Most bicarbonate secretion by Calu‐3 cells is mediated by CFTR and independent of pendrin

Abstract Bicarbonate plays an important role in airway host defense, however, its transport mechanisms remain uncertain. Here we examined the relative contributions of the anion channel CFTR (cystic fibrosis transmembrane conductance regulator, ABCC7) and the anion exchanger pendrin (SLC26A4) to HCO 3 − secretion by the human airway cell line Calu‐3. Pendrin and CFTR were both detected in parental Calu‐3 cells, although mRNA and protein expression appeared higher for CFTR than for pendrin. Targeting pendrin transcripts with lentiviral shRNA reduced pendrin detection by immunofluorescence staining but did not alter the rates of HCO 3 − or fluid secretion, HCO 3 − transport under pH‐stat conditions, or net HCO 3 − flux across basolaterally permeabilized monolayers. Intracellular pH varied with step changes in apical Cl− and HCO 3 − concentrations in control and pendrin knockdown Calu‐3 cells, but not in CFTR deficient cells. Exposure to the proinflammatory cytokine IL‐4, which strongly upregulates pendrin expression in airway surface epithelia, had little effect on Calu‐3 pendrin expression and did not alter fluid or HCO 3 − secretion. Similar results were obtained using air–liquid interface and submerged cultures, although CFTR and pendrin mRNA expression were both lower when cells were cultured under submerged conditions. While the conclusions cannot be extrapolated to other airway epithelia, the present results demonstrate that most HCO 3 − secretion by Calu‐3 cells is mediated by CFTR.


Introduction
The airways are lined by epithelia that secrete fluid and mucus which enable the clearance of inhaled substances. Fluid secretion is driven primarily by transepithelial Cl À transport, a two-step process in which basolateral Cl À loading occurs by cotransport with sodium and potassium and also by exchange with bicarbonate, followed by Cl À efflux through apical anion channels Shan et al. 2012). HCO 3 À is also secreted into the airway lumen (Smith and Welsh 1992) where it functions in mucin unpacking (Quinton 2008) and bacterial killing (Pezzulo et al. 2012). Despite the importance of bicarbonate secretion, the mechanisms of apical HCO 3 À efflux remain uncertain. cAMP-stimulated anion secretion has been studied extensively using the model human epithelial cell line Calu-3 as a model for serous cells of the submucosal glands, which mediate much of the secretion in the upper airways (Widdicombe and Wine 2015). Calu-3 cells resemble gland cells in forming polarized monolayers with tight junctions and transepithelial resistance >100 ΩÁcm 2 at the air-liquid interface (Shen et al. 1994) but have fewer dense granules and lower antimicrobial peptide expression compared to gland serous cells. Nevertheless, Calu-3 cells express the submucosal gland markers lysozyme (Duszyk 2001) and lactoferrin (Dubin et al. 2004), CFTR protein (Haws et al. 1994), and have robust cAMP-stimulated short-circuit currents (I sc ) that are stimulated by b-adrenergic agonists, vasoactive intestinal peptide, and adenosine (Shen et al. 1994).
Our goal was to examine the possible roles of pendrin and CFTR in HCO 3 À secretion by Calu-3 cells. We studied the contribution of pendrin to fluid and HCO 3 À secretion by control and pendrin knockdown Calu-3 cell lines. We also investigated the effects of IL-4 and culture conditions (air-liquid interface vs. submerged) on pendrin functional expression in Calu-3.

