Submucosal glands are the primary source of airway mucus, a critical component of lung innate defenses. Airway glands are defective in cystic fibrosis (CF), showing a complete absence of secretion to vasoactive intestinal peptide or forskolin, which increase intracellular cAMP concentration. This defect is attributed to gland serous cells, which express the cystic fibrosis transmembrane conductance regulator. Calu-3 cells, which mimic many features of serous cells, secrete Cl− and HCO3−, with HCO3− secretion predominating for forskolin stimulation and Cl− secretion predominating for stimuli that open basolateral K+ channels to hyperpolarize the cells. We used pH stat and ion substitution experiments to clarify the mechanisms and consequences of these two modes of secretion. We confirm that Calu-3 cells secrete primarily HCO3− in response to forskolin. Unexpectedly, HCO3− secretion continued in response to K+ channel openers, with Cl− secretion being added to it. Secretion of HCO3− from hyperpolarized cells occurs via the conversion of CO2 to HCO3− and is reduced by ∼50% with acetazolamide. A gap between the base equivalent current and short-circuit current was observed in all experiments and was traced to secretion of H+ via a ouabain-sensitive, K+-dependent process (possibly H+-K+-ATPase), which partially neutralized the secreted HCO3−. The conjoint secretion of HCO3− and H+ may help explain the puzzling finding that mucus secreted from normal and CF glands has the same acidic pH as does mucus from glands stimulated with forskolin or ACh. It may also help explain how human airway glands produce mucus that is hypotonic.
- mucus clearance
- carbonic anhydrase
submucosal glands are the primary source of airway mucus, a rich mixture of water, salts, mucins, and the “anti-” compounds: antimicrobials, antiproteases, and antioxidants. Airway mucus helps maintain sterility of the lungs, but in cystic fibrosis (CF), that function is severely compromised. CF results from loss-of-function mutations in CFTR, the gene for the cystic fibrosis transmembrane conductance regulator (CFTR), an ion channel that conducts Cl− and HCO3−. CFTR also influences the operations of many other proteins.
CFTR is expressed in gland serous cells (10), and its importance is demonstrated by the complete absence of vasoactive intestinal peptide (VIP)- or forskolin-mediated gland secretion in CF airways (17). The serous cell malfunction hypothesis proposes that CFTR-dependent Cl− and HCO3− secretion normally drives water movement across the serous cells of the gland, but in CF the loss of functional CFTR causes initial secretions to be underhydrated and lower in pH. We hypothesized that the resulting mucus is thicker and that there is reduced bioavailability of serous cell antimicrobials. Hence airway innate defenses, which depend on both mucus clearance and a chemical shield, are rendered inadequate (36). Tests of this hypothesis have been hampered by difficulties in studying the molecular mechanisms of gland secretion in situ.
The Calu-3 cell line was introduced as a serous cell model based on evidence that it expressed the same range of proteins as natural serous cells, including CFTR (11), and formed functional, polarized monolayers that secreted anions in response to elevations of intracellular cAMP concentration ([cAMP]i) and intracellular Ca2+ concentration ([Ca2+]i) (32). Calu-3 cells have since been studied extensively, and a great deal is now known about how they function.
A comprehensive model of Calu-3 anion secretion (Fig. 1, based on Ref. 8) has been proposed in which elevation of [cAMP]i produces only HCO3−-mediated secretion, whereas elevation of [Ca2+]i [or direct activation of K+ channels with 1-ethyl-2-benzimidazolinone (1-EBIO)] produces only Cl−-mediated secretion. This model (hereafter termed model 1) leads to several predictions about secretions from Calu-3 cells and, by extension, from airway glands. Model 1 predicts that CF gland secretions should be more acidic than normal, and gland mucus produced by elevations of [Ca2+]i should be much more acidic than mucus produced by elevation of [cAMP]i. However, the pH of pig gland mucus is ∼7.0 when stimulated via [Ca2+]i-elevating (acetylcholine, ACh) or [cAMP]i-elevating (VIP/forskolin) pathways (18), and pH is also ∼7.0 in mucus from ACh-stimulated glands of either CF or normal human subjects (16). To explain these results, it has been hypothesized that secondary processes in the gland modify secretions homeostatically as they pass from the serous acini along the mucous tubules and ducts, so that the pH of the final secretion is adjusted to near pH 7 regardless of the starting secretion (36).
What about the initial serous cell secretions? Here, our results agree with the model qualitatively, but not quantitatively. During preliminary tests of the pH of Calu-3 secretions produced by different stimuli (Krouse ME, unpublished observations), we did not see the extreme pH values expected by a switch between 150 and 0 mM HCO3− concentrations as predicted by model 1 (Fig. 1, Ref. 8). Also, Lee et al. (20) used isotopic flux measurements to determine that elevation of [Ca2+]i in Calu-3 cells resulted in no increased Na+ absorption, whereas Cl− secretion accounted for only a portion of the short-circuit current (Isc), and attributed the remaining component to HCO3− secretion (20). That interpretation is consistent with the effect of bumetanide, which spares a large component of thapsigargin-stimulated Isc (20, 26). To investigate these discrepancies, we have reinvestigated Calu-3 secretion using pH stat methods to determine whether additional mechanisms might account for our observations.
