3) (Novagen) as recombinant proteins fused with glutathione S-transferase (GST) and a His-tag at their N-terminus, according to standard protocol. After lysis of the bacteria, the soluble proteins were purified by glutathione-agarose beads (GE healthcare) for Rap1A and by Ni-NTA beads (Qiagen) for Epac1 and Epac1-Cat. Thrombin was used to cleave GST from Rap1 (11). All proteins used in this study were at least 80% pure, as judged by SDS-PAGE.

Measurement of in vitro activation of Epac1 -
GST-Rap1 was preloaded with a fluorescent derivative of GDP, and the Epac1-catalyzed nucleotide exchange was measured using a large excess of non-fluorescent GDP, by taking advantage of the spectroscopic difference between free and Rap1-bound fluorescently labeled GDP. The principle of the method is similar to that reported in particular by van den Berghe et al. (12), except that Bodipy FL-labeled GDP (bGDP) was used here, rather than 2′,3′bis(O)-N-methylanthraniloyl-GDP (mant-GDP). Indeed, the change in fluorescence between free and Rap1-bound GDP is broadly two-fold higher using bGDP as a ligand rather than mant-GDP. Furthermore, replacing the blue-fluorescent mant-GDP by the green-fluorecent bGDP results in a strong decrease of autofluorescence background generated by the medium and by some compounds under study (data not shown).
Concerning detailed kinetic studies, a "multipoint" (time-course) method was used. All components except the agonist (007 or cAMP) were mixed in a well and the exchange activity of Epac1 was initiated by injection of the agonist. The release of bGDP was measured in real time as the decay of fluorescence. A single exponential was fit to the data using the program Graphpad Prism. The initial velocity (V i ) of GDP exchange was calculated as K(Y0-P), where K, Y0, and P are the best fits for, respectively, the first order rate constant, the fluorescence value at t=0, and the fluorescence plateau value at infinite times.
A "two-points" method was used to screen the CE library. Each 384-well plate received 320 test compounds (final concentration: 6.67 µg/mL). On one side of the plate, 32 wells received no compound addition (as negative inhibition controls), whereas, on the other side of the plate, 32 wells received 009 at 50 µM final concentration (as positive inhibition controls). In all cases the final DMSO concentration was 0.67%. All the liquid handlings and plate operations were performed using a robotic workstation Biomek FX (Beckman Coulter) coupled to the multilabel plate reader. The fluorescence of bGDP was measured just before (t=0) the injection in each well of a 007 solution (2 µM final concentration), and 4 min after the injection. The variation in fluorescence (∆F) was directly considered as a measurement of the exchange activity of Epac1. The inhibition of the exchange activity induced by the test compounds was expressed as the percentage of the exchange activity promoted by 50 µM 009: % control = (∆Fmax -∆Fobs) / (∆Fmax -∆Fmin) x 100, where ∆Fmax = mean response (n = 32) obtained with 007 alone, ∆Fmin = mean response (n = 32) obtained with 007 plus 009, and ∆Fobs = observed response obtained with 007 plus test compound. The "two-points" method, allowed reading the 384 wells of a plate in less than 10 min.  (13). The CAMYEL construct was transferred to the pcDNA3 vector by usual restriction-ligation procedures. In the original CAMYEL construct, Epac1 contains two point mutations, T781A and F782A, which eliminate the guanine nucleotide exchange activity of Epac1. These mutations were reversed in pcDNA3-EBS by replacing the coding sequence between aminoacids 750 and 881 by that of wild-type Epac1, using conventional restriction-ligation procedures at BamH1 and EcoRI restriction sites. BRET signals were studied essentially as described by Jiang et al (13). Briefly, the pcDNA3-EBS plasmid was transiently expressed in HEK293 cells. Cells were seeded in 6-well plates (600 000 cells/well) and transfected (3 µg DNA/well) using Fugene HD (Roche Applied Science, Meylan, France) as a transfection reagent. Cells were lyzed 24 h post-transfection in 400 µL of buffer per well (20 mM Hepes pH 7.4; 50 mM KCl; 50 mM NaCl; 2.5 mM MgCl 2 ; 0.2% NP40; 5 mM DTT and protease inhibitor cocktail), and a supernatant was collected after centrifugation at 16,000 g for 15 min at 4°C. Test compounds and coelenterazine-h (2 µM final concentration) were added 7 min before injection of Epac1 agonists in wells of a white 384-well plate (50 µL final volume). The emission signals from Renilla luciferase and citrine-cp229 were recorded over time in a Perkin-Elmer Envision Xcite multilabel plate reader. The BRET ratio (mean ± SEM, n=3) was calculated as the ratio between the signal emitted by citrine-cp and that emitted by Renilla luciferase. The binding of agonists to the EBS was measured as a reduction in the BRET ratio.