Calu-3-derived cell lines
Calu-3 cells from American Type Culture Collection (ATCC, Rockville, MD) were cultured in Eagle's Minimum Essential Medium (Wisent Bioproducts Inc., Saint-Jean-Baptiste, QC). Basal medium contained NaHCO 3 , essential amino acids, and 2 mmol/L glutamine and was supplemented with 1% MEM nonessential amino acids (Wisent), 1% sodium pyruvate (Sigma, St. Louis MO) as recommended by ATCC Kim et al. 2014), and with 7% fetal bovine serum (FBS; Wisent). Growth rates, electrophysiological properties, and mRNA expression for transporters including CFTR and pendrin were similar when media contained 7% or 10% FBS (data not shown). Five pendrin knockdown Calu-3 cell lines (PDS-KD) were generated from the parental line using shRNA lentiviral vectors (see Table 1; Open Biosystems, Lafayette, CO) that target pendrin distal to nucleotide was generated using the same procedures. A CFTR knock down cell line (CFTR-KD) stably expressing shRNA was kindly provided by Dr. Scott O'Grady (Univ. Minnesota) and has been characterized previously (Palmer et al. 2006b;Shan et al. 2012). In most experiments, apical medium was removed after 24-48 h to establish an airliquid interface (ALI) and basolateral medium was replaced at 2 day intervals. To study submerged cultures, 200 lL medium was maintained on the apical surface. Transepithelial resistance was monitored using an epithelial volt-ohmmeter (World Precision Instr., Sarasota, FL). Monolayers were used 10-28 days after plating when the resistance had reached a plateau of~400 ΩÁcm 2 .

Quantitative real-time PCR
Total RNA was isolated from ALI cultures using the RNeasy Mini Kit (Qiagen) following the manufacturer's instructions. First-strand cDNA for real-time PCR was generated using 1 lg of total RNA and SuperScript VILO MasterMix (Invitrogen). The levels of CFTR, SLC26A4, SLC26A6, and SLC26A9 mRNA were assessed by Quantitative real-time PCR (qRT-PCR). Taqman primers (Life technologies; Assay ID: CFTR: Hs00357011_m1; SLC26A4: Hs01070620_m1; SLC26A6: Hs00370470_m1; SLC26A9: Hs00369451_m1) and the QuantStudio TM 7 Flex Real-Time PCR system (Life Technologies) were used. qRT-PCR was performed using the "Fast" program: 95°C for 20 sec, followed by 40 cycles of 95°C for 1 sec and 60°C for 20 sec. Data were normalized to GAPDH.

Immunocytochemistry
Cells cultured on Transwells were washed with PBS three times to remove apical secretions and fixed with 10% neutral-buffered formalin for 15 min at RT. After washing with PBS again, cells were permeabilized with 1% Triton X-100 then blocked with 2% BSA for 1 h. When staining for pendrin alone, samples were incubated with rabbit polyclonal antipendrin antibody (H-195; Santa Cruz) at 1:500-1:1000 dilution overnight at 4°C, followed by goat anti-rabbit IgG Alexa fluor 488 secondary antibody (Invitrogen, 1:1000). When staining both pendrin and ZO-1, goat polyclonal antipendrin (E20; 1:500; Santa Cruz) was used followed by donkey anti-goat IgG Alexafluor 488, 1:1000; Invitrogen) for 1 h at RT. ZO-1 was immunostained using rabbit anti-ZO-1 antibody (Life technology; 1:1000) followed by goat anti-rabbit Alexa 594 (Invitrogen; 1:1000) secondary antibody for 1 h at RT. Some samples were also exposed for 1 min to the nuclear stain DAPI (1 lgÁmL À1 ). The absence of staining when the primary antibody was omitted served as a negative control. Cell imaging was performed using an LSM-780 confocal microscope (Zeiss, Germany), equipped with a multiline argon laser (488 nm, 25 mW) for Alexa-488 excitation and a 560 nm line laser (15 mW Alexa-594 excitation. Imaging was performed using a Plan-Apochromat 639 oil immersion objective and digital gain setting of 900. Brief exposures to 5% laser power were used to maximize detection of pendrin in Calu-3 cells and 2% laser power used when quantifying expression. Imaging conditions were optimized to avoid pixel saturation. Relative fluorescence intensity measurements (arbitrary units [A.U.]) were analyzed using ImageJ (e.g., Fig. 1, (Rasband 2011)) or MetaXpress software (Molecular Devices, Sunnyvale CA).