The results of these studies have led us to a modified (and more complex) model of Calu-3 secretion. In contrast to model 1, we found that HCO3− is secreted during elevation of [Ca2+]i or in the presence of the direct K+ channel activator 1-EBIO despite decreased driving force on the Na+-HCO3− (NBC) cotransporter. We provide evidence that HCO3− is accumulated above its electrochemical equilibrium by carbonic anhydrase-catalyzed conversion of CO2 and H2O. More surprisingly, the expected alkalinization of the apical fluid during stimulation with 1-EBIO is partly masked by increased secretion of H+, which we trace to the activity of an apical process that is inhibited by apical ouabain and by removal of apical K+, suggesting an H+-K+-ATPase. Our evidence supports prior evidence that both HCO3− and Cl− secretion occur through the same apical conductance: CFTR.
We studied Calu-3 monolayers in Ussing chambers using two paradigms. Short-circuit (Isc) experiments, where the transcellular voltage is clamped to zero, have the advantage of maximizing anion transport. To obtain a separate measure of base (HCO3−) secretion, we also used the pH stat method in conjunction with Isc measurements.
The standard solution was Krebs-Henseleit (K-H) solution, which contains (in mM): NaCl, 128; KCl, 4.6; CaCl2, 2.5; MgSO4, 1.2; KH2PO4, 1.2; NaHCO3, 25; and glucose, 11.2. The K-H solution contains both Cl− and HCO3− anions. In Cl−-free solution, NaCl was replaced with Na-isethionate, all other Cl− salts were replaced with nitrate salts, and 3 M KCl bridges were replaced by 1 M Na-isethionate bridges. For HCO3−-free solution, NaHCO3 was replaced with Na-HEPES and bubbled with O2. For standard pH stat experiments, the apical solution was as follows (in mM): NaCl, 150; KCl, 5; CaCl2, 2.5; and MgSO4,1.2. For pH stat experiments with K+-free apical solutions, the K+ salts were replaced with Na+ salts, and 1 M NaCl was replaced with 3 M KCl in the agar bridges. All solutions were made up at 80% volume, and distilled water was added until the osmolarity was 320 mosM, as measured on a vapor pressure osmometer (Wescor 5500; Wescor, Logan, UT). The pH of each solution was adjusted to be 7.4 at 37°C when bubbled with the appropriate gas.
Monolayers of Calu-3 cells were grown on human placental collagen (HPC; Sigma, St. Louis, MO)-coated Snapwell inserts (1.13-cm2 area) for 4 wk (28 ± 12 days). The cells were fed from the basolateral side only and grown with an air interface.
Isc and pH Stat
3−-containing solutions were bubbled with 95% O2-5% CO2, and no-HCO3− and pH stat solutions were bubbled with 100% O2. The Ussing chambers were kept at 37°C with a temperature-controlled circulator (VWR Scientific, West Chester, PA). The voltage offset/junction potential and fluid resistance were measured and subtracted electronically before each experiment. Snapwells were mounted and allowed to reach a steady state (∼30 min) before the experiment was begun. A voltage pulse was passed across the monolayer every 20 s to measure monolayer resistance.
The pH stat method estimates the net acid or base secretion across a monolayer by measuring the amount of acid or base that must be added to the unbuffered apical solution to clamp its pH to 7.4. In our experiments, the unbuffered apical solution always alkalinized to varying degrees, so we added appropriate volumes of 3 mM acid (HCl or HNO3 in Cl−-free solutions) manually to keep the apical pH within the range 7.37–7.43. Net base secretion was then converted into an equivalent current (Ieq) on the assumption that one net unit of base carried one net negative charge. Figure 2 shows two raw pH traces from a Calu-3 monolayer that had been pretreated with bumetanide, an inhibitor of the Na+-K+-2Cl− cotransporter (NKCC). Acid was added near the start of the trace, and the subsequent slow alkalinization of the apical solution (dark line) was taken to reflect a low net level of secretion of base, which was assumed to be HCO3−. Stimulation with 1-EBIO caused the solution to alkalinize more quickly (dotted line), as seen by the increased slope of alkalinization and the increased rate of acid addition. This increase in acid addition rate is seen as an increase in Ieq.
For pH stat experiments, there can be no buffer in the apical solution, especially HCO3−. Therefore, we used K-H solution where all salts were Cl− salts and no buffers were present (NaHCO3 replaced by NaCl). The nominally unbuffered apical solution was further conditioned by bubbling with pure O2 for ∼1 h before the experiment to remove traces of CO2 and HCO3− introduced by the mounting of the monolayer. The basolateral solution is standard K-H buffer bubbled with 95% O2-5% CO2.
