Novel Epac fluorescent ligand reveals distinct Epac1 vs. Epac2 distribution and function in cardiomyocytes

Significance β-Adrenergic activation in heart has many important effects (physiological and pathological) that had been mainly attributed to cAMP and protein kinase A-dependent signaling. The discovery of exchange protein directly activated by cAMP (Epac1 and Epac2) has uncovered new parallel cAMP signaling pathways in β-adrenergic signaling, but Epac isoform localization and function were not well understood. Here we show, using a novel fluorescent Epac analog, that Epac1 and Epac2 are differentially concentrated at the nucleus vs. at Z lines in cardiomyocytes, and that this corresponds to functional roles in nucleus vs. calcium handling and arrhythmias, respectively. This finding may be essential to develop more targeted therapeutics for cardiomyopathy. Exchange proteins directly activated by cAMP (Epac1 and Epac2) have been recently recognized as key players in β-adrenergic–dependent cardiac arrhythmias. Whereas Epac1 overexpression can lead to cardiac hypertrophy and Epac2 activation can be arrhythmogenic, it is unknown whether distinct subcellular distribution of Epac1 vs. Epac2 contributes to differential functional effects. Here, we characterized and used a novel fluorescent cAMP derivate Epac ligand 8-[Pharos-575]-2′-O-methyladenosine-3′,5′-cyclic monophosphate (Φ-O-Me-cAMP) in mice lacking either one or both isoforms (Epac1-KO, Epac2-KO, or double knockout, DKO) to assess isoform localization and function. Fluorescence of Φ-O-Me-cAMP was enhanced by binding to Epac. Unlike several Epac-specific antibodies tested, Φ-O-Me-cAMP exhibited dramatically reduced signals in DKO myocytes. In WT, the apparent binding affinity (Kd = 10.2 ± 0.8 µM) is comparable to that of cAMP and nonfluorescent Epac-selective agonist 8-(4-chlorophenylthio)-2-O-methyladenosine-3′-,5′-cyclicmonophosphate (OMe-CPT). Φ-O-Me-cAMP readily entered intact myocytes, but did not activate PKA and its binding was competitively inhibited by OMe-CPT, confirming its Epac specificity. Φ-O-Me-cAMP is a weak partial agonist for purified Epac, but functioned as an antagonist for four Epac signaling pathways in myocytes. Epac2 and Epac1 were differentially concentrated along T tubules and around the nucleus, respectively. Epac1-KO abolished OMe-CPT–induced nuclear CaMKII activation and export of transcriptional regulator histone deacetylase 5. In conclusion, Epac1 is localized and functionally involved in nuclear signaling, whereas Epac2 is located at the T tubules and regulates arrhythmogenic sarcoplasmic reticulum Ca leak.

Indeed, in MIN6-insulin secreting cells, cNBD-A was required for targeting Epac2 near the plasma membrane (16) and in PC12 cells, the Epac1 catalytic region was suggested to be responsible for nuclear targeting (17). In other cell lines Epac1based cAMP FRET sensors or expressed Epac1 tagged with GFP (Ad-Epac1-GFP) expressed at the cytosol, plasma membrane, and mitochondria (16)(17)(18)(19)(20)(21)(22), and moved to the plasma membrane upon activation (23). In adult cardiomyocytes, subcellular localization and function of endogenous Epac1 and Epac2 are not yet clear.
