EPAC1 activation by cAMP stabilizes CFTR at the membrane by promoting its interaction with NHERF1

The cystic fibrosis transmembrane conductance regulator (CFTR) is an integral membrane protein of the ATP-binding cassette (ABC) transporter family that functions as a cAMP-activated Cl⁻ ion channel at the apical membrane of several fluid-transporting epithelia (Riordan, 2008), including those in the airways, where it has an active role in controlling airway surface liquid homeostasis through promotion of mucociliary clearance (Boucher, 2004). CFTR possesses two transmembrane domains (TMD1 and TMD2); two nucleotide-binding domains (NBD1 and NBD2) that bind to and hydrolyze ATP, regulating the gating of the channel; and a regulatory domain that contains multiple phosphorylation sites (Riordan, 2008). Cl⁻ secretion through CFTR depends on the net balance of channel density at the cell surface and the activity of each individual channel (Cihil et al., 2012). CFTR dysfunction leads to cystic fibrosis, the most common life-threatening recessive disorder in Caucasians.

Steady-state levels of CFTR at the plasma membrane rely on its biosynthetic processing, early and late secretory trafficking, as well as on its endocytic uptake, the latter being followed by either recycling to the plasma membrane or lysosomal degradation (Farinha et al., 2013b; Riordan, 2008). Because CFTR internalization at the cell surface is a rapid process compared to CFTR biosynthesis and anterograde trafficking, the recycling of internalized channels is considered to be a key process for sustaining a functional pool of CFTR at the plasma membrane (Farinha et al., 2013b; Haggie et al., 2006). Membrane trafficking of CFTR is stringently regulated by several protein interactors that bind to its C-terminus. Among these, the PDZ adaptor protein Na⁺/H⁺-exchanger regulatory factor isoform-1 (NHERF-1, also known as SLC9A3R1, EBP50, ezrin-binding protein with 50 kDa) is responsible for anchoring plasma membrane CFTR to the actin cytoskeleton (Wang et al., 1998). NHERF-1 links CFTR to the ezrin-radixin-moesin (ERM)-family protein ezrin and locks the channel in an immobile and actin-tethered complex that prevents its endocytosis (Reczek et al., 1997; Sun et al., 2000a). Therefore, CFTR surface anchoring and retention could constitute a major target pathway for cystic fibrosis pharmacotherapy (Farinha et al., 2013b; Young et al., 2009), especially considering that the rescued form of the most common cystic-fibrosis-causing CFTR mutant (F508del-CFTR, deletion of phenylalanine at position 508) still exhibits decreased stability at the plasma membrane (Amaral and Farinha, 2013).

CFTR channel gating at the plasma membrane involves cAMP-activation of protein kinase A (PKA), which phosphorylates the CFTR regulatory domain (Sheppard and Welsh, 1999). This PKA-mediated activation of CFTR relies on a local pool of cAMP near to the membrane that is regulated by the integrity of the actin cytoskeleton (Favia et al., 2010; Monterisi et al., 2012). Although it is well established that cAMP plays a crucial role in CFTR channel gating through PKA, this kinase is not the only cAMP effector within the cell, and the other cAMP-dependent signaling pathways might also impact CFTR.

EPAC1 (also known as RAPGEF3; UniProtKB accession number O95398), an exchange protein directly activated by cAMP, functions as a guanine nucleotide exchange factor (GEF) for both Rap1 and Rap2 (de Rooij et al., 2000), and is able to suppress Ras-mediated oncogenic transformation of cells (Gloerich and Bos, 2010). In response to cAMP, EPAC1 is targeted to the plasma membrane, where it tethers to phosphatidic acid or to phosphorylated ERM proteins to induce its downstream effectors (Schmidt et al., 2013).

Regulation of CFTR channel gating through EPAC1 activation has been addressed and is apparently tissue specific. One study proposes a role for EPAC1 in intestinal Cl⁻ secretion through a PKA-independent mechanism that does not involve CFTR (Hoque et al., 2010), but a more recent one postulates that, in bronchial cells, EPAC1 promotes amplification of the cAMP signal evoked by H₂O₂ through a mechanism that involves PKA (Ivonnet et al., 2015). However, EPAC1 localization at the plasma membrane, where it is potentially exposed to the same pool of increased subcortical cAMP that induces PKA-mediated activation of CFTR, and its ability to promote cortical cytoskeleton stabilization (Kooistra et al., 2005) suggest that EPAC1 might be involved in the regulation of CFTR membrane stability and/or anchoring to the plasma membrane. Although a more direct connection between CFTR and EPAC1 has never been reported, this cAMP effector is involved in the regulation of cell–cell and cell–matrix adhesion, cytoskeleton rearrangements and cell polarization, processes which have been described as affecting CFTR regulation and being dysregulated during cystic fibrosis (Monterisi et al., 2012). Herein, we explored the interaction between CFTR and EPAC1, and evaluated the impact of this cAMP effector on CFTR biogenesis, trafficking and plasma membrane anchoring. We show that EPAC1 and CFTR colocalize and interact through the protein adaptor NHERF1, and that EPAC1 activation promotes the NHERF1–CFTR interaction, stabilizing the latter at the plasma membrane. Altogether, these results reveal a new CFTR interacting protein that links cAMP signaling to cystic fibrosis modulation through a previously unreported mechanism. Furthermore, EPAC1 can be used as a novel therapeutic target to stabilize mutant CFTR at the plasma membrane.

