Effect of Annexin A5 on CFTR: regulated traffic or scaffolding?

Abstract

Previous studies have implicated annexins in regulating ion channels and in particular annexin A5 (AnxA5) in the traffic of the cystic fibrosis transmembrane conductance regulator (CFTR). In the present study, we further investigated the role of AnxA5 in regulating CFTR function and intracellular trafficking in both Xenopus oocytes and mammalian cells. Although we could confirm the previously reported CFTR/AnnxA5 interaction, we found that in oocytes AnxA5 inhibits CFTR-mediated whole-cell membrane conductance presumably by a mechanism independent of PDZ-binding domain at the C-terminus of CFTR but protein kinase C (PKC)-dependent and results from either endocytosis activation and/or exocytosis block. In contrast, in human cells, co-expression of AnxA5 augmented CFTR whole-cell currents, an effect that was independent of CFTR PDZ-binding domain. We conclude that annexin A5 has multiple effects on CFTR, so that the net effect observed is cell system-dependent. Nevertheless, both effects observed here are consistent with the described role of annexins forming scaffolding platforms at cell membranes, thus contributing to a decrease in their dynamics. Finally, we could not confirm that AnxA5 overexpression rescues traffic/function of the most frequent disease-causing mutant F508del-CFTR, thus concluding that AnxA5 is not a promising tool for correction of the F508del-CFTR defect.

Keywords::

• Cystic fibrosis
• CFTR
• membrane traffic
• exo/endocytosis
• Annexin A5
• ion channels

Introduction

Cystic Fibrosis (CF) is caused by mutations in the gene encoding the CF transmembrane conductance regulator (CFTR) protein, a cAMP-dependent and phosphorylation-regulated Cl⁻ channel, formed by two membrane-spanning domains (MSDs), two nucleotide-binding domains (NBD1 and NBD2) and a regulatory domain (RD). CFTR assembly initiates with synthesis and folding in the endoplasmic reticulum (ER) where it is core-glycosylated. Once correctly folded, this immature form of CFTR migrates through the Golgi where it is processed by glycosyltransferases being thereby converted into its mature form, until it reaches the cell surface. Depending on the cell type, only 30–70% of precursor wild-type (wt)-CFTR matures. F508del-CFTR, the most common CF-causing mutation occurring in 90% of CF patients worldwide, is a trafficking defect mutant which, due to misfolding, is mostly retained by the ER quality control (ERQC) and then rapidly ubiquitinated and targeted for proteasomal degradation. During its biogenesis and folding, multiple CFTR interacting proteins (CIPs), namely molecular chaperones, account for the efficiency of those processes. Moreover, these constitute the major intervenients of the ER retention which, in our current ERQC working model, act at distinct checkpoints (Roxo-Rosa et al. 2006).

Detailed knowledge also exists for peripheral trafficking of CFTR after the Golgi and distal secretory pathway of CFTR, as well as membrane retrieval and recycling, where additional checkpoints assess the folding status of secretory proteins (Gentzsch et al. 2004, Guggino & Stanton 2006, Gentzsch et al. 2007). A number of interacting proteins (SNAREs, Rab's, adaptors, cytoskeletal proteins as well as kinases) have also been described to affect the distal membrane traffic of CFTR. In fact, CFTR has been demonstrated to undergo regulated endocytosis in epithelial cells (Prince et al. 1994). Studies aimed at identifying the structural features of CFTR required for endocytosis revealed that Y1424 and I1427 act as essential signals favouring CFTR internalization (Prince et al. 1999, Peter et al. 2002, Swiatecka-Urban et al. 2002). Moreover, both the C-terminus (Gentzsch & Riordan 2001) and more recently the N-terminus (Ramalho et al. 2009) were shown to stabilize CFTR at the plasma membrane. This stabilizing effect of CFTR through its ends was shown to be mediated by cytoskeleton anchoring, namely to filamins for the N-terminus (Thelin et al. 2007) and to actin via NHERF through the PDZ binding domain (PDZ-BD) of CFTR (its last three residues – TRL), the latter also required for the polarized apical expression of CFTR. Indeed, it was shown that deletion of the PDZ-BD (ΔTRL-CFTR) reduced CFTR half-life at the apical membrane from ∼24 to 13 h, and this was shown to occur by reduction of CFTR endocytic recycling (Swiatecka-Urban et al. 2002). It is nevertheless believed that key proteins playing a role in CFTR traffic remain unidentified, namely those along the secretory pathway sensing the F508del-CFTR folding defect and targeting it off-pathway.

Annexins have been implicated in regulating ion channels (reviewed in Gerke & Moss 2002) and in particular CFTR was described to be regulated by annexin A2 (Borthwick et al. 2007). Therefore, we decided to investigate the regulation of CFTR traffic and function by AnxA5, which was reported to be overexpressed in CF tissues (Della et al. 1995). More recently, AnxA5 was reported to interact with CFTR (Trouvé et al. 2007) and even to rescue the membrane expression of the most frequent mutant in CF (Le Drevo et al. 2008).

Our goal here was to further characterize the previously reported functional interaction between AnxA5 and CFTR and possibly also to identify novel CIP's playing a role in CFTR traffic or function. By overexpressing AnxA5 in oocytes, we found an inhibition of CFTR function and that this effect is protein kinase C (PKC)-dependent and results from either endocytosis activation and/or exocytosis block. In contrast, co-expression of AnxA5 in HEK293 cells augmented CFTR whole-cell currents, an effect that was independent of CFTR PDZ binding domain (PDZ-BD) located at the C-terminus and acting via NHERF1. Such an effect, however, was no longer observed for the CFTR endocytic-incompetent (Y1424A,I1427A) CFTR mutant. Since we could detect neither a direct in vivo co-localization nor interaction between AnxA5 and CFTR that would be suggestive of regulated traffic, we propose that AnxA5 acts dually by inhibiting both exocytosis (in oocytes) and endocytic recycling (in mammalian cells), by reducing membrane protein mobility, as previously described for annexins in general and AnxA5 in particular.

Material and methods

Affinity chromatography

The first nucleotide binding domain (NBD1) of CFTR was used as ‘bait’ in affinity chromatography. To this end, NBD1 of CFTR (residues T389 – A655) was cloned into the HIS-tag pET28a bacterial expression vector (Qu & Thomas 1996), produced and purified in BL21 (DE3) bacteria. To ensure that NBD1 maintained a native-type fold (so as to capture native interactions) soluble NBD1 was purified directly from the bacterial supernatant (and not from refolded protein from the inclusion body pellet) and then purified by metal affinity chromatography. The purified p-His-NBD1 was present as a single protein that was recognised by anti-NBD1 antibody L12B4 (Chemicon, USA) and displayed an intrinsic tryptophan fluorescence emission spectrum consistent with folded NBD1 protein (peak emission at 342 nm). Following incubation of the NBD1 in urea (6 M), the emission intensity greatly decreased and the peak shifted to 355 nm, consistent with the presence of fully denatured human NBD1 (data not shown, see Scott-Ward & Amaral 2009).

