The human CFTR protein expressed in CHO cells activates aquaporin-3 in a cAMP-dependent pathway: study by digital holographic microscopy

The cystic fibrosis transmembrane conductance regulator (CFTR; also known as ATP-binding cassette subfamily C, member 7; ABCC7) protein is an apical membrane protein functioning as a chloride channel regulating anion transport in secretory epithelial cells. A dysfunction of this protein in cystic fibrosis affects the transport of chloride, bicarbonate, sodium and water in epithelial tissue, leading to thick and viscous secretions. Although the features of chloride transport by CFTR protein are well defined (review by Hanrahan et al., 1995), studies directly addressing water transport in epithelial physiology and in human pathologies (such as cystic fibrosis) are still at an early stage. One of the reasons for this area still being under-explored is the lack of reliable measurement techniques to directly quantify the transmembrane water flux, at a cellular scale, during a physiological (and pathological) situation. Among the transmembrane conduction pathways, chloride channels are generally poorly permeable to water molecules (Bormann et al., 1987) but it was proposed that in epithelia the movement of water could be directly linked to CFTR activity through a cAMP-stimulated aqueous pore (Hasegawa et al., 1992) or be dependent on water channels (Schreiber et al., 1997), in particular the aquaporin AQP3 (Schreiber et al., 1999; Schreiber et al., 2000).

Imaging techniques using the measurement of specific fluorophore concentration such as confocal microscopy (Crowe et al., 1995) or total internal reflection (TIR) microfluorimetry (Farinas et al., 1995) allow estimation of the variation of intracellular water content by sampling of fluorescence. Other imaging techniques defined as interferometry are also used to measure some biophysical parameters in relation to membrane water permeability (Farinas and Verkman, 1996; Farinas et al., 1997). However, all of these imaging techniques (fluorescence or interferometric) only detect large water movements in response to osmotic shocks of several hundred milliosmol, unrepresentative of fine water movements involved in many biological processes forming part of the normal cellular activity, such as transport of osmotically active substances, cell metabolism and cell volume regulation.

Recently, we have developed a new imaging technique called digital holographic microscopy (DHM), which can be used to non-invasively visualise cell structure and dynamics and/or study various biological processes without using dye or contrast agent [for reviews of the subject (Depeursinge et al., 2007; Marquet et al., 2013), and for some processes in relation to water transport (Rappaz et al., 2005; Jourdain et al., 2011; Jourdain et al., 2012; Boss et al., 2013)].

Here, we used the DHM quantitative phase signal to directly study the activity of the CFTR protein and, for the first time to analyse its impact on transmembrane water fluxes. Our data show that cAMP-dependent activation of the CFTR transport function is required to trigger water flux through the cAMP-stimulated aquaporin, AQP3, a mechanism of regulation that is inhibited in the presence of the selective CFTR blocker CFTR_inh172 and that is absent in CFTR-deficient cells or in cells expressing the ER-sequestered mutant form, F508del, of CFTR. Restoring the function of F508del-CFTR is also associated with a recovery of water transport. These observations are of particular relevance for our understanding of the molecular mechanisms controlling the passage of water and ions in healthy and dehydrated airway surfaces and for the screening of compounds potentially able to restore water transport in cells affected by the human disease cystic fibrosis.

DHM quantitative phase signal (QPS), resulting from its high sensitivity to the intracellular refractive index, directly measures the net transmembrane water fluxes accompanying the ionic movement (Rappaz et al., 2005; Jourdain et al., 2012; Boss et al., 2013). In CHO cells stably expressing CFTR protein (CHO_cftr) that were treated with the adenylate cyclase activator forskolin (1 µM; 5 minutes), which stimulates CFTR channel activity through an increase of cellular cAMP concentrations, there was a transient increase of the phase shift reflected by a peak amplitude of 5.48±0.5° (n_cell = 644; Fig. 1A; Table 1), whereas superfusion of control solution (0.1–1% DMSO) had no effect on the signal phase (Fig. 1B). The effect of forskolin on the QPS was concentration dependent between 500 pM and 10 µM with an EC₅₀ of 0.85±0.08 µM (n_cell = 126; Fig. 1C). In parallel, we performed iodide efflux experiments to monitor anionic transport activity of the CFTR protein (Norez et al., 2004). Our results also showed that increasing concentrations of forskolin stimulates the iodide efflux response in CHO_cftr cells with an EC₅₀ of 1 µM (data not shown), a value similar to that obtained by DHM.