The unbuffered apical solution in pH stat experiments creates gradients for HCO3− and CO2 across the monolayer. We have preliminary evidence that the paracellular pathway is equally permeable to Na+ and Cl− and minimally permeable to HCO3− (data not shown), which may explain why we observed little Isc offset in pH stat experiments. By convention, a positive or upward deflection in Isc indicates anion secretion or cation absorption and in this system is known to include both Cl− and HCO3− secretion and Na+ absorption via Na+-glucose cotransport (8, 20, 34). Amiloride-sensitive Na+ current was not observed (32). We eliminated Na+ absorption with apical phlorizin or removal of apical glucose. The percent change produced by inhibitors was corrected for a ouabain-sensitive pump current (usually ∼5 μA/cm2) present at the end of each experiment.
The stocks for thapsigargin (applied at 300 nM to both sides of the epithelium to inhibit the endoplasmic reticulum Ca2+-ATPase and thus elevate [Ca2+]i ), acetazolamide (100 μM), 4,4′-dinitrostilbene-2,2′-disulfonic acid (DNDS; 3 mM applied basolaterally to block the Na+-HCO3− cotransporter; Ref. 24), and ouabain (10 μM to block K+-ATPases) were made in water at ×1,000 (or ×100 for DNDS). The stocks for 1-EBIO (applied at 1 mM to both sides to open Ca2+-activated K+ channels directly; Ref. 8) and bumetanide (∼10 μM applied basolaterally to block the NKCC) were made in DMSO at ×1,000. Forskolin (10 μM applied basolaterally to activate adenylate cyclase and hence CFTR) was made at ×1,000 in ethanol.
Statistical significance was determined by use of the Student's t-test on Microsoft Excel, with either paired or unpaired tests as appropriate.
Ussing Chamber Experiments in Normal K-H (Cl− + HCO3−) Solutions
Figure 3A illustrates the main features of Calu-3 Isc that we are attempting to understand. In the absence of any apparent stimulation, Calu-3 monolayers develop a basal Isc that varies widely in different laboratories and sometimes within laboratories at different times, with mean values in the literature from 13 to 35 μA/cm2 (6, 8, 23). Mean basal Isc values in our experiments clustered in two groups of 56.3 ± 7.4 μA/cm2 (n = 11, high basal current) and 21.2 ± 2 μA/cm2 (n = 28, low basal current) when tested with phlorizin or zero apical glucose to eliminate the Na+-glucose cotransporter (34). In cells with high basal Isc, forskolin had a minimal effect (for example, Fig. 3A), showing on average a 6% increase to 58.2 ± 10.1 μA/cm2 (n = 23), whereas in cells with low basal Isc, forskolin caused a 100% increase in Isc to 43.6 ± 3.8 μA/cm2 (n = 17). Thus forskolin stimulation eliminated the significant difference between the groups for Isc (P > 0.05), resistance (250 ± 40 for high basal vs. 350 ± 50 Ω·cm2 for low basal, P > 0.05), and the proportion of Isc eliminated by bumetanide (20 ± 4% for high basal, n = 14, vs. 12 ± 3% for low basal, n = 10). We did not determine the source(s) of variation in basal Isc, but these results suggest that Calu-3 cell sheets with high basal Isc resemble those that have been forskolin stimulated, both having an Isc of ∼50 μA/cm2.
Stimulation with either thapsigargin or 1-EBIO, each of which results in open basolateral K+ channels (2, 7), should increase the driving force for anion secretion and produce a proportional increase in Isc. That was observed, with the responses peaking and then falling to a plateau (Fig. 4). Plateau increases were +177 ± 72% (n = 4) for high basal and +156 ± 61% (n = 7) for low basal (Table 1), with corresponding decreases in transepithelial resistance (∼190 Ω·cm2 for both high and low basal). Addition of bumetanide just after the peak Isc caused an immediate inhibition of most of the stimulated Isc (Fig. 3A; Table 1). However, note in Fig. 3A that the Isc after thapsigargin stimulation and bumetanide inhibition was larger than before stimulation. That was observed consistently; the ΔIsc to K+ channel openers in the presence of bumetanide was 11 ± 3 μA/cm2 for high basal (n = 12) and 12 ± 4 μA/cm2 for low basal (n = 17). It will be shown below that this ΔIsc represents increased HCO3− secretion.
To determine the basis of the bumetanide-insensitive Isc, acetazolamide was added to inhibit carbonic anhydrase, and DNDS was added to inhibit the basolateral NBC. Before 1-EBIO stimulation (but after bumetanide inhibition), DNDS (3 mM) inhibited 12.2 ± 2.5 μA/cm2 of the Isc (high basal, n = 10, Table 1), and after 1-EBIO stimulation, DNDS inhibited less of the Isc (7.8 ± 2.2 μA/cm2, high basal, n = 10). Under the same conditions (high basal cells), acetazolamide inhibited significantly more Isc before than after stimulation (16.2 ± 3.5 μA/cm2, n = 8, Table 1, vs. 30.5 ± 2.7 μA/cm2, P < 0.05,n = 11).