Purified His-NBD1 (∼1 mg) was then incubated with cell lysates from the respiratory epithelial human cell line Calu-3 cell lysates. Briefly, Calu-3 cells were rinsed twice with pre-warmed PBS and then immediately scraped from the plastic surface. Cells were then harvested by centrifugation and pellet re-suspended in lysis buffer (159mM NaCl, 50mM Tris, pH 8.5) containing protease inhibitor and 0.5% (w/v) NP40. After incubation at 4°C for 10 min, the sample was centrifuged for pre-clearing and supernatant incubated with His-NBD1 (3 h at 4°C) to capture proteins expressed in human respiratory epithelia. NBD1-CIP complexes were then captured from the cell extracts by incubation with Talon metal-affinity resin (3 h at 4°C). To minimize non-specific binding of proteins, the resin was previously incubated with a blocking buffer (5% milk+1% BSA). Controls were performed by incubating the cell lysate with the resin without NBD1. Bound proteins were eluted from the resin using firstly 1M NaCl, 0.1M glycine buffer (pH = 6.5), then 1M NaCl, 0.1M glycine and 20% glycerol buffer (pH = 6.5). Following elution, the recovered proteins were subsequently separated by 2D-electrophoresis.

2D-gels and protein identification

Amicon centrifugal Filter Devices (Milipore Corporation, 2005, USA) were used to concentrate and desalt the fractions eluted from the column. The samples were also further cleaned and precipitated using 2-D Clean-Up kit (Amersham Biosciences, New Jersey, USA) and, after addition of rehydration buffer (7M urea; 2M thiourea; 4% (w/v) CHAPS; 60 mM DTT; IPG buffer pH 3–10; 0.0002% (w/v) bromophenol blue), incubated for 1 h at room temperature. Samples were cleared by centrifugation at 12,000 g for 5 min and the supernatants were loaded onto 13 cm long Immobiline Dry Strips (Amersham Biosciences) with a nonlinear wide-range pH gradient (pH 3–10) for the first dimension (pI) separation. After gel strip re-hydration, IEF was run on an IPGphor IEF system for a total of 70kVh, during which the voltage was gradually increased up to 5000 V. Second dimension electrophoresis (SDS-PAGE) was as previously described (Roxo-Rosa et al. 2006). After electrophoresis protein visualization was carried out by silver staining, compatible with mass spectrometry (MS) analysis. By comparing the spot patterns on control and experimental gels (data not shown), seven candidate CIPs unique to the experimental gel were selected for commercial mass spectrometry (Maldi-Toff) and identified using the Mascott Program (Matrix Science Ltd). Five of these seven selected spots were identified as putative CIPs, being annexin A5 one of these (see Supplementary Table I – online only).

Cell culture

Human embryonic kidney (HEK293) cells and human submucosal gland (Calu-3) cells were grown in Dulbecco's modified Eagle's (DMEM)/Ham's F-12 medium (1:1) in a humidified CO₂ incubator at 37°C. The medium was supplemented with 100 units/ml penicillin, 100 μg/ml streptomycin (Pen/Strep) and 10% fetal calf serum (FCS). HEK293 cells were seeded on bovine plasma fibronectin (Invitrogen, Karlsruhe, Germany) and bovine dermal collagen (Cellon, Luxembourg) coated plastic dishes or glass cover slips. H441 cells were culture in RPMI-1640 media supplemented with 1% Pen/Strep, 1% Insulin-Transferrin-Selenium, 200 nM dexamethasone, 2% Ultrosa G. Baby hamster kidney (BHK) cells were cultured, seeded and used as described previously (Farinha & Amaral 2005).

Chemicals

PKC inhibitor BIM (bisindolylmaleimide I): 0.1 μM; Ca2+ chelator BAPTA-AM (1,2-bis(o-aminophenoxy)ethane-N,N,N,N′-tetraacetic acid): 50 μM, 1h; selective inhibitor of the Ca²⁺/calmodulin-dependent kinase kinase (CaMKK) STO-609: 10 μM; dynamin-mediated endocytosis inhibitor dynasore: 80 μM, 2h; Ca²⁺ ionophore ionomycin: 10 μM.

Downregulation of AnxA5 by Stealth^TM RNAi and overexpression of AnxA5

AnxA5-siRNA (human, Invitrogen, Germany, Cat No:10620318) was transfected into HEK293 and H441 cells. The sense strand of the RNAi used to silence human AnxA5 was 5′-GGGCUGAUGCAGAAACUCUUCGGAA-3′. A ‘scrambled’ RNAi ds-oligomer sequence, i.e. not homologous to any known gene (negative control low GC, Invitrogen) served as control. H441 cells were transfected with fluorescent RNA in order to determine the transfection efficacy, which was in the range of 50–80 %. Transfection of the HEK-293 cells was carried out one day after seeding on glass cover slips (Lipofectamine™ 2000, Invitrogen) in Opti-MEM 1. After 48 to 72 h, cells were used for patch clamping. pcDNA3.1 vector carrying cDNA for human CFTR construct and pT-Rex-Dest vector containing human AnxA5 cDNA were co-transfected (Lipofectamine™ 2000) in Opti-MEM into HEK-293 cells, together with pEGFP-1 (Clontech, Palo Alto, CA, USA) at a ratio of 10:1. Transfected cells were identified by EGFP fluorescence and used for patch clamp experiments. H441 cells were grown as polarized monolayers on permeable supports (Millipore, Schwalbach, Germany) for 10 days in air/liquid interface conditions and were then transfected (Lipofectamine™ 2000) with AnxA5-siRNA. Ussing chamber experiments were performed 72 h after transfection.

Patch clamping and Ussing chamber experiments

Patch-clamp experiments were performed on HEK-293 cells expressing CFTR/Anxa5 in the fast whole-cell configuration as described recently (Spitzner et al. 2008). At intervals membrane voltages (V_c) were clamped in steps of 10 mV from −50 to +50 mV relative to resting potential. The membrane conductance G_m was calculated from the measured current (I) and V_c values according to Ohm's law. For Ussing chamber experiments H441 cells were grown to confluence on permeable supports (Millipore, Schwalbach, Germany) and were mounted into a perfused micro Ussing chamber. The luminal and basolateral surfaces of the epithelium were perfused continuously with buffer solution at a rate of 5–10 ml/min (chamber volume 2 ml). All experiments were carried out at 37°C under open circuit conditions. Transepithelial resistance (Rte) was determined by applying short (1 s) current pulses (ΔI = 0.5 μA) and the corresponding changes in Vte (Vte) and basal Vte were recorded continuously. Values for the transepithelial voltage (Vte) were referred to the serosal side of the epithelium. Rte was calculated according to Ohms law (Rte = ΔVte/ΔI). The equivalent short-circuit current (Isc) was calculated according to Ohms law from Vte and Rte (Isc = Vte/Rte).