Epac1-based intramolecular BRET assay -
Cell Culture -All procedures for cardiac myocyte isolation were performed in accordance with the Guide for the care and use of laboratory animals and the veterinary committee has been informed of the myocyte isolation protocol used. Cardiac myocytes were isolated as previously described by Wollert and colleagues (14). HEK293 cells stably expressing the β 1adrenergic receptor (β 1 -AR) was a gift of respectively Dr. Shenoy (Duke University). HEK293 cells were maintained in MEM with FBS (Foetal Bovine Serum; 10%) and penicillin-streptomycin (1%). All media, sera and antibiotics used in cell culture were purchased from Invitrogen (Cergy Pontoise, France).
Plasmid and transfection -Epac1 and Rap1A plasmid constructs (Epac1 WT ) were generously provided by Drs J. L. Bos and J. de Gunzburg, respectively. The pQE30-CAMYEL plasmid was a gift of Dr Lily I. Jiang. Transient transfection experiments of HEK cells and primary cardiac myocytes were performed with respectively Xtreme GENE 9 reagent (Roche Applied Science) and Lipofectamine 2000 (Invitrogen Life Technologies) in the presence of various amounts of plasmid constructs according to the manufacturer's instructions.
Western blot analysis -Cells were rinsed once in cold PBS, scraped and lysed in RIPA buffer (PBS 1X; 1% Igepal CA-630; 0.5% sodium deoxycholate; protease inhibitors). Cell lysates were analysed by one-dimensional electrophoresis on 15% SDS-polyacrylamide gels. The proteins were transferred to Immobilon membranes (Millipore). The membranes were blocked for 1 h at room temperature in Tween-Tris buffer saline (TTBS) supplemented with either 5% of non-fat milk or 3% of bovine serum albumin (BSA), and then probed with primary antibodies raised against Rap1 or Epac1. The second antibody was coupled to horse radish peroxidase (Santa Cruz Biotechnology, USA). Membranes were revealed with Dura kit (Pierce).
PKA activity assay -PKA kinase activity assay (Biaffin GmbH ref [PKA-COOK]) was performed according to the manufacturer's instructions. Briefly, type I and II PKA holoenzyme (Biaffin GmBh) was added at 5 nM with or without CE3F4 (20 µM), H89 (20 µM, Sigma-Aldrich) or cAMP (0.5 µM, Biolog). The reaction was initiated by the addition of the Kemptide peptide and absorbance was measured at 340 nm every 30 s during 10 min on a spectrophotometer (Biorad). PKA activity was determinated in 3 steps: decreasing absorbance was plotted against time (min), then a linear regression of the plot was performed and the absolute slot value was used to calculate PKA activity.
Statistical analysis -All data are expressed as means ± standard error of the mean. Differences in quantitative variables were examined by oneway analysis of variance (ANOVA) or paired two-tailed t test. p value<0.05 (*), p value<0.01 (**) and p value<0.001 (***). All analyses were performed using GraphPad Prism.

RESULTS AND DISCUSSION
In vitro assay of Epac guanine nucleotide exchange activity toward Rap1 -To search for Epac specific inhibitors, we developed a variant of the assay described by van den Berghe et al.
