The Cystic Fibrosis Transmembrane Conductance Regulator Activates Aquaporin 3 in Airway Epithelial Cells*

Enhanced osmotic water permeability has been observed in Xenopus oocytes expressing cystic fibrosis transmembrane conductance regulator (CFTR) protein. Subsequent studies have shown that CFTR activates an endogenous water permeability in oocytes, but that CFTR itself is not the water channel. Here, we show CFTR-dependent activation of endogenous water permeability in normal but not in cystic fibrosis human airway epithelial cells. Cell volume was measured by novel confocal x-z laser scanning microscopy. Glycerol uptake and antisense studies suggest CFTR-dependent regulation of aquaporin 3 (AQP3) water channels in airway epithelial cells. Regulatory interaction was confirmed by coexpression of CFTR and AQP3 cloned from human airways inXenopus oocytes and of CFTR and rat AQP3 in Chinese hamster ovary cells. These findings indicate that CFTR is a regulator of AQP3 in airway epithelial cells.

Enhanced osmotic water permeability has been observed in Xenopus oocytes expressing cystic fibrosis transmembrane conductance regulator (CFTR) protein. Subsequent studies have shown that CFTR activates an endogenous water permeability in oocytes, but that CFTR itself is not the water channel. Here, we show CFTR-dependent activation of endogenous water permeability in normal but not in cystic fibrosis human airway epithelial cells. Cell volume was measured by novel confocal x-z laser scanning microscopy. Glycerol uptake and antisense studies suggest CFTR-dependent regulation of aquaporin 3 (AQP3) water channels in airway epithelial cells. Regulatory interaction was confirmed by coexpression of CFTR and AQP3 cloned from human airways in Xenopus oocytes and of CFTR and rat AQP3 in Chinese hamster ovary cells. These findings indicate that CFTR is a regulator of AQP3 in airway epithelial cells.
The cystic fibrosis transmembrane conductance regulator (CFTR) 1 protein is mutated and defective in cystic fibrosis (CF), a common lethal genetic disease. CFTR has been demonstrated to function in two ways: (i) as a protein kinase A-regulated Cl Ϫ channel and (ii) as a regulator of other membrane conductances (1)(2)(3). Thus, CF may not only affect cAMP-dependent Cl Ϫ conductance, but may also affect other membrane conductances normally regulated by wild-type CFTR. Such a defect in CFTRdependent regulation was found for the epithelial Na ϩ conductance (1,4). Thus, enhanced Na ϩ conductance was detected in the airways and intestinal epithelium of CF patients, which is very likely to contribute to enhanced absorption of electrolytes and water and to the altered mucociliary clearance as well as intestinal obstructions, respectively (5)(6)(7).
Apart from epithelial Na ϩ conductance, osmotic water permeability has been reported to be influenced by CFTR (8,9). These studies were performed in Xenopus oocytes and indicated enhanced osmotic cell swelling after expression and cAMPdependent activation of CFTR. Since water-injected control oocytes did not demonstrate such a cAMP-activated water permeability, the most likely explanation was that cAMP acted through activation of CFTR. In fact, it was suggested initially that water uses the same conductive pathway and moves together with Cl Ϫ through the CFTR Cl Ϫ channel. This pathway should therefore be formed by transmembrane domains (8). However, it was shown in a subsequent study (a) that both water and Cl Ϫ ions moved independently when CFTR was activated by an increase in intracellular cAMP. Independence was demonstrated by showing selective inhibition of CFTR Cl Ϫ conductance by glibenclamide and selective blockage of water permeability by mercurial compounds and phloretin (9). In addition, this water pathway activated by CFTR was permeable for glycerol, too. These results support the assumption that some sort of endogenous water and glycerol permeability must be present in Xenopus oocytes and that this is activated by CFTR.