cRNAs and double electrode voltage clamp

Oocytes were injected with cRNA (10 ng, 47 nl double-distilled water) encoding wt-CFTR, mutant CFTR and human AnxA5 and S100A8. The constructs were linearized with Notl or HpaI and in vitro transcribed using T7 or SP6 promoter and polymerase (Promega). Water-injected oocytes served as controls. Two to four days after injection, oocytes were impaled with two electrodes (Clark Instruments Ltd, Salisbury, UK), which had a resistance of <1 MΩ when filled with 2.7 mol/l KCI. DEVC experiments were performed as described previously (Bachhuber et al. 2008). Oocytes were continuously perfused throughout the experiment with ND96 media (in mM: 96 NaCl, 2 KCl, 1.8 CaCl₂, 2 H₂O, 1 MgCl₂, 6 H₂O, 5 HEPES, 2.5 sodium pyruvate, adjusted to pH 7.5). Using two bath electrodes and a virtual-ground head stage, the voltage drop across the serial resistance was effectively zero. Membrane currents were measured by voltage clamping (oocyte clamp amplifier, Warner Instruments LLC, Hamden CT) in intervals from −60 to +40 mV, in steps of 10 mV, each 1 s. The bath was continuously perfused at a rate of 5 ml/min. All experiments were conducted at room temperature (22°C).

Biochemical analyses

Co-immunoprecipitations (co-IPs) were carried out as previously described (Farinha & Amaral 2005). Briefly, cell extracts were obtained with lysis buffer [50 mM Tris; 150 mM NaCl; 10 mM (NH4)2MoO4; 0.09% (v/v) NP40; pH 7.4] and then incubated overnight with the first antibody (Ab): anti-CFTR 596 monoclonal (CFF, USA) or anti-AnxA5 polyclonal sc-8300 (Santa Cruz Biotechnology). Protein G-agarose beads (Roche) were added and incubation continued for a further 4-h period. After being washed with lysis buffer, the proteins were eluted from the beads in 1% (w/v) SDS for 1 h at 37°C. These eluates were then adjusted to the composition of RIPA lysis buffer for Western blot (WB). WB was performed as previously (Farinha et al. 2004), but using 25 U/ml of benzonase (Sigma) to shear chromosomal DNA, the same anti-CFTR or anti-AnxA5 Abs and the SuperSignal West Pico chemiluminescent substrate system (Thermo Scientific, Rockford, IL, USA). Filters were exposed to Fuji films (Fujifilm Europe GmbH, Düsseldorf, Germany).

Microscopy

HEK293 or HeLa Kyoto cells were seeded in 8-well Labtek chambered glass slide (Nunc-VWR International Holding Europe GmbH, Zaventem, Belgium). The following day, cells were co-transfected with 333 ng of pEGFP-N1- AnxA5 (Skrahina et al. 2008) and with either 667 ng of the novel mCherry-Flag-wt- or F508del-CFTR constructs cloned into pLenti vector (Invitrogen) using 1.5 μl of Fugene (Roche, Basel, Switzerland) in each well. The translocation of AnxA5 from the cytoplasm to the plasma and nuclear membranes was followed by incubating the cells (2–5 min, 37°C) with 5–10 μM of ionomycin (Sigma-Aldrich, Corp, St. Louis, MO, USA) and visualized by time-lapse fluorescence microscopy in single frames or fixed after 5 min using 4% paraformaldehyde (Polysciences Europe GmbH, Eppelheim, Germany). Images were acquired in a LSM 710 microscope (Zeiss, Jena, Germany). Immunofluorescence on BHK cells was performed as previously (Mendes et al. 2004). Briefly, stably CFTR-transfected BHK cells grown on chamber-slides (Nalgene Nunc, Roskilde, Denmark) were fixed in methanol, permeabilized and incubated for 1 h with 24-1 monoclonal anti-CFTR antibody (R&D Systems, Minneapolis, MN, USA) and AnxA5 sc-8300 (Santa Cruz Biotechnology, CA, USA) and after washing, incubated with FITC- or and rhodamine-conjugated secondary Abs, respectively (Amersham Biosciences Corp, Piscataway, NJ, USA) and mounted in Vectashield (Sigma-Aldrich) with DAPI (blue) for nuclear staining. Preparations were observed in different microscopes, as indicated in figure legends.

FRET experiments

Cells were co-transfected with vectors encoding for GFP-AnxA5 and mCherry-CFTR and imaged using LSM710 miscrocope (Zeiss). Förster Resonance Energy Transfer (FRET) was measured by acceptor photobleaching. The regions of interest in cells expressing both fluorophores were photobleached using the maximum power of 561 nm laser light to selectively photobleach the acceptor (mCherry). GFP and mCherry images were taken both before and after acceptor photobleaching. The unbleached areas were used as a control. The intensity of the GFP signal was quantified in bleached and unbleached areas using ImageJ. Absence of increase in fluorescence intensity of the donor (GFP) within the mCherry photobleached vs non-bleached areas, was taken as indication that the FRET between GFP-AnxA5 and mCherry-CFTR did not occur.

Materials and statistical analysis

All compounds used were of the highest available grade of purity and were from SIGMA (Taufkirchen, Germany) or Calbiochem (Germany). Student's t-test (for paired or unpaired samples as appropriate) was used for statistical analysis with p < 0.05 accepted as significant.

Results

Annexin A5 inhibits CFTR in Xenopus oocytes

We explored the potential effects of AnxA5 on CFTR by co-expressing wt-CFTR and AnxA5 in Xenopus laevis oocytes and determining the effect on whole-cell conductances in double-electrode voltage clamp (DEVC) experiments. Co-expression of AnxA5 reduced activation of CFTR by IBMX (1 mM) and forskolin (2 μM) (I/F) (Figure 1A,B). Inhibition of CFTR conductance by AnxA5 may be due to reduced membrane expression or reduced channel activity. We reported earlier that the CFTR R-domain mutant S737A/S768A-CFTR exhibits an enhanced channel activity even under control conditions, i.e., in the absence of I/F, due to the inhibitory role that these residues play in regulating the CFTR channel pore (King et al. 2009, Kongsuphol et al. 2009). Here, co-expression of AnxA5 still reduced I/F-induced whole-cell currents generated by S737A/S768A-CFTR but does not inhibit the baseline, i.e., spontaneously active CFTR currents (Figure 1B). These data suggest that inhibition of CFTR activity by AnxA5 does not result from a direct effect on closure of an open CFTR channel through these two channel inhibitory residues (Figure 1C). This result led us to explore other mechanisms, namely a possible effect of AnxA5 on CFTR traffic.