(12) that is based on the ability of Epac to catalyze the nucleotide exchange activity of Rap1 (see methods for details). Rap1 was preloaded with Bodipy-GDP (bGDP), a fluorescent derivative of GDP, and the exchange for non-fluorescent GDP, present in a large excess in the incubation medium, was measured due to the spectroscopic difference between free and Rap1-bound labeled GDP. The exchange of guanine nucleotide bound to Rap1 was measured in real time as a decrease in fluorescence (Fig. 1). Epac1 induced a rapid decrease in fluorescence in the presence of 2 µM 007, a strong Epac agonist (6) , whereas it had no effect in the absence of 007. When we compared the GDP exchange properties of thrombin-cleaved and uncleaved GST-Rap1 (Fig. 1A), the calculated initial velocity (Vi) of exchange induced by Epac1 and 007 was very similar using either preparations of Rap1 (2.8 x 10 4 RLU/second with GST-Rap1 versus 2.7 x 10 4 RLU/second with cleaved Rap1). The EC 50 of 007 was also measured using thrombin-cleaved Rap1 (not shown) and found to be 2 µM. This EC 50 is in agreement with the value of 1.8 µM reported by Rehmann et al. (11). This observation showed that the GST tag did not impair the exchange reaction and therefore experiments were performed with uncleaved GST-Rap1 (referred to as Rap1). The cGMP analog 8-(4-chlorophenylthio)guanosine-3',5'-cyclic monophosphate (referred here to as "009") has been previously reported to behave as an inhibitor of cAMP-induced Epac activation in vitro (16,17). Fifty micromolar 009 strongly reduced the release of Rap1-bound fluorescent bGDP induced by 2 µM 007 (Fig.  1B). The recombinant catalytic domain of Epac1 (Epac1-Cat), lacking Epac1 cAMP-binding domain, is known to behave as a constitutive activated form of Epac1 (18). Fig. 1B shows that Epac1-Cat had constitutive exchange activity in the absence of 007 and that 009 (50 µM) had no inhibitory effect on Epac1-Cat, as expected for an inhibitor that would compete with agonists at the cAMP binding site of Epac1. Therefore, the combined use of recombinant Epac1 and Epac1-Cat allows discriminating in vitro Epac1 inhibitors that would exert their effect on the Epac1 regulatory domain from those that would inhibit the interaction of Epac1 with Rap1 and/or directly inhibit nucleotide exchange on Rap1.
Dose-response analysis of cyclic nucleotide analogs on Epac1 activation -To further define the experimental conditions allowing screening for Epac1 pharmacological inhibitors, the kinetic characteristics of the in vitro reconstituted Epac1-Rap1 system were studied in more detail. Measurement of initial velocities of exchange at increasing concentrations of 007 (Fig. 1C) gave an EC 50 value of about 2 µM. This concentration of 007 was therefore used to study the antagonism promoted by increasing concentrations of 009 (Fig. 1D). The IC 50 of 009 for inhibition of 2 µM 007-stimulated Epac1 activity was ≈17 µM, suggesting that 009 had a 7-8-fold lower affinity for the cAMP binding site of Epac1 compared to 007. Using Graphpad Prism, the IC 80 (i.e. the concentration of 009 that promoted 80% inhibition of Epac1 activity) was by guest on March 24, 2020 http://www.jbc.org/ Downloaded from CE3F4 inhibits Epac activation found to be around 50 µM. Finally, the inhibitory effect of 25 µM 009 was completely suppressed by 100 µM excess of 007 (data not shown), indicating the competitive nature of Epac1 inhibition by 009. A sub-saturating concentration of 007 (2 µM) was therefore chosen for subsequent screening experiments in order to avoid that an excess of this strong Epac1 agonist could prevent inhibition by relatively weak competitive inhibitory compounds.
Validation of the assay in the 384-well plate format -These studies were performed using a "two-points" robotized method (see Methods for details). In brief, the variation in fluorescence of bGDP in the presence of Epac1 was measured 4 min after the injection of a 007 solution, and was directly considered as a measurement of the exchange activity of Epac1. In the context of the present inhibition assay, the "max" value was obtained with 2 µM (EC 50 ) of the agonist 007, and the "min" value was obtained with the same concentration of 007 plus 50 µM (the IC 80 ) of the standard in vitro antagonist 009. A mean max value and a mean min value were determined from 64 wells of a 384-well plate in each condition (not shown), allowing the calculation of a Z'-factor. The calculated Z'-factor was ≈ 0.7, fulfilling the accepted criterion of a value ≥ 0.5 (19) and indicating that the assay was appropriate for screening.
