Hypertonic Saline Therapy in Cystic Fibrosis

Recent data indicate the clinical benefit of nebulized hypertonic saline in cystic fibrosis lung disease, with a proposed mechanism involving sustained increase in airway surface liquid volume. To account for the paradoxical observation that amiloride suppresses the beneficial effect of hypertonic saline, it has been previously concluded (Donaldson, S. H., Bennett, W. D., Zeman, K. L., Knowles, M. R., Tarran, R., and Boucher, R. C. (2006) N. Engl. J. Med. 354, 241-250) that amiloride-inhibitable aquaporin (AQP) water channels in airway epithelia modulate airway surface liquid volume. Here, we have characterized water permeability and amiloride effects in well differentiated, primary cultures of human airway epithelial cells, stably transfected Fisher rat thyroid epithelial cells expressing individual airway/lung AQPs, and perfused mouse lung. We found high transepithelial water permeability (Pf, 54 ± 5 μm/s) in airway epithelial cells that was weakly temperature-dependent and inhibited by >90% by reduced pH in the basal membrane-facing solution. Reverse transcription-PCR and immunofluorescence suggested the involvement of AQPs 3, 4, and 5 in high airway water permeability. Experiments using several sensitive measurement methods indicated that amiloride does not inhibit water permeability in non-cystic fibrosis (non-CF) or CF airway epithelia, AQP-transfected Fisher rat thyroid cells, or intact lung. Our data provide evidence against the mechanism proposed by Donaldson et al. to account for the effects of amiloride and hypertonic saline in CF lung disease, indicating the need to identify alternate mechanisms.

brane conductance regulator (CFTR) chloride channel. Morbidity and mortality in CF result primarily from chronic airway infection, which results in progressive deterioration of lung function.
Two recent clinical studies have demonstrated short and long term benefits of nebulized hypertonic saline in improving lung function in CF (1,2). Prior clinical studies also support the efficacy of various inhaled hyperosmolar agents in CF (3)(4)(5)(6). The recent study by Donaldson et al. (1) concludes that hypertonic saline produces sustained elevation in airway surface liquid (ASL) volume, which improves mucociliary clearance in the airways. However, a paradoxical effect was found when amiloride was administered together with hypertonic saline. Rather than improving lung function because of its inhibitory effect on ENaC and consequent prevention of ASL absorption, amiloride negated the beneficial effect of hypertonic saline. To account for these findings, Donaldson et al. (1) postulated the involvement of airway AQPs in establishing ASL volume (Fig. 1) and reported strong amiloride inhibition of osmotic water permeability in airway epithelial cells. In their model, ENaC hyperactivity dehydrates the ASL in CF airways (Fig. 1, left panel), and hypertonic saline restores ASL volume (middle panel). If amiloride acts only on ENaC, then amiloride is predicted to increase ASL volume ( Fig. 1, right, top panel). To account for the clinical data, they hypothesize that amiloride inhibits airway AQPs, which prevents water from entering the airways (Fig. 1, right, bottom panel).
The proposed involvement of AQPs in ASL regulation is surprising, as is the inhibition of AQP water permeability by amiloride. Osmotically induced water transport across airways has been shown to be high and likely AQP-dependent in microperfused small airways (7) and later in cultured human tracheal epithelial cells (8) and in spheroids composed of airway epithelial cell monolayers (9). Despite this, the reduction of airway epithelial water permeability by deletion of the various lung/ airway aquaporins (AQPs 1, 3, 4 and 5) in transgenic mice did not affect ASL volume or ionic composition (10). Inhibition of AQP water permeability by amiloride is surprising, because amiloride has no aquaretic effect at concentrations that inhibit ENaC in the renal distal tubule.
Here, we tested the hypothesis of Donaldson et al. (1) by characterizing water permeability and amiloride effects in human bronchial epithelial cell cultures, stably transfected epithelial cells expressing airway/lung AQPs, and intact lung using a variety of biophysical methods to quantify water transport. Water permeabilities of airway cells and intact lung surface were found to be high and AQP-dependent, although amiloride did not inhibit water permeability nor did it inhibit water permeability of the individual AQPs. These findings have important implications for understanding the benefit of hypertonic saline in CF and for future development of related strategies for the therapy of CF lung disease.