The contribution of DNDS- and acetazolamide-sensitive transport to the bumetanide-insensitive Isc varied with stimulation (Fig. 3B). After forskolin, similar proportions of the Isc were inhibited by DNDS (20%) or acetazolamide (27%), but after 1-EBIO, there was a large disparity in the proportion of the Isc inhibited by DNDS (10%) vs. acetazolamide (37%). The DNDS-sensitive component of the Isc decreased after 1-EBIO stimulation (P = 0.03), and the acetazolamide-sensitive component of the Isc increased after 1-EBIO stimulation.
These results suggest four provisional conclusions. 1) Elevation of [Ca2+]i or stimulation with 1-EBIO stimulates a large increase in Cl− secretion and a small increase in HCO3− secretion. 2) Most of the HCO3−-dependent secretion after stimulation is mediated by carbonic anhydrase-generated HCO3−. 3) The contribution of NBC is relatively minor. 4) After 1-EBIO stimulation, NBC is not working in reverse, because if it were DNDS should have increased the Isc.
To determine whether prior treatment with bumetanide influenced the DNDS results, we compared the inhibitory effect of DNDS on the 1-EBIO-stimulated Isc in paired experiments with and without bumetanide (low basal cells). DNDS alone inhibited 14.6 ± 1.0% of 1-EBIO-stimulated Isc, and this was reduced to 9.1 ± 0.5% after bumetanide (n = 10, P < 0.001, Fig. 3B). The significant reduction in DNDS inhibition after bumetanide inhibition of NKCC might arise if the diminished intracellular Cl− results in hyperpolarization of the cells, which would then reduce the transport of HCO3− by the NBC (see discussion).
A substantial Isc (32% of forskolin-stimulated Isc, Table 1) remained after combined treatment with bumetanide, DNDS, and acetazolamide. Evidence to be presented below suggests that this Isc includes a component of HCO3− secretion derived from the uncatalyzed conversion of CO2 + H2O to HCO3−, which still proceeds relatively rapidly in the absence of carbonic anhydrase, with a half-time of ∼5 s (12).
Analysis of Isc with pH Stat Experiments
Measurement of Isc under pH stat conditions is designed to maximize ionic currents while providing an additional measure of acid/base movement, and no other method can provide the same kind of information. However, pH stat results have been questioned because the technique creates an artificially large HCO3− gradient across the cells and across the paracellular space. If any of the drugs caused a changed in the paracellular pathway's HCO3− conductance, this might lead to a noncellular response. To estimate the magnitude of any artifact that might have arisen from the HCO3− gradient, we measured the Isc in paired Ussing (apical and basolateral solutions clamped to 25 mM HCO3−) and pH stat experiments (zero apical HCO3−). Any change of HCO3− movement across the paracellular pathway will produce an increase or decrease in Isc. For each condition tested (7 different conditions), there was no significant difference between the Isc measured with the Ussing technique and that with the pH stat technique (data not shown). For example, the maximum HCO3− current after 1-EBIO and bumetanide was 41 ± 3 μA/cm2 in the pH stat experiments and 52 ± 6 μA/cm2 in conventional Ussing experiments (n = 11, P > 0.08). As a second example, we can compare the magnitude of HCO3− current in each condition that was blocked by acetazolamide + DNDS. These agents blocked 26 ± 2 μA/cm2 of the HCO3− Isc in pH stat experiments and 32 ± 3 μA/cm2 in conventional Ussing chamber experiments (n = 6, P > 0.1).
Neither of these measures showed a significant increase in HCO3− secretion during the pH stat method; in fact, the trend was toward a smaller Isc when the apical solution lacked HCO3−. This independence from the HCO3− gradient can be explained if there is minimal movement of HCO3− across the tight junctions, and if the limiting factor for HCO3− Isc is not dependent on either the apical HCO3− conductance or on gradient-sensitive mechanisms for HCO3− accumulation (e.g., NBC). Instead, the rate-limiting step in maximally stimulated cells appears to be the gradient-insensitive and voltage-insensitive production of HCO3− from CO2. The CO2 inside the Calu-3 cells in pH stat experiments is likely to be <5%, because the basolateral CO2 is 5% and the apical CO2 is 0%. Because CO2 and HCO3− are in equilibrium inside the cell (assuming constant intracellular pH), the lower CO2 level leads to a lower HCO3− level and less HCO3− current.
The power of the pH stat technique is that it provides an estimate of the net base (HCO3−) movement across the monolayer. The assumption underlying these experiments was that Calu-3 cell Isc is the sum of NKCC-mediated Cl− secretion plus HCO3− secretion, but it quickly became apparent that the assumption was not valid. Instead, an unidentified process was producing a discrepancy between Isc and Ieq measurements.