In neonatal and adult cardiomyocytes, Ad-Epac1-GFP expresses at T tubules, cytosol, and nuclear membrane (6,7). Epac1 localization at perinuclear membranes of hypertrophic neonatal cardiomyocytes suggests its role in maladaptive hypertrophy (24,25). Consistent with that, Epac1 mRNA is up-regulated in human heart failure and animal models of hypertrophy (induced by pressure overload or catecholamines). Epac1 also seems to be part of a nuclear mAKAP/PKA/PDE4/ErK5 complex to regulate cAMP signaling (24). Indeed, Epac activation by adenoviral Epac1 expression or the specific Epac agonist 8-(4-chlorophenylthio)-2-O-methyladenosine-3-,5-cyclic monophosphate (OMe-CPT), can activate hypertrophic transcription regulators MEF2 and NFAT, via phospholipase C (PLC)/IP 3 /Ca/Ca-calmodulin kinase II (CaMKII)/ HDAC4-HDAC5 (5,11) and H-Ras/Ca/calcineurin pathways (6, 7), respectively. Epac activation in myocytes influences both nuclear and cytosolic Ca signaling modifying transcription and EC coupling and arrhythmias. For example, sustained Epac activation increases expression of calmodulin, leading to Ca mishandling and Significance β-Adrenergic activation in heart has many important effects (physiological and pathological) that had been mainly attributed to cAMP and protein kinase A-dependent signaling. The discovery of exchange protein directly activated by cAMP (Epac1 and Epac2) has uncovered new parallel cAMP signaling pathways in β-adrenergic signaling, but Epac isoform localization and function were not well understood. Here we show, using a novel fluorescent Epac analog, that Epac1 and Epac2 are differentially concentrated at the nucleus vs. at Z lines in cardiomyocytes, and that this corresponds to functional roles in nucleus vs. calcium handling and arrhythmias, respectively. This finding may be essential to develop more targeted therapeutics for cardiomyopathy. enhancing susceptibility to arrhythmia (12). Epac-dependent arrhythmia is also observed during acute Epac activation through CaMKII-dependent activation of diastolic Ca leak from the sarcoplasmic reticulum (SR) due to ryanodine receptor (RyR) phosphorylation and involving PLC, Rap, and IP 3 receptors (4,(8)(9)(10). We recently showed that activation of Epac2, but not Epac1, induces β1-AR-induced arrhythmias through CaMKIIdependent diastolic SR Ca release (14). We hypothesize that cardiac Epac1 and Epac2 mediate different functions, in part because of distinct subcellular localization. However, available Epac antibodies may lack specificity in immunolocalization studies, rendering information on Epac isoform localization equivocal.
Here, we characterize a novel commercially available fluorescent Epac ligand that has not been previously described, 8-[Pharos-575]-2′-O-methyladenosine-3′,5′-cyclic monophosphate (Φ-O-Me-cAMP). We characterized and used it to clarify Epac isoform localization in adult ventricular myocytes, in mice in which one or both Epac isoforms were knocked out [Epac1-KO, Epac2-KO, and double knockout (DKO) mice] (14). We found that Epac2 is highly concentrated at Z lines in cardiomyocytes, consistent with our study implicating Epac2 selectivity in SR Ca-dependent arrhythmogenesis. In contrast, Epac1 was more concentrated at the nuclear envelope. This is consistent with our findings here that Epac-dependent nuclear export of histone deacetylase 5 (HDAC5) is abolished in Epac1-KO mice and mediated by β1-AR. Thus, differential localization of Epac isoforms strongly influences their functional roles in cardiac myocytes.
Φ-O-Me-cAMP Binding and Effects on Nucleotide Exchange. We used purified proteins to measure Epac-dependent nucleotide (GDP) exchange on Rap1, a known proximal Epac effector. Φ-O-Me-cAMP partially activated Epac1, but only at concentrations above 10 μM, and to much lower activity than for 100 μM cAMP (Fig. 1D). When both ligands were combined (at 100 μM), Φ-O-Me-cAMP suppressed Epac1 activity vs. cAMP alone, consistent with a weak partial agonist effect of Φ-O-Me-cAMP on Epac1 (Fig. 1E). For Epac2, 100 μM Φ-O-Me-cAMP produced negligible activation, but could limit activation by cAMP (Fig. 1E). The degree to which equimolar Φ-O-Me-cAMP inhibits cAMPinduced Epac activity suggests that Φ-O-Me-cAMP has a slightly lower Epac affinity vs. cAMP.
Thus, most Φ-O-Me-cAMP binding in myocytes is due to Epac, but it also binds specifically to a non-Epac target in the myocyte (conceivably a phosphodiesterase). Correcting for this non-Epac component reduces the B max and apparent Epac affinity, so those values are approximations. Φ-O-Me-cAMP is lipophilic (29) and   (14). Φ-O-Me-cAMP binding was reduced by ∼75% in DKO and also by pretreatment with 100 μM of the Epac agonist OMe-CPT ( Fig. 3 A and B). However, as in intact cells, a non-Epac residual signal was seen. photobleach ( Fig. 4 A and B). Both Φ-O-Me-cAMP and OMe-CPT (at 10 μM) reduced FRET in ICUE1 (Fig. 4C), confirming that they both bind to the cAMP domain of Epac1 and cause conformational changes.