Screening for inhibitors of Epac1 activity using the CE sub-library from the French National Chemical Library -The "two-point" robotized assay validated in the 384-well format was used to screen the 640 compounds of the CE library for their potency to inhibit Epac1 exchange activity. Compounds were blind screened, the properties, structures, and molar concentrations of the compounds being unknown at this stage. Fig. 2A shows the scatter plot analysis of singleconcentration (6.67 µg/mL) screening data. The potency of each library compound was expressed as the % of inhibition promoted by 50µM 009 and the positive inhibition control was set at 100%. One compound, identified as CE3F4 strongly deviated from the mean response of the other compounds, showing >100 % of the inhibition promoted by 50 µM 009 (Fig. 2A). The inhibition promoted by CE3F4 differed by ≈10 standard deviations from the mean inhibition promoted by all of the 640 compounds (i.e. Z score was ≈10). Its formula was obtained from the French National Chemical Library, and is shown in Fig. 2B. The name of this tetrahydroquinoline analog is 5,7-dibromo-6fluoro-2-methyl-1,2,3,4-tetrahydroquinoline-1carbaldehyde and its molecular mass is 351,01 Da, indicating that CE3F4 was tested at ≈19 µM in the screening assay. To exclude a possible false positive response, the activity of CE3F4 was confirmed using the "multipoint" (timecourse) assay (see Methods), in triplicate and at the same concentration that was used in the screening process, first from a "daughter" library plate and further from a new batch of the crystallized compound provided by the French National Chemical Library (data not shown).

Kinetic characteristics of Epac1 inhibition by CE3F4 -
The effects of CE3F4 on the exchange activity of Epac1 and Epac1-Cat were studied manually using the "multipoint" assay. Fig. 3A showed that the exchange reaction induced by Epac1 activated by 007 (2 µM) was inhibited to a similar extent by CE3F4 (20 µM, close to the concentration used in the screening) and by 009 (25 µM). In contrast, neither CE3F4 nor 009 had any inhibitory effect on the constitutive exchange activity of Epac1-Cat (Fig. 3B). The exchange activity of Epac1 induced by 007 (2 µM), was reduced by CE3F4 in a concentrationdependent manner (IC 50 = 23 ± 3 µM), whereas the dose-response study confirmed the lack of inhibition of Epac1-Cat by CE3F4 (Fig. 3C). This result indicated that CE3F4 inhibited neither the interaction of Epac1 with Rap1 nor directly the GDP exchange on Rap1.
The Epac1 regulatory domain was therefore necessary for CE3F4 to inhibit Epac1 activity. One may speculate that CE3F4 could impair the binding of 007 to the regulatory domain of Epac1, either by direct competition or by an allosteric mechanism. Alternatively, it might impede the agonist-induced conformational changes that are necessary to relieve the auto-inhibitory mechanisms that allow Epac1 to adopt its active conformation (11,16), without interference with agonist binding. This latter hypothesis was supported by the results reported in Fig. 4A, showing that, in contrast to 009, the residual exchange activity of Epac1 in the presence of 20 µM CE3F4 was independent of the concentration of 007 (2 µM or 20 µM) used to activate Epac1. In other words, an excess of agonist did not relieve the inhibition promoted by CE3F4. CE3F4 had also a similar inhibitory effect on the activation of Epac1 by 50 µM cAMP (not shown), even if by guest on March 24, 2020 http://www.jbc.org/ Downloaded from CE3F4 inhibits Epac activation cAMP at this concentration had only half the potency of 2 µM 007 to activate Epac1. Fig. 4B shows that increasing 007 concentration from 2 to 20 µM resulted in a decrease of CE3F4 IC 50 from 22.5 ± 1 µM to 15 ± 2 µM (means ± SEM of 3 independent experiments). This suggests that 007 binding was required to stabilize CE3F4 binding to Epac1, a property analogous to uncompetitive inhibition that takes place when an enzyme inhibitor binds only to the complex formed between the enzyme and the substrate. Since both V max and K m are reduced in enzymatic uncompetitive inhibition (20), we studied the kinetic characteristics of Epac1 activation by 007 in the absence and presence of CE3F4 (Fig. 4C). CE3F4 (20 µM) decreased both the V max of GDP exchange and the EC 50 for 007 by approximately 80%. At 50 µM CE3F4, V max and EC 50 were too small to be computed. Taken together, these data suggest that CE3F4 likely acted uncompetitively with respect to the Epac1 agonist 007.