Forskolin-stimulated monolayers in the pH stat experiments had an Isc of 65 ± 5 μA/cm2 (n = 15, high basal) or 44 ± 2 μA/cm2 (n = 15, low basal), and the Ieq was approximately two-thirds of the Isc in both high and low basal conditions (Figs. 4 and 5). Bumetanide blocked 18 ± 3% (n = 11) of the Isc and none of the Ieq, leaving 13% of the forskolin-stimulated Isc unexplained (Fig. 5). Stimulation with 1-EBIO after forskolin produced a further large increase in Isc (110 ± 21%, n = 6), but, surprisingly, the Ieq remained stable (increase of 2.5 ± 5%, n = 6, Fig. 4). We had expected an increase in Ieq on the basis of the evidence just presented that 1-EBIO increased the bumetanide-insensitive and acetazolamide-sensitive Isc, consistent with increased driving force on carbonic anhydrase-produced HCO3−. On the other hand, if HCO3− secretion depended entirely on an NBC with 2:1 stoichiometry, it should have been eliminated under these conditions (8). Neither expectation was met.
Furthermore, in the presence of bumetanide, 1-EBIO produced a significant increase in Isc but not in Ieq. This is shown most strikingly in Fig. 5, where, in the presence of bumetanide, 1-EBIO stimulates a 20 μA/cm2 increase in Isc but a slight decrease in Ieq. On average, in the presence of bumetanide 1-EBIO increased the Isc by 27 ± 9% (n = 13) and the Ieq by 11 ± 7% (n = 9, not significantly different from 0), further increasing the discrepancy between Isc and Ieq so that Ieq was on average only 70 ± 4% of Isc (n = 9, Figs. 4 and 5).
Bovine tracheal epithelial cells (28), Calu-3 cells (14), and cultured human bronchial epithelium (4) each express H+-K+-ATPase, and, at least in the human cells, expression in the apical membrane is only of the nongastric form. These results suggest the possibility that some of the HCO3−-based Ieq was being neutralized by K+-dependent H+ secretion. To test this idea, ouabain (10–20 μM), a nonspecific H+-K+-ATPase inhibitor, was added apically after 1-EBIO stimulation and bumetanide inhibition (Fig. 5). Apical ouabain had no effect on the Isc (n = 9), showing that it did not have access to the basolateral bath but instead caused a dramatic increase in Ieq from 70 ± 4 to 95 ± 1% of the Isc (n = 9, P < 0.0001, paired; see Figs. 5 and 6B).
As an additional test of a role for a putative H+-K+-ATPase, we omitted K+ from the apical solution. With a nominally K+-free apical solution, the Ieq and the Isc after 1-EBIO in the presence of bumetanide were equivalent (94 ± 4%, n = 6, P < 0.0001 vs. control), and apical ouabain had no further effect on Ieq (n = 6, Fig. 6A). Similarly, when cells were stimulated with 1-EBIO in the presence of apical ouabain and bumetanide, the Isc and Ieq rose together, and Ieq was 100 ± 4% of Isc (n = 9, P < 0.001 vs. control, Fig. 6B). These results indicate that a ouabain-sensitive, K+-dependent nonelectrogenic process of acid secretion is stimulated in Calu-3 cells by 1-EBIO in parallel with increased Cl− and HCO3− secretion. One possible mechanism for such secretion is the nongastric form of H+-K+-ATPase, which others have shown to be expressed in Calu-3 cells (14, 28).
To determine whether ouabain-sensitive H+ secretion was also stimulated by forskolin, ouabain was added apically during forskolin-stimulated Isc but before 1-EBIO addition. Ouabain did not alter either Isc or Ieq in forskolin-stimulated cells (n = 9, Fig. 6B), suggesting that H+-K+-ATPase is not activated by forskolin stimulation.
Ussing Chamber Experiments in HCO3−-Free Media (Cl− as the Only Permeant Anion)
The results so far indicate that HCO3− is secreted by Calu-3 cells under all of the conditions we tested and is derived to a large extent from conversion of CO2 to HCO3−, because Isc is inhibited by acetazolamide. However, acetazolamide might have other effects including inhibition of NKCC. To study the NKCC transporter in relative isolation, we performed Ussing chamber experiments on Calu-3 cells in HEPES-buffered HCO3−-free media, where Cl− is the only CFTR-permeant anion (Table 2).
Figure 7 shows a typical experiment under these conditions. In unstimulated (high basal) or forskolin-prestimulated cells, the initial Isc was 45 ± 4 μA/cm2 if glucose was present in the apical solution and was decreased to 35 ± 5 μA/cm2 by addition of phlorizin (Table 2). 1-EBIO caused a large sustained increase in Isc of 64 ± 4 μA/cm2 (n = 4) (Fig. 7 and Table 2). Thapsigargin increased the Isc by 35 ± 4 μA/cm2 (n = 5). Bumetanide inhibited ∼90% of the unstimulated Isc (n = 12, Table 2) and 100% of the Isc stimulated by either 1-EBIO or thapsigargin (n = 7). Thus NKCC is the major pathway for transporting Cl− into the cell under all the conditions we tested.
