Disruption of cytokeratin-8 interaction with F508del-CFTR corrects its functional defect

Abstract

We have previously reported an increased expression of cytokeratins 8/18 (K8/K18) in cells expressing the F508del mutation of cystic fibrosis transmembrane conductance regulator (CFTR). This is associated with increased colocalization of CFTR and K18 in the vicinity of the endoplasmic reticulum, although this is reversed by treating cells with curcumin, resulting in the rescue of F508del-CFTR. In the present work, we hypothesized that (i) the K8/K18 network may interact physically with CFTR, and that (ii) this interaction may modify CFTR function. CFTR was immunoprecipitated from HeLa cells transfected with either wild-type (WT) CFTR or F508del-CFTR. Precipitates were subjected to 2D-gel electrophoresis and differential spots identified by mass spectrometry. K8 and K18 were found significantly increased in F508del-CFTR precipitates. Using surface plasmon resonance, we demonstrate that K8, but not K18, binds directly and preferentially to the F508del over the WT human NBD1 (nucleotide-binding domain-1). In vivo K8 interaction with F508del-CFTR was confirmed by proximity ligation assay in HeLa cells and in primary cultures of human respiratory epithelial cells. Ablation of K8 expression by siRNA in F508del-expressing HeLa cells led to the recovery of CFTR-dependent iodide efflux. Moreover, F508del-expressing mice topically treated with K8-siRNA showed restored nasal potential difference, equivalent to that of WT mice. These results show that disruption of F508del-CFTR and K8 interaction leads to the correction of the F508del-CFTR processing defect, suggesting a novel potential therapeutic target in CF.

INTRODUCTION

Cystic fibrosis (CF) is attributed to mutations in the gene coding for the chloride channel CFTR (cystic fibrosis transmembrane conductance regulator) (1,2). This protein is composed of two transmembrane domains, two nucleotide-binding domains and one regulatory domain (3). Deletion of phenylalanine 508 (F508del) in the nucleotide-binding domain-1 (NBD1) is present in >70% of mutated alleles (Cystic Fibrosis Mutation Database, http://www.genet.sickkids.on.ca/app). F508del produces a partially folded, core-glycosylated immature form that is largely retained and degraded in the endoplasmic reticulum (ER) (4), but partially functional when it reaches the plasma membrane (5). This has prompted an intense search for small molecular correctors or protein substitution therapy (6–15). We have reported increased expression of cytokeratins 8/18 (K8/K18), which belong, respectively, to groups II and I of intermediate filaments (IF) (16) and are representative of one-layered epithelia (17), in HeLa cells expressing the F508del mutation (18). CFTR and K18 colocalization in the vicinity of the ER is reversed by curcumin, resulting in the rescue of F508del-CFTR (19). Likewise, the K18 network is altered by resveratrol, another molecule that rescues F508del-CFTR (20).

Two recent studies have revealed conformational changes within mutated NBD1. The first one shows experimentally that NBD1 destabilization occurs as a consequence of three solubilizing mutations, namely V510D, F494N and Q637R (21). The second study on NBD1 structure has revealed an increased conformational flexibility due to F508del (22). This may result in the exposure of buried hydrophobic region(s) that become(s) available for interaction with other molecules. We hypothesize that some of these interactions may contribute to F508del-CFTR retention and degradation and comprise the K8/K18 network. In accordance with these results, a very recent report shows that the mild KRT8 allele (K8) is associated with CFTR-mediated residual chloride secretion in F508del-CFTR homozygote patients, suggesting KRT8 as a modifier gene for CF severity CFTR mediated residual chloride secretion (23).

Based on these observations suggesting a determinant role for cytokeratin network in mutant CFTR misprocessing, we hypothesized that K8 and/or K18 interact physically with CFTR. The aim of the present work was to confirm this interaction and to study its biochemical basis and consequences on F508del-CFTR localization and function. Our results show that K8 preferentially binds to F508del-NBD1 over its wild-type (WT) counterpart, and that preventing this interaction in vivo leads to the functional plasma membrane expression of F508del-CFTR, both in cell lines and in F508del mice.

RESULTS

K8 and K18 form a complex with CFTR in F508del-expressing cells

To evaluate the impact of F508del on CFTR interactome, we performed 2D-gel electrophoresis of CFTR immunoprecipitates in HeLa cells stably transfected with either WT-CFTR or F508del-CFTR. Representative gels are shown in Figure 1A. Most of the proteins differentially expressed (shown in detail in Supplementary Material, Table S1) were identified as cytoskeletal, including K8, K18, β-actin and vimentin. Colocalization of K8 and CFTR was confirmed by immunofluorescence at the cytosol in WT-CFTR cells, and perinuclearly in F508del-CFTR cells (Fig. 1B). There was no apparent colocalization of CFTR, confined to the base of cilia, in nasal cells from healthy individuals (Fig. 1C, upper panel). In contrast, significant colocalization was detectable in the perinuclear area in nasal cells obtained from CF patients (Fig. 1C, lower panel).

Figure 1.