EXPERIMENTAL PROCEDURES
Cell Culture-Surface bronchial epithelial cells were obtained from non-CF human lung specimens that were not suitable for lung transplantation. CF bronchial epithelial cells were obtained from patients at the time of transplantation. The cells were dissociated by enzymatic digestion, as described previously (11). Isolated cells were suspended in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium containing 5% fetal calf serum (Hyclone, Logan, UT), gentamicin (50 g/ml), penicillin (100 units/ml), streptomycin (1 mg/ml), and fungizone (2.5 g/ml) and seeded as passage 0 (P0) cultures at a density of 10 6 cells/cm 2 onto 12-mm Transwell polycarbonate inserts (0.4 m pore size; Costar, Corning, NY) overlaid with a thin coat (15 g/cm 2 ) of human placental collagen (Sigma-Aldrich). The next day, the cells were rinsed with PBS, and the medium was replaced with air-liquid interface medium (12) containing gentamicin, penicillin, and streptomycin at the concentrations given above. The medium was changed daily. When the cell sheets became confluent and actively absorbed mucosal fluid, as indicated by the loss of apical fluid in the central portion of the insert (generally 3-5 days), the mucosal surface was rinsed with PBS, and medium was added only to the basal side of the insert.
For some experiments, bronchial epithelial cells (non-CF and CF) were obtained after expansion in human placental collagen-coated tissue culture flasks. After 80 -90% confluence, cells were released using 0.05% trypsin and 0.2% EDTA in 0.9% NaCl at 37°C. The trypsin was neutralized with Dulbecco's modified Eagle's medium/Ham's F-12 medium containing 20% fetal calf serum, and propagated cells were resuspended in plating medium and seeded onto cell culture inserts as described above but designated as passage 1 (P1) cultures. By day 15 after seeding, both P0 and P1 cultures formed tight layers with transepithelial resistance of 1-2 k⍀⅐cm 2 , at which time they were used for experiments. Cultures were confirmed to express cAMPactivated CFTR by measurement of short circuit current, as reported previously (13). Protocols were approved by the University of California, San Francisco Committee on Human Research.
Transepithelial Water Permeability-Osmotic water permeabilities across human bronchial epithelial (HBE) and FRT cell layers were determined using a dye dilution method, as depicted in Fig. 2A. The dilution of a cell-impermeant, photostable, inert dye (Texas Red TM -dextran, 10 kDa; Molecular Probes, Eugene, OR) was used as a measure of transcellular osmotic water flux. The basal surface of cells on the porous filter was bathed in 1 ml of isosmolar PBS. The apical surface was bathed in 200 l of hyperosmolar PBS (PBS ϩ 300 mM D-mannitol) containing 0.25 mg/ml Texas Red-dextran. In some experiments, 100 -500 M amiloride (Sigma-Aldrich), dissolved freshly from powder, was added to the apical buffer, as done by Donaldson et al. (1), or to both the apical and basal bathing buffers. Cultures were placed in a 5% CO 2 tissue culture incubator (at 27 or 37°C), and 5 l samples of dye-containing apical fluid were collected at specified times. The samples were diluted in 2 ml of PBS, and fluorescence was measured by cuvette fluorimetry (Fluoro Max-3; Horiba, Tokyo, Japan). In some experiments, the osmotic gradient was reversed by the addition of mannitol to the basal rather than the dye-containing apical solution. In other experiments, the pH of either the basal or apical solutions was adjusted to 5.0 using buffers containing 137.6 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 6 mM D-glucose, and 10 mM HEPES (for pH 7.4) or MES (for pH 5.0). In some experiments, HgCl 2 (0.2 or 1 mM) was added to the apical and basolateral solutions 5 min prior to the osmotic gradient.