Fluid secretion
Apical fluid that was present initially was removed and 1/ 1000 DMSO control (v/v) or 10 lmolÁL À1 forskolin + 200 lmolÁL À1 cAMP was added to the basolateral side. A microelectrode (9826BN, Orion) was used to measure the pH of secretions in humidified 5% CO 2 / 95% air at 37°C. The volume of the fluid measured at 24 h intervals by aspiration yielded reproducible measurements of cAMP-stimulated fluid secretion.
Apical anion exchange assayed in Ussing chambers using pH-stat Snapwell â filters were mounted in Ussing chambers and current-clamped using a VCCMC6 amplifier (EasyMount, Physiologic Instruments, San Diego CA). Basolateral Krebs-Henseleit solution contained (in mmolÁL À1 ) 120 NaCl, 25 NaHCO 3 , 3.3 KH 2 PO 4 , 0.8 K 2 HPO 4 , 1.2 CaCl 2 , 1.2 MgCl 2 , and 10 glucose, and was gassed with 95% O 2 /5% CO 2 . Apical pH-stat solution contained (in mmolÁL À1 ): 150 NaCl, 1.2 CaCl 2 , 1.2 MgCl 2 , and 5 KCl and was bubbled with 100% O 2 . pH-stat experiments were performed at 37°C. Equivalent short-circuit current (I eq ) was calculated at 100 sec intervals using the spontaneous transepithelial potential (V t ) and the resistance (R t ), which was determined by injecting small constant-current pulses (1 lA, 1 sec duration). A mini-pH electrode (pHG200-8, Radiometer Analytical) connected to a titration workstation (TitraLab 854, Radiometer) delivered 1 lL aliquots of 10 mequivÁL À1 HCl automatically to maintain the pH at 7.000 AE 0.002 and the volume of HCl was used to calculate net HCO 3 À secretion. To assay apical ion exchange, Cl À was replaced bilaterally with gluconate, the basolateral membrane was permeabilized for 15 min using nystatin (360 lgÁmL À1 ), and either NaCl or NaI solution was added to the apical side to create a 30 mmol/L apicalto-basolateral gradient for Cl À or I À , respectively . The secretory flux of HCO 3 À was monitored by apical pH-stat as described above.

Intracellular pH (pH i ) assays of anion exchange
Intracellular pH was measured using an inverted fluorescence microscope (IX81, Olympus, Center Valley PA) and imaging system (Photon Technology International, Edison NJ). Confluent cell monolayers on inserts were loaded with the pH-sensitive dye BCECF (2 0 ,7 0 -bis-(2-carboxyethyl)-5-(and 6)-carboxyfluorescein by removing the medium, washing for 30 min in HCO 3 À -buffered solution, then incubating cells with 5 lmolÁL À1 BCECF-AM in HCO 3 À -buffered 5% CO 2 solution at 37°C for ≥45 min. Cells were then rinsed and mounted in a chamber that allowed independent perfusion of the apical and basolateral surfaces. Flow was maintained at 1.5 mLÁmin À1 using a four-channel peristaltic pump (205S, Watson Marlow, Wilmington MA), which also removed solutions by suction. The temperature was maintained at 37°C using a thermostatically controlled platform (FC-5, Live Cell Instr., Seoul, Korea).