The K8/K18 network interacts with CFTR preferentially in F508del-expressing cells. (A) K8/K18 coprecipitates with CFTR preferentially in F508del-expressing HeLa cells. Cell extracts were subjected to immunoprecipitation by the 24.1 monoclonal anti-CFTR antibody. Precipitates were analyzed by 2D-electrophoresis and differentially displayed spots identified by MALDI-TOF. Protein spots designed by numbers are defined in Supplementary Material, Table S1. The gels shown are representative of three independent experiments. (B) Colocalization of K8 and CFTR in HeLa cells. HeLa cells stably transfected with either WT-CFTR or F508del-CFTR were subjected to CFTR and K8 immunodetection and analyzed by confocal microscopy (scale bar = 10 µm). Representative of three independent experiments. (C) Immunodetection of CFTR and K8 localization in cells obtained from nasal scrapings from a healthy control and from a CF patient homozygous for the F508del mutation (scale bar = 10 µm). Representative of three independent experiments.

K8, but not K18, interacts physically with NBD1

To assess K8/K18 direct binding to CFTR in vitro, we used surface plasmon resonance (SPR). Recombinant human NBD1, either WT or bearing the F508del mutation, was covalently bound to a sensor chip and the binding pattern of K8 and K18 was tested at 20/25°C (Fig. 2A). K8 (Fig. 2A, upper panels) increased the resonance signal in both WT- and F508del-NBD1, indicating a direct interaction, unlike K18 (Fig. 2A, bottom panels)—results confirmed in mouse NBD1 (Supplementary Material, Fig. S1). Nevertheless, K8 binding was significantly stronger on F508del-NBD1, as shown by the kinetic parameters and apparent dissociation constants (Supplementary Material, Table S2). These results suggest that K8-NBD1 binding is responsible for the interaction between CFTR and IF, and that this interaction is stronger when NBD1 is mutated. Since temperature influences the conformational stability of NBD1 (5), the experiment was repeated at 7°C, at which F508del-NBD1 should present normal folding. No differences in binding affinity were found between human WT- and F508del-NBD1 at 7°C (Fig. 2B), strongly suggesting that temperature-sensitive destabilization is responsible for the higher affinity of K8 for mutated NBD1.

Figure 2.

K8 interacts physically with NBD1. (A) SPR analysis of K8 and K18 binding to human NBD1 (WT or F508del-NBD1). K8, but not K18, interacts directly with NBD1, and more strongly in the presence of the F508del mutation. Experiments were performed at 20 or 25°C. Different analyte (K8 or K18) concentrations are shown. The kinetic constants were calculated using the solution affinity model from the Biacore BIAEVALUATION 3.1 software. Representative of three independent experiments. (B) SPR analysis of K8 binding to human NBD1, bearing or not the F508del mutation, at 7°C. K8 interacts directly with both NBD1 with the similar affinity. values are shown in both cases. Different analyte (K8) concentrations are shown. Representative of three independent experiments.

To search for K8–CFTR interaction in vivo, we used the proximity ligation assay (PLA) in transfected HeLa cells. A <40 nm proximity between both proteins was found in F508del-CFTR, but not in WT-CFTR-expressing cells (Fig. 3A), strongly suggesting that detectable physical interaction only prevails for F508del mutation. To confirm that the interaction takes place at endogenous expression levels of mutated CFTR, we performed PLA on primary human respiratory epithelia from either healthy individuals or F508del-homozygous patients (MucilAir™). Molecular proximity of K8 and CFTR was detectable in patients cells only (Fig. 3B).

Figure 3.

Differential interaction between K8 and CFTR in cells. (A) PLA of K8 and CFTR in HeLa cells transfected with either WT- or F508del-CFTR. The absence of primary antibody was used as a negative control (scale bar = 10 µm). Results are representative of at least six independent experiments. (B) PLA of K8 and CFTR in human primary respiratory epithelial cells (MucilAir™) from one healthy control and one F508del/F508del patient. A confocal image of a transversal section of the epithelium (y-, z-axes) is shown (scale bar = 10 µm). Resuts are representative of at least three independent experiments.

K8 silencing restores F508del-CFTR function in HeLa cells and in the nasal epithelia of F508del mice

We hypothesized that inhibiting the K8–CFTR interaction by K8 silencing may restore F508del-CFTR plasma membrane localization and function. We tested the effect of K8-siRNA on PLA outcome in transfected HeLa cells using a combination of four siRNA sequences, one of which (Seq. 3) decreased K8 considerably and increased the expression of both B and C bands of CFTR (Fig. 4A and B). K8-siRNA (Seq. 3) treatment abolished the K8–CFTR PLA signal in F508del-expressing cells (Fig. 4B).

Figure 4.

K8 silencing disrupts interaction with F508del-CFTR and increases F508del-CFTR expression. (A) Effect of siRNA treatment on K8 and CFTR protein expression in HeLa cells transfected with F508del-CFTR. Western blot on protein extracts from non-treated cells (ctrl) or treated with scramble siRNA, or with Seq. 3 of K8-siRNA). The membrane was hybridized using a monoclonal anti-K8 antibody and a polyclonal anti-CFTR antibody. A polyclonal anti-GAPDH antibody was used as a loading control. Note that CFTR expression was increased by siRNA-K8 treatment (see two bands in the top panel corresponding to B, lower, and C, upper, of CFTR). The lower panel shows the semi-quantification by densitometric analysis (representative of three independent experiments). (B) Immunofluorescence images of CFTR (green) and K8 (red) in HeLa cells expressing F508del-CFTR in control conditions and after treatment with either Seq. 3 K8-siRNA or with scramble siRNA. (C) PLA of K8 and CFTR in HeLa cells transfected with either WT- or F508del-CFTR after K8-siRNA treatment and after treatment with control scramble siRNA (scale bar = 10 µm).