ISO and OMe-CPT can induce ICUE1 translocation to the plasma membrane in HEK293 cells (18,23). We see this also in cardiac myocytes for treatment with ISO, OMe-CPT, and ISO + PKA inhibition, but that effect is blocked by pretreatment with Φ-O-Me-cAMP (Fig. S3). We also detect OMe-CPT-induced nuclear export of Epac1-YFP in myocytes (Fig. 4D). This was not replicated by Φ-O-Me-cAMP and was prevented by Φ-O-Me-cAMP. OMe-CPT also induces a robust increase in Ca spark frequency in myocytes, which we have shown requires Epac2 (14). That effect is blocked by Φ-O-Me-cAMP, which by itself did not activate Ca sparks (Fig. 4E). Epac activation by O-Me-CPT induces nuclear export of the transcriptional regulator HDAC5 (11), which was blocked and not mimicked by Φ-O-Me-cAMP (Fig. 4F). Thus, for all four of these known downstream Epac functional effects in cardiac myocytes (5,11,21,23), Φ-O-Me-cAMP behaves like an antagonist (i.e., partial agonist effects were not apparent).
Φ-O-Me-cAMP fluorescence in Epac1-KO mice (i.e., Epac2 location) was similar to the striated signal from di-8-ANEPPS, with peak intervals near 1.8 μm characteristic of T tubules [Φ-O-Me-cAMP: 1.75 ± 0.02 μm, n = 11 vs. di-8-ANEPPS 1.75 ± 0.02, n = 11; P = not significant (N.S.)] (Fig. 5B). Similar widths at half maximum peak (full width at half maximum, FWHM) suggest a similarly narrow spatial focus for both (Fig. 5B). This result agrees with our previous functional findings that Epac2 is sufficient to fully recapitulate the Epac-mediated activation of SR Ca leak (14), because the RyR2 release sites are mainly along the T tubules. In contrast, Epac1 (in Epac2-KO cells) did not show sarcomeric striations (Fig. 5D), but showed concentration at the nuclear envelope, compared with either WT or Epac1-KO myocytes (Fig. 5 E and F). For WT and Epac2-KO there is clear enhancement of Epac at the nuclear edge compared with the nuclear interior [shown as nuclear envelope (NE) to nuclear ratio]. In contrast, when only Epac2 is present, fluorescence was low throughout the nucleus and NE, rising only when it reaches the sarcomeric Epac2 location outside that NE region. We did not detect Epac1 to be highly concentrated at either the sarcolemma or mitochondria.
Epac1 Is at the Nucleus to Regulate Transcriptional Pathways. In both adult and neonatal cardiomyocytes, Epac1 overexpression mediates β-AR-dependent hypertrophic signaling and activation of gene transcription via CaMKII/HDAC5/MEF2 and calcineurin/NFAT pathways (5,11). In isolated cardiac myocytes, we used the FRET sensor Camui (full-length CaMKII flanked by CFP and YFP at the N and C termini) where FRET is maximal in the basal autoinhibited state (31). Epac stimulation by OMe-CPT activated CaMKII (Fig. 6 A and B), nearly as well as ISO in our previous study (31). However, this effect was abolished in a mutant Camui lacking the T286 autophosphorylation site (T286A), suggesting that Epac-dependent activation requires CaMKII autophosphorylation ( Fig. 6 A and B). Because cytosolic CaMKII (and Camui) is concentrated along the T tubules (like Epac2), this cytosolic CaMKII activation is dominated by Epac2 activation, as previously described (14). Nuclear CaMKII activity was also significantly increased by OMe-CPT exposure (increased F CFP /F YFP; P < 0.001; Fig. 6C) and was abolished by 2-APB [an inositol-1,4,5-trisphosphate (IP 3 ) inhibitor]. This is consistent with nuclear IP 3 receptor-dependent activation of nuclear CaMKII (32). Nuclear CaMKII activation also requires autophosphorylation (Fig. 6C). Previous work showed that OMe-CPT-induced HDAC5 nuclear export was entirely CaMKII and IP 3 receptor dependent (5,11). Consistent with this, we found that HDAC5 nuclear export was blocked by CaMKII inhibitor KN-93 and by 2-APB (Fig. 6D). We also find here that the OMe-CPT-induced nuclear export of HDAC5 was completely prevented in Epac1-KO mice (Fig. 6D), consistent with perinuclear Epac1 vs. Epac2. Both Epac and HDAC5 translocation were also prevented by β1-AR block by CGP20712A (Fig. 6 E and F). Altogether, these results suggest that Epac1 is sufficient (in part because of its localization) in nuclear signaling via the IP 3 /CaMKII/HDAC5 transcriptional regulation pathway and that this pathway is selectively mediated by Epac1 and β1-AR stimulation.