Apical and basolateral sides were initially perfused with HCO 3 À -buffered solution for 5-10 min or until pH i became stable. The HCO 3 À solution contained (mmolÁL À1 ): 116 NaCl, 25 Na HCO 3 À , 5 KCl, 1 CaCl 2 , 1 MgSO 4 , 2.8 NaHEPES, 2.2 HEPES, and 10 D-glucose. Changes in pH i were measured during brief exposure of the apical or basolateral sides to nominally Cl À -free solution containing ( , 25 sodium gluconate, 1 MgSO 4 , 2.8 NaHEPES, 2.2 HEPES, and 10 D-glucose and remained at pH = 7.5 when bubbled with 95% O 2 /5% CO 2 . When HCO 3 À concentration was reduced by half, the low-HCO 3 À solution contained: 100 NaCl, 12.5 Na HCO 3 À , 37.5 sodium gluconate, 1 MgSO 4 , 2.8 NaHEPES, 2.2 HEPES, and 10 D-glucose and the pH was allowed to decrease to~7.2, where it remained for the duration of the experiment when bubbled with 95% O 2 /5% CO 2 . Bicarbonate was not directly measured therefore the calculated concentrations are considered estimates. All solutions were maintained at 37°C. BCECF was excited alternately at 440 and 490 nm for 1 s at each wavelength, with 5 s between each pair of measurements. Since individual cells could not be resolved in the confluent monolayers, unbiased sampling of pH i was performed as follows: Fluorescence was measured in five regions of interest (ROIs) per field, one at the outer edge of each of four quadrants and one near the center. These ROIs covered $ 2/3 of the total area in the field of view using a 409 objective. Fluorescence intensities in the five ROIs were averaged to obtain one value. Fluorescence ratios (F490/F440) were recorded and displayed continuously using Easy Ratio Pro software. To calculate pH i , F490/F440 was calibrated using high-K + -nigericin solutions containing (mmolÁL À1 ): 140 KCl, 1 CaCl 2 , 1 MgSO 4 , 20 HEPES, and 20 lmolÁL À1 nigericin, adjusted to different pH values. Graphs and statistical analyses were performed using Microsoft Excel and Prism 5 software.

Statistics
Data are means AE standard error of mean (SEM) of n observations. Datasets were compared using the Student's t test or two-way analysis of variance (GraphPad Prism) with P < 0.05 considered significant unless otherwise indicated.

Results
Generation of pendrin knockdown and scrambled control Calu-3 cells We began by confirming that pendrin and CFTR are expressed in Calu-3 cells. Transcripts encoding CFTR, SLC26A4 (pendrin) and the related exchangers SLC26A6 and SLC26A9 were readily detected in lysates of parental Calu-3 cells from ATCC (Fig. 1A). Transcript levels were normalized to those of GAPDH before determining relative levels in transduced versus parental cell lines. Control cells transduced with scrambled shRNA (Scr-KD) did not have significantly altered mRNA levels for the four transporters examined. Lentiviral transduction of parental Calu-3 cells with shRNAs targeting five different pendrin sequences (see Table 1) identified one sequence (sh-PDS-5) that greatly reduced PDS mRNA (to 10.7 AE 9%, mean AE SEM, n = 5, P < 0.0001; Fig. 1A). All experiments were performed using low-passage, nonclonal cells transduced with sh-PDS-5 to minimize selection of atypical variants. Pendrin protein was not detected in parental Calu-3 cell lysates by immunoblotting using several different antipendrin antibodies although they recognized heterologous pendrin overexpressed in HEK cells, thus it was not possible to convincingly detect pendrin protein in Calu-3 by immunoblotting, as in previous studies (Garnett et al. 2011). Nevertheless, PDS knock down was confirmed at the protein level by immunostaining (green) and confocal imaging with settings identical to those used when imaging immunstained pendrin in Parental and Scr-KD cells (Fig. 1B). Tight junctions were also visualized by immunostaining ZO-1 (red). Pendrin was detected by immunostaining and was absent when the primary antipendrin antibody was omitted. Confocal images confirmed pendrin protein expression in WT, Scr-KD and CFTR-KD cells, which was reduced by 80-85% in PDS-KD cells ( Fig. 1B and C), similar to the mRNA decrease observed using qPCR (~90%; Fig. 1A). Interestingly, pendrin expression was also reduced in a CFTR knockdown cell line, suggesting there is a positive interaction between these proteins ( Fig. 1B and C). Pendrin knockdown did not alter AE2, NBCe1, NKCC1, or CFTR expression by immunoblotting (Fig. 1D-H). Similar results were obtained whether a-tubulin, Na + /K + -ATPase or b-actin was used as the loading control. Pendrin knockdown remained constant throughout the experiments according to qPCR performed at the beginning and end of the project.