To evaluate the functional impact of CFTR–K8 interaction, we measured cAMP-dependent iodide efflux, as an index of CFTR-mediated plasma membrane halide transport activity (24), before and after K8-siRNA treatment in WT- and F508del-CFTR-expressing HeLa cells (Fig. 5A and B). CFTR-dependent iodide efflux was consistently increased in K8-siRNA-treated F508del-CFTR cells (Fig. 5B), while marginal elevation was detectable in WT-CFTR cells. The percent variation between t = −1 and t = 0 (where t = −1 indicates addition of CFTR-activating cocktail) in K8-siRNA-treated versus control F508del cells in three independent experiments was 40.9 versus 11.4, 16.4 versus 0.3 and 63.3 versus 9.7 (13.5, −12.5 and 3.7 for scramble siRNA, respectively). We have previously shown the potential therapeutic effect of siRNA targeting of K18, which activated a cAMP-dependent anion conductance consistent with CFTR activity in HeLa cells expressing F508del-CFTR (18). This seminal result pointed at the cytokeratin network as a potential therapeutic target in CF.

Figure 5.

K8-siRNA increases iodide efflux in cells. Iodide efflux measurements in control conditions and after siRNA-K8 treatment in HeLa cells transfected with either WT-CFTR (A) or F508del-CFTR (B). FS-GS indicates the treatment of cells with CFTR activators 10 μm forskolin plus 50 μm genistein. The results are representative of at least three independent experiments.

Our results in cells suggested that K8 ablation may correct the F508del-CFTR functional defect in the respiratory epithelium of mice. Consequently, nasal potential difference (NPD) (▵V_TE) was monitored following 4 and 11 days after intranasal delivery of either lipofectamin-complexed scrambled or K8-specific siRNA. Representative experiments are shown in Figure 6 and the data presented in Supplementary Material, Table S3. As expected, basal V_TE values, as well as the changes induced by both the Na⁺-blocker amiloride and a low Cl⁻-containing solution, were within the range observed previously (25). A typical recording of ▵V_TE in WT-CFTR is shown in Supplementary Material, Figure S2. The CFTR-related ▵V_TE was unmasked by the effect of CFTR inhibitor, I_Inh172. The fact that both B and C bands of F508del-CFTR are increased after K8-siRNA treatment (Fig. 4A) suggests the presence of the mutated channel at the plasma membrane.

Figure 6.

K8-siRNA restores NPD in mice expressing F508del-CFTR. NPD measurements in F508del-CFTR homozygous mice: (A) before application of siRNA, (B) treated with 30 µl of 20 nm scramble-siRNA, (C) 3 days after application of 30 µl of 20 nm K8-siRNA to nasal epithelium and (D) 7 days after application of K8-siRNA to nasal epithelium. Addition of Na⁺-blocker amiloride depolarized V_TE, unmasking Na⁺ conductance. Subsequent perfusion with a low-Cl⁻-containing solution hyperpolarized V_TE, showing participation of Cl⁻ conductances in V_TE. Subsequent addition of CFTR inhibitor I_Inh172 decreased hyperpolarization by ∼30%, suggesting that, in addition to CFTR, other Cl⁻ conductances participate in overall transepithelial Cl⁻ fluxes. (E) NPD values before and after scramble siRNA (left panel) and K8-siRNA (right panel) treatment in six mice.

In F508del mice, at day 0, basal V_TE values were more negative than in WT mice as previously reported (25). Consequently, ▵V_TEamil was larger, indicating the increase in Na⁺ conductance, whereas ▵V_TElowCl was abolished. Comparison of results from WT (Supplementary Material, Fig. S2) and F508del mice suggests a significant contribution of CFTR to V_TE in WT. Changes in ▵V_TE before K8-siRNA treatment are shown in Figure 6A. Scramble-siRNA induced no changes either in ▵V_TEamil or in ▵V_TElowCl (Fig. 6B and E, left). Three days after K8-siRNA application to F508del mice, ▵V_TEamil slightly diminished, and ▵V_TElowCl was significantly increased by the treatment, above a cut-off value of −2.15 mV that we have established (manuscript in preparation). This cut-off value corresponds to a ▵V_TElowCl threshold beyond which the nasal epithelium of ▵F508 displays a WT-like Cl⁻ conductance. In the experiments described here, ▵V_TElowCl reached roughly three-fourths (78%) of WT mice values, as shown in Figure 6C, in three out of six mice (Fig. 6E, right panel). ▵V_TElowCl was partially inhibited by CFTR inhibitor, I_Inh172 (Fig. 6C), supporting the conclusion that K8-siRNA treatment restored CFTR-dependent conductance. Measurement of ▵V_TE 6 days after K8-siRNA showed a return of V_TE to the original values, demonstrating a transient effect of the treatment (Fig. 6D).