To address whether an increase in intracellular cAMP affects CFTR levels at the plasma membrane, cystic fibrosis bronchial epithelial (CFBE) cells expressing wild-type (wt)-CFTR (CFBE-wt) were treated with increasing concentrations of forskolin (Fsk) to induce adenylate cyclase activation and hence cAMP. Levels of CFTR at the plasma membrane were assessed by using cell surface biotinylation, followed by streptavidin pull down and CFTR detection by western blotting (Fig. 1A,B). Results show an increase in CFTR plasma membrane levels with increased Fsk concentrations, detectable from 1 µM and higher concentrations, with peaks observed at 10 µM and 25 µM (Fig. 1B). This, however, does not correspond to an increase in total CFTR levels or in its steady-state processing levels, as evidenced by similar levels of mature CFTR (band C) and total CFTR levels across these treatments (Fig. 1A, middle panel, CFTR WCL). The saturating concentration of Fsk was 50 µM because there was no further increase in levels of CFTR at the plasma membrane when this Fsk concentration was combined with 100 µM of the phosphodiesterase inhibitor IBMX (Fig. 1C).

Because it has been known for a long time that an increase of subcortical cAMP levels activates PKA, thus promoting CFTR function, we assessed whether the observed increase in CFTR at the plasma membrane was also dependent on PKA activation. To test this hypothesis, cell surface levels of CFTR were assessed in CFBE-wt cells that had been treated with the cAMP analogue 6-Bnz-cAMP-AM, which selectively activates PKA (Christensen et al., 2003), or transfected with a PKA inhibitor (the peptide inhibitor PKI) (Tkachenko et al., 2011). Treatment with this agonist (Fig. 1D,E) or the inhibitor (Fig. 1F,G) did not affect CFTR plasma membrane levels. To further clarify the mechanism for the observed increase in CFTR plasma membrane levels, we then assessed the effect of activation of EPAC1. To this end, we used a cAMP analogue that selectively activates EPAC1 [8-(4-chlorophenylthio)-2′-O-methyladenosine-3′,5′-cyclic monophosphate acetoxymethyl ester, hereafter named 007-AM] (Christensen et al., 2003; Schwede et al., 2015).

Firstly, to assess the EPAC1 selectivity of this compound in CFBE cells, we measured sensitized emission from two different Förster resonance energy transfer (FRET) reporters: one based on EPAC1 (cAMP sensor; hereafter referred to as ‘camps’) (DiPilato et al., 2004; Ponsioen et al., 2004) and another based on a PKA substrate [A-kinase activity reporter 4 (AKAR4)] (Liu et al., 2011). In camps, which is an EPAC-based FRET sensor, the cyclic-nucleotide-binding domain of EPAC is sandwiched by yellow fluorescent protein (YFP) and cyan fluorescent protein (CFP). The binding of cAMP generates a conformational change, moving apart the two fluorophores, thereby decreasing FRET between these two. In AKAR4, a PKA substrate sequence is inserted between the cerulean and venus FRET pair. The phosphorylation of this sensor generates a conformational change, which promotes the convergence of these two fluorophores and the increase of FRET signal. The 007-AM agonist induced a large FRET change for camps but only a minor FRET change in AKAR4 (Fig. S1A–C), indicating that 007-AM is an EPAC1-specific agonist in CFBE cells. Conversely, the PKA agonist 6-Bnz-cAMP-AM is able to activate only the AKAR4 but not the camps sensor (Fig. S1D–F). Additionally, PKA inhibition with PKI precluded activation of the AKAR4 but not of the camps sensor (Fig. S1G,H). Specificity of EPAC1 activation with 007-AM was confirmed by assessing the activation of Rap1 using both the Raichu-Rap FRET sensor (Mochizuki et al., 2001) and a Rap activity assay (Franke et al., 1997). We observed that, after treatment with 007-AM, CFBE-wt cells showed a FRET increase at the plasma membrane for this sensor (Fig. S1I,J) and also to a trend towards an increase in active Rap1A levels (Fig. S1K,L).

We then assessed the effect of treatment with 007-AM on levels of CFTR at the plasma membrane. We observed an increase in plasma membrane CFTR levels after stimulation with 1 µM 007-AM for 2 h (Fig. 1D,E) and that this time point corresponded to that of the maximal effect (Fig. S2A,B). To further validate the role of EPAC1 in regulating CFTR plasma membrane levels, we assessed the effect of reducing EPAC1 levels with small interfering (si)RNAs. EPAC1-specific siRNA knockdown decreased its total levels by 50.4% compared to control siRNA and, in agreement with our results with the selective EPAC1 activator, this decrease in EPAC1 levels correlated with a 82.5% decrease in the levels of CFTR at the membrane plasma compared to that in control (Fig. 1H–J). Interestingly, treatment with 007-AM or transfection with a specific siRNA targeting EPAC1 did not affect the total levels of CFTR nor its processing at steady state (ratio of mature CFTR – band C – over total CFTR) in these cells (Fig. 1D,E, WCL). These data suggest that the observed effects are not caused by changes in CFTR synthesis, turnover or ER-to-Golgi trafficking but rather that they result from either enhanced channel delivery from the trans-Golgi network (TGN) to the plasma membrane or its increased retention at the plasma membrane – i.e. decreased endocytosis or increased recycling.

To assess whether the increased levels of CFTR at the plasma membrane also correspond to an increase in overall function, we tested wt-CFTR channel activity in CFBE cells by determining iodide efflux. We observed higher levels of CFTR activity in cells that had been treated with 1 µM 007-AM or 25 µM Fsk for 2 h because they exhibited a decreased iodide concentration that remained inside the cells (56.9%±13.7 or 58.3%±18.6, respectively; mean±s.e.m.) in comparison to DMSO-treated cells (Fig. 1K). Interestingly, pre-incubation of the cells with the PKA agonist 6-Bnz-cAMP-AM for 2 h did not lead to increased CFTR activity, and the combined effect of this agonist with 007-AM was equivalent to the effects with 007-AM alone (54.4%±14.8 or 56.9%±13.7, respectively). This result confirms that EPAC1 activation increases the levels of functional CFTR at the plasma membrane in CFBE-wt cells.