Volume and pH of cAMP-stimulated secretions
To examine pendrin function we began by measuring fluid secretion volume and pH when parental, Scr-KD Forskolin-stimulated I eq and net HCO 3 À flux under pH-stat conditions Net HCO 3 À secretion and electrogenic anion transport were monitored in Ussing chambers under open-circuit, pH-stat conditions Shan et al. 2012). Equivalent short-circuit current (I eq ; blue symbols) and net HCO 3 À flux (red symbols) time courses are shown for Scr-KD (Fig. 3A) and PDS-KD monolayers (Fig. 3B). Steady-state HCO 3 À transport rates estimated by pH-stat are summarized in Figure 3C. Forskolin stimulated net HCO 3 À fluxes of 0.4 lequivÁcm À2 Áh À1 in both Scr-KD and PDS-KD monolayers, and these fluxes were abolished by CFTR inh -172. These results suggest that forskolin-stimulated HCO 3 À transport is CFTR-dependent and unaffected by pendrin knockdown, consistent with the results described above when Calu-3 cells were bathed apically with their own secretions.
Apical Cl À /HCO 3 À exchange in basolaterally permeabilized monolayers To investigate anion exchange independently of the basolateral membrane, monolayers were mounted in Ussing chambers, bathed symmetrically with nominally Cl À -free solution, and the basolateral membrane was permeabilized by adding 360 lgÁmL À1 nystatin. The appearance of HCO 3 À on the apical side was monitored using pH-stat when an apical-to-basolateral Cl À gradient was imposed by apical addition of 30 mmolÁL À1 NaCl. We showed previously that the small osmotic gradient that results from asymmetrical addition of NaCl does not affect I eq or HCO 3 À flux significantly ). The Cl À gradient produced a small inward current as expected with basal Cl À conductance. Forskolin stimulated a much larger negative I eq across parental, Scr-KD, and PDS-KD Calu-3 monolayers, consistent with activation of CFTR channels in the presence of an apical-to-basolateral Cl À gradient ( Fig. 4A-C). We showed previously that the forskolin-stimulated negative I eq measured under these conditions is abolished in CFTR-KD cells . No HCO 3 À flow in the opposite (secretory) direction was detected under basal conditions but it appeared after forskolin was added, suggesting efflux through CFTR. The outward HCO 3 À fluxes were similar in parental Calu-3 cells (0.554 AE 0.049 lequivÁcm À2 Áh À1 , n = 6, P > 0.2), Scr-KD (0.414 AE 0.037 lequivÁcm À2 Áh À1 , n = 5), and PDS-KD (PDS-KD: 0.369 AE 0.057 lequivÁcm À2 Áh À1 , n = 9) indicating that they are independent of pendrin (Fig. 4D).
Effect on pH i of removing extracellular Cl À Fluorescence imaging was also used to study apical anion exchange. pH i was measured when cells were challenged with nominally Cl À -free apical solution containing 25 mmol/L HCO 3 À . Cl À removal from the apical side during forskolin stimulation increased pH i in Scr-KD cells (DpH i = 0.46 AE 0.04 units, n = 4) consistent with inward HCO 3 À flow through activated CFTR channels, whereas basolateral Cl À removal had little effect (0.06 AE 0.03 units, n = 5, P = 0.0016; Fig. 5A). To test if the alkalinization requires pendrin or CFTR, Scr-KD cell pH i responses were compared with those obtained using PDS-KD (Fig. 5B) and CFTR-KD (Fig. 5C) cells. Forskolin-stimulated