Figure 1. AnxA5 inhibits CFTR in Xenopus oocytes. (A) Original traces of the whole-cell current and (B) Summary of whole-cell conductances activated upon stimulation with IBMX and forskolin (I/F, 1mM, 2μM) of oocytes expressing CFTR in the absence (left) or presence (right) of co-expression AnxA5. A reduction in whole-cell conductance was observed when AnxA5 was co-expressed. (C) The inhibitory effect of AnxA5 on whole-cell currents is equally observed in oocytes expressing the CFTR-mutant S737A/S768A. Open bars indicate the conductance before stimulation, i.e., in the absence of I/F. Mean ± SEM, n = number of experiments. Indicates significant activation of whole-cell conductances by I/F (paired t-test). ^#Indicates significant difference from cells lacking expression of AnxA5 (unpaired t-test).

AnxA5 inhibits membrane trafficking of CFTR in Xenopus oocytes

Since the family of annexin proteins has been described to influence protein trafficking as well as to regulate interactions between membranes and the cytoskeleton (Gerke et al. 2005), we examined whether compounds that are known to interfere with trafficking and the cytoskeleton are able to abolish the effects of AnxA5 in Xenopus oocytes. Thus, AnxA5 may control CFTR trafficking from the sub-membrane vesicle pool to the plasma membrane upon cAMP/PKA activation, which is known to be actin-dependent (Cantiello 2001). The actin-depolymerizing compound cytochalasin D (Cyto D) interferes with protein trafficking by inhibiting both endo- and exocytosis (Hug et al. 1995, Cantiello 2001). Here, we observe that Cyto D (10 μM, 2h) significantly reduced activation of CFTR by I/F (Figure 2A, left). Moreover, in the presence of Cyto D, AnxA5 had no further inhibitory effect on CFTR (Figure 2A, right), suggesting that AnxA5 may inhibit activation of CFTR by an actin-dependent mechanism.

Figure 2. AnxA5 inhibits CFTR function in Xenopus oocytes independently of the PDZ domain. (A) Summary of whole-cell conductances activated upon stimulation with I/F (1 mM, 2 μM) of oocytes expressing CFTR with and without AnxA5 co-expression. Effects of pre-incubation with cytochalasin D (Cyto D, 10 μM, 2h) on activation of CFTR. (B) Effects of pre-incubation with dynasore (80 μM, 2 h) on activation of CFTR. (C) Summary of whole-cell conductances activated upon stimulation with I/F (1 mM, 2 μM) in oocytes expressing CFTR mutant E1474X (lacking the PDZ-BD) with and without co-expression of AnxA5. (D) Effects of co-expressing the protein S100A8 on whole-cell conductance of oocytes expressing CFTR and AnxA5. Open bars indicate the conductance before stimulation, i.e., in the absence of I/F. Mean ± SEM, n = number of experiments. Indicates significant activation of whole-cell conductances by I/F (paired t-test). ^#Indicates significant difference from cells lacking expression of AnxA5 (unpaired t-test).

The cell-surface density of CFTR is determined by the rates of exocytosis and endocytosis, the latter being dependent on dynamin, a component of clathrin-coated vesicles (Doherty & McMahon 2009). We examined whether AnxA5 interferes with CFTR endocytosis by using the dynamin inhibitor dynasore (80 μM, 2h) previously described to affect CFTR endocytosis (Young et al. 2009, Cholon et al. 2010). As a positive control for the effects of dynasore we show here that dynasore inhibits Nedd4-2-dependent endocytosis of ENaC induced by phenformin. Phenformin is known to induce Nedd4-2/dynamin mediated endocytosis of ENaC and inhibition of the membrane conductance (ΔG = 11.5 ± 1.7 nS). Dynasore largely reduced inhibition of ENaC by blocking dynamin-mediated endocytosis (ΔG = 2.4 ± 0.4 nS). Already in the absence of AnxA5, dynasore did not cause a significant difference in I/F currents (Figure 2B, left) indicating that endocytic retrieval of CFTR does not seem to play a major role in oocytes. However, in the presence of dynasore, CFTR there was no longer an inhibition of CFTR by AnxA5 (compare two ‘dynasore’ black bars in left and right graphs of Figure 2B). These results suggest that AnxA5 inhibits CFTR, at least partially, by increasing endocytic retrieval from the cell membrane.

Moreover, this inhibitory effect of AnxA5 on CFTR in oocytes does not appear to rely on known retrieval sequences in CFTR (Peter et al. 2002) since mutation of essential amino acids within these motifs (Y1424A/I1427A-CFTR) do not abrogate the inhibitory effect of AnxA5 in oocytes (Supplementary Figure 1A – online only).

On the other hand, endocytosis can be augmented by protein kinase C (PKC) which is potently inhibited by 0.1 μM bisindolylmaleimide I (BIM, by interacting with the PKC kinase domain), and prolonged activation of PKC has been shown to inhibit membrane retrieval of CFTR (Broughman et al. 2006, Alvi et al. 2007). In the presence of BIM, AnxA5 was unable to inhibit CFTR (Supplementary Figure 1B – online only), while blocking of the Ca²⁺/calmodulin dependent (CaM) kinase with STO-609 (10 μM), or chelating Ca²⁺ with BAPTA-AM (50 μM, 1h), did not affect AnxA5-dependent inhibition of CFTR (Figure S1C,D, respectively). These data suggest that in oocytes the endocytic retrieval of CFTR by AnxA5 is mediated by PKC, but independent of CAMK or of lowering the intracellular concentration of Ca²⁺ ([Ca²⁺]_i).

The PDZ-binding domain of CFTR is important for its stabilization at the plasma membrane in mammalian cells, but not in Xenopus oocytes (Boucherot et al. 2001, Schreiber et al. 2004). To test whether CFTR lacking the PDZ-binding domain located at the C-terminus (ΔPDZ-BD) is still inhibitable by AnxA5, the E1474X-CFTR construct was co-expressed in oocytes together with AnxA5. Results clearly show that CFTR-ΔPDZ-BD was still inhibited by co-expression of AnxA5 (Figure 2C).

Because the S100A8 protein is known to interact with AnxA5 and it was also captured in our affinity chromatography assay (Supplementary Table I – online only) we also tested the effect of overexpressing it together with AnxA5, on CFTR currents in oocytes (Figure 2D). Results show that the inhibition of CFTR currents caused by AnxA5 was the same in the presence or absence of S100A8.

Taken together, these experiments suggest that in Xenopus oocytes, CFTR is inhibited by AnxA5 through enhanced retrieval from the plasma membrane in a PKC-dependent fashion and that this endocytosis proceeds independently of the Y1424/1427 motifs and PDZ-BD of CFTR. Data are also compatible with some inhibition of CFTR exocytosis in oocytes.

Regulation of CFTR by AnxA5 in mammalian cells

Interestingly, and in contrast to the above data obtained in oocytes, when co-expressed in HEK293 cells we found that AnxA5 increased CFTR-mediated currents by about 30% (Figure 3), consistent to what has been described previously (Trouvé et al. 2007).