DISCUSSION

The present study delineates a previously unrecognized mechanism that contributes to the F508del-CFTR intracellular retention and strongly suggests that interaction between K8 and F508del-CFTR plays an important role in the ER retention of mutant CFTR. Compelling evidence supports our model. First, we demonstrate the interaction of K8 with CFTR, using immunoprecipitation and proteomics. Second, the direct interaction of purified K8 and NBD1 is reconstituted in vitro using SPR. Third, the in vitro interaction is stronger in the presence of F508del mutation and observable in F508del- but not in WT-CFTR-expressing cells by PLA. Fourth, inhibition of K8–CFTR interaction restores the F508del-CFTR Cl⁻ channel function at the cell surface both in cultured cells and in the nasal respiratory epithelia of the F508del-CF mouse. Altogether, these four findings suggest that the network of IF plays a major role in the trafficking of CFTR between the different intracellular compartments.

Most of the effort in studying the functional link between cytoskeleton and CFTR has been focused so far on those proteins that participate in the actin network. Multiple immunoglobulin domains of filamin-A have been proven to interact with the N-terminus of CFTR (26,27), and suggested to play a role in the trafficking and membrane stability of the channel (28). Myosin VI, which plays a crucial role in CFTR endocytosis (29,30), establishes its interaction via the adaptor protein alpha-AP-2 (31). Likewise, a complex including the Na(+)/H(+) exchanger regulatory factor 1 (NHERF1) RhoA, Rho kinase, ezrin and actin would stabilize F508del-CFTR at the plasma membrane (15). Moreover, we have found ezrin expression increased in nasal epithelial cells (Jeanson et al., manuscript under evaluation), and therefore available for this role. It is thus tempting to hypothesize that a dynamic cross-talk between the two cytoskeletal networks—actin and IF—and their respective interactions with CFTR could play a key role in determining CFTR membrane localization and functionality.

When addressing the presence of CFTR at the membrane, one crucial aspect is degradation and quality control. This ubiquitin-dependent mechanism involves chaperones, co-chaperones and ubiquitin ligases (32). Since cytokeratin degradation is mediated by a ubiquitin-proteasome machinery (33), it would be interesting to establish whether CFTR–K8 interaction plays a role in CFTR targeting towards the proteasome degradation mechanism.

One main issue in CF pathophysiology is the response of cells to inflammation. The CFTR network shown in Supplementary Material, Figure S3 and Table S4 gives a condensed view of the complexity underlying the links between inflammation, protein trafficking and ion transport processes. Only physical interactions are represented, which limits the number of proteins in the network, compared with a pioneering study on CFTR interactome (34). Our analysis shows that under normal conditions, CFTR is a node linking the eicosanoid inflammatory pathway to trafficking and ion transport (e.g. SLC26A3, PDZ-proteins, SNPA23, etc., Supplementary Material, Fig. S3A) through direct interaction with S100A10. In the case of F508del-CFTR (Supplementary Material, Fig. S3B), when most of the CFTR is in the ER and other intracellular vesicles, and assuming that all the interactions presented in Supplementary Material, Figure 3SA still occur, it can be speculated that K8–F508del-CFTR interaction may determine a preponderance of ‘inflammatory connections’ over ‘trafficking connections’. Intriguingly, one of the reported direct interactions of K8/K18 is that with TNFR1. We have recently reported that TNFR1 is recruited in detergent-resistant microdomains (DRM) upon TNF-α stimulation in MDCK cells along with CFTR (35). This suggests that TNFR1 and CFTR would be present in the same macromolecular complex under certain conditions, along with other proteins that regulate the release of inflammatory mediators, such as ANXA1 and cPLA2 (36). We have recently shown a decreased expression of ANXA1 in cftr-knockout mice and in cells from CF patients (37), and an indirect interaction of this protein with NBD1 via the adaptor S100A10 (36). Interestingly, TNFR1 recruitment does not occur when CFTR lacks the TRL extremity, which contains the PDZ domain (35). Nevertheless, we have found K8 increased in DRM preparations from CFBE cells expressing F508del-CFTR when compared with DRM from cells expressing WT-CFTR (unpublished data). The latter raises the point of K8 playing a stabilizing role within the complex, and a potential role of this interaction in the regulation of the inflammatory signaling pathway initiated by TNF-α. Another link between K8 and inflammation in CF is brought about autophagy. In fact, upregulation of phosphorylated K8 is one of the markers of defective autophagy (38), and defective autophagy has been recently reported to drive lung inflammation in CF (39).

Supplementary Material, Figure S3C represents the protein network in the absence of CFTR. This network may reflect a situation at the plasma membrane for cells expressing F508del/F508del and/or stop mutations, and suggests the hypothesis that CFTR is a hub protein linking inflammation and ion transport pathways. In the absence of CFTR, the two networks are not connected (compare Supplementary Material, Fig. S3A, B and C). Moreover, other nodes (STX1-SNPA23, PRKAA, ANXA5-ACTG1, DNAJA1, DNAJB1) are completely isolated in the absence of CFTR. This analysis represents, to our knowledge, the first clear-cut hypothesis linking inflammation, ion transport and trafficking in CF.