Such an increased QPS corresponds to an exit of water accompanying the exit of an ion (Jourdain et al., 2012). To determine the nature of the involved ion, we recorded the membrane current in parallel. Application of forskolin (1 µM, 5 minutes) activated an inward current in CHO_cftr cells (I = −0.22±0.04 nA; n_cell = 25; Fig. 1D). Furthermore, analysis of the curve I/V relationship for the current activated by forskolin indicated a reversal potential of −27.4±2.3 mV (Fig. 1D), a value similar to the theoretical reversal potential of Cl⁻ (around −30 mV) calculated from the Nernst equation. Thus, this first set of experiments identified an optical phase response resulting from an activation of a cAMP-dependent Cl⁻ conductance mediated by forskolin.

To further characterize the Cl⁻ conductance activated by forskolin, we applied various Cl⁻ channel inhibitors. First we tested 4,4′-diisothiocyanato-stilbene-2,2′-disulfonic acid (DIDS; 100 µM), a broad-spectrum anion transport blocker, which inhibits chloride conductances, except for those mediated by CFTR. Our data show that there was no significant difference between the optical response triggered by forskolin alone and that in the presence of DIDS (+28%; P>0.05; n_cell = 112; Fig. 1E). In contrast, in the presence of 10 µM CFTR_inh172, a selective CFTR blocker (Ma et al., 2002), the optical response evoked by forskolin was dramatically reduced (−80%; P<0.05; n_cell = 252; Fig. 1E; Table 1). Application of 8Br-cAMP (1 mM), a membrane permeant and non-degradable analogue of cAMP, to CHO_cftr cells triggered a strong increase in the QPS, which reached a plateau value of ∼18.55±3.29° (n_cell = 252; Fig. 2A; Table 1). In parallel, iodide efflux experiments showed that increasing concentrations of 8Br-cAMP (100 µM–3 mM) stimulated the efflux response in CHO_cftr cells (Fig. 2B) clearly confirming the cAMP dependence of the QPS observed in CHO_cftr cells. The blockade of QPS (Fig. 2A; Table 1) or iodide efflux (not shown) by CFTR_inh172 was also found when we stimulated CHO_cftr cells with 8Br-cAMP. Finally, in mock-transfected CHO cells, which contained the empty pNUT vector (CHO_pNUT), we observed a residual QPS when we applied forskolin (1.85±0.13°; n_cell = 448; Fig. 1B). Its amplitude was significantly weaker than in CHO_cftr cells (P>0.005; Fig. 1B) and not modified by CFTR_inh172 (Table 1), suggesting a non-CFTR origin for this residual optical response. These findings were also confirmed by experiments with 8Br-cAMP on CHO_pNUT cells (Fig. 2C; Table 1). In conclusion, the optical response observed in CHO_cftr cells after application of forskolin or 8Br-cAMP is mainly due to activation of CFTR protein.

To determine whether the optical signals recorded by DHM and associated with the activation of CFTR in CHO_cftr cells could involve a water channel, we applied mercuric chloride (HgCl₂) as a broad-spectrum blocker of aquaporins (Savage and Stroud, 2007). We added 5 µM HgCl₂ to the extracellular medium and measured the optical response after the application of forskolin (1 µM, 5 minutes). The response was significantly lower than that of CHO_cftr cells (1.86±0.77°; n_cell = 126; P<0.005; Fig. 3A) stressing the involvement of a water channel in the genesis of the forskolin-associated optical signal.

However, it has been previously shown that mercury affects mitochondrial metabolism inducing an ATP depletion (Chen et al., 2010), finally leading to the reduction in cAMP production by adenylate cyclase. In these conditions, we can hypothesize that the decrease in the optical signal (reflecting a decrease of CFTR activity) in the presence of mercury is not directly linked to a blockade of aquaporin but associated with a decrease in cAMP production. The above experiments involving 8Br-cAMP treatment suggested that the effect of HgCl₂ on CFTR activity is independent of the cAMP production. Indeed when we treated CHO_cftr cells with HgCl₂ (5 µM), the optical response recorded during 8Br-cAMP application (1 mM) was also strongly and significantly decreased when compared to the untreated CHO_cftr cells (5.0±1.6°; n_cell = 126; P<0.005; Fig. 3B; Table 1) confirming the involvement of water channels in the genesis of the optical signal and of the interrelationship between CFTR and water channels.