Acetazolamide and DNDS had negligible effects on Isc (Fig. 7 and Table 2). The lack of inhibition by these agents is an important result because it indicates that their ability to inhibit Isc when HCO3− is present (34) (Fig. 2 and Table 1) is not a nonspecific effect. This further supports a role for CO2 conversion as well as basolateral HCO3− uptake in HCO3− secretion (20).
The Isc properties of Calu-3 cells when Cl− is the only CFTR-permeant anion resemble the properties of T84 cells in normal media, where inhibition of NKCC with bumetanide eliminates essentially all Isc, as reported in a long series of experiments by Dharmsathaphorn et al. (9) and others (3).
Isc in Cl−-Free Solutions (HCO3− is the Only Permeant Anion)
In these experiments, we attempted to isolate the pathways responsible for HCO3− secretion and further assess the relative contributions of CO2 conversion and Na+-HCO3− cotransport.
When all Cl− in the bathing solutions was replaced with isethionate, leaving 25 mM HCO3− as the only CFTR-permeant anion, a high basal current of 43 ± 3 μA/cm2 remained (Fig. 8 and Table 3). Apical phlorizin inhibited ∼25% of this basal Isc, indicating that the electrogenic Na+-glucose cotransporter was highly active under these conditions (Table 3). A possible reason for the increased activity of Na+-glucose cotransport is that the driving force for Na+ entry is increased because the cells are hyperpolarized, and intracellular Na+ concentration ([Na+]i) is lowered. Removal of extracellular Cl− will drive intracellular Cl− concentration ([Cl−]i) to near zero and should inactivate NKCC, which should lower [Na+]i. With HCO3− and K+ as the main permeant ions, and a relatively low amount of cytosolic HCO3− relative to intracellular K+ concentration ([K+]i), the [K+]i gradient will dominate the membrane potential and move it toward the K+ reversal potential (EK).
However, evidence that the cells are still far from EK is provided by the response to thapsigargin or 1-EBIO. These agents stimulated increases in Isc of 107% (n = 4) and 163% (n = 5), respectively, and corresponding large increases in transepithelial conductance, indicating that, although the membrane potential may have been hyperpolarized by removal of Cl−, it was still depolarized relative to EK before stimulation (Fig. 8 and Table 3).
With HCO3− as the only CFTR-permeant anion, bumetanide was without effect, as expected (Fig. 8, Table 3), whereas acetazolamide inhibited 26 ± 3% of the basal Isc (n = 9), and DNDS inhibited 25 ± 4% (n = 9, Table 3). After stimulation with 1-EBIO, the acetazolamide inhibition increased to 45 ± 6% (significant increase, P = 0.005), and the DNDS inhibition decreased to 11 ± 8% (decrease not significant). The substantial Isc that remained after inhibition with bumetanide, acetazolamide, and DNDS (30 ± 6%, n = 6) of stimulated Isc could represent Na+ absorption via unknown pathways or continued secretion of HCO3− from unknown sources. To decide among these possibilities, glibenclamide and N-phenylanthranilic acid (DPAC), which are frequently used to inhibit CFTR (but have other actions), were tested for their ability to inhibit Isc in Cl−-free media. Each produced nearly complete inhibition of Isc and increased the monolayer resistance (n = 8), implying that all the Isc was current through CFTR.
The results in zero Cl− make three points. 1) Because DNDS inhibited the Isc, the NBC was still causing an influx of HCO3− under these conditions. This could be because [Na+]i was reduced secondary to NKCC inactivity, but the magnitude of DNDS inhibition did not differ from that observed in K-H solution. 2) The large effect of acetazolamide emphasizes the importance of CO2 conversion in HCO3− secretion. 3) The substantial Isc that remains after both pathways for generating intracellular HCO3− concentration are inhibited probably represents continued secretion of HCO3− produced by the uncatalyzed conversion of CO2 to HCO3−, rather than Na+ absorption, because the residual Isc was blocked by glibenclamide and DPAC.
Evidence for a Single Exit Pathway for HCO3− and Cl−
CFTR conducts both Cl− and HCO3− (15, 21), and glibenclamide blocks Isc carried by either ion (30, 33). Although glibenclamide is not specific for CFTR (e.g., see Refs. 1, 22), its effect on CFTR might be sufficient to test the hypothesis that both Cl− and HCO3− exit Calu-3 cells via CFTR. Glibenclamide blocked the Isc with identical dose-inhibition curves (IC50 of 120 μM) in K-H solution or when Cl− or HCO3− was the only permeant anion (Fig. 9). The conductance dose-inhibition curves mirrored the Isc dose-inhibition curves in all three solutions. The Hill slope of these curves was 1, suggesting that one molecule of glibenclamide blocks the relevant channel. (Because the effect of glibenclamide on intracellular ATP concentration occurs at concentrations >1 mM, this action should not affect a process with an IC50 of 120 μM.)