Because treatment with either 007-AM or EPAC1-specific siRNA did not significantly impact total CFTR levels (see above), we aimed to dissect how EPAC1 activation increases CFTR plasma membrane levels. First, we performed a modified cell surface biotinylation assay to determine the CFTR endocytosis rate. Results showed that EPAC1 activation with 007-AM decreased CFTR internalization over time (Fig. 2A,B). These results suggest that EPAC1 promotes CFTR stability at the plasma membrane by attenuating its endocytosis. To further understand this mechanism, we assessed the effect of blocking endocytic pathways using a potent inhibitor of dynamin-dependent endocytic pathways – Dynasore – thus blocking CFTR internalization (Macia et al., 2006; Young et al., 2009), and again assessed CFTR plasma membrane levels by determining cell surface biotinylation. Under conditions that inhibit endocytosis, this approach assesses CFTR delivery to the cell surface. The results show that treatment with Dynasore promoted an overall 3.4-fold increase in the levels of CFTR at the plasma membrane. However, there was no statistically significant difference in the effects of treatment with DMSO, 007-AM or Fsk in combination with Dynasore (Fig. 2C,D), indicating that EPAC1 activation does not interfere with CFTR delivery from TGN to the plasma membrane but rather with its retention at the plasma membrane.

To determine if the observed effects upon CFTR internalization are due to either decreased endocytosis or increased recycling, we then performed a modified cell surface biotinylation protocol to assess CFTR recycling rate and thus distinguish between these two hypotheses. We observed that EPAC1 activation with 007-AM did not significantly alter CFTR recycling rates at 2.5, 5 and 10 min in comparison to recycling rates upon DMSO treatment (Fig. 2E,F). Altogether, these results indicate that EPAC1 promotes CFTR stability at the plasma membrane through attenuation of endocytosis.

To assess the specificity of this effect of EPAC1 activation on CFTR trafficking, we tested the effect of 007-AM on another cell surface protein, epidermal growth factor receptor (EGFR). Results from endocytosis assays showed that EPAC1 activation with 007-AM did not significantly change EGFR plasma membrane levels or its endocytosis rate at the 10 min time point in comparison to that with DMSO treatment (Fig. S2C–E), suggesting that EPAC1-mediated CFTR stabilization is not a mechanism that is broadly involved in the general regulation of plasma membrane proteins.

To characterize the mechanism through which EPAC1 stabilizes CFTR at the plasma membrane, we characterized the cellular location of EPAC1 and CFTR upon EPAC1 activation.

To this end, we first assessed the effect of 007-AM on EPAC1 localization by performing fluorescence confocal microscopy. CFBE parental, -wt or -F508del cells were transfected with GFP–EPAC1. EPAC1 is mainly located in the cytosol and around the nucleus under basal conditions, and it translocates to the plasma membrane after activation with 1 µM 007-AM. This effect was detected both in CFBE and HEK293T cells; the latter cells being used as a control because translocation of EPAC1 to the plasma membrane in those cells after treatment with 1 µM 007-AM has been previously reported (Consonni et al., 2012; Gloerich et al., 2010; Ponsioen et al., 2009) (Fig. 3A–C; Fig. S3). The translocation occurred within the first 3 min of incubation with 007-AM and did not depend on the presence of CFTR (as it was detected in all three types of cells). This relocalization of activated EPAC1 suggests that this protein and CFTR might colocalize or even interact.

To assess the colocalization between CFTR and EPAC1, CFBE parental cells were co-transfected with mCherry–CFTR (the wt or F508del protein) and GFP–EPAC1, and live-cell imaging was performed (Fig. 3D,E). As expected, wt-CFTR was located at the plasma membrane, whereas F508del-CFTR showed a perinuclear location. In CFBE-wt and CFBE cells expressing the F508del CFTR protein (CFBE-F508del cells) under pre-stimulus conditions, EPAC1 localized to the cytosol and translocated to the plasma membrane after activation with 1 µM 007-AM. Thus, treatment with this cAMP analogue increased colocalization between CFTR and EPAC1, as assessed by determining the Pearson's coefficient (Fig. 3F) in CFBE-wt cells, whereas it decreased colocalization of the two proteins in CFBE-F508del cells.

To assess whether the observed colocalization corresponded to the presence of CFTR and EPAC1 in the same protein complexes, co-immunoprecipitation experiments were performed in CFBE and A549 cells (Fig. 4).

The results show that GFP–EPAC1 co-immunoprecipitated with mCherry–wt-CFTR under basal conditions in CFBE parental cells that had been transiently transfected with both constructs, suggesting that although the degree of colocalization under these conditions was close to 0.5, these proteins could still interact with each other (Fig. 4A). The same experiment was performed in A549 and CFBE cells that expressed mCherry–wt- or –F508del-CFTR (Fig. 4B,D) and endogenously expressed EPAC1. Besides confirming that EPAC1 co-immunoprecipitates with wt-CFTR, results from this experiment also produced the same outcome for F508del-CFTR. This might indicate that EPAC1 can interact with CFTR in the early stages of its trafficking – i.e. even if the protein is not at the plasma membrane. Co-immunoprecipitation experiments were repeated in the same cell lines under EPAC1 activation with 1 µM 007-AM. The treatment increased the association between CFTR (both wt and F508del) and EPAC1 (Fig. 4B–F), suggesting that, at least for wt-CFTR, the interaction is promoted when EPAC1 is active and near to the plasma membrane. After being activated, EPAC1 translocated from the cytosol and perinuclear region to the vicinity of the plasma membrane, which could facilitate the interaction with CFTR. The effect of 007-AM in promoting the interaction between EPAC1 and CFTR appears to be more pronounced on the wild-type protein than for F508del-CFTR (Fig. 4E,F). Taken together, the co-immunoprecipitation of CFTR with endogenous EPAC1 and their localization at the plasma membrane in human airway epithelial cells indicate that CFTR and EPAC1 are present in the same protein complexes.