So far, it was unclear whether CFTR-activated water permeability is unique to Xenopus oocytes or whether a similar interaction can also be observed in mammalian cells. This question was addressed in this study by measuring cell volume changes and radioactive glycerol uptake in human airway epithelial and Chinese hamster ovary (CHO) cells. We found CFTR-dependent activation of osmotic water permeability in normal respiratory cells and identified one member of the aquaporin family (AQP3) as the interacting partner. Since CFTR-dependent stimulation of osmotic water permeability is absent in airway cells derived from CF patients, we speculate that this has a pathophysiological impact on the lung disease in CF.

EXPERIMENTAL PROCEDURES
Cell Culture-Human non-CF, CF, and CF airway epithelial cells transfected with 6REP-wtCFTR have been described in previous studies (10,11) and were kindly provided by Dr. D. C. Gruenert (University of California, San Francisco). CHO-K1 cells and CHO cells stably expressing wtCFTR (CHO-wtCFTR) or ⌬F508-CFTR (CHO-⌬F508) were kindly provided by Dr. X.-B. Chang (Mayo Clinic, Scottsdale, AZ). Cells were cultured on tissue culture plastic dishes or glass coverslips and kept in an atmosphere of 5% CO 2 under conditions described elsewhere (10,12). CHO cells were transfected with rat AQP3 cDNA (kindly provided by M. Echevarría, University of Sevilla, Sevilla, Spain) inserted into expression vector pZeoSV (Invitrogen) using standard techniques. Isolated colonies were expanded and assayed for AQP3 expression by measurement of glycerol uptake and detection of AQP3 mRNA (RT-PCR).
Cell Volume Measurements by Confocal Microscopy-Cells grown on coverslips were loaded with the fluorescent dye calcein/AM (10 mol/ liter for 30 min at room temperature), and repetitive x-z line scans at 488 nm excitation of one scanning line through three to six cells were performed using an LSM 410 apparatus (Zeiss, Germany) (13). A water immersion lens (Zeiss C-APO 63/1.2w) was used to avoid optical distortions due to refractive index mismatches. The z-line scan distance was set to result in square voxel; usually these were in z-steps of 0.33 m. The size of the confocal pinhole was set to give a full-width halfmaximum in the z-direction of 0.4 m. The z-line scans were stacked in a two-dimensional image giving a x-z slide through the cells. A customized macro allowed the recording and storage of time series of x-z slides (usually 200 -300 images). Cell area and the mean intensity of the cell area at the cutting z-line were analyzed with Metafluor software (Universal Imaging Corp., West Chester, PA). Initial area increase and initial intensity decrease induced by hypotonic bath solution due to omission of mannitol (72.5 mmol/liter NaCl, 0.4 mmol/liter KH 2 PO 4 , 1.6 mmol/liter K 2 HPO 4 , 1 mmol/liter MgCl 2 , 1.3 mmol/liter calcium gluconate, and 5 mmol/liter D-glucose, pH 7.4) were determined in the same cells before and after stimulation and in the presence of IBMX (0.5 mmol/liter) and forskolin (10 mol/liter).
Osmotic Water and Glycerol Permeability in Xenopus Oocytes-Isolation and microinjection of oocytes have been described in a previous report (15). Oocytes of identical batches were injected with 10 -50 ng of cRNA (wild-type CFTR, G551D-CFTR, and hAQP3 cloned from human airway cells). Water-injected oocytes served as controls. The osmotic water permeability coefficient (P f ) was calculated from the rate of hypotonic volume increase, measured by gravimetric techniques at 22°C (9). Hypotonicity was induced by omission of 120 mmol/liter mannitol from ND96 normotonic medium. P f was calculated according to Ref. 16: P f ϭ dV/dt⅐[1/S⅐⌬], with S being the oocyte surface area, V being the volume, and ⌬ being the osmolality gradient at zero time. The ratio dV/dt was calculated from the weight change 1 min after exposure to hypotonic buffer. Glycerol permeability (P gly ) was calculated from the initial rate of [ 14 C]glycerol uptake per oocyte during incubation with [ 14 C]glycerol (2 Ci/ml) and 1 mmol/liter glycerol. Oocytes were lysed overnight in 100 g/liter SDS at room temperature, and radioactivity was counted.