The importance of the K8/K18 network in the regulation of CFTR is strengthened in the context of the cell response to oxidative stress, which constitutes one of the major defects associated with CF (39,40). It has been recently reported that hypoxia induces the degradation and redistribution of cytokeratin intermediary filaments, which may be prevented by the superoxide dismutase/catalase mimetic EUK-134 (41). Because of the significance of oxidative stress responses in CF, it is tempting to hypothesize that ROS production in CF cells may be linked to the reorganization of K8 network, that modulation of ROS production may be helpful in the disruption of the unwanted interaction between F508del-CFTR and K8, and, in fine, in the correction of the functional defect of F508del-CFTR.

Silencing of genes has been attempted as a therapeutic strategy to ameliorate glomerulopathies (42,43) and skin disorders (44–46), to inhibit hepatitis virus infection (47,48) and Ebola virus infection (49), and suggested as an alternative strategy to treat cancer (50,51) and sepsis (52). Silencing of K8-mRNA has been tested experimentally. In a nasopharyngeal carcinoma cell line, a role of K8 in the resistance to the proapoptotic effect of cisplatin has been revealed (53). Likewise, K8 knockdown has been shown to attenuate resistance to cadmium in rat lung epithelial cells (54). In the cancer cell line MCF-7, siRNA of K8 leads to decreased migration and cell adhesion, and a reversal of their multidrug-resistant phenotype (55–57). However, to our knowledge, the current study represents the first attempt to modify K8 expression outside the cancer field. We have shown in a previous report the potential therapeutic effect of siRNA targeting of K18 in cells expressing a mutated form of CFTR (18). In that study, also performed on transfected HeLa cells, siRNA treatment against K18 revealed a cAMP-dependent anion conductance, consistent with CFTR activity, in F508del-expressing cells. This seminal result prompted the suggestion that the cytokeratin network could represent a potential therapeutic target in CF, which is the original hypothesis of the present work. We have also demonstrated that K8-siRNA leads to the recovery of CFTR function when the protein is mutated, and also points at the K8/K18 network as a target. The fact that silencing of either of the two cytokeratins leads to functional correction of the CFTR defect prompts the question of the role of K8/K18 interaction dynamics as a crucial subject in future research. Further research will be necessary to define the nature and sites of interaction between K8 and CFTR, and its functional implications in those manifestations of CF not directly linked to CFTR channel dysfunction, such as inflammation and oxidative stress.

In summary, our results provide new insights into the nature of the interaction of CFTR with the cytokeratin network. More importantly, our study suggests a potential therapy in CF consisting of the use of an siRNA-based approach targeting the interaction between K8 and F508del-CFTR. This is inferred from data showing a restoration of iodide efflux due to CFTR activity in transfected HeLa cells expressing the mutated form, and most notably, a restoration of ▵V_TE changes due to the activation of Cl⁻ conductances in the nasal epithelium of mice expressing homozygously the F508del mutation. Nevertheless, as delivery of siRNA to the lungs represents a considerable challenge, further research will address pharmacological targeting of K8–CFTR interaction as well as the exploration of delivery vectors.

MATERIALS AND METHODS

Protein purification

Purification of human NBD1 was performed as follows. A 70 ml of inoculum of BL21-DE3 cells containing the SMT-3 fusion in the pET 28 expression system was grown overnight in LB medium at 37°C with kanamycin (50 μg/ml working concentration) present. Cells were induced with 0.179 g of IPTG, cooled to 15°C overnight, then harvested and pelleted at 4000 r.p.m. at 4°C for 30 min. Each pellet was resuspended in 15 ml of lysis buffer (50 mm Tris, 100 mm l-arginine, 50 mm NaCl, 5 mm MgCl₂ hexahydrate, 12.5% glycerol, 0.25 IGEPAL CA630, 2 mm 2-mercaptoethanol, 2 mm ATP, pH 7.6). Suspensions were combined into 50 ml conical vials and lysed by sonication after adding lysozyme and incubating on ice for 30 min. The lysate was centrifuged at 40 000g for 45 min to separate the soluble and insoluble fractions, and loaded into a pre-equilibrated 5 ml bed of Ni Sepharose 6 Fast Flow resin (GE Amersham). The column was equilibrated with five column volumes (CV) of loading buffer. During this step, the elution tubes were prepared with 2 mm ATP and 2 mm 2-mercapto ethanol. The sample was loaded and bound to the column, and washed with 5 CV of washing buffer (20 mm Tris, 500 mm NaCl, 60 mm imidazole, 12.5% glycerol, pH 7.6). The sample was eluted in 5 CV of elution buffer (20 mm Tris, 250 mm NaCl, 400 mm imidazole, 12.5% glycerol, pH 7.6). Samples were taken for SDS–PAGE analysis, and pooled together for concentration in a Beckman Coulter Allegra 6R centrifuge with a swinging bucket rotor, using the Amicon Ultra 15 30,000 MWCO centrifugal filters (Millipore). The protein was concentrated using 10 min spins at 4000 r.p.m. at 4°C. The SMT-3 fusion was cleaved off of NBD1 by using a 1:1000 dilution of ubiquitin-like protease on ice for 1 h. The protein was filtered using a Nalgene 0.22 μm syringe filter and injected onto a Hi Load 16/60 Superdex S200 prep grade gel filtration column (GE Amersham), and ran in S200 buffer (50 mm Tris, 150 mm NaCl, 5 mm MgCl₂ hexahydrate, 2 mm ATP, 2 mm 2-mercapto ethanol, 12.5% glycerol, pH 7.6). The void volume fractions were rejected and the protein was loaded back onto the Ni affinity column to remove the His-tagged SMT-3. The flow through was collected and concentrated in the same manner as before, except in a 10 000 MWCO Amicon Ultra 15 (Millipore). The protein was filtered again and injected onto the Superdex gel filtration column for buffer exchange. The flow through was collected and analyzed in a 10% SDS–polyacrylamide gel to check for purity The sample was prepared by making a dilution of protein in S200 buffer into phosphate buffer (20 mm Na₂PO₄, 150 mm NaCl, 12.5% glycerol, 1 mm DTT, pH 7.4) to give a 5 μm final concentration. Mouse NBD1 was purified according to Schmidt et al. (58).