Because EPAC1 has been previously reported to interact with NHERF1 and ezrin – two proteins also known to interact with CFTR (Farinha et al., 2013b) – or to regulate processes dependent on those two proteins (Gloerich and Bos, 2010), we hypothesized that CFTR and EPAC1 could interact through one of those scaffold proteins. In fact, EPAC1 does not contain a PDZ domain within its sequence, supporting the need of an adaptor protein between CFTR and EPAC1.

To test this hypothesis, we assessed the interaction of CFTR with EPAC1 upon downregulation of either ezrin or NHERF1. This was performed in CFBE-wt cells that had been transiently transfected with specific siRNAs (Fig. 5A–C) and in Calu3 cells that had been stably transduced with lentiviral particles expressing small hairpin (sh)RNAs against each protein (Fig. 5D–F). As observed, NHERF1, but not ezrin, knockdown abolished the interaction between EPAC1 and CFTR. Results in Calu3 cells show that the above-described CFTR–EPAC1 interaction (Fig. 4) also occurred when both proteins were endogenously expressed (Fig. 5D). These results suggest that NHERF1, but not ezrin, mediates the EPAC1–CFTR interaction.

As the disruption of the CFTR–EPAC1 interaction could also result from a change in the subcellular localization of EPAC1 when NHERF1 levels are decreased, we performed fluorescence live-cell imaging in Calu3 cells that had been transfected with GFP–EPAC1 (Fig. S3A,B). Our results show that neither the subcellular localization of EPAC1 nor the translocation to the plasma membrane after treatment with 1 µM 007-AM was affected by knockdown of ezrin or NHERF1.

Additionally, to further characterize the mechanism of CFTR increase at the plasma membrane under EPAC1 activation, we used cell surface biotinylation to assess the levels of plasma membrane CFTR after treatment with 007-AM in cells in which either NHERF1 or ezrin had been knocked down. The results show that knockdown of NHERF-1, but not of ezrin, prevented the stabilization of CFTR at the plasma membrane after EPAC1 activation (Fig. 5G,H). These observations are in agreement with the hypothesis that the EPAC1–CFTR interaction is mediated by NHERF1.

To further dissect the role of NHERF1 in the EPAC1-mediated stabilization of CFTR at the plasma membrane, we used a co-immunoprecipitation assay to assess if EPAC1 and NHERF1 interact in both CFBE and HEK293T cells (Fig. 6). As observed, EPAC1 and NHERF1 co-immunoprecipitated, and EPAC1 activation also promoted the interaction between these proteins (Fig. 6C,D), strengthening our hypothesis that the CFTR–EPAC1 interaction is mediated by NHERF1.

To better characterize the EPAC1–NHERF1 interaction, we aimed to identify specific domains involved in the interaction. We observed that the interaction was disrupted after deletion of the PDZ1 domain of NHERF1, suggesting that this could be the domain mediating that interaction (Fig. S4A). Additionally, deletion of the N-terminal region of EPAC1 impaired the interaction between EPAC1 and NHERF1 (Fig. S4B), suggesting that it might be mediated by this region. Overall, the interaction seems to be mediated by the NHERF1 PDZ1 domain and EPAC1 N-terminus.

Lastly, we assessed whether EPAC1 activation is also able to increase the rescue of F508del-CFTR that is caused by the drug VX-809, which has been recently approved by the Food and Drug Administration. To this end, CFBE-F508del cells were incubated with 3 µM VX-809 (to promote the rescue of the mutant protein) and 1 µM 007-AM (to activate EPAC1) (Fig. 7A,B). The levels of mature and immature CFTR were assessed by western blotting. The results show that the combined effects of VX-809 and 007-AM increased the processing levels of F508del-CFTR from 11 and 13% in VX-809-only-treated cells at 24 and 48 h, respectively, to 17 and 19% in cells that were treated simultaneously with both compounds. Additionally, we assessed the effect of EPAC1 activation on the channel activity of VX-809-rescued F508del-CFTR in CFBE cells. We detected a lower amount of iodide remaining inside the cells (20% decrease relative to cells treated only with VX-809) – i.e. higher levels of CFTR activity – in cells that had been treated simultaneously with both compounds (Fig. 7C). These findings indicate that targeting CFTR membrane stability through EPAC1 activation is an approach that could be used in combination with treatment with VX-809.

Ion transport through CFTR is stimulated when levels of cAMP in the subcortical compartment are increased, leading to PKA activation, thus triggering CFTR phosphorylation and channel opening. However, besides PKA, EPAC1, an exchange protein that is directly activated by cAMP, is another more recently discovered cAMP effector (Gloerich and Bos, 2010). Here, we describe a PKA-independent cellular mechanism that links cAMP to regulation of CFTR in CFBE cells.

Our data demonstrate that an increase in cAMP levels promotes CFTR stability at the plasma membrane (Fig. 1). A maximal effect was observed under stimulation with 10–25 µM Fsk, whereas no significant increase in CFTR plasma membrane levels was observed for Fsk concentrations in the range 0.1–1.0 µM, which have been reported previously to maximally induce CFTR Cl⁻ currents in CFBE-wt cells (Bebok et al., 2005). This suggests that cAMP signaling can regulate CFTR across a broad range of concentrations. PKA-mediated cAMP signals regulate CFTR function at lower concentrations of cAMP, whereas EPAC1-dependent regulation is in place at higher cAMP levels. This is in agreement with the fact that cAMP exhibits a lower affinity for EPAC1 (K_d 2.8 µM) than for PKA (K_d 0.1–1.0 µM) (Christensen et al., 2003). Moreover, the effect of PKA on CFTR function is fast but short in time, whereas EPAC1-mediated effects occur at later time points and for higher levels of cAMP. Thus, this cAMP-dependent dual control of CFTR allows for precise temporal and spatial control of Cl⁻ secretion. In fact, our observations are in agreement with published work suggesting that treatment with Fsk and/or IBMX enhances CFTR membrane trafficking (Chang et al., 2002). Here, we demonstrate that this effect on CFTR relies on cAMP-dependent activation of EPAC1 and does not involve PKA. This EPAC1-dependent effect was further confirmed through knockdown of EPAC1 with siRNA, which led to a decrease in plasma membrane levels of CFTR (Fig. 1). This effect led to an increase in CFTR function (Fig. 1), an observation that is in agreement with a recent report, according to which EPAC1 is involved in signal amplification that occurs upon H₂O₂-mediated CFTR activation (Ivonnet et al., 2015). Our study now shows that in bronchial cells, this increase is likely to occur through stabilization of CFTR at the membrane.