Finally, to be sure that the cAMP-dependent Cl⁻ conductance was not modified by HgCl₂, we performed a first set of experiments using the patch clamp technique. Our results indicated that current density in CHO_cftr cells at −60 mV was around (35±7 pA/pF; n = 6) when we applied forskolin (10 µM) to the medium, a value significantly higher than in the basal condition (7±2 pA/pF; P<0.005; n = 6; Fig. 4). In the presence of HgCl₂ (5 µM), this current density was slightly decreased but significantly persistent in comparison with the control condition (23±3 pA/pF; P<0.05; n = 6) and further blocked by CFTR_inh172 (10 µM, n = 6; Fig. 4), suggesting that HgCl₂ decreased the water transport and not the Cl⁻ flux. This result is in agreement with the study of Weber et al. in which it was shown that the human CFTR protein was weakly sensitive to mercury (Weber et al., 2006).

We then focussed on the involvement of cAMP in the interaction between CFTR protein and the water channel. To understand whether the activation of CFTR by cAMP is crucial for water channel activation, we used genistein, a direct activator of CFTR protein with a cAMP-independent mechanism of action (Illek et al., 1995). Treatment of CHO_cftr cells with genistein (30 µM; 5 minutes) triggered an inward current (−0.13±0.04 nA; n_cell = 13; Fig. 5A), indicating activation of CFTR channels. However, genistein was unable to modify the QPS of CHO_cftr cells (0.2±0.5°; n_cell = 204; Fig. 5A), highlighting the important role of cAMP in the activation of the water channel. To confirm the key role of cAMP in the interaction between CFTR protein and water channels, we also used MPB-91, a benzo[c]quinolizinium compound activating the chloride channel activity of CFTR through a cAMP-independent mechanism (Becq et al., 1999; Marivingt-Mounir et al., 2004). Similarly to genistein, a transient application of MPB-91 (50–100 µM; 5 minutes) was unable to induce a phase response in CHO_cftr cells (0.08±0.4°; n_cell = 126) compared with that to forskolin (1 µM; 5 minutes; Fig. 5B). Therefore, activation of water channels cannot be solely triggered by opening the CFTR pores but requires cAMP-dependent activation of the CFTR channel.

We next determined whether the activated state of CFTR protein is required to maintain the activation of the water channel. The fact that a transient application of forskolin (1 µM; 5 minutes) triggered a transient increase in the QPS and not a sustained plateau response, as shown in Fig. 1, is in favour of a dependence on the CFTR activity. Nevertheless, to clarify this, we applied CFTR_inh172 to CHO_cftr cells pre-stimulated with 8Br-cAMP (1 mM). We reasoned that if such delayed application of CFTR_inh172 has no effect on the optical response amplitude, the activity of water channels would be independent of CFTR. However, our data show that activity of water channels requires a constant activation of CFTR protein because delayed application of CFTR_inh172 (10 µM; delay 3 or 5 minutes) also stopped the optical signal normally obtained by an application of 8Br-cAMP (1 mM; Fig. 5C).

The above data with genistein and MBP-91 indicated that Cl⁻ and water molecules are transported by two distinct channels. Thus, we could have two distinct situations: either a functional coupling between the CFTR protein and an endogenous aquaporin, or a cAMP-stimulated aqueous pore within the CFTR protein. Quantitative information concerning the expression of endogenous proteins by CHO cells is scarce (Wlaschin and Hu, 2007). However, we have focussed our study on two aquaporins, AQP3 and AQP9, because it has been shown they are both cAMP-dependent ‘aquaglycoporins’, indirectly and positively activated by the CFTR protein (Schreiber et al., 1999; Pietrement et al., 2008).