Cells, cell culture and mouse models

Stably transfected HeLa cells (with pTracer alone as a control or coding for either WT-CFTR or F508del-CFTR) were grown as described before (59). Human nasal cells were obtained after informed consent by direct mucosa scraping following the protocols approved by the board of Hopital Necker. Fully differentiated human air–liquid-interface cultures (MucilAir™) were purchased from Epithelix SÀRL (Geneva, Switzerland) and cultured at 37°C and 5% CO₂ in a humidified atmosphere in a standard tissue culture incubator according to the manufacturer's recommendation. Briefly, epithelial cells were isolated from human nasal polyps, trachea or bronchus. Then, they were amplified and seeded at a high density onto microporous filters, and maintained at the air–liquid interface. The basolateral culture medium (MucilAir culture medium, Epithelix) was replaced every 2–3 days (60). Male and female homozygous for F508del-CFTR mutation (CF) on 129/FVB backgrounds and their WT control littermates were obtained from the Centre de Distribution, de Typage et d'Archivage Animal (Orléans, France) and housed at the Animal Care Facility at Necker Faculty of Medicine. After weaning, mice were fed a fiber-free diet, and Colopeg (17.14 g/l; Bayer Santé Familiale, France) was administered in water to prevent intestinal obstruction. Experiments were performed on 8–16-week-old mice, in conformity with the Public Health Service Policy on Humane Care and Use of Laboratory Animals, and in accordance with Necker Faculty of Medicine Animal Care and Use Committee (Comité Regional d'Etique pour l'Expérimentation Animale, P2.AE.099.09).

Keratin-8 silencing

A K8-siRNA pool was provided by Thermo Scientific (Villebon-sur-Yvette, France). The sequences of the regions targeted are: 5′-GAAGCAACATGGACAACAT-3′ (Seq. 1)/5′-TGGAAGGGCTGACCGACGA-3′ (Seq. 2)/5′-GCACAAAGACTGAGATCTC-3′ (Seq. 3)/5′-GCCCATGCCTCCAGCTACA-3′ (Seq. 4). HeLa cells at 70% confluence were transfected with the Lipofectamine 2000 Reagent (Invitrogen, Carlsbad, CA, USA). The final concentration of siRNA in the culture medium was 20 nm. Cells were subjected to other assays after 72 h transfection. Control cells were incubated in the same conditions with the transfection reagent alone. Control experiments were carried out with the non-specific siRNA control (a mixture of four scrambled sequences: 5′-UGGUUUACAUGUCGACUAA-3′, 5′-UGGUUUACAUGUUGUGUGA-3′, 5′-UGGUUUACAUGUUUUCUGA-3′ and 5′-UGGUUUACAUGUUUUCCUA-3′) (Thermo Scientific) under siRNA conditions or with Lipofectamine alone. Treatment of mice by siRNA was done by intranasal instillation of 30 µl of 20 nm K8-siRNA or scrambled-siRNA solution.

Immunoprecipitation

For good reproducibility, cells were plated at the same time at a density of 1 million per 60 cm². Cells were grown at 80% confluence and starved overnight with serum. Cells were washed twice with ice-cold PBS, scraped in PBS and counted in a Malassez cell. Cell pellets were resuspended in lysis buffer (20 mm HEPES, pH 7.5, 150 mm NaCl, 1 mm EDTA, 1% IgePal in 1× RIPA), solubilized for 40 min under continuous rotation. After that, the mixture was sonicated at 47 Hz for 5 min. To eliminate non-specific binding, a preclearing step was performed with protein A/G beads for 1 h. Then, the sample was centrifuged for 5 min at 2000g. The supernatant was quantified with the RCDC protein assay kit (Bio-Rad, Hercules, CA, USA) and 14 mg of protein for each HeLa cell extract was used. On these extracts, immunoprecipitation with anti-CFTR antibodies [clone 24-1 (R&D, Folsom, CA, USA) and clone mm13-4 (Abcam, Cambridge, MA, USA)] and protein A/G beads was carried out as described elsewhere (61,62). After 3 h of incubation under continuous rotation, the mixture was centrifuged. Immunoprecipitating proteins were eluted from the beads with 500 µl of 2D-electrophoresis buffer [7 m urea, 2 m thiourea, 4% w/v CHAPS, 50 mm DTT, 1% ampholytes, pH 3–10 (GE Healthcare, Orsay, France)].