PDS-KD cells showed a similar large increase in pH i during challenge with low apical Cl À (0.38 AE 0.05 pH unit, n = 4; Fig. 5B) and there was no change in pH i upon removal of basolateral Cl À (0.05 AE 0.04 pH unit, n = 4, P = 0.0055). The inability of basolateral Cl À removal to alkalinize Scr-KD or PDS-KD cells despite the presence of basolateral AE2 anion exchangers would be explained if the HCO 3 À that is taken up by basolateral exchange escapes through apical CFTR channels during forskolin stimulation (Kim et al. 2014). To test that interpretation we also measured pH i responses during basolateral Cl À removal in CFTR-KD cells and forskolin stimulation (Fig. 5C). In CFTR-deficient cells, intracellular alkalinization induced by apical Cl À substitution was attenuated (0.1 AE 0.02 pH unit, n = 4; Fig. 5C) while that produced by basolateral Cl À removal was greatly increased (0.34 AE 0.05 pH unit, n = 4, P = 0.042). Figure 5D summarizes these pH i responses to extracellular Cl À removal in all three cell lines. We conclude that forskolin normally prevents pH i alkalinization during exposure to low basolateral Cl À because basolateral HCO 3 À taken up by the cells exits through apical CFTR channels. were bathed initially in symmetrical Cl À -free solutions and permeabilized basolaterally using nystatin (360 lgÁmL À1 ). Apical HCO 3 À efflux was monitored using pH stat. 30 mmol/L NaCl was added on the apical side to establish an apical-to-basolateral Cl À gradient and 10 lmol/L forskolin was added to stimulate CFTR. (D) summary of HCO 3 À secretion rates under each condition (mean AE SEM, n = 4-9). HCO 3 À secretion by Calu-3 Scr-KD and PDS-KD monolayers was negligible in unstimulated monolayers with Cl À gradient, and similar after forskolin addition (ns, not significant, P > 0.2). Apical HCO 3 À /Cl À exchange could potentially be missed due to HCO 3 À efflux through basolateral AE2 exchangers.
To test this possibility, intracellular Cl À was depleted by Cl À removal from both sides for ≥15 min, then Cl À was restored only on the apical side while anion exchange-mediated apical HCO 3 À efflux was monitored as the rate of re-acidification ( Fig. 5E-H). Under these conditions, apical HCO 3 These results, which are summarized in Figure 5H, provide further evidence that CFTR is the predominant apical pathway for HCO 3 À in Calu-3 cells and that pH i responses to apical Cl À removal are due to the Cl À efflux that is electrically coupled to HCO 3 À entry through CFTR channels.

Pharmacological inhibitor effects on pH i
A large alkalinization (0.8 pH unit) has been reported during apical Cl À removal while apical CFTR and basolateral HCO 3 À transport are both inhibited pharmacologically, and this provided strong evidence for pendrin-mediated HCO 3 À transport (Garnett et al. 2011).
We performed similar experiments using both control and pendrin knockdown cell lines (Fig. 6A-C). pH i was measured in unstimulated cells while they were challenged with nominally Cl À -free apical solution and continuously exposed to basolateral H 2 DIDS (500 lmol/L; to inhibit HCO 3 À flux through AE2 and Na + -HCO 3 À cotransporters) and apical CFTR inh -172 (10 lmol/L; to inhibit HCO 3 À flux through CFTR).