We quantified by qPCR analysis the mRNA copy numbers for the endogenous aquaporins (AQP3 and AQP9) simultaneously with the human form of the CFTR gene in CHO_cftr and CHO_pNUT cells. As expected, only CHO_cftr cells expressed the mRNA for the human form of CFTR (1.56±0.04×10⁸ copies; Fig. 6A). For CHO_pNUT cells, we did not detect CFTR mRNA (Fig. 6A), which is correlated to the fact that CHO_pNUT cells contained only the empty pNUT vector. Concerning the water channels, CHO_cftr and CHO_pNUT cells were found to express more than 6×10⁵ copies per ng of mRNA for AQP3 (respectively 6.95±0.9×10⁵ versus 6.20±0.19×10⁵; P>0.05; Fig. 6A), which is not significantly different in terms of mRNA expression (Fig. 6A). The mRNA copy number for AQP9 was below the level required for selection (<5000 copies per ng of mRNA; Fig. 6A), indicating than these cells are not expressing AQP9 significantly. For this reason, the rest of our study was focussed only on AQP3 expression.

To determine a putative functional relationship with the CFTR protein, we measured and compared the amplitude of the optical response triggered by a transient application of forskolin (1 µM; 5 minutes) from four categories of CHO_cftr cells cultures: (1) untreated; (2) transfected with a specific siRNA against AQP3 (active siRNA); (3) transfected with scramble siRNA (inactive siRNA); (4) treated with Lipofectamine (vehicle control). In untreated CHO_cftr cells, application of forskolin (1 µM; 5 minutes) triggered a strong optical response (7.08±0.67°; n_cell = 128). With a similar application of forskolin, CHO_cftr cells transfected with siRNA against AQP3 displayed an optical response (4.68±0.47°; n_cell = 174) significantly lower than for normal CHO_cftr cells (P<0.006; Fig. 6B). In parallel, the quantification of AQP3 mRNA from the same culture indicated a significant decrease of mRNA copy number per ng of mRNA for AQP3 between untreated CHO_cftr cells (6.95±0.09×10⁵ copies; n_cult = 15) and CHO_cftr cells treated with siRNA against AQP3 (1.51±0.7×10⁵ copies; n_cult = 11; P<0.005).

This reduction in AQP3 copy number expression (1.51±0.7×10⁵ copies; n_cult = 11) needs also to be compared with cells treated using same conditions, namely, with scramble siRNA (3.92±1.96×10⁵ copies; n_cult = 8; P<0.001) or with Lipofectamine alone (4.57±0.98×10⁵ copies; n_cult = 7; P<0.001). In addition, CHO_cftr cells treated with scramble siRNA or with Lipofectamine alone were not significantly different from untreated CHO_cftr cells for both the optical response (Fig. 6B1) (scramble mRNA: 6.18±0.25°; n_cell = 130; P>0.05. Lipofect: 6.22±0.50°; n_cell = 141; P>0.05) (Fig. 6B) and the quantity of mRNA for AQP3 (scramble siRNA: 3.92±1.96×10⁵; n_cult = 8; P>0.05. Lipofect: 4.57±0.98×10⁵; n_cult = 7; P>0.05; Fig. 6B2).

All of these results identify that a part of movement of water associated with activation of CFTR protein is linked to the presence of AQP3.

Among the numerous mutated forms of the CFTR protein, the most common one is the deletion of phenylalanine at position 508 (F508del). According to the literature, this deletion results in a processing defect leading to ER sequestration of the F508del-CFTR protein at the origin of the observed epithelial Cl^– impermeability measured in CF cells (Cheng et al., 1990). Application of forskolin (1 µM; 5 minutes) to CHO cells expressing this mutated form of CFTR protein (CHO_F508del), triggered a QPS with an amplitude (2.31±0.54°; P<0.005; n_cell = 784; Fig. 1B; Table 1) significantly lower than that measured in CHO_cftr cells using the same experimental protocol. This difference was even more pronounced with 8Br-cAMP (Fig. 2C). The peak amplitude of the optical signal was 5.62±0.38° (n_cell = 154; Fig. 2C; Table 1), a 70% decrease compared with optical signals measured for CHO_cftr cells (P<0.005; Fig. 2C). In the presence of CFTR_inh172 in the bath medium, the amplitude of the optical response in CHO_F508del cells was not significantly altered following either forskolin (1.81±0.72; n_cell = 126; P = 0.42; Table 1) or 8Br-cAMP (4.92±0.44; n_cell = 140; P = 0.70; Table 1) application indicating that the residual optical responses recorded in CHO_F508del cells are not associated with CFTR activity. Furthermore, the fact that there were no significant differences between the optical responses measured in CHO_F508del and CHO_pNUT cells after application of forskolin (Fig. 1B) or 8Br-cAMP (Fig. 2C) confirmed that the optical responses developed by CHO_F508del cells are not due to CFTR. Finally, HgCl₂ (5 µM) did not modify the residual phase shift recorded from CHO_F508del cells (Fig. 3; Table 1) indicating that water channels are also not responsible for the genesis of this residual optical response.