Compounds and Statistics-All chemicals used were the highest grade of purity available. Data are given as means Ϯ S.E. (n), where n refers to the number of experiments. Paired and unpaired t tests were used for statistics with a p value of Ͻ0.05, indicating statistical significance.

Water and Glycerol Permeability Is Increased by cAMP in
Normal Airway Cells, but Not in Cells Derived from CF Patients-Water permeability in airway epithelial cells was assessed by measuring osmotically induced cell swelling using the novel microscopic techniques described above (13). Rapid change to hypotonic bath solution resulted in an increase in cell area and a decrease in fluorescence intensity, as shown for a non-CF cell in Fig. 1. Linear regression was fitted to the data points obtained during the first 30 s after exposure to the hypotonic bath solution (Fig. 1B). Stimulation with IBMX and forskolin augmented the area increase and the fluorescence pixel (calcein concentration) decrease elicited by hypotonic bath solution. These measured volume changes necessarily underestimate the real rate of hypotonic cell swelling and the effects of CFTR on osmotic water permeability because the cells activate their volume regulatory decrease mechanisms, which will be even augmented when CFTR Cl Ϫ conductance is activated (9). The summary shown in Fig. 2A indicates that cell swelling was significantly enhanced after activation of CFTR (solid bars). Such a cAMP-dependent increase in hypotonic cell swelling was not observed in airway epithelial cells derived from CF patients (Fig. 2B).
Previous studies have demonstrated that osmotic water permeability activated by CFTR in Xenopus oocytes is also permeable for glycerol. Therefore, we measured [ 14 C]glycerol uptake in airway epithelial cells that expressed endogenous wild-type CFTR (non-CF), mutant CFTR (CF), or exogenous wild-type CFTR (CF-wtCFTR) (Fig. 3). After stimulation of non-CF cells with IBMX and forskolin, [ 14 C]glycerol uptake (closed circles) was enhanced when compared with control uptake (open circles). Such an activation of [ 14 C]glycerol uptake by cAMP was not observed in CF cells. Here, [ 14 C]glycerol uptake was even attenuated when the cAMP-dependent pathway was activated.
However, expression of exogenous wild-type CFTR in CF cells (CF-wtCFTR) restored the effects of IBMX and forskolin. Pretreatment of the cells with HgCl 2 (5 mol/liter, 15 min) completely inhibited the effect of cAMP on [ 14 C]glycerol uptake in non-CF cells as well as in CF cells expressing exogenous CFTR (closed squares). At this concentration, CFTR Cl Ϫ currents were not affected (data not shown). These results suggest, in agreement with what has been observed in Xenopus oocytes (9), that, also in human airway epithelial cells, a glycerol-permeable water channel is activated by CFTR.
AQP3 Is Responsible for CFTR-dependent Glycerol Uptake-Water permeability in mammals is facilitated in many cells by specific water channels, so-called aquaporins (17). Expression of various types (AQP1, AQP3, AQP4, and AQP5) in non-CF and CF airway epithelial cells was performed by RT-PCR. No expression of AQP4 could be detected in airway epithelial cells. However, expression of AQP1, AQP3, and AQP5 was found (data not shown). Message for AQP3 was detected in freshly isolated airway epithelial cells from nasal polyps as well as in various cultured CF and non-CF airway epithelial cells (Fig.  4B). Since AQP3, in contrast to AQP1 or AQP5, has been described to transport glycerol (18,19), these results suggest that AQP3 is the water pathway responsible for CFTR-dependent water and glycerol uptake. This was further supported by staining of AQP3 in these cells. Immunocytochemical localization of AQP3 demonstrated abundant expression of this protein throughout the plasma membrane (Fig. 4A). AQP3 was also immunolocalized in CHO cells expressing AQP3 and in CF airway cells, but not in control CHO cells or in the absence of the primary antibody (Fig. 4A).