Chemical specificity of Epac1 towards CE3F4 analogs -Several analogs or derivatives of CE3F4, whose formulas are shown in Fig. 5A, were studied for their potential to prevent Epac1 activation. Epac1 was stimulated by 20 µM 007 in the absence or presence of 20 µM of each compound, and the initial velocity of GDP exchange on Rap1 was measured (Fig. 5B). CE3F4 (Compound 1) showed the strongest inhibitory effect. Compound 2, which lacks the formyl group on position 1 had no inhibitory effect, as it was the case for Compound 3 and Compound 4 which lack the bromine atoms on both positions 5 and 7: Compound 3 is substituted by a formyl group on carbon 5, whereas Compound 4 has two cyano groups on positions 5 and 7. Finally, the removal of only one bromine atom on position 7 (Compound 5) reduced significantly but did not suppress inhibition compared to CE3F4, whereas the selective removal of the bromine atom on position 5 (Compound 6) completely hampered the inhibitory potential of the original molecule. This preliminary structure-activity study shows that compounds having the formyl group on position 1 and the bromine atom on position 5 seem to be the most efficient molecules to block Epac activation. This does not exclude that other derivatives at these positions may influence Epac1-GEF activity. CE3F4 and its derivatives are tetrahydroquinoline analogs which are chemically unrelated to the series of pyrimidine analogs identified by Chen et al (9), based on their direct competition with a fluorescent cAMP derivative in binding to Epac1 and Epac2. Effect of agonists and antagonists on an Epac1based BRET sensor -Next we investigated the inhibitory effect of CE3F4 on an Epac1-based BRET sensor (EBS), consisting of human Epac1 inserted between citrine and Renilla luciferase as the BRET pair (see Methods for details). Binding of agonists to the EBS induces a conformational change in the Epac1 part, resulting in a decrease of energy transfer from the luciferase to citrine. HEK293 cells were transfected with an EBSexpressing construct, and the changes of BRET ratios were then measured in cell extracts upon binding of Epac1 agonists and antagonists. Fig.  6A showed that the BRET ratio abruptly dropped by ≈2-fold after addition of 10 µM 007. To further test whether the EBS assay was a valuable tool to evaluate conformational changes undergone by Epac1 upon binding of agonists, BRET ratios were measured in cell extracts in response to increasing concentrations of either 007 or cAMP (Fig. 6B). Both nucleotides induced remodelling of Epac1 conformation reflected by dose-dependent variations in the measured BRET ratio. The estimated EC 50 for 007 and cAMP were 0.6 µM and 50 µM, respectively. Both 009 (50 µM) and CE4F4 (40 µM) inhibited the BRET variations induced by 10 µM 007 (≈ 25% decrease) or 100 µM cAMP (≈ 50% decrease). Compound 2, a CE3F4 analog devoid of antagonistic activity towards Epac1 guanine nucleotide exchange activity (Fig. 6B), had no inhibitory effect on conformational changes induced by either 007 or cAMP (Fig.  6C). These results constitute an independent confirmation that CE3F4 specifically inhibits Epac1 guanine nucleotide exchange activity without interference with Rap1 activity or Epac1-Rap1 interaction. Finally, we found that CE3F4 did not influence PKA activity in the presence or absence of cAMP (Fig. 7) indicating that this compound was specific of the cAMPbinding protein Epac.
Effects of CE3F4 on Epac1 downstream effector in living cultured cells -As a next step, we tested the ability of CE3F4 to block Epac1induced Rap1 activation in living cultured HEK293 cells. For these experiments, we used a membrane-permeant Epac agonist, Sp-8-pCPT-2′-O-Me-cAMPS (Sp-007) (21). We found that Sp-007 (10 µM) had almost no effect on Rap1 activation due to low expression level of Epac1 in this cellular system (Fig. 8A). However, Sp-007 induced a robust activation of Rap1 in cells overexpressing Epac1 compared to control cells transfected with the empty vector (Fig. 8A). Consistent with the data obtained in vitro on Epac exchange reaction, CE3F4 (20 µM) prevented the increase in the amount of Rap1-GTP following 10 µM Sp-007 treatment (Fig.  8B). Similar findings were observed in primary cardiac myocytes as Epac-induced Rap1 activation was blocked by CE3F4 (20 µM) (Fig.8C). Finally, as β-adrenergic receptors (β-ARs) activate Epac in cardiac myocytes and other cell types (15,22), we tested the effects of CE3F4 on Epac1-induced Rap1 activation following stimulation of β 1 -adrenergic receptor (β 1 -AR) by isoprenaline (Iso, 10 µM), a nonselective β-AR agonist. As shown in Fig. 8D, we found that CE3F4 decreased Epac1-induced Rap1 activation in β 1 -AR overexpressing HEK293 cells treated with Iso (Fig. 8D). Altogether these data showed that CE3F4 was efficient in preventing Epac1-induced Rap1 activation in cultured cells.