Figure 3. AnxA5 activates CFTR-currents in HEK293 cells. (A) I/F (100 μM; 2 μM) activated whole-cell currents in HEK293 cells expressing wt-CFTR or co-expressing wt-CFTR and AnxA5. Partial removal of Cl^- from the bath solution (32 Cl⁻) inhibits whole-cell currents, indicating activation of CFTR Cl⁻ currents. (B) Summary of the whole-cell conductances activated upon stimulation with I/F. Patch-clamp experiments performed in the fast whole-cell configuration. Open bars indicate the conductance before stimulation, i.e., in the absence of I/F. Mean ± SEM, n = number of experiments. Indicates significant activation of whole-cell conductances by I/F (paired t-test). ^#Indicates significant difference from cells lacking expression of AnxA5 (unpaired t-test).

We then tested the effect of the above Y1424A/I1427A-CFTR mutant previously reported to reduce retrieval of CFTR from the cell membrane (Peter et al. 2002). It is observed that in HEK cells expressing this variant, the baseline CFTR-mediated Cl⁻ conductance (in the absence of IBMX/Fors), is already significantly larger than the one in wt-CFTR expressing cells. Indeed, an increase in the net current (at Vc = +50 mV) by the Y1424A/Y1427A-CFTR variant is observed vs wt-CFTR (from 2.2 ± 0.3 to 3.4 ± 0.2 nA), similarly to what has previously been reported (Peter et al. 2002) (1.07 ± 0.09 to 1.65 ± 0.2 nA). Thus, the Y1424/Y1427A –mutations show a relatively small increase in IBMX/Fors – induced whole cell conductances (Figure 4A, left).

Figure 4. Enhancement of CFTR function by AnxA5 and membrane trafficking in mammalian cells. (A) Summary of the whole-cell conductances induced by I/F in HEK293 cells expressing the CFTR membrane retrieval-mutant Y1424A/I1427A in the absence or presence of AnxA5. (B) Effects of co-expression of AnxA5 on Y1424A/I1427A-CFTR in the presence of the dynamin-inhibitor dynasore (40 μM). (C) Summary of the effects of co-expression of AnxA5 on E1474X-CFTR mutant (ΔPDZ-BD). (D) I/F activated whole-cell conductances generated by the double retrieval/ΔPDZ-BD mutant Y1424A/I1427A/E1474X-CFTR in the absence or presence of AnxA5. Open bars indicate the conductance before stimulation, i.e., in the absence of I/F. Mean ± SEM, n = number of experiments. Indicates significant activation of whole-cell conductances by I/F (paired t-test). ^#Indicates significant difference from cells lacking expression of AnxA5 (unpaired t-test).

We also observed that, in contrast to oocytes, co-expression of AnxA5 no longer enhanced whole-cell currents produced by Y1424A/I1427A-CFTR (Figure 4A, black bars in left and right graphs). These data suggest that AnxA5 enhances membrane currents by interfering with CFTR-internalization. If Y1424A/I1427A-CFTR already shows maximal membrane expression, it is logical to assume that AnxA5 will no longer enhance it. So, while the internalization signal appears to be irrelevant in oocytes, it is important in mammalian cells. The trafficking (including endocytosis) of CFTR in oocytes is remarkably different when compared to mammalian cells, as previously noted (Weber et al. 2001).

Dynamin-dependent endocytosis can be potently blocked by dynasore (Almaça et al. 2009). As observed for the retrieval mutant Y1424A/I1427A-CFTR, inhibition of endocytosis by dynasore did not augment whole-cell conductance generated by wt-CFTR, but abrogated the effects of AnxA5 (Figure 4B).

In contrast to Xenopus oocytes, stability of CFTR at the cell membrane is largely dependent on the PDZ-BD, being CFTR-ΔPDZ-BD highly unstable in mammalian cells in comparison to Xenopus oocytes (Sun et al. 2000, Boucherot et al. 2001, Schreiber et al. 2004, Guggino & Stanton 2006). So, we further examined whether the PDZ-BD at the CFTR C-terminus is relevant to AnxA5-mediated upregulation of CFTR-currents, by expressing the CFTR-ΔPDZ-BD mutant E1474X-CFTR. Whole-cell conductances generated by I/F-dependent stimulation of E1474X-CFTR (24.1 ± 2.1 nS) were reduced when compared to those of wt-CFTR (34.2 ± 2.8 nS) but were still further augmented through co-expression of AnxA5 (Figure 4C). These data indicate that the stimulatory effect by AnxA5 on CFTR is independent of its PDZ-BD and hence of NHERF-binding. Both, the PDZ-BD and the putative retrieval sequences seem to be of little relevance for oocyte-expression of CFTR (c.f. above). We then generated the triple mutant Y1424A/I1427A/ E1474X – CFTR to ‘mimic’ the situation in oocytes, i.e., to abolish endocytic retrieval of CFTR. Notably, the whole-cell conductance of Y1424A/I1427A/E1474X –CFTR was unchanged (33.4 ± 4.5 nS) when compared to that of wt-CFTR, and was no longer augmented but rather inhibited (21.0 ± 3.9 nS) by co-expression of AnxA5, thus unmasking an inhibitory effect of AnxA5 on CFTR expressed in mammalian cells (Figure 4D), similar to the one observed in Xenopus oocytes.

Effect of AnxA5 on F508del-CFTR

We also examined the effect of AnxA5 on F508del-CFTR, since it has been reported that AnxA5 increases cell surface expression and the Cl⁻ channel function of the F508del-CFTR (Le Drevo et al. 2008). When F508del-CFTR was expressed in the absence of AnxA5, a small whole-cell conductance of 0.9 ± 0.2 nS (n = 7) was activated by I/F that was indistinguishable from background activation in non-expressing HEK293 cells (Supplementary Figure 2 – online only). Co-expression of AnxA5 did not increase whole-cell conductance activated by I/F (1.0 ± 0.2 nS; n = 7), suggesting that AnxA5 is unable to increase surface expression and activation of F508del-CFTR (Supplementary Figure 2 – online only). Taken together AnxA5 has multiple indirect effects on CFTR causing slightly enhanced Cl^- conductance but does not correct the transport defect in F508del-CFTR (Le Drevo et al. 2008).

Do AnxA5 and CFTR interact in vivo?