Moreover, we found that EPAC1 activation promotes an increase in CFTR levels at the plasma membrane by decreasing endocytosis of the channel (Fig. 2). Interestingly, we also show that this stabilization effect is not observed for the unrelated membrane protein EGFR, suggesting that this is not a general regulation mechanism that nonspecifically affects any membrane protein (Fig. S2C–E). Because we detected no changes in either delivery to the cell surface or endocytic recycling, the increased stabilization might impair recognition of endocytic signals or simply potentiate its binding to NHERF1, an adaptor protein that is essential for anchoring of CFTR to the actin cytoskeleton (Favia et al., 2010; Loureiro et al., 2015).

Several approaches support the notion that EPAC1 regulates CFTR. First, we evaluated the effect of treatment with 007-AM on EPAC1 subcellular localization, tertiary structure and activity (Fig. 3; Figs S1 and S3), and confirmed that this molecule is an EPAC1-specific agonist; we also validated the concentration of this compound used for this study (1 µM) as correct in being able to promote maximal EPAC1 activation in this cellular model.

EPAC1 translocation to the plasma membrane after 007-AM treatment was observed in CFBE cells and found to be independent of CFTR (Fig. 3). In fact, this translocation occurs in CFBE-wt, CFBE-F508del and parental CFBE cells. This is in agreement with previous reports that have shown that EPAC1 translocates to the plasma membrane – a process that is dependent on structural changes in the protein and binding to phosphatidic acid or ERM proteins (Consonni et al., 2012; Gloerich et al., 2010). EPAC1 translocation to the plasma membrane, together with the observed stabilization of CFTR at the plasma membrane after treatment with 007-AM, suggests that EPAC1 and CFTR colocalize or interact with each other, mainly when EPAC1 has been activated with 007-AM.

In fact, our results show that EPAC1 colocalizes with both wt- and F508del-CFTR in CFBE cells without stimulus or under 007-AM treatment (Fig. 3). Additionally, the two proteins co-immunoprecipitated from CFBE, A549 and Calu3 cells (Figs 4 and 5). The fact that an interaction of EPAC1 was also detected with F508del-CFTR suggests that this interaction might also occur at the early stages of CFTR trafficking and, thus, before CFTR reaches the plasma membrane, despite the interaction being stronger when EPAC1 is active and mostly located at the plasma membrane. As CFTR–EPAC1 co-immunoprecipitation is detected in the absence of EPAC1 activation with 007-AM, endogenous levels of cAMP might be enough to promote some EPAC1 activation and thus its interaction with CFTR. Nevertheless, our data also suggest that active EPAC1 has an increased association with CFTR compared to that of inactive EPAC1.

This increased CFTR–EPAC1 interaction under EPAC1 activation could result from different molecular events that are not mutually exclusive: (1) EPAC1 undergoes a conformational change, (2) EPAC1 translocates to the plasma membrane and/or (3) EPAC1 becomes catalytically active. After being activated, EPAC1 translocates from the cytosol and perinuclear region to the vicinity of the plasma membrane, which might facilitate the interaction with CFTR, namely with wt-CFTR. This could explain why the promotion of the interaction between EPAC1 and wt-CFTR by 007-AM is stronger than that with F508del-CFTR. Moreover, EPAC1 activation does not affect the ratio between band C and band B (i.e. the ratio between mature and immature CFTR), a readout for the proportion of CFTR that has been processed, thus indicating that the role of EPAC1 in the early stages of CFTR biogenesis is less relevant than it is at the later stages.

Interestingly, knockdown of NHERF1 (through siRNA or shRNA transfection), but not of ezrin, prevented the CFTR–EPAC1 interaction without disrupting EPAC1 subcellular localization (Fig. 5; Fig. S3A,B), suggesting that EPAC1 might interact with CFTR through the mediator NHERF1. The fact that EPAC1 does not contain a PDZ domain within its sequence supports the need for an adaptor protein in order for the CFTR–EPAC1 interaction to occur. Our data suggest that such a role is played by NHERF1, which generally functions as a protein adaptor and is highly expressed in epithelial tissues (Voltz et al., 2001). Furthermore, our data clearly show that EPAC1 co-immunoprecipitates with NHERF1, with this interaction being potentiated under EPAC1 activation (Fig. 6; Fig. S4).

Our results further clarify that the interaction between EPAC1 and NHERF1 could be mediated by the NHERF1 PDZ1 domain. In the absence of ezrin, the NHERF1 PDZ2 domain is inhibited. Binding of ezrin to NHERF1 promotes a conformational change, exposing the PDZ2 domain (Morales et al., 2007). If the binding of EPAC1 to NHERF1 were mediated by the PDZ2 domain, then the absence of ezrin would abolish this interaction, but our results show that ezrin knockdown does not affect the CFTR–EPAC1 interaction. Therefore, according to our data, EPAC1 binds CFTR through NHERF1, an observation that is in agreement with the previously established binding of PKA to CFTR through ezrin.