Furthermore, normal respiratory epithelial cells expressing endogenous CFTR (non-CF) and CF epithelial cells expressing exogenous wild-type CFTR (CF-wtCFTR) were incubated with AQP3 sense and antisense oligonucleotides, respectively. After 48 h of incubation, CFTR-dependent, i.e. forskolin/IBMX-induced, glycerol uptake was abolished in non-CF and CF-wtCFTR cells. No such effects on forskolin/IBMX-activated glycerol uptake were observed in cells incubated with sense oligonucleotides for AQP3 or antisense oligonucleotides for AQP1 and AQP5 (Fig. 5), suggesting that AQP3 is responsible

FIG. 2. Activation of CFTR enhances water permeability in non-CF but not CF airway epithelial cells.
A, shown is a summary of cell volume measurements in non-CF cells. Stimulation by IBMX and forskolin (black bars) significantly enhanced initial area increase and initial fluorescence decrease in cells compared with control conditions before and after stimulation (n ϭ 18). CFTR-activated water permeability led to an increase in the absolute area, whereas fluorescence intensity was decreased (n ϭ 18). B, in CF epithelial cells, stimulation by IBMX and forskolin had no effect on cell volume change (n ϭ 12). Data are means Ϯ S.E. Asterisks indicate significant difference from control (p Ͻ 0.05).

FIG. 3. Activation of CFTR increases glycerol permeability. [ 14 C]Glycerol uptake was measured in the initial 3 min after exposure (open circles) in non-CF and CF airway epithelial cells and in transfected CF cells (CF-wtCFTR). Stimulation of non-CF cells with IBMX and forskolin
increased [ 14 C]glycerol uptake, whereas a decrease in [ 14 C]glycerol uptake was observed for CF cells (closed circles) (n ϭ 3 each). cAMP-dependent inhibition of glycerol uptake in CF cells is currently poorly understood and is due to another CFTR-independent mechanism. Transfection with wtCFTR (CF-wtCFTR) restored the effects of IBMX and forskolin on [ 14  for CFTR-dependent activation of osmotic water permeability in airway epithelial cells.
CFTR Is Required for Activation of AQP3 by cAMP in Xenopus Oocytes-Human AQP3 was cloned from human airway epithelial cells used here by RT-PCR. The obtained nucleotide coding sequence was identical to the sequence published in a previous report (20). Expression of the cloned hAQP3 cDNA in Xenopus oocytes induced an enhanced osmotic glycerol permeability coefficient (P gly ) (9) (11.0 Ϯ 1.2 ϫ 10 Ϫ6 cm/s; n ϭ 13) compared with control oocytes (1.7 Ϯ 1.7 ϫ 10 Ϫ6 cm/s; n ϭ 11). When CFTR and hAQP3 were coinjected into Xenopus oocytes, enhanced basal P gly was further augmented by stimulation of the oocytes with IBMX (Fig. 6C). No increase was observed when hAQP3 was expressed alone (data not shown). Moreover, in hAQP3/CFTR-coexpressing oocytes, activation of CFTR by IBMX in the presence of an isotonic bath solution induced initial cell shrinkage due to Cl Ϫ exit from the oocytes (Fig. 6A). Subsequent exposure to an extracellular hypotonic solution in the continued presence of IBMX augmented cell swelling and increased the osmotic water permeability coefficient (P f ) significantly, indicating activation of hAQP3 water channels by CFTR (Fig. 6, A and C). As described for the experiments performed in airway cells, the detected volume changes will even underestimate the impact of CFTR on osmotic water permeability because of the volume regulatory decrease that is augmented by activation of CFTR Cl Ϫ conductance (9). A mutation in the first nucleotide-binding fold of CFTR (G551D-CFTR) abolished the effects of IBMX on P f (Fig. 6B), confirming impaired regulation of hAQP3 by mutant forms of CFTR.