To determine whether the increased CFTR activity caused by AnxA5 overexpression in mammalian cells results from interaction of the two proteins, as previously proposed (Tzima et al. 2000) we tested their in vivo co-localization by confocal microscopy. To this end, we co-expressed GFP-AnxA5 and mCherry-CFTR (both wt- and F508del) in either HEK293 or HeLa cells. Our data shows that wt-CFTR was mostly detected at the cell membrane (Figure 5A, upper row, left panel) and F508del-CFTR at its typical intracellular localization (lower row, left panel), whereas AnxA5 showed an unstructured cytoplasmic pattern and nuclear distribution, but not significantly at the plasma membrane (Figure 5A, middle panels in both rows), thus not significantly co-localizing with wt-CFTR. Co-localization between AnxA5 and F508del-CFTR was not observed either. Importantly, no rescue of F508del-CFTR to plasma membrane occurred as a result of AnxA5 overexpression (Figure 5A, right panels, lower row). Similar results were obtained in HEK293 cells (data not shown). To exclude that co-localization is prevented by the presence of the GFP-tags which is fused to AnxA5, we also assessed the intracellular localization of endogenously expressed (non-tagged) AnxA5 in HeLa cells using an anti-AnxA5 Ab, showing the lack of membrane localization for untagged AnxA5 (Figure 4S). Indeed, AnxA5 localizes both on the cytoplasm and nucleus (Supplementary Figure 4 – online only), although less visible than GFP-AnnxA5 (Figure 5) due to poor penetrance of the Ab into the nucleus. Nevertheless, nuclear localization of untagged AnxA5 is confirmed by its clear shift to the nuclear membrane upon ionomycin treatment (Supplementary Figure 4 – online only) just like GFP-AnxA5, as shown here (Supplementary Figure 3 – online only) and by others (Skrahina et al. 2008).

Figure 5. CFTR and AnxA5 do not co-localize in vivo. (A) Confocal images of HeLa cells transiently co-transfected with GFP-AnxA5 and either mCherry-flag-wt-CFTR (upper row) or mCherry-flag-F508del-CFTR (lower row). Columns from left to right: mCherry-CFTR; GFP-AnxA5; overlay of mCherry-CFTR and GFP-AnxA5. (B) Confocal images as in A but after treatment with 10 μM ionomycin for 5 min: HeLa cells transiently co-transfected with GFP-AnxA5 alone (upper row); GFP-AnxA5 and mCherry-flag-wt-CFTR (middle row); or GFP-AnxA5 and mCherry-flag-F508del-CFTR (lower row). Columns from left to right: DAPI nuclear staining (in AnxA5 alone cells) or mCherry-CFTR; GFP-AnxA5; overlay of the two previous columns. Scale bar: 10 μm. Cells were observed using a LSM710 Zeiss confocal microscope. Images are representative of at least n = 50 cells from n = 5 independent transfections.

Since AnxA5 is a Ca²⁺-binding protein described to change its location upon changes in [Ca²⁺]_i, and because in previous co-IP studies, CFTR-AnxA5 interaction was described to be Ca²⁺-dependent, we increased the latter by addition of 10 μM ionomycin and tested AnxA5 intracellular localization by immunocytochemistry (Figure 5B) and time-lapse imaging (Supplementary Figure 4 – online only). Interestingly, we observed a translocation of AnxA5 to the plasma membrane and also to the nuclear envelope (Figure 5B, middle panels in all rows). However, even under these high [Ca²⁺]_i conditions, AnxA5 did not co-localize in vivo with either wt-CFTR or with F508del-CFTR (Figure 5B, right panel in middle and lower rows, respectively).

Finally, we performed acceptor bleaching Förster resonance energy transfer (FRET) experiments, which did not demonstrate an in vivo interaction between CFTR and AnxA5 (data not shown). Altogether, the lack of co-localization shown by these results, leads to the conclusion that AnxA5 does not interact in vivo with CFTR, even under high [Ca²⁺]_i.

Because we initially identified AnxA5 in an affinity capture in vitro assay using CFTR-NBD1 as a ‘bait’ and also because a putative interaction of AnxA5 and CFTR was recently described by co-immunoprecipitation (Trouvé et al. 2007), we also tested this approach here. First, we performed co-immunoprecipitations in BHK cells stably expressing wt-CFTR and transiently transfected with AnxA5, by pulling-down CFTR and a band for AnxA5 could indeed be detected by Western blot (Figure 6A, upper panel, lane 4, upper band), albeit very faint despite AnxA5 being present at significantly high levels in transfected (lane 2) vs non-transfected cells (lane 1).

Figure 6. CFTR and AnxA5 do not co-immunoprecipitate. (A) Western blot (WB) of AnxA5 (upper panel) or Hsc70 (a known CFTR interactor as a positive control, lower panel) in total lysates as control (lanes 1,2) or after CFTR co-immunoprecipitation (lanes 3,4) from stable CFTR-expressing BHK cells either non-transfected (lane 2) or transiently transfected (lanes 1,3,4) with AnxA5. In the total lysate controls for WB detection (lanes 1,2) specific bands for both AnxA5 and Hsc70 can be observed at the correct apparent molecular weight. The higher amount of AnxA5 levels detected in lane 2 shows that the transient transfection of AnxA5 was efficient. For samples undergoing co-immunoprecipitation (lanes 3,4) the lysates from cells transiently transfected with AnxA5 were incubated without antibody, ‘beads-only’ (lane 3) or with (lane 4) anti-CFTR primary Ab (IgG) plus the beads to perform the CFTR pull-down. The CFTR-co-immunoprecipitated proteins were then analyzed by WB for AnxA5 (upper panel) or Hsc70 (lower panel). Although AnxA5 was present in the CFTR pull-down (lane 4, upper panel), it was similarly present in the absence of the anti-CFTR Ab (lane 3, upper panel). The lower band, marked ‘*’ corresponds to the IGg light chain. In contrast, Hsc70 was only present in the CFTR immunoprecipitate (lane 4, lower panel), but not in the absence of the anti-CFTR Ab (lane 3, lower panel). In (B) the CFTR-immunoprecipitated material was analyzed by WB for CFTR as the ‘input’ control of the experiment in (A). In (C), the opposite IP experiment of (A) was performed, i.e., AnxA5 was co-immunoprecipitated from AnxA5-transfected, CFTR-expressing BHK cells using AnxA5 Ab and samples were immunoblotted for CFTR. For enhanced sensitivity, the same WB was exposed for longer (left panel) and shorter periods (middle panel). Presence of CFTR in the AnxA5 pull-down can only be observed in the left panel (high exposure time). As a control, a pull-down of CFTR was performed in parallel (right panel) showing that CFTR could be efficiently immunoprecipitated.

However, the same band, and of equivalent intensity, was also observed in pull-downs performed with beads only, i.e., no anti-CFTR Ab (Figure 6A, upper panel, lane 3, upper band), thus indicating that this band is not specific. As an additional control, we could specifically detect Hsc70, a known interactor of CFTR, in CFTR pull-downs (Figure 6A, lower panel), i.e., only when the CFTR Ab was added (lane 4) but never in the presence of the beads only (lane 3). When the opposite experiment was performed, i.e., co-immunoprecipitation of AnxA5 followed by anti-CFTR immunoblot, in some experiments, but not reproducibly, a very faint CFTR band could be detected in co-immunoprecipitated AnxA5 (Figure 6C, left panel). The results from co-IPs regarding AnxA5 and CFTR were thus inconclusive in demonstrating a direct interaction between these two proteins.