As expected, removing apical Cl À under these conditions caused a small alkalinization in Scr-KD and PDS-KD cells (DpH i = 0.09 AE 0.02 and 0.08 AE 0.04 pH uni-tsÁmin À1 , respectively, n = 4; Fig. 6A and B), about 25% of that observed without inhibitors present (compare with Fig. 5 and B). Basal CFTR activity was apparently responsible for the intracellular alkalinization because it was further reduced in CFTR-KD cells. Similar results were obtained using AE2-KD cells ) that were exposed to CFTR inh -172 (DpH i = À0.02 AE 0.03, 0.06 AE 0.07 pH unitsÁmin À1 , n = 6-7, respectively, Fig. 6C and D). Thus, pH i is not increased significantly by low apical [Cl À ] in any of the cell lines when CFTR and basolateral anion exchangers are inhibited. These results differ from those reported previously using Calu-3 cells that had been cultured under submerged conditions rather than at the ALI, therefore the experiments were repeated using cells that had been cultured under submerged conditions. When parental Calu-3 cells were stimulated using forskolin to activate CFTR, robust alkalinization was observed during apical Cl À -free exposure regardless of whether they were cultured at the ALI (Fig. 6G) or submerged (6H); compare with Figures 6E and F, respectively. These responses were inhibited by CFTR inh -172 or GlyH-101 and were not restored by inhibiting potential basolateral HCO 3 À efflux with H 2 DIDS (Fig. 6I). We conclude from these inhibitor studies that the apparent anion exchange at the apical membrane is mediated by electrical coupling in CFTR channels and is similar whether cells are cultured under ALI or submerged conditions. ] to 50 mmol/L produced a small, reversible alkalinization. Similar results were obtained using Scr-KD, PDS-KD, and CFTR-KD cell lines (DpH i = 0.1 AE 0.06, 0.09 AE 0.02, and 0.11 AE 0.05 pH unitsÁmin À1 , respectively, n = 4). During forskolin stimulation (Fig. 7A, traces on the right side), the alkalinization induced by 50 mmol/L HCO 3 À was enhanced by forsko-  Fig. 7A). As summarized in Figure 7B, these results with knock down cell lines and CFTR inh -172 indicate that most HCO 3 À flows through CFTR.

IL-4 does not increase pendrin expression or secretion by Calu-3
Pendrin expression is strongly upregulated in primary surface airway epithelial cells during inflammation therefore we studied Calu-3 cells exposed to the cytokine IL-4 (10 ngÁmL À1 ) for 48 h. qPCR revealed a slight increase in pendrin mRNA which was not statistically significant (P = 0.18, n = 3, Fig. 8A). Likewise, pendrin protein level was not increased significantly after IL-4 treatment in scr-KD control (13.1 AE 15.2%, ns, n = 5) or PDS-KD (12.9 AE 16.1%, ns, n = 5) cell lines according to immunostaining ( Fig. 8B and C). IL-4 alone did not alter fluid secretion by parental Calu-3 monolayers, however, there was a trend toward higher forskolin-stimulated secretion that did not reach statistical significance at P < 0.05 (but did at P < 0.1, Fig. 8D; n = 4, P = 0.08). The apparent increase in forskolin-stimulated fluid secretion after IL-4 was not statistically significant in parental cells and was completely absent in PDS-KD cells. Forskolin increased the pH of secretions by~0.5 pH units regardless of the presence of IL-4 ( Fig. 8E and F). In summary, IL-4 did not increase pendrin expression significantly in Calu-3 cells and had little effect on the secretion of fluid and HCO 3 À , and pendrin knockdown did not alter the pH or volume of secretions (n = 6-9, P > 0.2).

Effect of culture conditions on anion transporter expression
Abnormally low pendrin expression or high CFTR expression under our ALI conditions could potentially explain the negligible role of pendrin in this study, therefore we compared cells that had been cultured under ALI versus submerged conditions, which are known to impact Calu-3 differentiation (Kreft et al. 2015). Figure 9A-

Discussion