For computation of transepithelial osmotic water permeability coefficients (P f , in cm/s), the time course of fluorescence in FIGURE 1. Proposed mechanism of hypertonic saline and amiloride effects on airway surface liquid (ASL) in cystic fibrosis. In CF, periciliary fluid is depleted because of impaired Cl Ϫ secretion through CFTR and increased Na ϩ absorption through ENaC (left panel). Hypertonic saline is proposed to increase ASL volume (middle panel) via AQP-mediated and/or transmembrane water movement (blue arrows). By inhibition of ENaC, amiloride is postulated to increase ASL volume by blocking Na ϩ and fluid absorption (right, top panel). Amiloride inhibition of AQP water channels was proposed as a mechanism to account for the amiloride effect of negating the beneficial clinical effects of hypertonic saline (right, bottom panel). response to solution osmolarity changes, F(t), was fitted to a single exponential time constant, F(t)/F o ϭ B ϩ A⅐e Ϫt/ , where F o is initial fluorescence, A is amplitude, and is the exponential time constant. Because of dye binding to cells and/or the support, data used for exponential regression was collected at 5 min and later after establishing the osmotic gradient. Water flux was computed as the rate of dye dilution by dV(0)/ dt ϭ V o ⅐A/, which follows from the relation: . V o is the initial volume of dye-containing apical buffer (0.2 cm 3 ), and V(t) is the apical solution volume at time t. From these equations, P f was computed as dV(0)/ dt ϭ P f ⅐S⅐v w ⅐(⌽ 1 Ϫ ⌽ 2 ), where S is the tissue surface area assuming a smooth surface (1.13 cm 2 ), v w is the partial molar volume of water (18 cm 3 /mol), and (⌽ 1 Ϫ ⌽ 2 ) is the transepithelial osmotic gradient (3 ϫ 10 Ϫ4 mol/cm 3 ).
Plasma Membrane Water Permeability-Plasma membrane osmotic water permeability was measured using a calceinquenching method (shown in Fig. 3A) that has been applied previously to in vitro and in vivo cell systems (15,16). Cells were loaded with calcein by incubation with PBS containing 10 M calcein-AM (Molecular Probes) for 30 min at 37°C. HBE cultures grown on collagen-coated supports were studied by cutting the flat supports from the plastic inserts and mounting them in a custom-built perfusion chamber with support bottoms pressed against a coverslip and apical cell surfaces exposed to the perfusate. FRT cells grown on glass coverslips were inserted in a similar perfusion chamber but with cells facing upwards. Solutions were exchanged between PBS (290 mOsm) and hyperosmolar PBS (590 mOsm, PBS with added D-mannitol) using a gravity pinch valve system (ALA Scientific Instruments, Westbury, NY). Perfusates contained either Me 2 SO vehicle (0.1%) or amiloride (500 M) added from a 1000ϫ Me 2 SO stock solution. Fluorescence was excited using an X-cite 120 mercury lamp (EXFO Life Sciences, Ontario, Canada) and focused through an inverted epifluorescence microscope through a long working distance 25ϫ air objective (numerical aperture 0.35, Leitz). Excitation and emission light was filtered using a custom cube (Chroma, Rockingham, VT), and fluorescence was collected using a 14-dynode photomultiplier and then amplified, digitized, and recorded using custom software (written in LabView, National Instruments, Austin, TX).
To compute plasma membrane P f , relative fluorescence (F(t)/F o ) was fitted to a single exponential function (as above) for FRT cell studies or a biexponential function (F(t)/ F o ϭ A⅐e Ϫt/1 ϩ B⅐e Ϫt/2 ϩ C) for HBE studies. Plasma membrane P f was computed as described above, where surface area S was taken as the apical surface area of the cell monolayer, assuming a flat, smooth surface. The time-varying cell layer height, h(t), was related to normalized amplitudes according to the predicted 50% final decrease from initial height (h o ) under hypertonic conditions. For FRT and HBE cells, h o was taken as 5 and 20 m, respectively. P f was then computed from dh(0)/dt ϭ P f ⅐v w ⅐(⌽ 1 Ϫ ⌽ 2 ).
Airspace-Capillary Water Permeability in Intact Lung-Wild-type and AQP5-deficient mice in a CD1 genetic background (age 8 -14 weeks) (17) were sacrificed using an overdose of 2,2,2-tribromoethanol (avertin, Sigma-Aldrich) intraperito-neally. Protocols were approved by the University of California, San Francisco Committee on Animal Research. The trachea was transected and cannulated, the pulmonary artery was cannulated, and the heart and lungs were moved en bloc to a perfusion chamber as described previously (18). After ensuring that no air leaks were present, 0.5 ml of fluorescein isothiocyanate-dextran solution (10 kDa, 0.5 mg/ml in PBS; Sigma-Aldrich), with or without freshly dissolved amiloride (500 M), was instilled into the tracheal catheter. The pulmonary artery was perfused at constant pressure (ϳ60 cm H 2 O) with PBS, with or without 500 M amiloride, for at least 5 min before measurements. The pleural surface was washed continuously with PBS, and perfusate and wash effluent were continuously withdrawn from the chamber by suction (perfusate flow 5 ml/min). The time course of fluorescence intensity from a 3-5-mm spot on the lung pleural surface, in a similar location for all samples, was continuously monitored while perfusate fluid was exchanged between isosmolar and hyperosmolar PBS (PBS ϩ 300 mM D-mannitol) using the microscope setup, detection system, and gravity pinch valve system described above.