It has also been suggested that ezrin can positively regulate the cooperative binding of NHERF to the CFTR C-terminus (Li et al., 2005). As a result of ezrin binding, a specific ternary complex (CFTR)₂–NHERF1–ezrin with 2:1:1 stoichiometry is formed, in which two CFTR molecules are anchored to NHERF1 (Li et al., 2005). Nevertheless, one possibility is that NHERF1 interacts simultaneously with CFTR and EPAC1 through its PDZ domains, and with ezrin through its C-terminus (the latter might also tether PKA). Moreover, ezrin is a kinase-anchoring protein, bringing PKA, phosphatases, phosphodiesterases and other signaling proteins within the vicinity of CFTR (Monterisi et al., 2012; Sun et al., 2000b). The generation of a macromolecular complex involving cytoskeleton proteins might play a key role in fine-tuning the regulation of CFTR stability and function (Monterisi et al., 2013). In fact, our data support a model of a macromolecular complex where CFTR, EPAC1 and NHERF1 interact, the latter acting as a mediator, possibly with proteins like ezrin (Fig. 7).

Interestingly, modulation of such a macromolecular complex (and its stabilizing effect upon CFTR at the plasma membrane) can be used to improve the functional rescue of F508del-CFTR (Fig. 7). Recent evidence shows that the correction of the F508del-CFTR trafficking defect by using VX-809 (Van Goor et al., 2011) has modest therapeutic efficacy (Clancy et al., 2012; Wainwright et al., 2015), suggesting that a combinatorial approach is needed to achieve correction levels that translate into clinical benefit (Amaral and Farinha, 2013; Farinha et al., 2013a). Our results show that the combined effect of VX-809 and 007-AM improves F508del-CFTR rescue relative to that of VX-809 alone by an additional 6% (from 11% to 17%). Altogether, these results enforce the hypothesis that modulation of the EPAC1 pathway could constitute an additional strategy to correct the function of F508del-CFTR. Interestingly, this effect seems to be specific to respiratory epithelia, as it has been previously shown that EPAC1 also mediates Cl⁻ secretion in the intestine, but without the involvement of CFTR (Hoque et al., 2010). Thus, EPAC1 activation is a putative novel target in the organ that is responsible for most of the morbidity and mortality associated with cystic fibrosis. These observations are in agreement with the previous use of compounds that elevate intracellular levels of cAMP as therapeutic options in respiratory diseases, such as asthma and chronic obstructive pulmonary disease (COPD) (Schmidt et al., 2013). Given that EPAC1, as well as its effectors Rap1 and Rap2, is expressed in human bronchial epithelial cells, targeting of EPAC1 might also be used to regulate epithelial integrity in diseased cells. In fact, EPAC1 is involved in the inhibition of cell proliferation and is thus related to a more differentiated phenotype (Roscioni et al., 2009). Interestingly, EPAC1, the expression of which in the adult lung is dominant over that of EPAC2 (Ulucan et al., 2007), also interacts with one of the receptors for TGF-β – a pro-inflammatory cytokine, the levels of which are increased in individuals with cystic fibrosis. This interaction leads to the subsequent inhibition of Smad-dependent TGF-β signaling (Conrotto et al., 2007). These observations further strengthen the relevance of EPAC1 in the context of CFTR trafficking and in the overall context of the normal versus cystic fibrosis cells.

In this study, we revealed that cAMP signaling promotes stabilization of CFTR at the plasma membrane through activation of EPAC1. These findings demonstrate the existence of a previously unidentified mechanism of CFTR regulation by cAMP. In addition to PKA-mediated activation of Cl⁻ transport, which operates under lower concentrations of cAMP, we have identified a new mechanism through which higher levels of cAMP activate EPAC1 to promote its interaction with NHERF1, thus stabilizing CFTR at the plasma membrane by decreasing the rate at which it is endocytosed. This work constitutes an important characterization of a new CFTR-interacting protein that links cAMP to modulation of cystic fibrosis through a previously unreported mechanism.

The following constructs were used: GFP–EPAC1, GFP–Δ49-EPAC1 and GFP–Δ148-EPAC1, with EPAC1 lacking the first 49 or 148 amino acids, as indicated (Gloerich et al., 2010); GFP–NHERF1 (NHERF1 tagged at the N-terminus), NHERF1–GFP (NHERF1 tagged at the C-terminus), Myc–NHERF1, GFP–PDZ2-ERM (lacks PDZ1 domain), GFP–ERM (lacking both the PDZ1 and PDZ2 domains) (Castellani et al., 2012; Loureiro et al., 2015); Raichu sensor (Nakamura et al., 2006), with the Rap1 sequence between YFP and CFP; pGEX-RalGDS-RBDGST plasmid (Franke et al., 1997); mCherry and PKI–mCherry (Lefkimmiatis et al., 2013).

CFBE cells (CFBE4^1o) that stably overexpressed either wt- or F508del-CFTR (CFBE-wt and CFBE-F508del, respectively) (Bebok et al., 2005), or that did not express CFTR (parental CFBE) (Gruenert et al., 1995), and A549 cells overexpressing mCherry–wt- or mCherry–F508del-CFTR (Almaca et al., 2011), and Calu3 cells (Calu3-wt) (Fogh et al., 1977) were cultured as described previously. Calu-3 cells that had been transduced to express a control shRNA (Calu3 shControl), or shRNAs against ezrin (Calu3 shEzrin) or NHERF1 (Calu3 shNHERF1) were generated in this work (see below) and cultured with puromycin (5 µg/ml). HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) FBS. Cells were transiently transfected with plasmid DNA or siRNA using Lipofectamine2000 (Life Technologies), and were analyzed 24 or 48 h later. All cells were tested for contamination, including mycoplasma contamination.

