Large pH oscillations promote host defense against human airways infection

A microscopic film of airway surface liquid (ASL) containing water, salts, mucins, and antimicrobial factors secreted by the airway surface epithelium and submucosal glands covers the airway mucosa and serves as the first line of defense against inhaled pathogens. The airway mucosa is a unique microenvironment that experiences large (>140-fold) variations in CO₂ every few seconds due to tidal breathing. Normally, CO₂ in the airway lumen is ~5.6% during expiration of alveolar air and falls to 0.035% when room air is inspired. ASL is an open pH buffer system with an enormous surface area–to-volume ratio in which CO₂ can diffuse rapidly in its thin aqueous layer (Thomas and Adams, 1965). The half-time for CO₂ equilibration in thin aqueous films comparable to ASL is <219 ms (Uchida et al., 1983; Wu et al., 1998).

To investigate the mucosal microenvironment and its role in host defense, we began by directly measuring pH on human nasal mucosae in vivo using ultrafine capillary electrodes. Nasal epithelium has long served as a surrogate for distal airways in cystic fibrosis (CF) research, and nasal potential difference remains the most accepted in vivo assay for airway epithelial ion transport in this disease (Knowles et al., 1981). Recently, it has been shown that CF drugs have similar effects on nasal potential difference in vivo and in nasal epithelial cell cultures and that cells from nasal and bronchial brushings respond similarly to CF drugs and are predictive of clinical efficacy (Pranke et al., 2017). Most people with CF (~95%) have chronic rhinosinusitis or polyposis, and their nasal and bronchial microbiomes are highly correlated, suggesting the bronchi may be colonized by bacteria that migrate from the paranasal sinuses (Hansen et al., 2012). Thus, despite some differences in cell composition, the nasal surface provides a useful model of lower airway ion transport and microbiology (Brewington et al., 2018). In this study, we confirmed the results obtained on nasal mucosae using well-differentiated primary cultures of human bronchial epithelial (HBE) cells.

Flexible, double-barreled capillary electrodes were used to measure nasal pH (O’Donnell, 1992). One capillary contained H⁺-selective liquid ion exchange resin while the other contained 3 M KCl in agar and served as a reference (Fig. 1 A). The electrode tip had a minimum width of 150 µm and a maximum width of 550 µm, both of which exceed ASL height. However, the polypropylene surface of the capillaries created a meniscus that immersed both capillaries when they were placed on bronchial cell cultures (Fig. 1, B and C). ASL height may be further increased in vivo due to mechanically induced secretion by submucosal glands. Although we could not observe the meniscus while recording in vivo, electrical continuity during nasal pH recordings confirmed that both barrels were in contact with the ASL. A new electrode was used for each subject and was calibrated in a series of buffer solutions before being placed on the inferior nasal turbinate of a healthy volunteer under visual control. We recorded the electrical potential difference between the pH-sensitive and reference barrels continuously using a high-impedance differential amplifier while subjects took 5–10 slow breaths. Breathing caused oscillations in the voltage that had amplitudes >75 mV (Fig. 1 D). Electrodes were returned to the calibrating solutions immediately after each recording (e.g., Fig. 1 D) to confirm the electrode response and correlate measured voltages with pH (Fig. 1 E). A typical pH recording is shown in Fig. 1 F. The pH on nasal mucosae of healthy subjects oscillated between ~7.5 during exhalation and pH 8.5–9.0 during inhalation and declined during breath holding, presumably due to an accumulation of CO₂ (Fig. 1, F and G). The mean amplitude of excursions in pH (peak to peak) in six healthy subjects ranged between 0.7 and 1.4 pH units (s1–s6; Fig. 1 E). These results confirm that CO₂ in the airway lumen equilibrates rapidly with ASL and that CO₂ hydration–dehydration reactions are sufficiently rapid to generate large pH changes during the breathing cycle.

To study ASL pH in vitro, we used well-differentiated primary HBE cells, which had spontaneously beating cilia and mucus-secreting goblet cells (Fulcher et al., 2005; Kim et al., 2019). We washed the cultures with PBS containing 5 mM N-acetylcysteine to remove accumulated mucus, then coated the cells with a thin (~10 µm) layer of artificial ASL containing both the fluorescent indicator 2',7'-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein (BCECF)–dextran to monitor ASL pH and the lipid soluble red styryl dye FM 1–43 to visualize the epithelium (Fig. 2 A). A computer-controlled gas-mixing pump (Fig. S1) delivered a time-varying, humidified CO₂/air mixture onto the cells with a waveform resembling that in the trachea during tidal breathing (Cochrane et al., 1982). The oscillation frequency was reduced from ~0.2 to 0.017 Hz to ensure CO₂ mixing and efficient washout from the perfusion chamber, which was verified by comparing CO₂ levels at the gas inlet and exit of the chamber (Fig. 2 B, solid black and dashed red lines, respectively). CO₂ oscillations produced reciprocal pH changes in the artificial ASL despite weak buffering by 5 mM Hepes (Fig. 2 C). Basal pH during exposure to 5% CO₂ (to simulate breath exhalation) and peak pH during exposure to room air (to simulate inhalation) were both shifted to more alkaline values when the HCO₃⁻ concentration was increased from 3.2 mM (Fig. 2 C, red trace) to 10 mM (Fig. 2 C, blue trace), concentrations estimated based on the pH of CF and normal airway gland secretions, respectively (Song et al., 2006). Measurements calculated from BCECF-dextran fluorescence were confirmed using a macroscopic pH electrode (Fig. 2 D).