Proteomics

Precipitates were analyzed by 2D-gel electrophoresis by a method adapted from Bensalem et al. (37). Prior to isoelectrofocusing (IEF), precipitates were filtered through 0.45 mm Vectaspin microfilters (Whatman, Maidstone, UK). Then, 450 µl of aliquots of protein precipitates were subjected to in-gel rehydration (50 V, 10 h), and IEF was performed on immobilized pH gradient strips (pH 3–10 non-linear, GE Healthcare). IEF was performed for a total of ∼50 kVh on an IPGphor system (GE Healthcare). Prior to SDS–PAGE on a 10% polyacrylamide gel, strips were incubated at room temperature in equilibration buffer (50 mm Tris–HCl, pH 8.8, 6 m urea, 2% SDS) with 2% DTT for 10 min and for another 10 min with 2.5% iodoacetamide. Gels were fixed, washed and proteins were visualized by the MS-compatible silver staining method as previously described (37,63,64). Spots were excised and proteins of interest digested in situ with trypsin (sequencing grade; Promega, Madison, WI, USA). Digested proteins were analyzed using an Autoflex MALDI-TOF mass spectrometer (Bruker Daltonics, Bremen, Germany). Proteins were identified using in-house Mascot server.

Immunocytochemistry

Cells were fixed with cold acetone, permeabilized with 0.25% Triton and immunostained with rabbit polyclonal antibodies against C-terminal CFTR (pCterB was a gift from Genzyme, Cambridge, MA, USA) at a 1:100 dilution, and against K8 (Progen, Germany) at a 1:10 dilution. The secondary antibodies were coupled to Alexa fluor 594 anti-mouse immunoglobulin G (IgG) (goat anti-mouse, Alexa; Molecular Probes, Eugene, OR, USA) and Alexa fluor 488 anti-rabbit IgG (goat anti-rabbit; Alexa) and used at 1:1000. The analysis was performed with a Leica TCS SP5 Confocal Microscopy System using 640 oil objectives.

Western blotting

Proteins were extracted as described previously (41). Protein extracts were mixed with 1 volume of 5× Laemmli buffer. Samples were processed as previously described (19). Briefly, they were heated at 37°C for 15 min, resolved by 8% SDS–PAGE, transferred onto PVDF membranes and blocked for 1 h with 5% non-fat milk diluted in TBS/Tween (0.1%). Blot membranes were hybridized using primary monoclonal anti-K8 (1:1000) (Progen), rabbit polyclonal anti-CFTR (1:500) (Alomone) and rabbit polyclonal anti-GAPDH (1:1000) (Santa Cruz) antibodies. Bands were visualized by incubation with IR dye-coupled secondary antibodies and analysis by Odyssey infrared imager (LI-COR Biosciences, Cambridge, UK).

Surface plasmon resonance

SPR was performed in a Biacore-2000 system (GE Healthcare). Covalent immobilization of NBD1 via primary amino groups to a CM5 sensor chip surface was performed at 7°C according to the modified protocol described by Borot et al. (36), using an immobilization buffer composed of 10 mm sodium acetate (pH 4.5), 3 mm Mg-ATP and 0.005% (w/v) surfactant P20. WT or mutated-NBD1 (80 μg/ml) was injected at a flow rate of 5 μl/min pendant for 4 min. A separate flow channel on the same sensor chip was subjected to blank immobilization without NBD1. Interactions were monitored by injecting different concentrations of K8 or K18 in running buffer [50 mm Tris–HCl at pH 7.4, 150 mm NaCl, 5 mm MgCl₂, 1 mm DTT and 0.005% (w/v) surfactant P20]. Binding was assessed at 7 and 20/25°C at 30 μl/min flow rate. Association phase was followed by increasing refractive index at the sensor surface. Subsequent dissociation phase was followed by injecting running buffer alone. Between injections, surfaces were regenerated by two washes with 5 µl of 0.05% SDS (20°C) and two washes with 5 µl of 10 mm HCL (7°C). Before measurements, the recombinant domain was dialyzed in the buffer, and centrifuged immediately before runs to minimize non-specific aggregation. Association and dissociation curves were corrected for non-specific binding by subtraction of control curves obtained by injection of different analyte concentrations through the blank channel. Kinetic constants, k_on, k_off and , were calculated using the Biacore BIAEVALUATION 3.1 software (Biacore AB), assuming a simple two-component model of interaction.