Histology and Immunocytochemistry-Paraffin sections of P0 HBE cell culture inserts and fixed human trachea were prepared. For paraffin sections, inserts were fixed for at least 30 min in buffered formalin (10%) and stored in 0.1 M phosphate buffer (pH 7.4). Standard tissue dehydration and paraffin infiltration was performed in a Tissue-Tek VIP processor (Sakura Finetek, Torrance, CA). Sections were cut at 4 m on a rotary microtome and stained with hematoxylin and eosin. Specimens were viewed and photographed on an Olympus light microscope (Olympus America, Inc., Melville, NY) equipped with a digital imaging system (QImaging, Burnaby, Canada). Immunostaining was done by standard procedures using polyclonal antibodies for AQP3, AQP4, and AQP5 (Chemicon, Temecula, CA) incubated for 2 h and Cy3-conjugated goat anti-mouse IgG (1:200; Sigma-Aldrich) incubated for 30 min.
Fluorescence-based real-time RT-PCR was done to compare relative AQP3, AQP4, and AQP5 mRNA expression in non-CF versus CF cells using the LightCycler TM and with the LightCycler FastStart DNA Master PLUS SYBR Green I kit (Roche Diagnostics) according to the manufacturer's instructions. Surface epithelial cells were dissociated from human bronchi, quickfrozen, and stored at Ϫ80°C until thawed for RNA isolation. PCR primers were as described above. Results were reported as a normalized, calibrated ratio with all samples normalized to ␤-actin. Concentration ratios for each CF and non-CF sample were calibrated to calibrator samples (pooled cDNA from non-CF subjects), with the results reported as a normalized ratio with the calibrator sample as the denominator as follows: relative mRNA level ϭ ratio of sample (target/reference)/ratio of calibrator (target/reference).

RESULTS
Water transport in airway surface epithelium was studied using well differentiated HBE cell cultures. Transepithelial osmotic water permeability was measured using a dye dilution method that assays water flux across tight epithelia cultured on porous filters ( Fig. 2A). The fluorescence of an apical solution volume marker provided a quantitative readout of osmotically driven water transport across the cell layer. An induced osmotic gradient caused transepithelial water movement from which osmotic water permeability (P f ) was deduced. The P0 non-CF cultures were first characterized. Fig. 2B (top) shows fluorescence dilution at 37°C, in which a 300 mM gradient of D-mannitol induced osmotic water flux into the apical, dye-containing solution. Data in the absence versus the presence of amiloride (500 M) were compared. Fig. 2C summarizes transepithelial P f values. There was no significant difference in P f upon exposure of P0 HBE cultures to high dose amiloride (control versus amiloride, 54 Ϯ 5 versus 52 Ϯ 7 m/s). Lower concentrations of amiloride (100 and 250 M) also did not change P f (not shown).
When the direction of the osmotic gradient was reversed by the addition of mannitol to the basal bathing solution, the fluorescent volume marker became concentrated over time (Fig.  2B, middle). The deduced transepithelial P f (46 Ϯ 15 m/s) did not differ significantly from that in control studies (mannitol in apical solution), indicating symmetrical water transport. To investigate the involvement of AQPs in HBE water permeability, experiments were done at reduced temperature (27°C) and with reduced pH in the apical or basal bathing solutions (Fig.  2C, bottom). P f at 27°C (47 Ϯ 9 m/s) was similar to that at 37°C, indicating a low Arrhenius activation energy, as expected for AQP-facilitated water transport. Reducing apical solution pH to 5 did not change P f significantly, whereas a basolateral solution pH of 5 decreased P f by Ͼ90%, providing functional evidence for the involvement of basolateral membrane AQP3 in water permeability; the water permeability of AQP3 is inhibited at low pH, whereas that of other AQPs is not (19).