Antibodies against the following proteins were used as recommended by the manufacturers: α-tubulin (Sigma-Aldrich, T5168), β-tubulin (SCBT, sc-9104), calnexin (BD Biosciences, 610523), CFTR (Cystic Fibrosis Foundation, 596), EPAC1 (Aviva, ARP52140_P050), ezrin (BD Biosciences, 610602), GAPDH (SCBT, sc-166574), GFP (SCBT, sc-9996), Myc (SCBT, sc-789), NHERF1 (BD Biosciences, 611160), polyubiquitin (BIOMOL, PW8805) and Rap1A (SCBT, sc-1482); horseradish-peroxidase-conjugated goat anti-mouse or anti-rabbit secondary antibodies were also used (Bio-Rad; dilution 1:3000). All primary antibodies were used at 1:1000 dilution, except those against CFTR (1:3000) and Rap1A (1:500). The following reagents were used: 8-pCPT-2′-O-Me-cAMP-AM (007-AM) (BioLog); VX-809 (Selleckchem); Forskolin and IBMX (Sigma-Aldrich); and H89 (Enzo Life Sciences). The following siRNAs were used: Silencer^® Select Negative Control No. 2 siRNA and against RAPGEF3 (Life Technologies, 4390846 and 4392420, respectively); siGENOME SMART pool non-targeting siRNA and against human EZR and human SLC9A3R1 (Thermo Scientific, D-001206-13, EG:7430 and EG:9368, respectively).

To produce lentiviral particles containing specific shRNAs, HEK293T cells were transiently transfected with the packaging (pCMV-dR8.74psPAX2), envelope (pMD2.G) and hairpin-pLKO.1 vector (control, SHC001; NHERF1, SHCLND SLC9A3R1; Ezrin, SHCLND EZR; Sigma-Aldrich) plasmids using X-tremeGENE9 DNA transfection reagent (Sigma-Aldrich). Medium was removed after 18 h incubation at 37°C and replaced with high-concentration bovine serum albumin (high-BSA) growth medium [DMEM supplemented with 10% (v/v) inactivated FBS (iFBS), 1.1 g/100 ml BSA and penicillin-streptomycin]. The next day, the medium was collected and stored at −20°C. The cells were incubated again with high-BSA growth medium and harvested again 24 h later.

For transduction, Calu3-wt cells were washed with HBSS. EMEM supplemented with 10% (v/v) FBS, polybrene (8 µg/ml) and 5-20% (v/v) lentivirus-containing medium was added. and incubated at 37°C for 24 h. Medium was then replaced by Eagle's minimum essential medium (EMEM) supplemented with 10% (v/v) FBS and puromycin (5 µg/ml), and this was repeated for, at least, five consecutive days.

Cells were lysed and extracts analyzed as described previously (Farinha et al., 2002). Signal was developed with ImmunStar Western C (Bio-Rad) or enhanced chemiluminescence substrate (Thermo Scientific). Detection was performed with the Chemidoc XRS+ analyzer (Bio-Rad) or by using Compact X4 Automatic Processor (Xograph). Quantification was performed using the Image Lab software (Bio-Rad) or ImageJ (http://imagej.nih.gov/ij/).

Cells were lysed with PD buffer [50 mM Tris-HCl, 0.1 M NaCl, 1% (v/v) NP40, 10% (v/v) glycerol, pH 7.5] supplemented with protease inhibitor cocktail (Roche) at 4°C and collected with a scraper. Lysate was cleared by centrifuging, and the supernatant was pre-cleared through incubation with Protein-G agarose beads (Invitrogen). The supernatant was then incubated overnight with the appropriate antibody at 4°C. For mCherry and Myc pull down, anti-RFP-antibody-conjugated beads (Chromotek, Planegg-Martinsried, Germany; RFP-Trap_A sta-20) and anti-c-Myc-antibody-linked beads (Sigma-Aldrich, A7470), respectively, or untagged beads (Chromotek bab-20), as a control, were used. Beads were washed three times with wash buffer [Tris-HCl 0.1 M, NaCl 0.3 M, Triton X-100 1% (v/v), pH 7.5], which was followed by elution with 1× sample buffer, separation on SDS-PAGE gels and analysis by western blotting.

Plasma membrane CFTR levels were determined by performing cell surface biotinylation using membrane-impermeable EZ-Link™ Sulfo-NHS-SS-Biotin (Pierce), followed by lysis in 25 mM HEPES, pH 8.0, 1% (v/v) Triton, 10% (v/v) glycerol and protease inhibitor cocktail (Roche), as described previously (Moyer et al., 2000). Biotinylated proteins were isolated with streptavidin–agarose beads, eluted with sample buffer and separated by SDS-PAGE.

Endocytic and recycling assays were performed as described previously (Cihil et al., 2012; Swiatecka-Urban et al., 2002). For both assays, plasma membrane proteins were first biotinylated at 4°C using membrane-impermeable and cleavable EZ-Link™ Sulfo-NHS-SS-Biotin (Pierce). For endocytosis, cells were warmed to 37°C for different time periods after biotinylation and, subsequently, the disulfide bonds on Sulfo-NHS-SS-biotinylated proteins remaining at the plasma membrane were reduced with L-glutathione (GSH; Sigma-Aldrich) at 4°C. At this point of the protocol, biotinylated proteins reside within the endosomal compartment. Subsequently, cells were lysed, and biotinylated proteins were isolated with streptavidin–agarose beads, eluted into SDS-sample buffer and separated by performing SDS-PAGE. The amount of biotinylated CFTR at 4°C and without the 37°C warming was considered 100%. The amount of biotinylated CFTR remaining at the plasma membrane after GSH treatment at 4°C and without the 37°C warming was considered background and subtracted from the levels of biotinylated CFTR after warming at each time point. CFTR endocytosis was calculated after subtraction of the background and was expressed as the percentage of biotinylated CFTR at each time point after warming compared to the amount of biotinylated CFTR present before warming.