Proximity ligation assay

Cells were fixed with cold acetone and analyzed using the Duolink™ kit (Eurogentec, Angers, France) according to manufacturer's instructions. Briefly, slides were pre-washed in PBS, and incubated with blocking solution. Samples were then incubated with primary antibodies against C-terminal CFTR (pCterB, Genzyme), and K8 (Progen), at 1:100 and 1:10, respectively. Secondary antibodies conjugated with oligonucleotides (PLA probe MINUS and PLA probe PLUS) were added to the reaction tube and incubated. The oligonucleotides contained in the hybridization solution hybridized to the two PLA probes if they were at <40 nm. A ligase (Ligation Solution), nucleotides and polymerase were added sequentially, allowing formation of a rolling-circle amplification product detected by labeled oligonucleotides in case of proximity. The signal was visible as a distinct fluorescent dot and analyzed by fluorescence microscopy (excitation at 557 nm and emission at 563 nm) using 640 oil objectives (Leica TCS SP5 Confocal Microscopy System).

Iodide efflux test

To assess CFTR Cl⁻ channel activity, iodide efflux was measured in transfected HeLa cells treated either with 10 μm forskolin plus 50 μm genistein or 20 μm CFTR Inh172, according to the protocol described by Hughes et al. (24). Briefly, HeLa cells were cultured in 60 mm dishes at 80–90% confluence. Cells were washed five times with 4 ml of loading buffer (136 mm NaI, 3 mm KNO₃, 2 mm Ca(NO₃)₂, 20 mm HEPES, 11 mm glucose, pH 7.4) and incubated with this buffer for 1 h at room temperature. Cells were gently washed 15 times by adding 4 ml of efflux buffer (136 mm NaNO₃, 3 mm KNO₃, 2 mm Ca(NO₃)₂, 20 mm HEPES, 11 mm glucose, pH 7.4). Then, cells were incubated with 4 ml of fresh efflux buffer (drug-free or not) and iodide efflux was recorded. An iodide-selective electrode (ISE251, Radiometer Analytical SAS, France) connected to a pH meter (PHM250, Ion Analyzer, Radiometer Analytical SAS) was used to measure the amount of iodide released by cells at 1min intervals.

Nasal potential difference

The method was adapted and miniaturized from the technique developed for young children (65,66). Mice were anesthetized by intraperitoneal injection of ketamine (133 mg/kg; IMALGENE 1000, MERIAL, France) and xylazine (13.3 mg/kg; Rompun 2%, BayerPharma, France). Mice were positioned on a 45° tilt board, and a paper pad was placed under the nose to avoid mice suffocation. A subcutaneous needle was connected to an Ag⁺/AgCl reference electrode by an agar bridge. A double-lumen polyethylene catheter (0.5 mm diameter) was inserted into one nostril to a depth of 4 mm. One lumen perfused by Ringer solution (140 mm NaCl, 6 mm KCl, 10 mm HEPES, 10 mm glucose, 1 mm MgCl₂, 2 mm CaCl₂, pH 7.4) at 0.15 ml/h is connected to a measuring Ag⁺/AgCl electrode. The two Ag⁺/AgCl electrodes were connected to a high-impedance voltmeter (LOGAN Research Ltd, UK). The second lumen perfused solution with the following sequence: (i) Ringer solution, (ii) Ringer solution containing amiloride (100 µm), (iii) low-chloride Ringer solution (140 mm Na-gluconate, 6 mm K-gluconate, 10 mm HEPES, 10 mm glucose, 1 mm MgCl₂, 6 mm Ca-gluconate, pH 7.4), (iv) low-chloride Ringer solution containing CFTR inhibitor I_inh172 (5 µm, Calbiochem, Germany). Each solution was perfused for at least 3 min, and 30 s stability was required before perfusion switch. Steady-state transepithelial potential, V_TE, ▵V_TEAmil (difference between V_TE and transepithelial potential recorded after perfusion of amiloride-containing solution), ▵V_TElowCl (difference between V_TE and transepithelial potential recorded after perfusion with low Cl⁻ plus amiloride-containing solution) and ▵V_{TElowClInh-172} (difference between V_TE and transepithelial potential recorded after addition of CFTR inhibitor to the previous solution) were the means of 30 values recorded during stability. Data were not corrected for junction potential, which was negligible (−0.5 mV).

Statistical analysis

In NPD experiments, a cut-off value was established for −2.15 mV, after measuring NPD in 21 WT and 47 F508del mice, and evaluating ROC curves. This corresponds to a threshold V_TE value beyond which mice present a WT-like Cl⁻ conductance. The increase in ▵V_TElowCl was considered significant or not significant according to this threshold. NPD results are expressed as median ± interquartile range. SPR constant data are expressed as means ± SE.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG online.

Conflict of Interest statement. The authors declare no competing interests.

FUNDING

This work was supported by European Commission grant ‘NEUPROCF’ (FP6) (to A.E.); by French Agence Nationale de la Recherche grant ‘EICOCF’ (to A.E., M.O.); by the associations Vaincre la Mucoviscidose (to A.E., M.O.) and Mucoviscidose: ABCF2 (to A.E., I.S.-G.); Legs Poix-University of Paris 5 (to A.E., M.O.); the Canadian Institutes of Health Research; and NIH-NIDDK and Canadian Cystic Fibrosis Foundation (to G.L.L.).