We conclude that the most likely effect of glibenclamide in these experiments is blockade of a single type of channel that conducts both Cl− and HCO3−. Presumably, that conductance is CFTR. Because glibenclamide blocks >90% of the Isc, there is no need to propose additional apical Cl− or HCO3− conductances or electrogenic transporters. However, these results cannot eliminate electrically silent transport processes, such as an apical anion exchanger, because an electroneutral HCO3− exchanger produces no Isc.
Estimation of HCO3−/Cl− Conductance
Direct comparisons of the Isc in Cl−-only and HCO3− solutions cannot be made because the driving forces in each condition are not known. However, a comparison is possible if we make the simplifying assumption that, at the peak of the 1-EBIO response, the cell voltage goes to EK. In HCO3−-only solutions, the initial cell voltage is determine by the HCO3− reversal potential (EB), the HCO3− conductance (GB), the EK, and the initial K+ conductance (GK). The fold increase in the driving force after 1-EBIO stimulation is proportional to (GB)/(GK). Thus the ratio of the peak fold increase stimulated by 1-EBIO in HCO3−-only solutions and Cl−-only solutions should reflect the HCO3−-to-Cl− permeability ratio (Fig. 10, A and B).
The peak fold increase in the HCO3−-only solution was 2 (Fig. 10C, open bar ) vs. a peak fold increase of 7 in the Cl−-only solution (Fig. 10C, solid bar). [The peak fold increase in K-H solution that contains both HCO3− and Cl− current was intermediate (Fig. 10C, gray bar), consistent with 1-EBIO stimulation causing both Cl− and HCO3− efflux through CFTR.] The ratio of the fold increases was 2/7, giving a HCO3− conductance that is 29% of the Cl− conductance, a value similar to estimates for CFTR arrived at by other methods (15, 21, 27). It is possible that the cell voltage does not actually reach EK because of the apical anion conductance through CFTR, and this effect will be greatest in Cl−-only solutions. Thus 29% of Cl− conductance is an upper limit for HCO3− conductance through CFTR.
Calu-3 cells secrete a HCO3−-rich fluid when stimulated with agents that elevate [cAMP]i and a Cl−-rich fluid when stimulated with agents that open basolateral K+ channels (8). The main new result from this paper is that K+ channel openers also stimulate HCO3− secretion, but the effect is masked by concomitant stimulation of H+ secretion via an apical mechanism that has the properties of a nongastric H+-K+-ATPase. A similar mechanism has been proposed by Furukawa et al. (13), where H+ secretion via Na+/H+ exchanger-3 masks some of the HCO3− secretion by CFTR in rat intestine. It is not yet known whether serous cells in submucosal glands, for which Calu-3 cells are models, possess a similar apical mechanism.
Evaluation of Model 1
We began these studies to examine perceived inconsistencies between model 1 (Fig. 1) and results from experiments in our laboratory. We considered three features of the model. 1) Is most HCO3− derived from the activity of an NBC? 2) Does the NBC have a stoichiometry that causes it to be reversed by cellular hyperpolarization, i.e., does it export HCO3− basolaterally after stimulation? 3) As predicted by these features of the model, does activation of basolateral K+ channels with 1-EBIO cause Calu-3 cells to secrete a highly acidic fluid, essentially devoid of HCO3−, with Cl− as the only anion (8)? Our results lead us to answer “no” to all three questions.
We found that most HCO3− was not derived from NBC-mediated transport but is instead generated by a carbonic anhydrase-dependent mechanism. In our experiments, the DNDS-sensitive pathway contributed ∼20% of forskolin-stimulated HCO3− secretion and ∼10% of 1-EBIO-stimulated HCO3− secretion (Fig 3B). Most of the HCO3− current was acetazolamide sensitive, suggesting that conversion of CO2 to HCO3− is the predominant mechanism for generating intracellular HCO3− concentration (Fig. 3B). An acetazolamide-insensitive, DNDS-insensitive source for HCO3− was also found, which could represent uncatalyzed conversion of CO2 to HCO3−, a reaction that proceeds rapidly with a half-time of only ∼5 s (12). The unidentified source for HCO3− may also include a contribution from anion exchanger-2, which exists in the basolateral membrane of Calu-3 cells (23). Because none of these pathways are voltage sensitive, they are well suited to mediating HCO3− transport in hyperpolarized cells.
We found that the NBC was not reversed by cellular hyperpolarization, because DNDS continued to inhibit Isc after 1-EBIO stimulation. DNDS inhibition was significantly less after 1-EBIO. These results are consistent with recent experiments on two different NBC [rkNBC (31) and NBC4 (35)] that found that both transport HCO3− 2:1 over Na+. In both cases, the predicted reversal potential for the NBC was more negative than −100 mV in normal Krebs ringer. If every cation in the cytosol were K+ (∼150 mM [K+]i), the potential with a 5 mM extracellular K+ concentration can only reach a maximum negativity (Vm = EK) of −90 mV [log (5/150) × 61 mV], which is not enough to reverse an NBC with 2:1 stoichiometry but will reverse an NBC with a 3:1 stoichiometry. Because reversal was not seen, we propose that the NBC in Calu-3 cells has a stoichiometry not greater than 2:1.