Discussion
Epac2 was recently identified as the isoform responsible for Epac-mediated SR Ca leak and arrhythmia (14), suggesting isoform-specific signaling pathways. However, Epac1 and Epac2 localization in cardiac myocytes, critical for clarifying Epac function in cardiac pathologies such as hypertrophy and arrhythmia, were not clear. Here, we provide evidence, using a novel fluorescent Epac ligand Φ-O-Me-cAMP, that Epac1 and Epac2 are compartmentalized. The permeability, specificity, and binding affinity of Φ-O-Me-cAMP make it a useful tool to define endogenous Epac1 and Epac2 distribution in cardiomyocytes. We found that endogenous Epac2 is concentrated in T tubules, facilitating its ability to influence CaMKII-dependent RyR phosphorylation and raise SR Ca leak to trigger arrhythmias. In contrast, endogenous Epac1 was most clearly concentrated in the perinuclear region, positioned well to regulate nuclear signaling and β1-AR-mediated gene transcription involved in cardiac hypertrophy.
Φ-O-Me-cAMP, a Novel Epac Fluorescent Indicator. Until now, nonfluorescent tools have helped clarify the biological roles of Epac vs. PKA (26,33,34). Φ-O-Me-cAMP is a novel fluorescent Epac ligand, which we find to be a partial agonist, but at concentrations used in myocytes here (≤10 μM), it acts mainly as a competitive antagonist on Epac signaling, as it blocks all OMe-CPT-mediated effects studied. As in OMe-CPT, this fluorescent cAMP analog has specificity for Epac over PKA. The Pharos vs. pCPT group results in somewhat lower Epac affinity vs. OMe-CPT (26)(27)(28)30), which agrees with the pCPT group's importance for OMe-CPT affinity (33). Φ-O-Me-cAMP excitation in the visible range is advantageous vs. fluorescent nucleotides that require UV excitation, which can harm cells.
Φ-O-Me-cAMP, a Fluorescent Tool for Epac Localization. Epac1 and Epac2 share high homology, and the Epac1 antibodies we tested were not suitable for immunolocalization because of strong signals observed in the total absence of Epac (DKO). In contrast, Φ-O-Me-cAMP signal was greatly reduced both in DKO myocytes and by competition with the Epac-selective agonist OMe-CPT, indicating the specificity that is key for a useful fluorescent probe (35). Despite homology between the Epac and PKA binding domains (36), we find no evidence for Φ-O-Me-cAMP binding to PKA. Moreover, neither OMe-CPT nor Φ-O-Me-cAMP, at the concentrations used, could activate PKA. We could detect a residual and saturable signal when Φ-O-Me-cAMP was used in the DKO (Fig. 3A). This signal seems not to be due to Φ-O-Me-cAMP degradation by PDE (29), but because many cAMP analogs bind to phosphodiesterase isoforms (37), that might explain the non-Epac binding.
In addition, Φ-O-Me-cAMP has the advantage of being membrane permeable, unlike most fluorescent nucleotides, presumably due to the highly lipophilic Pharos dye (29). Indeed, the hydrophobic 8 substitution on cAMP analogs enhances their membrane permeability (33). In our conditions, Φ-O-Me-cAMP reached maximal binding within 30 min. Functionally, Φ-O-Me-cAMP binds to Epac without activating HDAC5 translocation, Epac1 translocation, and Epac2-mediated SR Ca leak, suggesting that Φ-O-Me-cAMP binding does not activate Epac signaling to these downstream targets in myocytes. Moreover, Φ-O-Me-cAMP prevents OMe-CPT-dependent activation of these pathways. So, although Φ-O-Me-cAMP is a partial agonist (particularly on Epac1 GDP exchange), it behaves mainly as an Epac antagonist in cellular signaling. However, caution should be used in interpreting Φ-O-Me-cAMP inhibition effects on Epac, keeping in mind the partial agonist effect, especially for Epac1. Φ-O-Me-cAMP permeability, specificity, affinity, and functional effects make it an effective tool to study basal endogenous Epac distribution (and serve in some cases as a functional antagonist).