Non-CF and CF P1 cell cultures were also studied to mimic the culture conditions employed by Donaldson et al. (1). As summarized in Fig. 2D, P1 cultures had ϳ50% lower water permeability than P0 cultures but without significant differences comparing non-CF versus CF cultures or control cultures versus cultures exposed apically to 500 M amiloride (or to 100 M amiloride, not shown). Finally, as an additional attempt to look for amiloride-sensitive water transport, water permeability was measured in P0 non-CF cultures in the presence of both apically and basolaterally added amiloride (500 M). As seen from the raw fluorescence data in Fig. 2E, osmotic water flux was not inhibited by high-dose amiloride, indicating the lack of amiloride inhibition of apical or basolateral AQPs.
Plasma membrane water permeability in P0 HBE cultures was determined using a calcein-quenching method, which is based on rapid changes in cytoplasmic calcein fluorescence in response to changes in the concentration of cytoplasmic anionic proteins and hence to changes in cell volume (Fig. 3A) (15). Supports with calcein-loaded HBE cells were mounted in a perfusion chamber for continuous measurement of cytoplasmic calcein fluorescence. Switching between isosmolar and hyperosmolar perfusates produced reversible changes in fluorescence with biexponential kinetics (Fig. 3B), probably reflecting the heterogeneous distribution of AQPs and the complicated geometry of bronchial epithelia. P f was quantified from the initial kinetics of fluorescence change in response to osmotic gradients as described under "Experimental Procedures." As summarized in Fig. 3C, the plasma membrane P f was not significantly affected by inclusion of 500 M amiloride in the perfusates.
AQP expression was characterized in P0 HBE cells. Hematoxylin and eosin-stained transverse paraffin sections confirmed the differentiation of 2-week-old cultures into a ϳ20-m thick pseudostratified epithelium with numerous cilia (Fig. 4A). Cultures were scraped for RT-PCR analysis, which detected mRNA for several AQPs (Fig. 4B). Transcripts of AQP3, AQP4, and AQP5 were found in highest abundance, consistent with previous studies (see "Discussion"). Lesser but non-zero expression of AQP1, AQP6, and AQP7 mRNA was detected. Quantitative real-time RT-PCR was done to compare expression of the principal AQPs in P0 HBE cultures from non-CF versus CF airways. As summarized in Fig. 4C, no significant differences were found in AQP transcript expression in non-CF versus CF cells. Immunofluorescence done on paraffin sections of HBE cultures and intact human bronchus demonstrated AQP3 and AQP4 protein expression on basolateral membranes, whereas AQP5 protein expression was not detected. Positive control staining of kidney for AQP3 and AQP4 is shown and of salivary gland for AQP5.
We next investigated the possible effects of amiloride on the major lung AQPs, including AQP1 (expressed in microvascular endothelia), using AQP-transfected epithelial cells and intact mouse lung. Transepithelial and plasma membrane water permeabilities were measured in FRT cells expressing either yellow fluorescent protein (control), AQP1, AQP3, or AQP4. Transepithelial water permeability of FRT cell lines grown on Transwell supports was measured after the cultures formed electrically tight monolayers. Water permeabilities of AQP-expressing FRT cells were greater than that in control cells and not sensitive to amiloride (Fig. 5A, top). Fig. 5A (bottom) summarizes transepithelial P f values. Plasma membrane water permeability was measured on the FRT cells, with calcein fluorescence curves fitted to single exponential functions (representative traces for FRT-AQP3 and FRT-AQP4 cells shown in Fig. 5B, top). In agreement with the measurements of transepithelial water permeability, plasma membrane P f was increased with AQP expression and not sensitive to amiloride (Fig. 5B, bottom).
Last, the effect of amiloride was investigated in intact, perfused mouse lung using a pleural surface method to measure airspace-capillary osmotic water permeability (18). The airspace compartment was filled with isotonic fluid containing fluorescein isothiocyanate-dextran as a volume marker. The kinetics of pleural surface fluorescein isothiocyanate-dextran fluorescence was measured in response to changes in osmolality of the pulmonary artery perfusate. There was rapid osmotic equilibration across the airspace-capillary barrier, which was not affected by the inclusion of 500 M amiloride in the airway instillate and pulmonary artery perfusate (Fig. 6A, top). In agreement with previous results (20), osmotic equilibration was ϳ10-fold slowed in lungs of AQP5-deficient mice (Fig. 6A, bottom), indicating the involvement of AQP5 in airspace-capillary osmotic water transport. The rates of osmotic water equilibration in lungs from wild-type mice (with and without amiloride) are summarized in Fig. 6B.