Only CHO Cells Coexpressing Both CFTR and Rat AQP3 Demonstrate cAMP-activated [ 14 C]Glycerol Uptake-The above results obtained in Xenopus oocytes have been verified in a mammalian expression system by coexpressing rAQP3 and wtCFTR or ⌬F508-CFTR, respectively, in CHO cells. [ 14 C] Glycerol uptake was measured before and after stimulation with forskolin and IBMX. As shown in Fig. 7, control cells or cells expressing only wtCFTR or ⌬F508-CFTR demonstrated base-line [ 14 C]glycerol uptake, which was not influenced by stimulation with forskolin and IBMX. Control, ⌬F508-CFTR-, and wtCFTR-expressing CHO cells were transfected with rAQP3, and expression was detected by RT-PCR analysis (data not shown). Additional expression of rAQP3 enhanced base-line Control staining was carried out without primary antibody (lower panel). Bars ϭ 30 m. B, total RNA of airway cells derived from human nasal polyps and cultured non-CF and CF airway epithelial cells as well as cultured CF airway cells transfected with wtCFTR was reversetranscribed, and expression of AQP3 was detected by PCR amplification. RTϮ indicates PCR of RNA samples that were reverse-transcribed (ϩ) or that were used without previous reverse transcription (control) (Ϫ). Experiments were performed in triplicates. bp, base pairs.

FIG. 5. cAMP-dependent activation of glycerol uptake is abolished in AQP3 antisense oligonucleotide-treated non-CF cells and in CF cells expressing wtCFTR (CF-wtCFTR).
After 48 h of treatment with antisense oligonucleotides for AQP3, CFTR-activated glycerol uptake was abolished. Incubation with AQP1 or AQP5 antisense oligonucleotides or AQP3 sense oligonucleotides was without any significant effects. Data are expressed as means Ϯ S.E. (n ϭ number of measurements). Asterisks indicate significant difference from control (p Ͻ 0.05).
glycerol uptake in all three cells lines to different degrees. Control and ⌬F508-CFTR-expressing cells did not further enhance glycerol uptake upon exposure to IBMX and forskolin, whereas it was significantly increased in wtCFTR-expressing cells (Fig. 7). Both the basal [ 14 C]glycerol uptake caused by expression of rAQP3 and the uptake caused by activation of wtCFTR were inhibited by HgCl 2 (5 mol/liter). These experiments clearly indicate CFTR-dependent activation of hAQP3 when both proteins are expressed heterologously in CHO cells. DISCUSSION The present data show compelling evidence for a novel type of CFTR-dependent regulation of yet another epithelial mem-brane permeability. They demonstrate that CFTR, when stimulated by protein kinase A, activates a water permeability in respiratory epithelial cells. These results suggest coupling between Cl Ϫ transport as performed by the CFTR Cl Ϫ channel and water transport, thereby adjusting water permeability to electrolyte transport in the airway epithelium. In that sense, this mechanism would help to secure proper hydration of the surface fluid as well as mucociliary clearance (21,22).