Effect of AnxA5 on endogenous CFTR expressed in physiologically relevant cellular systems

To confirm whether AnxA5 also affects CFTR in human respiratory cells where it is endogenously expressed, we used human lung adenocarcinoma cell line H441. These cells express low but detectable levels of CFTR (Faria et al. 2009). However, due to the difficulties in efficiently transfecting epithelial cells with cDNAs, we employed an siRNA strategy instead. Thus, when endogenous AnxA5 was knocked down by siRNA in polarized grown H441 cells, the reverse effect on CFTR conductance was observed (Figure 7). These data further support that in physiologically relevant cells grown to polarization; AnxA5 also has an activating role on CFTR.

Figure 7. siRNA-AnxA5 suppresses whole-cell conductance generated by wt-CFTR: (A) Representative original recording from an open-circuit Ussing chamber experiment with H441 cells grown on permeable supports showing the negative deflection of the transepithelial voltage upon stimulation with I/F (100 μM/2 μM), indicating activation of CFTR-dependent Cl⁻ secretion. Incubation of the cells with siRNA for AnxA5 reduced I/F-induced Cl⁻ secretion when compared to cells treated with scrambled (scrbld) siRNA. (B) Calculated equivalent short-circuit current (Isc) activated by I/F (100 μM/2 μM) in cells treated with scrambled or AnxA5-siRNA. Mean ± SEM, n = number of experiments. ^#Indicates significant difference from cells treated with scrambled siRNA (unpaired t-test).

Discussion

AnxA5 inhibits CFTR in oocytes

Annexin A5 (AnxA5) is a member of the annexin family, an evolutionary conserved multiprotein group. All annexins contain 4–8 repeats of the endonexin domain (a non-EF hand Ca²⁺-binding motif) which can function as organizers of membrane domains and as membrane-recruitment scaffolds for proteins they associate with, by interacting with negatively charged phospholipids in a Ca²⁺-dependent manner (Gerke & Moss 2002). AnxA5, in particular interacts with phosphatidylserines, but it was also shown to interact with the hydrocarbon chains of membrane lipids (Gerke et al. 2005).

Here, we have studied the effect of AnxA5 overexpression on CFTR function both in Xenopus oocytes and in mammalian cells. Our functional data in oocytes showing an inhibitory effect by AnxA5 on CFTR, which disappears under cytochalasin D (Cyto D), clearly suggest that AnxA5 inhibits activation of CFTR by actin-dependent mechanisms, either by inhibiting exocytosis or enhancing endocytosis. We next tested whether in oocytes this effect would result from the described role of annexins in endocytosis, by using dynasore, a known inhibitor of dynamin-dependent CFTR endocytosis (Young et al. 2009, Cholon et al. 2010). Strikingly, still in the absence of AnxA5, dynasore did not significantly enhance I/F currents, suggesting that endocytic retrieval of CFTR plays a minor role in oocytes. However, as I/F currents under dynasore were no further inhibited by AnxA5, this argues that the actin-dependent inhibitory effect by AnxA5 on CFTR in oocytes acts, at least partially, by enhancing endocytosis. Moreover, our data also indicate that in oocytes such a stimulatory endocytic effect occurs independently of the Y1424/I1427 CFTR internalization motifs. The fact that inhibition of CFTR-currents by AnxA5 did not occur in oocytes in the presence of dynasore but was still observed for the endocytosis-deficient Y1424A/I1427A-CFTR variant, might be due to the absence in oocytes of the internalization machinery required to recognize these residues. Nevertheless, possible side-effects of dynasore cannot be fully excluded.

Results shown here also indicate that in oocytes this stimulatory endocytic effect occurs independently of the channel gating inhibitory residues S737/S768 and of the CFTR-PDZ binding domain (PDZ-BD). The latter efficiently mediates CFTR stabilization at the plasma membrane in mammalian cells by counteracting its endocytosis (Short et al. 1998, Benharouga et al. 2003, Guerra et al. 2005). However, our results indicate that in oocytes the endocytic retrieval of CFTR by AnxA5 is mediated by PKC, a major promoter of internalization (Alvi et al. 2007). These results are consistent with recent data demonstrating that AnxA5 acts as a PKC activator (Kheifets et al. 2006), despite previous contrasting results (Schlaepfer et al. 1992, Wang & Kirsch 2006). Nevertheless, given that CFTR endocytosis does not seem to be a major pathway in oocytes, some of the inhibitory effect by AnxA5 may also result from a block in actin-mediated exocytosis. Consistent with this possibility, it has previously been shown that activation of Cl^- currents in Xenopus oocytes is partly due to membrane exocytosis of CFTR (Cunningham et al. 1992, Takahashi et al. 1996, Weber et al. 1999).

Enhancement of CFTR function by AnxA5 in mammalian cells

Recent findings support a role for some annexins in regulating membrane traffic of various ion channels, both exocytosis (AnxA2; AnxA7; AnxA13) and endocytosis (AnxA1; AnxA2; AnxA6) (Futter & White 2007). Notably, by overexpressing AnxA5, Trouvé and co-workers demonstrated recently that AnxA5 is necessary for CFTR function and plasma membrane localization in airway epithelial cells, concluding that annexin A5 is necessary for normal CFTR Cl⁻ channel activity (Della et al. 1995, Trouvé et al. 2007). Our results in mammalian cells here partially support these data, as co-expression of AnxA5 and CFTR resulted in a significant increase of I/F stimulated Cl⁻ currents. Moreover, since this enhancement of CFTR function is abolished when cells are incubated with the dynamin-inhibitor dynasore, we conclude that this effect results from AnxA5 interference with endocytosis, as generally reported for annexins. Since these proteins also regulate the dynamic assembly/ disassembly of monomeric actin into filamentous (F−) actin (Stossel et al. 1985, Lajoie & Nabi, 2007) and AnxA5, in particular was shown to interact directly with F-actin (Tzima et al. 2000), actin could indeed be the link to the role of AnxA5 in CFTR endocytosis. Consistently, previous studies have reported that actin-skeleton disruption decreases Cl⁻ secretion (Kheifets et al. 2006), a finding which was also confirmed more recently (Penmatsa et al. 2010). Furthermore, it was proposed that annexins shuttle actin monomers with actin monomer-sequestering proteins, like profilins, (Hayes et al. 2004) and remarkably, in our affinity capture, we also found profilin 2.

Nevertheless, while in mammalian cells, the C-terminal PDZ-BD of CFTR and interacting proteins such as NHERF1 are required for proper expression and membrane stabilization through actin-skeleton tethering, this is not essential in oocytes (Boucherot et al. 2001, Schreiber et al. 2004). Such an explanation can be the reason for the apparently opposing results that we obtained here in oocytes and mammalian cells, arguing that CFTR in oocytes is quite efficiently exocytosed but very little endocytosed, whereas this is not the case in mammalian cells. Indeed, by analyzing the E1474X-CFTR mutant, we reconfirmed in the present study previous findings demonstrating that PDZ-interactions are essential for CFTR stability/function in mammalian cells. However, we found that the positive effect that AnxA5 exerts on CFTR function is independent of the PDZ-BD, thus indicating that the putative stabilization CFTR by AnxA5 mimics it, but is not mediated by NHERF. Curiously, membrane expression of CFTR in oocytes is known to be independent of PDZ-domains and we found that AnxA5 did not enhance CFTR currents in oocytes.