For the recycling assay, cells were warmed to 37°C for 5 min after biotinylation and cooled to 4°C, and the disulfide bonds on plasma membrane proteins were reduced with GSH. Following this, cells were either lysed or warmed again to 37°C for different time periods (to allow endocytosed biotinylated CFTR to recycle to the plasma membrane). Cells were then cooled again to 4°C, and the disulfide bonds on recycled proteins were reduced with GSH. The amount of recycling of endocytosed CFTR was calculated as the difference between the amount of biotinylated CFTR after the first and second GSH treatments.

The active fraction of Rap was measured with a pulldown assay using the selective interaction of the Rap-binding domain (RBD) of RalGDS with the active GTP-bound form of Rap, as described previously (Franke et al., 1997). This domain was expressed tagged with glutathione S-transferase (GST) from bacteria transformed with pGEX-RalGDS-RBDGST and isolated with glutathione agarose beads (Thermo Scientific). Proteins bound to GST-RalGDS-RBD-coupled beads were analyzed by SDS-PAGE, confirming that the protein was effectively captured.

For the active Rap1 pull down, cells were lysed on ice in Rap1A lysis buffer [25 mM Tris-HCl, 1% (v/v) NP-40, 5 mM MgCl₂, 150 mM NaCl, 0.1 mM DTT, 5% (v/v) glycerol, protease inhibitor cocktail, pH 7.5] for 15 min. Lysates were centrifuged, and the supernatant was incubated with GST-RalGDS-RBD-coupled beads for 2 h at 4°C. Levels of active Rap1A were assessed by western blotting.

Fluorescence imaging was performed 24–48 h after co-transfection of cells with plasmids encoding GFP–EPAC1 and mCherry–CFTR. Cells were kept at room temperature in PBS and imaged on a Fluoview FV1000 microscope – an inverted IX81 confocal system (Olympus, Tokyo, Japan) and 60× NA 1.35 oil immersion UPlanSApo objective (Olympus). Images were acquired using FluoView FV10-ASW software (Olympus) and processed using ImageJ. Intracellular localization of EPAC1 and CFTR in live cells in the absence or presence of 007-AM was monitored, and the extent of overlap between the two was quantified using Pearson's correlation coefficient (ImageJ, JACoP plugin) (Zinchuk and Zinchuk, 2008). The overlap between the images of both proteins was analyzed with ImageJ. For quantification of the fluorescence intensity in a specific region of interest (plasma membrane), MetaFluor software (Molecular Devices) was used.

Real-time FRET imaging experiments were performed as described previously (Monterisi et al., 2012). cAMP sensor (camps) (DiPilato et al., 2004; Ponsioen et al., 2004), AKAR4 (A-kinase activity reporter 4) (Liu et al., 2011) or Raichu-Rap1 FRET (Mochizuki et al., 2001) sensors were used. FRET imaging experiments were performed 24–48 h after transfection. Cells were maintained at room temperature or at ∼34°C in PBS and imaged on an inverted microscope (Olympus IX81) using a PlanApoN 60× NA 1.42 oil immersion objective, 0.17/FN 26.5 (Olympus). The microscope was equipped with coolSNAP HQ monochrome camera system (Photometrics), white- and 505-nm-light-emitting diode (LED; Cairn Research) and a beam-splitter optical device (Dual-view simultaneous imaging system, DV2 mag biosystem, Photometrics, ET-04-EM). Images were acquired using MetaFluor or MetaMorph software (Molecular Devices) and processed using ImageJ. FRET changes were measured as changes in the background-subtracted 545/480 nm (AKAR 4 and Raichu-Rap1) or 480 nm/545 nm (camps) fluorescence emission intensity on excitation at 430 nm and expressed as R/R0, where R is the ratio at time t and R0 is the ratio at time=0 s.

The CFTR-mediated iodide efflux assay was performed as described previously (Mendes et al., 2011). Cells that had been grown in 6-well plates were treated with 1 µM 007-AM or 25 µM Fsk (or DMSO as control) for 2 h, in duplicates. After that, cells were loaded with iodide in the loading buffer for 30 min at 37°C, thoroughly washed with iodide-free efflux buffer and equilibrated for 10 min in the same buffer. Cells were then incubated for 5 min either in the presence of iodide-free efflux buffer or in the presence of CFTR stimulators (10 μM Fsk and 50 μM IBMX, Sigma-Aldrich). Cells were lysed, and the iodide concentration in each sample was determined using an iodide-sensitive electrode (Orion 96–53; Thermo Scientific) with a pH/mV meter and normalized to the amount of protein. In this assay, increased channel activity corresponds to more iodide released from the cells and thus a decreased iodide concentration remaining within the cells.

Data are presented as the mean±s.e.m., as indicated in figure legends. Data were analyzed using Student's t-test, with P<0.05 accepted as the level of statistical significance. The number of biological replicates is indicated in figure legends. Error bars reflect independent experiments.

We are grateful to the following individuals for their kind gifts of plasmids: Dr Kees Jalink (NKI, The Netherlands) for the EPAC1 constructs, Prof. Michiyuki Matsuda (Kyoto University, Japan) for the Raichu sensors, Dr Paulo Matos (INSA, Portugal) and Prof. Valeria Casavola (University Bari, Italy) for the NHERF1 constructs, and Prof. David Altschuler (University of Pittsburgh School of Medicine, PA) for the pGEX-RalGDS-RBDGST plasmid. We thank Konstantinos Lefkimmiatis (University of Oxford, UK) for the PKI-mCherry plasmid and for helpful comments, and Inna Uliyakina, José Múrias, Dr Stefania Monterisi and Dr Andreas Koschinski for assistance.