In conclusion, using a validated fluorescence-based high-throughput screening assay we have identified a pharmacological tetrahydroquinoline analog, named CE3F4, that displays Epac1 antagonistic activity. CE3F4 behaves as an uncompetitive antagonist of 007 and cAMP with inhibition working best when 007 (or cAMP) concentration is high. CE3F4 blocks Epac1 guanine nucleotide exchange reaction on Rap1 in intact cells. Because CE3F4 did not exert its antagonism by directly competing with cAMP for binding to Epac1, it is difficult to predict whether it could also inhibit Epac2. More work is necessary to explore this possibility. At the moment, our data show that CE3F4 and related compounds may provide valuable pharmacological tools for determining the biological functions of Epac1 and for better understanding the involvement of Epac1 in the manifestation of diseases, such as cardiac hypertrophy, tumour invasion and inflammation.

ACKNOWLEDGMENTS
F. Lezoualc'h was supported by grants from Agence Nationale de la Recherche (« HyperEpac » Genopath09), Région Midi-Pyrénées, A.F.M and Fondation pour la Recherche Médicale (Programme Cardiovasculaire). J.P. Blondeau was supported by a grant from Centre National de la Recherche Scientifique (PIR Innovation Thérapeutique 2010). We are indebted to Dr Lily I. Jiang for providing the pQE30-CAMYEL plasmid.    Fluorescence was also recorded in the absence of Epac1-Cat (squares). (C) Initial velocities of nucleotide exchange induced by Epac1 together with 2 µM 007 (circles) or by Epac1-Cat (squares) were measured in the presence of increasing concentrations of CE3F4. IC50 was calculated using Graphpad Prism.  The initial velocity of Epac1-catalyzed GDP exchange was measured in the absence (Ctrl-) or presence (Ctrl+) of 20 µM 007, and the latter value was set at 100% exchange activity. CE3F4 and Compounds 2-6 were added just before the agonist 007, and the resulting exchange activity was expressed as the percentage of that measured under the Ctrl+ condition. Values are mean ± SEM (n=3). Student t tests : (*) p < 1% versus Ctrl + ; (ns) p > 5% versus Ctrl +; ( ¶) p < 1% versus CE3F4.  (2), or vehicle (no inh.) were added to cell extract before injection of 007 (10 µM) or cAMP (100 µM), and BRET ratios (mean ± SEM, n=3) were measured as in (B) and plotted as % variations in BRET ratios relative to each noinhibitor control value. p < 1% versus control value without 007 (* mark) or control value without cAMP ( ¶ mark).   HEK293 cells (A,  B) and rat neonatal cardiac myocytes (C) were transfected with Epac1 expression vector. HEK293 cells were also transfected with pcDNA3 control vector. 24 h after transfection, cells were preincubated or not with CE3F4 for 30 min and were then treated or not with Sp-007 (10 µM) for 10 min. (D), HEK293 cells overexpressing β1AR and transfected with Epac1 were pretreated or not with CE3F4 as in (B) and stimulated or not with Iso (10 µM) for 10 min. Amounts of Rap1-GTP were determined by pull-down assays. A control for total Rap expression is shown. Expression of Epac1 was measured in cell lysates. The bar graph represents the mean ± S.E.M. of 5 (A, B) or 3 (C, D) independent experiments. ***p<0.001, **p<0.01, *p<0.05 between indicated conditions, paired twotailed t test. Results are expressed as the percentage of unstimulated control cells.