The results above suggested that water movement is abnormal in CHO_F508del cells. Because it is known that several transport properties of CF cells are also affected, including calcium (Antigny et al., 2009), sodium (Sheridan et al., 2005) and bicarbonate (Choi et al., 2001) transport, we further studied the effect on water movement of correcting the abnormal trafficking of F508del-CFTR. We used the corrector miglustat, known to restore a functional cAMP-dependent F508del-CFTR activity (Norez et al., 2006; Norez et al., 2009). After treatment of CHO_F508del cells with miglustat (100 µM; 4–6 hours), application of forskolin (1 µM, 5 minutes) triggered an increase of phase shift (4.5±0.58°; n_cell = 462) significantly higher than for untreated CHO_F508del cells (P<0.05; Fig. 1B). Comparable results were obtained when we substituted 8Br-cAMP (1 mM) for forskolin, with a peak of maximum amplitude reaching 14.15±3.62° (n_cell = 154), a value again significantly higher than that obtained for untreated CHO_F508del cells (P<0.005; Fig. 2C; Table 1). This rescue of optical response is dependent on CFTR activity, because the optical responses stimulated by either forskolin or 8Br-cAMP were both fully blocked by the selective CFTR inhibitor CFTR_inh172 (Table 1). Finally, after perfusing miglustat-treated CHO_F508del cells with an extracellular medium containing HgCl₂ (5 µM), the optical responses normally obtained after stimulation with forskolin or 8Br-cAMP were also strongly decreased (Fig. 3; Table 1).

Results reported here indicate that the QPS from CHO_cftr cells stimulated with the cAMP-agonists forskolin and 8Br-cAMP are linked to CFTR activity, highlighting a functional coupling between the CFTR protein and the water channel. Evidence for this functional coupling is supported by results obtained in CHO_F508del cells, because the absence of operative CFTR protein at the plasma membrane (due to the deletion F508del) did not trigger optical responses sensitive to mercury whereas the reinsertion of CFTR proteins into the plasma membrane after treatment with miglustat (Norez et al., 2006; Norez et al., 2009) restored this mercury sensitivity. Otherwise, the coupling mechanism between CFTR protein and a water channel is cAMP-dependent but not functional when CFTR is activated by cAMP-independent agonists (genistein, MPB-91) or inhibited by the selective CFTR blocker CFTR_inh172. Importantly, the rescue of F508del-CFTR activity by the pharmacological corrector miglustat leads to the re-establishment of the CFTR_inh172-sensitive cAMP-dependent functional coupling between CFTR and the water channel. In conclusion, we propose that a functional CFTR protein activated through a cAMP-dependent pathway is required for water molecules to pass out of the cell.

Another central question of this paper is whether this water channel is an integral part of the CFTR protein as suggested by the study of Hasegawa et al. (Hasegawa et al., 1992) or a separate entity of the CFTR protein, similar to an endogenous aquaporin (Schreiber et al., 1999; Schreiber et al., 2000) expressed by CHO cells, but that would be functionally related to the human form of the CFTR protein. Indeed, the hypothesis of a cAMP-dependent pore for water as an integral part of the CFTR protein may explain the water efflux during activation of CFTR protein. This is the main conclusion of the work by Hasegawa et al. performed on Xenopus oocytes that expressed the human form of the CFTR protein (Hasegawa et al., 1992). However, another study, also performed on Xenopus oocytes expressing the human form of the CFTR protein, has shown that both Cl^– conductance and water channels were inhibited through different pathways (Schreiber et al., 1997). The authors, in contrast to the conclusion of Hasegawa et al. concluded that the two fluxes, water and Cl^–, are associated with two different proteins, respectively, an endogenous aquaporin and the CFTR protein (Hasegawa et al., 1992). Even though several reports have shown that wild-type CHO cells express few or no endogenous aquaporins (Farinas et al., 1997; Schreiber et al., 1999), our results are clearly in favour of the second hypothesis because the experiment using siRNA against the endogenous AQP3 led to a significant decrease in both the forskolin-induced phase signal and the quantity of AQP3 mRNA. At the same time, the scrambled siRNA (or Lipofectamine treatment) was inefficient at decreasing both the optical signal induced by forskolin and the quantity of AQP3 mRNA. Thus, we can conclude that the Cl⁻ and water fluxes associated with CFTR activity involve two distinct proteins, the CFTR protein (for Cl⁻ transport) and AQP3 (for water transport). However, the nature of this interaction remains to be determined. Although a direct physical interaction between CFTR protein and the epithelial sodium channel has been shown (Berdiev et al., 2007), such interaction between CFTR protein and AQP3 has, however, not yet been demonstrated.