DISCUSSION
The primary purpose of this study was to investigate a proposed AQPdependent mechanism to account for the clinical effects of nebulized hypertonic saline (HS) and amiloride on lung function in CF. The validity of the proposed mechanism of clinical benefit involving amiloride-sensitive airway AQPs has important implications regarding the use of HS versus non-salt hyperosmolar agents in CF therapy, as well as in the proposed use of amiloride-type ENaC inhibitors in preventing ASL dehydration.
HS is being used increasingly as a chronic therapy in CF, particularly following the recent publication of two clinical trials in the New England Journal of Medicine demonstrating clinical benefit (1,2). The larger of the two studies reported that 4 ml of 7% HS nebulized twice daily for 48 weeks decreased exacerbation rate by 56% in CF subjects (2). Exacerbations were defined by the need for intravenous antibiotics or using a symptom score involving increased sputum production, lethargy, dyspnea, and fever. In previous smaller studies of CF subjects, a single nebulized dose of 6 -7% HS increased mucociliary clearance at 60 and 90 min, as measured by radionucleotide imaging (3,5). Increased mucociliary clearance was dose-dependent up to 12% HS in 10 adults treated with a single dose and tested at 90 min (4). In a 2-week study in 52 CF subjects, 10 ml of 6% HS improved lung function as measured by a 12% increase in forced expiratory volume in 1 s (FEV1) (6). HS has also been shown to increase mucociliary clearance in healthy subjects and in subjects with asthma or chronic obstructive lung disease (21,22).
Various non-salt hyperosmolar agents have been posited to improve mucus clearance in CF and other chronic inflammatory lung diseases. The best-studied non-salt hyperosmolar agent is mannitol. In 12 CF subjects, single dose administration of 300 mg of mannitol as a dry powder improved bronchial mucus clearance at 60 min, although there was a small decline in FEV1 immediately after inhalation that was unresponsive to bronchodilator premedication (5). The same dose of mannitol improved mucus clearance 2-fold in healthy, asthmatic, and bronchiectatic subjects (21,23). A 12-day trial of 400 mg of mannitol daily in nine non-CF bronchiectatic subjects gave no change in FEV1 but improved various sputum characteristics, such as surface tension and spinnability (24). Another inhaled hyperosmolar agent, xylitol, reduced ASL salt concentration in CF and non-CF airway epithelia in vitro (25). Pretreatment of human airway explants for 1 h with 60 -80 mg/ml xylitol reduced binding of a Burkholderia cepacia isolate (26).
As part of the present study, we characterized AQP expression and function in water transport across human bronchial epithelial cells. As discussed in the Introduction, a central prediction of the mechanism proposed by Donaldson et al. (1) is the inhibition of AQP water permeability by amiloride. To support their mechanism, the authors measured transepithelial flux across HBE cultures in response to an apically directed mannitol gradient, as we have done here ( Fig. 2A), with two differences. First, in their study, the concentration of Texas Red-dextran was measured in the basolateral (rather than the apical) solution. Second, in their study, fluorescence was measured by confocal microscopy without any buffer sampling.
They detected membrane transport rates as changes in cell height over time, with cytoplasmic calcein fluorescence imaged using reconstructed x-z confocal stacks. From both approaches, they reported a Ͼ70% reduction in osmotic water transport by 100 -400 M amiloride in non-CF and CF HBE cultures. Amiloride also prevented the transient increase in ASL volume upon acute addition of perfluorocarbondispersed NaCl to the mucosal surface of epithelial cultures, although this approach does not provide a direct measure of osmotic water permeability and was thus not replicated here. HgCl 2 , a toxic and nonspecific inhibitor of AQPs, dramatically inhibited this ASL volume response, although this finding appears at odds with their prior report of a much milder (ϳ30%) reduction of transepithelial and membrane water permeability by apically added HgCl 2 (8). Based on these data, the authors proposed that inhibition of airway AQPs by amiloride accounted for the clinical effect and supported the hypothesis of ASL dehydration in CF.