In this study we have examined apical anion transport in the Calu-3 cell line using converging approaches and found that CFTR is the predominant pathway for HCO 3 À efflux. Pendrin mRNA was expressed in Calu-3 cells as expected (Garnett et al. 2011) but its levels were low compared to both CFTR and the basolateral anion exchanger AE2, in agreement with a previous study (Kim et al. 2014), and knocking down pendrin did not alter HCO 3 À secretion. Pendrin protein was not detected reliably on immunoblots but was clearly observed by immunostaining as reported previously (Garnett et al. 2011). Pendrin mRNA and immunofluorescence were both reduced~80% by shRNA stably expressed using a lentivirus. The proinflammatory Th2 cytokine IL-4, which strongly induces pendrin expression in bronchial epithelial cells (Galietta et al. 2002), did not increase pendrin mRNA levels significantly in marked contrast to surface airway cells. The Calu-3 cell line may lack some molecule in the IL-4 receptor signaling pathway which is present in surface airway epithelia and essential for upregulating pendrin expression. Thus our conclusion regarding the lack of pendrin-mediated bicarbonate secretion is restricted to the Calu-3 cell line. It is reasonable to expect pendrin to mediate significant bicarbonate flux in surface airway epithelial cells especially when upregulated by cytokines, as has been reported by others (Gorrieri et al. 2016). We compared anion and fluid transport by control and pendrin knock down cells and found no evidence for pendrin-dependent anion exchange activity using pH-stat or fluorescence assays of pH i . We studied apical Cl À / HCO 3 À exchange with pH-stat by imposing an apical-tobasolateral Cl À gradient and permeabilizing the basolateral membrane. Net HCO 3 À secretion was only detected under these conditions if forskolin was added, and the HCO 3 À flux was similar in control and pendrin knockdown cells suggesting it occurs through CFTR. Control and pendrin knockdown cells had similar rates of intracellular reacidification when cells were HCO 3 À loaded by exposure to apical low-Cl À solution. Although at first glance this might seem to be evidence for apical anion exchange, inhibitor studies and the effects on pH i of manipulating extracellular HCO 3 À were consistent with CFTR-mediated HCO 3 À flux. Together the results suggest that opposing Cl À and HCO 3 À fluxes are coupled electrically in the CFTR channel pore rather than chemically through pendrin-mediated Cl À /HCO 3 À exchange (Kim et al. 2014). We cannot exclude the possibility that some CFTR-dependent HCO 3 À flux occurs through another type of anion channel that may be regulated by CFTR transporters, rescaled to enable comparison. Note that after normalization to GAPDH, the qRT-PCR signals for SLC26A6 and SLC26A9 were >100-fold higher than for pendrin.  (Bertrand et al. 2009), however, sensitivity to both CFTR inh -172 and GlyH-101 suggests that most HCO 3 À flux is mediated by CFTR channels.
The present results differ from a previous study in which pendrin knockdown caused a reduction in the pH of secretions whereas silencing CFTR had no effect (Garnett et al. 2011). The reason for the discrepancy is not known, however, different cell lines were used in the two studies. The extent of pendrin knockdown was probably similar in the two studies as residual pendrin mRNA expression was 8.5% in cells used previously (Garnett et al. 2011) versus 10.7 AE 9% in the cells used here. CFTR knock down was less complete in the previous study (28 AE 5% residual CFTR) compared to the cell line used here (~5% residual CFTR) which might reduce the apparent role of CFTR (Palmer et al. 2006a;MacVinish et al. 2007;Garnett et al. 2011). Another technical difference was our use of air-liquid interface culture whereas submerged cultures were used previously. However, we found here that this had little effect on relative anion transporter expression (Fig. 9) or apical anion exchange activity ( Fig. 6E-H), thus further work is needed to reconcile the results of these studies.
Replacing basolateral Cl À with gluconate increased pH i in all Calu-3 cell lines examined. This was anticipated since there is robust AE2 mediated Cl À /HCO 3 À exchange at the basolateral membrane (Loffing et al. 2000;Huang et al. 2012). However, the alkalinization observed during forskolin stimulation was >6-fold larger in CFTR-KD cells than in Scr-KD (control) and PDS-KD cells. This strongly suggests that intracellular HCO 3 À accumulates during exposure to basolateral low-Cl À solution as long as apical CFTR is not functional, presumably because HCO 3 À taken up escapes through apical CFTR channels when they are activated by forskolin.
In summary, the present results indicate that most HCO 3 À secretion by Calu-3 cells occurs via CFTR Kim et al. 2014). This is consistent with the HCO 3 À permeability of CFTR channels (Gray et al. 1990;Poulsen et al. 1994;Linsdell et al. 1997) and the large contribution of CFTR to apical membrane conductance in Calu-3 cells (Tamada et al. 2001). Pendrin may play a more important role in HCO 3 À secretion by surface airway epithelial cells where it is strongly upregulated by the proinflammatory cytokine IL-4 (Gorrieri et al. 2016).