An important point also concerns the pharmacological corrector miglustat and its mechanism of action. We have seen that the mRNA copy number for both CFTR and AQP3 in CHO_F508del cells was significantly lower compared with the levels in CHO_cftr cells (not shown) indicating that the machinery for Cl⁻ and water transport was less abundant in CHO_F508del than in CHO_cftr cells. This is consistent with the fact that forskolin and 8Br-cAMP triggered a phase response significantly lower than that of CHO_cftr cells. Although the treatment of CHO_F508del cells with miglustat (100 µM; 4–6 hours) did not modify the mRNA copy number for both CFTR and AQP3 (not shown), forskolin (and 8Br-cAMP) did significantly restore an optical response, both indicating that the action of miglustat did not involve synthesis of new proteins but rather mobilized proteins most probably residing in the endoplasmic reticulum.

In conclusion, taking advantage of the DHM technique to monitor net flux of water during CFTR activity coupled to ion flux and electrophysiological experiments, we explored the hypothesis that CFTR and water transport could be coupled. Here, we provided evidence for the existence of a direct functional coupling between the epithelial chloride channel CFTR, an endogenous HgCl-sensitive water channel and/or the aquaporin AQP3. These results open the field for new types of pharmacological investigation (notably the screening of drugs acting on water transporter molecules) and therapeutic approach such as rehydration of airway surface of F508del-CFTR-expressing cells.

This study was carried out in compliance with the Public Health Service (PHS) Policy on Human Care and Use of Laboratory Animals (Animal Welfare Assurance no. A5692-01). Experimental procedures were approved by the Cantonal Veterinary Authorities (Vaud, Switzerland).

CHO cells stably transfected containing wild-type CFTR (CHO_cftr), cultured at 37°C in 5% CO₂ were maintained in MEM containing 7% foetal bovine serum, 0.5% antibiotics (50 IU/ml penicillin and 50 µg/ml streptomycin) and 150 µM, 100 µM or 20 µM methotrexate respectively for CHO_F508del, CHO_cftr and CHO_pNUT cells. For detailed procedures see elsewhere (Bulteau et al., 2000; Dérand et al., 2001).

For optical and electrical recordings, all cell cultures were perfused in a fluid containing (in mM): NaCl 150, KCl 3, D-glucose 5, HEPES 10, CaCl₂ 3, and MgCl₂ 2 (pH 7.4; room temperature). All drugs were dissolved and applied by bath perfusion (for 5 minutes) after a minimum of 2 minutes of stable baseline recording of the optical signal.

The optical and electrophysiological recordings were analysed using MATLAB 7.6 (Mathworks Software, Natick, MA) and all curves were fitted using ORIGIN 7.5 (Microcal Software, Northampton, MA). Forskolin concentration–response profiles were fitted to the following logistic equation: ϕ/ϕ _max = 1/[1+(EC₅₀/[forskolin])ⁿ], where ϕ and ϕ_max are the normalized forskolin-induced phase shift at a given concentration and the maximum phase shift induced by a saturating forskolin concentration, EC₅₀ is the half-maximal effective forskolin concentration (or time application) and n is the slope factor. All data are presented as means ± s.e.m. Student's t-test (paired or unpaired) was used to determine statistical significance (P<0.05).