Here, using essentially the same airway epithelial cell culture model used by Donaldson et al., we found no effect of amiloride on transepithelial or plasma membrane osmotic water permeability. Our fluorescence-based methods were capable of detecting small water permeability differences of Ͻ10%. Also, we found no effect of amiloride on the water permeability of individual airway/lung AQPs in transfected epithelial cells or in AQP5-dependent airspace-capillary water permeability in intact perfused mouse lung. We conclude that amiloride does not inhibit osmotically driven water transport in the airways or lung, indicating the need to identify other mechanism(s) to explain the paradoxical clinical observations reported by Donaldson et al. The discrepancy between our findings and those of Donaldson et al. may be related to differences in measurement techniques, as discussed above. The techniques of Donaldson et al. (1) were first employed by Matsui et al. (8) in characterizing HBE water permeabilities. This earlier study reported in both non-CF and CF P1 cultures a plasma membrane P f (ϳ90 m/s) ϳ2-fold less than transepithelial P f (ϳ160 m/s), which is a physical impossibility for a tight epithelium, because the apical and basolateral membranes are barriers in series. Here, our results for P0 cultures are very different, a higher membrane P f (ϳ200 m/s), which is ϳ4-fold greater than transepithelial P f (ϳ50 m/s). For non-CF and CF P1 cultures, we found a lower transepithe- lial P f (ϳ25 m/s). Our results are consistent with the predicted lower transepithelial permeability of a multilayered cell culture.
Without confirmed inhibition of AQP function, it remains difficult to understand the reported deleterious clinical effect of amiloride. Several small studies, including that by Donaldson et al. (1), have shown that amiloride does not reverse the HSmediated acute increase in mucociliary clearance. For example, in 16 healthy subjects, a single dose of inhaled amiloride plus 7% HS did not increase or decrease mucociliary clearance compared with 7% HS alone (27). There is some evidence to suggest that inhaled amiloride alone, not paired with HS, may have a beneficial effect on mucociliary clearance in healthy and CF subjects (27,28). However, it is reported that amiloride should not remain in lung tissue long enough to have significant long term effects on lung function (29). Amiloride is known to inhibit Na ϩ /H ϩ exchange, so it is possible that the apparently negative effects of amiloride in the presence of a large hyperosmotic load are related to action at other site(s) on airway epithelial cells via alteration of cell pH regulation or ion balance.
Prior immunolocalization studies have reported the expression of AQP4 at the basolateral membrane surface in epithelial cells throughout large and small airways in mice, rats, and humans (reviewed in Ref. 30). AQP3 was found to have a more limited expression pattern in basolateral membranes in the basal cells of large and central airways and nasopharnyx. Several studies have shown the expression of AQP1 in microvascular endothelia through the airways/lung but not in epithelial cells (31,32). AQP5 is expressed strongly in type I alveolar epithelial cells in humans and rodents and likely in some airway epithelial cells in human large airways (32,33). In well differentiated primary cultures of human airway epithelial cells and in fixed human bronchi, we found here the expression of AQP3 and AQP4 in the basolateral membranes of airway epithelia but did not detect AQP5 protein expression. However, the lack of AQP5 immunostaining probably reflects a lack of antibody specificity rather than the absence of functional protein, given the demonstrated presence of AQP5 in nasal and bronchial epithelial cells in previous studies (33).
Prior functional studies have suggested the involvement of AQPs in airway water permeability. Our initial measurements in microperfused small airways from guinea pigs showed high transepithelial water permeability (P f ϳ 50 m/s) that was weakly temperature-sensitive and was mercury-insensitive, suggesting the involvement of AQP4 (7). Recent data from spheroid explants of human nasal polyps provided evidence for AQP5 function in airway epithelia. In these large, spherical airway cell monolayers, with the ciliated apical membrane facing outward, Pederson et al. (9) reported high membrane P f (ϳ150 m/s) that was sensitive to brief exposure to low concentration mercury. Here, our high P f values for HBE cultures and the weak temperature dependence of permeability suggest AQPmediated water movement (Fig. 2). Additionally, the reduced transepithelial osmotic water permeability under low pH in the basal (but not apical) membrane-facing solution provides indirect evidence for the function of AQP3 (the only pH-regulated airway AQP) in HBE basolateral membranes (19). The addition of HgCl 2 to the cultures used here resulted in marked leakiness with reduced transepithelial resistance, precluding conclusions about HgCl 2 effects. CF cultures had similar water permeability to non-CF cultures, consistent with our RT-PCR data (Fig. 3C) and with previous reports (8,9).