Epac2 Is Distributed at the T Tubule to Mediate Arrhythmia. Φ-O-Me-cAMP binding in Epac1-KO and Epac2-KO cardiomyocytes revealed isoform compartmentalization. Epac2 was concentrated in T tubules, whereas Epac1 was concentrated around the nucleus. Previous work suggested that Epac1 appeared at plasma membranes and mitochondria (17)(18)(19)(20)(21)(22)(23), and we cannot exclude that some Epac1 is at these sites. Some previous work used exogenous expression of Epac1 constructs, which could alter localization. In Epac1-KO, Epac2 was not detected around the nucleus but was especially concentrated in the T tubules, the site of RyR and Cainduced Ca-release during EC coupling (38,39). That finding is consistent with a regulatory role for Epac2 on RyR function during β-AR stimulation. Indeed, β1-AR stimulation activates Epac2mediated SR Ca leak and triggers arrhythmia by CaMKIIdependent phosphorylation of RyR in cardiac myocytes and intact murine hearts (4, 8-10, 12, 14). This propensity for arrhythmias could also be exacerbated by an Epac-mediated down-regulation of slow delayed rectifier potassium channels (40,41) mediated by Epac translocation, which could prolong action potential duration and increase arrhythmogenic risk.
Epac1, a Key Player in Gene Transcription. Our perinuclear Epac1 localization agrees with work showing that Epac1 is part of a nuclear complex with a PKA anchoring protein mAKAP in hypertrophied neonatal cardiomyocytes (24). Moreover, Epac1-KO fully prevented OMe-CPT-dependent HDAC5 translocation, consistent with its involvement in maladaptive hypertrophy. Indeed, Epac1 is up-regulated in human heart failure and animal models of hypertrophy (6,13). Epac activation can also elevate nuclear Ca, activating CaN/NFAT and CaMKII/HDAC hypertrophic pathways. These signaling pathways seem to be mediated only by β1-AR (Fig. 6 E and F). Moreover, Epac induced activation of nuclear CaMKII and HDAC5 translocation (Fig. 6), which can alter gene transcription by derepression of MEF2. Under sustained activation, Epac enhanced nuclear Ca release by IP 3 receptors, leading to up-regulation of calmodulin, which may also influence arrhythmogenesis (11,12). Epac1 may also be part of a complex with CaMKII and β-arrestin (42), where local activation and translocation of Epac1 could influence CaMKII activation elsewhere in the cell. Finally, our work excludes participation of Epac2 in HDAC5 dependent hypertrophy, because HDAC5 nuclear export was abolished by Epac1 deletion.
In conclusion, our work demonstrates the importance of Epac1 and Epac2 distribution in adult ventricular myocytes and how that contributes to distinct functional roles for each isoform. Epac2 localized at T tubules and can mediate cardiac arrhythmias, whereas Epac1 at the nuclear envelope can regulate β1-AR-mediated gene transcription. Our characterization of the novel Epac ligand Φ-O-Me-cAMP, instrumental here, should allow this tool to become more broadly useful for understanding Epac localization and signaling in other cell types. Our work also raises the possibility that Epac1 activation can mediate rather than be a consequence of cardiac hypertrophic remodeling, but will require further study. Thus, the two Epac isoforms may present functionally distinct therapeutic targets or ISO + H89 (n = 7). *P < 0.05, **P < 0.01, ***P < 0.001. to be explored in the treatment of cardiac hypertrophy and arrhythmia.

Methods
Additional methods are described in SI Methods. Myocyte isolation was performed in adult Wistar rats and Epac1-KO, Epac2-KO, DKO, and C57BL6 mice using retrograde Langendorff perfusion and type-2 collagenase (Worthington) (37°C) as previously described (14). All procedures were approved by the University of California Davis Institutional Animal Care and Use Committee. Data were expressed as mean ± SEM. Significance was evaluated by using unpaired or paired Student's t test, or two-way ANOVA as appropriate.