Opening of K+ channels with 1-EBIO did not produce HCO3− -free secretion; instead, it increased HCO3− secretion. The result is a primary secretion consisting of Cl−, HCO3−, Na+, and K+, with the K+ then exchanged for H+ and the H+ then capable of combining with HCO3− to form CO2 and water. We did not attempt to measure K+ secretion, but evidence favors an apical K+ channel in Calu-3 cells (6), and theoretical considerations suggest that optimal secretion requires 10–20% of the K+ conductance to be apical (5). We have shown that when K+ is present in the apical solution, it is exchanged for H+, and, under the conditions of our pH stat experiments, it neutralizes part of the HCO3−.
In summary, our present results for hyperpolarized Calu-3 cells supply a mechanism to reconcile the increased HCO3− secretion measured by Lee et al. (20) with the lack of HCO3− secretion proposed by Devor et al. (8). As shown in Figs. 5–6, stimulation with 1-EBIO or thapsigargin in the presence of bumetanide increases Isc by increasing both HCO3− secretion and H+ secretion. H+ secretion was blocked either with apical ouabain or elimination of apical K+, suggesting that H+-K+-ATPase is responsible. An apical H+-K+-ATPase has been reported in Calu-3 cells (14) and in human surface airway epithelium (4). In the human airway, the activity of H+-K+-ATPase does not appear to be regulated (4). Furthermore, its sensitivity to ouabain (500 μM was used) may be less than we observed; the inhibitory curve for the nongastric H+-K+-ATPase protein expressed from ATP1AL1/βH-K (25) can accommodate both our observations and those in surface airway epithelium. Thus the ouabain-sensitive mechanism described here could also be a variant nongastric H+-K+-ATPase or yet another K+-dependent mechanism.
Role of CFTR
Model 1 proposes that both Cl− and HCO3− are conducted through CFTR, with no provision made for apical anion exchangers. Our results strongly support that aspect of model 1: either both anions are conducted through CFTR, or their secretion is mediated by a transporter that is completely dependent on CFTR. We show two CFTR channels in Fig. 1 for clarity and also because of evidence that Cl− and HCO3− conductances mediated by CFTR can be independently controlled (29).
Implications for Airway Submucosal Gland Mucus Secretion
The gland serous cells that are modeled by Calu-3 cells contribute ∼50% of the secretory apparatus of glands, with the other 50% consisting of mucous cells that do not express CFTR. Secretions from both types of cells pass through a ductal system comprising at least two cell types, which presumably modifies the secretions in as-yet-unknown ways. We noted in the introduction that predictions from model 1 do not fit with observations of secretions from actual glands, which invariably have less than one-half as much HCO3− as the bath (mucus pH ∼7.0 vs. bath pH of 7.4) under all stimulation conditions, across species and in both CF and control subjects (16, 18). As a further paradox, secretions from human airway glands are hypotonic (16, 19). To explain these results, we originally hypothesized that cells in the gland (mucous cells?) secrete HCO3− via a non-CFTR-dependent pathway and that secondary processes in the gland modified the secretions homeostatically so that the pH of the final secretion is adjusted to near pH 7, regardless of the starting secretion. We had no explanation for how gland secretions might become hypotonic.
A New Role for HCO3−: the Ion Subtraction Hypothesis
The discovery of a regulated apical pathway for H+ secretion (H+-K+-ATPase) provides one possible mechanism for adjusting the pH of the submucosal gland secretions. Furthermore, concordant H+ and HCO3− secretion provides a potential explanation for 1) pH regulation, 2) hypotonicity of gland secretions, and 3) the discordance between expected high levels of HCO3− secretion from gland serous cells and the observation that the final secretions from glands are essentially neutral. Unlike in the large volumes of an Ussing chamber, the secretions of gland cells completely determine the composition of their luminal fluid. The importance of controlling pH and osmolarity is likely to be crucial given the abundant quantities of diverse macromolecules that are also secreted by the gland cells. If the two-step conversion of K+-HCO3− to H+-HCO3− and then to CO2 + H2O that has been demonstrated here can be shown to apply to intact glands, it will provide a new control mechanism for both the tonicity and pH of airway surface liquid. It is apparent that this mechanism also alters the expected consequences of lost CFTR function.
This study was funded by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-51817-06 (to J. J. Wine) and Cystic Fibrosis Foundation Grants KROUSE04GO (to M. E. Krouse) and WINE03PO (to J. J. Wine).
We thank Kim Winges and Dennis Lee for technical assistance and William Reenstra for discussions of uncatalyzed interconversion of CO2 and H2O to HCO3−.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 the American Physiological Society