Whole-cell recordings were made, and signals were amplified using Multiclamp 700B amplifiers (Axon Instruments, Union City, CA) and digitized by means of an ITC-1600 interface (Instrutech, Great Neck, NY) to a PC computer running Igor Pro (Wavemetrics, Portland, OR). All currents (sampling interval, 5 kHz) were low-pass filtered (2 kHz). They were recorded with pipettes containing 95 mM potassium gluconate, 40 mM KCl, 10 mM Hepes and 2 mM MgCl₂ (pH 7.3). The pipettes were pulled with a DMZ universal puller.

To measure the effect of mercury on CFTR channel activity, ionic currents were recorded using the patch-ruptured whole-cell variant of the patch-clamp technique and measured with an Axopatch 200B amplifier (Molecular Devices, Union City, CA). Currents were filtered at 5 kHz (−3 dB; 4-pole Bessel filter). The holding potential was −40 mV in all whole-cell experiments. Current/voltage (I/V) relationships were built by clamping the membrane potential to −40 mV and by pulses from −100 to +100 mV in 20 mV increments. Pipettes were prepared by pulling borosilicate glass capillary tubes (GC150-T10; Harvard Apparatus, Edenbridge, UK) using a two-step vertical puller (Narishige Tokyo, Japan). They were filled with the following solution (in mM): 120 N-methyl-D-glucamine, 86 L-aspartic acid, 3 MgCl₂, 1 CsEGTA, 5 TES, and 3 MgATP ex temporane (titrated with NaOH to pH 7.2, the osmolarity was 285±5 mOsmol). The pipette solution was always hypotonic (with respect to the bath solution) to prevent cell swelling and activation of the volume-sensitive chloride channels. The stilbene derivative 4,4′-diisothiocyanatostilbene-2,2′-disulfonic acid disodium salt hydrate (DIDS) was also used to block non-CFTR Cl⁻ conductance. Pipettes were connected to the head of the patch-clamp amplifier through an Ag–AgCl pellet (pipette resistance of 10–20 MΩ). The external bath solution contained (in mM): 140 NaCl, 1.2 CaCl₂, 1 MgSO₄, 10 dextrose, and 10 TES (titrated with NaOH to pH 7.4, the osmolarity was 315±5 mOsmol). Pipette capacitance was electronically compensated in cell-attached mode. Membrane capacitance and series resistances were measured in the whole-cell mode by fitting capacitance currents obtained in response to a hyperpolarisation of 40 mV, with a first-order exponential and by integrating the surface of the capacitance current. Voltage clamp signals were recorded using a microcomputer equipped with an analogue/digital–digital/analogue conversion board (Digidata 1440A interface, Molecular Devices). All experiments were performed at room temperature (20–25°C) and drugs were applied by using a gravity-fed perfusion system. All Cl⁻ currents were analysed with the pCLAMP version 10.2 package software (Molecular Devices).

The CFTR Cl⁻ channel activity was assayed by measuring the rate of iodide (¹²⁵I) efflux from CHO cells as described previously (Norez et al., 2004). The CFTR activator benzo[c]quinolizinium compound MPB-91 (5-butyl-10-chloro-6-hydroxybenzo[c]quinolizinium chloride) was synthesised in our laboratory as previously described (Marivingt-Mounir et al., 2004). Forskolin was purchased from LC Laboratories (PKC Pharmaceuticals, Woburn, MA). The F508del-CFTR corrector, miglustat (or NB-DNJ) was purchased from Toronto Chemical Research and dissolved in water (Norez et al., 2006). CFTR_inh172 (3-[(3-trifluoromethyl)-phenyl]-5-[(4-carboxyphenyl)methylene]-2-thioxo-4-thiazolidinone) a CFTR inhibitor (Ma et al., 2002) was from VWR International (Fontenay-sous-bois, France). All other products were from Sigma (St. Louis, MO). Stock solutions of MPB-91 and forskolin (100 mM), CFTR_inh172 (10 mM) were prepared in DMSO. Miglustat was prepared freshly in water. Results are expressed as means ± s.e.m. of n observations. Sets of data were compared with a Student's t-test. Differences were considered statistically significant when P<0.05; ns, non significant difference, *P<0.05, **P<0.01, ***P<0.001. All statistical tests were performed using GraphPad Prism version 4.0 for Windows (Graphpad Software) and Origin version 5.0.