Several different types of aquaporins are expressed in the airways, including AQP1, AQP3, AQP4, and AQP5 (23). AQP expression in the lung shows a complex pattern suggesting functional specialization of the different AQPs, which seems to be required in the postnatal period as well as in the adult lung (23). AQP3 is expressed primarily in basolateral membranes of acinar cells forming the submucosal glands (24). These glands are regarded as the predominant site for electrolyte secretion (24 -26). On the other hand, AQP3 is expressed abundantly in nonpolarized basal cells of the trachea and in basolateral membranes of nasal surface epithelial cells (23,24) involved in reabsorption of electrolytes (6,21). In this study, expression of AQP3 and its CFTR-dependent regulation were detected in cultured airway epithelial cells, which demonstrate features of cell differentiation such as expression of amiloride-sensitive Na ϩ currents (10,27). However, expression of AQP3 does not seem to be limited to the basolateral membranes of these cells, thereby reflecting the properties of basal cells rather those that of surface cells. Within the respiratory tract, CFTR is expressed predominantly in apical membranes of submucosal gland cells and, to a lesser degree, also in those of surface epithelial cells. CFTR Cl Ϫ currents can also be detected in basal cells (28), but there is no evidence for CFTR expression in basolateral membranes of either cell type (29). This suggests that, except for the basal cells, CFTR and AQP3 are expressed on the different poles of airway cells. We may therefore currently only speculate in what cell type of the highly differentiated respiratory tract the CFTR-AQP3 interaction actually takes place. Further studies have to clarify whether AQP3 contributes to both airway absorption and secretion and how it participates in formation of the airway surface liquid, which is FIG. 6. wtCFTR activates human AQP3 in Xenopus oocytes. A, shown is the volume increase in oocytes under hypotonic conditions as measured by gravimetric techniques (9). Hypotonic (Hypo) volume changes in water-injected oocytes were independent of the presence of IBMX (open circles; n ϭ 9). Hypotonic cell swelling was augmented by IBMX in AQP3/CFTR-coexpressing oocytes (closed circles; n ϭ 8). After initial osmotic swelling, cells were allowed to recover in isotonic bath and were then exposed to forskolin/IBMX, which led to cell shrinkage due to activation of CFTR Cl Ϫ conductance and Cl Ϫ exit. B, the calculated osmotic water permeability coefficient (P f ) in Xenopus oocytes was larger after stimulation with IBMX due to CFTR-dependent activation of the endogenous water channel in Xenopus oocytes (9). P f was increased in AQP3/CFTR-coexpressing oocytes and was further enhanced by IBMX, whereas oocytes coexpressing AQP3 and G551D-CFTR did not show an increase in P f upon stimulation by IBMX. C, the osmotic glycerol permeability coefficient (P gly ) (9) was enhanced by IBMX in CFTR-expressing oocytes. In hAQP3/wtCFTR-coexpressing oocytes, base-line P gly was enhanced and was further increased by IBMX. Data are means Ϯ S.E. (n). Asterisks indicate significant difference from control (p Ͻ 0.05). essential for proper mucociliary clearance. CFTR-dependent activation of AQP3 does not take place in airway cells carrying the CF defect or in cells overexpressing both AQP3 and mutant forms of CFTR. We may therefore speculate that regulation of AQP3 by CFTR is impaired in the respiratory tract of CF patients, which may contribute to the imbalance between secretion and absorption detected in CF airways (6,26,30).
Activation of AQP3 by CFTR is yet another example for CFTR-dependent regulation of other epithelial membrane conductances. Several other ion conductances have been reported to be regulated by CFTR, like the epithelial Na ϩ conductance and intermediate conductance Cl Ϫ channel (outwardly rectified Cl Ϫ channel and intermediate conductance outwardly rectifying chloride channel) (1,4,31). Furthermore, the affinity of K ϩ channels like Kir6.1 and ROMK2 for the inhibitory compound glibenclamide is enhanced by CFTR (32). The mechanisms of this regulation are currently under examination (3). They are probably examined in most detail for CFTR-dependent inhibition of the epithelial Na ϩ conductance, which was shown only recently to take place in the native respiratory tissue (7). CFTRdependent regulation of epithelial Na ϩ conductance depends on the presence of Cl Ϫ ions and the existence of an intact first nucleotide-binding domain of CFTR. It takes place even in isolated membrane patches, suggesting a rather direct interaction of both proteins (33)(34)(35)(36). Although the actual mechanism of this interaction still remains obscure, results from these studies may also guide further experiments to unmask CFTRdependent regulation of AQP3. The results shown here add a new example to the growing list of CFTR-dependent interactions with other membrane proteins, thereby emphasizing the function of CFTR as a conductance regulator. The results may also help to understand the complex pattern of the common disease cystic fibrosis.