Does the stabilizing effect of AnxA5 on CFTR result from their direct interaction?

In their study of the effects of AnxA5 on CFTR, and based on co-staining in human native tissues, co-IP, overlay and surface plasmon resonance experiments, Trouvé and co-workers state that the observed enhancement of CFTR function results from a direct interaction between the two proteins (Della et al. 1995, Trouvé et al. 2007). However, when we thoroughly investigated the intracellular localization of Anx5 and CFTR in vivo, no significant co-localization could be observed between AnxA5 and wt-CFTR or F508del-CFTR, neither at the cell membrane nor intracellularly. Indeed, we observed that both tagged and untagged AnxA5 localizes in the cytoplasm and in the nucleus. Only upon ionomycin treatment (raising [Ca²⁺]_i to non-physiological levels), did AnxA5 move to the respective compartment membranes (cytoplasmic and nuclear, respective), consistent with previous studies (Barwise & Walker 1996, Skrahina et al. 2008). Since these data are inconsistent with those described by Trouvé and colleagues in human native bronchial tissues (Trouvé et al. 2007), a possible explanation could be that AnxA5 requires polarization to be (apically) membrane localized and thus interact with CFTR, as described for AnxA2 (Gerke et al. 2005). However, our data show that the enhancement of CFTR function by AnxA5 occurs similarly in non-polarized and in polarized epithelial cells, thus hinting that the functional activation by AnxA5 does not require cell polarization. As membrane localization of AnxA5 is known to occur in apoptotic cells (being membrane-localized AnxA5 a common biomarker of apoptosis (Gerke et al. 2005) or after cell wounding/injury (Skrahina et al. 2008), a possible explanation for the previously described co-localization of CFTR and AnxA5 could be due to occurrence of either situation in those native tissues, causing AnxA5 to translocate to the plasma membrane. Since in the present study we did not observe AnxA5 at the plasma membrane, we can assume that the cells analyzed here were not apoptotic.

Because the previous AnxA5-CFTR association was described to be Ca²⁺-dependent (Trouvé et al. 2007) and given that AnxA5 is known to change its location after an increase in [Ca²⁺]_i which enhances its phospholipid binding (Gerke et al. 2005, Skrahina et al. 2008), we further tested the possible co-localization of CFTR and AnxA5 in vivo after rising [Ca²⁺]_i by addition of ionomycin. Under these conditions, we observed a major translocation of AnxA5 to the plasma membrane (also to the nuclear envelope) as previously described (Gerke et al. 2005, Skrahina et al. 2008). However, AnxA5 still did not significantly co-localize with either wt-CFTR or F508del-CFTR, and the latter remained in its typical ER localization. Furthermore, although relocation of AnxA5 from the cytosol to the plasma membrane under high Ca²⁺ could be suggestive of a role in secretory trafficking, its intranuclear relocation is inconsistent with such a role, given that this effect is more compatible with its well-described Ca²⁺-mediated membrane lipid binding (Gerke et al. 2005, Skrahina et al. 2008).

Our biochemical data could equally not confirm the previously described co-IP interaction between AnxA5 and CFTR (Trouvé et al. 2007). In fact, although we could detect AnxA5 in CFTR co-IPs, the same result was obtained in the absence of anti-CFTR antibody, suggesting that AnxA5 also binds to the beads. Of note, in the co-IPs of the above study, this ‘beads only’ control was never shown (Trouvé et al. 2007). Moreover, this study reported that the presence of AnxA5 in CFTR co-IPs was more significant when Ca²⁺ was added in vitro to the immunoprecipitated complex, thus probably reflecting a Ca²⁺-mediated membrane lipid binding and not so much a real protein-protein interaction occurring in vivo. Such an association between AnxA5 and CFTR found to occur in lysates could thus be part of a larger complex, mediated by lipids and/or other proteins (e.g., actin binding proteins). Moreover, data from the NBD1-AnxA5 obtained by surface plasmon resonance and NBD1-overlay in the same study (Trouvé et al. 2007), as well as our in vitro affinity capture reported here, may reflect an in vitro interaction whereby NBD1 may expose molecular surfaces which are hidden in vivo in the context of full-length CFTR. However, such interaction does not demonstrate that the two proteins interact in vivo, consistent with the lack of in vivo co-localization between AnxA5 and CFTR in results shown here.

In summary, the present results indicate a dual role of AnxA5 in regulating CFTR function: (i) by stabilizing the channel at the plasma membrane in a PDZ-BD independent manner, through tethering to the actin-skeleton underlying the plasma membrane, as previously shown for AnxA2/A4/A6 regarding other ion channels (Gerke & Moss 2002); and (ii) by activating endocytic retrieval and blocking exocytosis in a PKC-dependent fashion. While the first mechanism prevails in mammalian cells, where endocytic retrieval predominates, the second occurs in oocytes. We conclude that both roles are consistent with the described role of annexins forming scaffolding platforms at cell membranes binding to phospholipids in a Ca²⁺-dependent way, hence contributing to a decrease in membrane dynamics and, in mammalian cells, organization of membrane microdomains. Although AnxA5 was identified as a binding partner of NBD1-CFTR, this interaction appears to occur predominantly in vitro, where possibly this domain is more exposed or/and through additional lipid/protein intermediates. In vivo, despite AnxA5 clearly enhancing CFTR function, the two proteins do not significantly co-localize. Moreover, we could not observe the rescue of F508del-CFTR plasma membrane traffic or function by AnxA5, in contrast to a previous observation (Le Drevo et al. 2008). Because of its limited (presumably unspecific) effects on wt-CFTR and lack of correction of F508del-CFTR, we conclude that AnxA5 is not a promising target for F508del-CFTR correction.

Acknowledgments

We acknowledge the expert technical assistance by Ms P. Seeberger. We thank C. Schultz (EMBL, Heidelberg) for the GFP-annexin A5 construct and the US Cystic Fibrosis Foundation (CFF) for the 596 anti-CFTR antibody.

Declaration of interest: This work was supported by DFG SFB699 A6, DFG KU 756/8-2; POCTI/SAU/MMO/58425/2004 (FCT/FEDER Portugal); TargetScreen (EU-LSH-2005-1.2.5-3-037365). LA, TS-W and DF were recipients of fellowships Move-ERDISU (Trieste, Italy) and SFRH/BPD/18989/2004, SFRH/BD/43313/2008 (FCT, Portugal) respectively.