RNA was extracted from 90–95% confluent CHO cell cultures grown in 35 mm dishes. We used the Maxwell 16 LEV simplyRNA kit (Promega) to purify the mRNA following the provided protocol. We modified the beginning of the procedure by adding the 400 µl of lysis solution (1∶1 homogenisation solution and thioglycerol + lysis solution) directly to each culture. Each dish was agitated slowly to lyse the cells. Homogenates were then transferred to cartridge as described. With this procedure we obtained between 2 and 8 µg of mRNA/dish of very pure RNA (OD 260/280 above 2.1 and OD 260/230 above 2.0).

500 ng of mRNA was used in each reverse transcription reaction. cDNA was made using the High Capacity RNA-to-cDNA kit (Life Technologies) following manufacturer's instructions.

Quantitative polymerase chain reactions (QPCR) were performed using an ABI 7900 system (Life Technologies). 10 ng of cDNA was used for each reaction, performed in triplicate (technical replicates). Settings of the machine were not modified using a temperature of hybridisation and elongation of 60°C. We used the Power Sybr Green Master Mix (Life Technologies) containing both Taq polymerase and the fluorophore. To compare the expression difference between genes, we converted the cycle threshold (C_T) in copy number by applying the rule that a C_T = 10 is equivalent to 1E11 copies and a C_T = 37.34 is equivalent to 625 copies. A normalisation factor for each reference gene was calculated by dividing each copy number obtained for the reference gene by the geomean calculated with each sample of this reference gene. This normalisation factor is centred on 1. We obtained a normalisation factor for each reference gene used (actin beta and Cog1). The final calibration number was obtained by calculating the average of these two normalisation factors (actin beta and Cog1). We divided the copy number calculated for each sample for CFTR, UCP3 and UCP9 by this final calibration number. The result is a normalized value of gene expression for the different genes that keeps the difference between genes intact. The concentration of primers in the final reaction was 200 nM.

Reference genes were chosen from Bahr et al. (Bahr et al., 2009). We selected as reference genes actin beta and Cog1 (component of oligomeric Golgi complex I). ChoActbFo (forward) 5′-GCTCTTTTCCAGCCTTCCTT-3′; choActbRe (reverse) 5′-GAGCCAGAGCAGTGATCTCC-3′; choCog1Fo 5′-ACTAGCCTCCAGCCAGATCA-3′; choCog1Re 5′-GCAGGTGAGTCGTCTTCCT-3′.

For human CFTR (RefSeq NM_000492) we designed the following primers: hCFTR_WT_Fo1628 5′-CTGGCACCATTAAAGAAAATATCATC-3′; hCFTR_WT_Fo1628 5′-CTGGCACCATTAAAGAAAATATCATC-3′.

For CHO sequences we designed primers sequences as follows: choAQP3Fo167 5′-TCCTCACCATCAACTTGGCT-3′; choAQP3Re317 5′-AGTGTGTAGATGGGCAGCTT-3′; choAQP9Fo260 5′-TAAGAATTGGGGTGGGTGGG-3′; choAQP9Re379 5′-CAGTCAGTGCTGTTCTTGGG-3′.

CHO genomic sequences were found in the NCBI database (NW_003615295.1). Stealth siRNA was designed using the BLOCK-iT™ RNAi Designer (Lifetechnologies). Duplexes of siRNAs were made with the following sequences: SiRNA against CHO AQP3: 5′-CAUGGUGGCUUCCUCACCAUCAACU3′; 5′-AGUUGAUGGUGAGGAAGCCACCAUG-3′. Scramble siRNA: 5′-AGUCCUAGGUUGAGGGAACGACAUG; 5′-CAUGUCGUUCCCUCAACCUAGGACU-3′.

Lipofectamine 2000 (Life Technologies) was used for transfections. For each 35 mm dish, 9 µl of Lipofectamine 2000 and 300 nmol of siRNA duplex was used for each transfection. DHM measurements were made 24 to 48 hours after transfection, after which the CHO cells were immediately lysed with the RNA lysis buffer.

The authors thank Corinne Moratal and Sandra Borel, Mathilde Jolivet and Lyncée Tech SA (www.lynceetec.com) for technological assistance. We also thank Anne Cantereau for critical comments to the present manuscript.