On theoretical grounds the functioning of airway AQPs is not expected to be a significant determinant of ASL volume, composition, or other properties. Because the lipid bilayer in plasma membranes has substantial water permeability, the presence of AQPs generally increases plasma membrane water permeability by only 5-10-fold, which is important for rapid fluid movement, such as in kidney or salivary gland, but not for the many orders of magnitude slower fluid movement in the lung (reviewed in Ref. 34). Indeed, despite 10-fold slowed osmotic water permeability in mouse lung by the deletion of AQP1 and AQP5, isosmolar fluid absorption was unaffected (20,35). Also, we reported previously that deletion of each of the airway/lung aquaporins (AQP1, AQP3, AQP4, and AQP5), individually and in combinations, did not affect ASL volume or ionic composition and had only a minimal effect of airway hydration at maximal respiratory rates (10).
The discussion above indicates the difficulty in ascribing aquaporin function, and increased ASL volume in general, to account for the effects of nebulized HS. Another potential problem with the hypothesis of increased ASL volume is that the excess fluid added to the intact airways/lung is likely to dissipate over minutes or tens of minutes rather than produce a sustained increased in ASL volume over hours between hypertonic saline treatments. Because of the large surface-to-volume ratio of the alveoli and airways and their high water permeability, added hypertonic fluid becomes nearly isotonic within 1 FIGURE 6. Airspace-capillary water permeability in perfused mouse lung. A, representative traces of pleural surface fluorescein isothiocyanate-dextran fluorescence from wild-type mouse lung without (top, left) and with (top, right) 500 M amiloride and from untreated lung of AQP5-null mouse (bottom). Pulmonary artery perfusate osmolality is indicated. B, summary of rates of capillary-airspace osmotic equilibration from experiments as described for A. (four mice/condition Ϯ S.E.). Difference was not significant. min due to osmotic water efflux into the airspaces. We showed rapid osmotic equilibration initially in sheep lung (36) and later in other mammals (18), including mouse as seen in Fig. 6. Because of rapid isosmolar fluid absorption in mammalian lung, including human lung (reviewed in Ref. 37), the ϳ40 ml of isosmolar fluid produced by nebulizing 5 ml of 7% HS is likely to be absorbed in Ͻ15-30 min. However, the exact clearance rate of HS would depend on its distribution in small versus large airways following nebulization. Rapid isosmolar fluid absorption has also been shown in the CF lung despite absent CFTR, because basal cAMP-independent lung fluid absorption does not involve CFTR (38). Thus, it is difficult to reconcile the clinical benefit of HS with the expected transient increase in ASL volume; however, the exact location of excess fluid, in terms of airways versus alveoli and periciliary versus mucus fluid, is not known. The increased ASL volume mechanism would also predict a better outcome for inhaling non-absorbed solutes, such as mannitol or xylitol, which from clinical studies is not the case (5).
Alternative mechanisms have been proposed to account for the apparent clinical benefit of inhaled hyperosmolar agents. One mechanism involves the direct effects of NaCl and mannitol on mucus rheology (5). HS has been proposed to reduce mucin entanglement by shielding fixed negative charges on the mucin backbone; mannitol has been proposed to compete with oligopolysaccharides on mucin macromolecules, thereby disrupting hydrogen bonds and reducing entanglements (39 -41). A proposed mechanism for the benefit of mannitol is based on the hypothesis that a relatively salty ASL in CF decreases innate immune defenses, such as lysozyme, lactoferrin, and ␤-defensins. If correct, a nonionic osmotic agent such as mannitol might be of benefit by reducing ASL salt concentration (5,25,42). Another proposed mechanism is the induction of cough in subjects inhaling hyperosmolar agents, which are quite irritating to the airways, although some studies conclude that changes in the mucociliary clearance are unlikely to be attributed to increased cough (3,4,22). Another proposed mechanism involves the direct or indirect action of hyperosmotic agents on mucus-secreting cells or sensory nerves in stimulating mucus secretion, whereby increased mucus secretion and ciliary action are proposed to "flush out" sticky, bacteria-laden secretions (5). We would add to this list the possibility of transient osmotic flow-induced convection of mucus into the airways as a major factor in the clinical benefit of HS. A critical re-examination of the mechanism(s) by which HS improves lung function in cystic fibrosis might lead to a better, mechanism-based approach for inhaled osmolar therapy.