Mucus secreted onto the airways is removed by mucociliary transport in vivo but accumulates over the course of several weeks when cells are cultured at the air–liquid interface. This endogenous mucus reduced the amplitude of pH oscillations (Fig. 2 E) and the rate of alkalinization after a step decrease in PCO₂ relative to cultures that had been washed with PBS containing 5 mM N-acetylcysteine (Fig. 2 F). Since mucins buffer the pH of airway epithelial secretions (Kim et al., 2014), we examined the effect on the oscillations of adding exogenous mucins at concentrations expected in non-CF and CF airways (2.5% and 8% solids, respectively; Hill et al., 2014; Martens et al., 2011). Crude mucin preparations (80% pure) were used, as the properties of highly purified mucins can differ from those of native mucus and because we were interested in physiological effects on pH rather than properties of the mucins. When cultures were washed and covered with a small volume of artificial ASL sufficient to produce a layer ~10 µm high (verified using image z-stacks obtained with a confocal microscope), the pH oscillation amplitude decreased to ~1.1 pH units with 2.5% mucins (25 mg/ml) and was decreased another 40% by the addition of 8% mucins (80 mg/ml; Fig. 2 G). These results with exogenously added crude mucins resemble those obtained when mucus secreted by the cells accumulated on the cultures (compare Fig. 2 E). In both instances, mucin buffering had little effect during exposure to 5% CO₂, with pH remaining near 7. By contrast, the maximum pH during air exposure was progressively reduced, and the rate of alkalinization after a step decrease in PCO₂ from 5% to 0.035% was slowed (Fig. 2 H). Strong buffering in the oscillating pH range was confirmed by microtitration, which indicated a buffer capacity β = 60 mM/pH with 8% solids (Fig. 2 I; gray indicates oscillatory pH range) in agreement with previous studies (Kim et al., 2014).

Computer simulations predict that large pH oscillations can occur only if CO₂ hydration and dehydration reactions (CO₂ + H₂O ⇔ H⁺ + HCO₃⁻) are rapid compared with the breathing cycle, and this would require catalysis by carbonic anhydrase (Fig. 2, J and K). ASL does not contain soluble carbonic anhydrase (Candiano et al., 2007; Thornell et al., 2018), and we observed large pH oscillations when artificial ASL lacking this enzyme was placed on washed bronchial cultures; therefore, we hypothesized that carbonic anhydrase activity is associated with the epithelial surface. As a test, we added the broad-spectrum carbonic anhydrase inhibitor acetazolamide (100 µM) or topiramate (200 µM), an inhibitor that is moderately selective for carbonic anhydrase 12 (CA12), to the artificial ASL on washed cultures and measured the rate at which pH increased after stepping PCO₂ from 5% to 0.035% (Fig. 2 L). Both carbonic anhydrase inhibitors reduced the rate of alkalinization >50% compared with controls, indicating carbonic anhydrase activity on the apical surface of HBE cells (Fig. 2 M). Also shown for comparison are the alkalinization rates when mucus had accumulated (without wash) and after addition of exogenous carbonic anhydrase type 2 (1 unit/ml; Sigma-Aldrich). To examine the effect of mucus and the role of carbonic anhydrase in vivo, we sprayed nasal mucosae of non-CF subjects with sterile control saline (Fig. 3, A and B) or saline containing sterilized crude mucins (8%; Fig. 3, C and D) or 500 µM acetazolamide (Fig. 3, E and F). Mucins and the carbonic anhydrase inhibitor both reduced the amplitude of breathing-induced pH excursions by >50% (Fig. 3 G).

Four of the 14 human carbonic anhydrase isozymes are membrane associated (CA4, CA9, CA12, and CA14) and could potentially catalyze changes in ASL pH. We compared their transcript levels (normalized to GAPDH) in well-differentiated primary HBE cells from five non-CF donors using quantitative PCR (qPCR). CA12 was the most abundant enzyme (Fig. 4, A and B). We also compared CA12 mRNA levels in freshly isolated bronchial epithelial cells obtained by brushing, well-differentiated HBE cultures, and human embryonic kidney (HEK) cells expressing endogenous CA12 (as a positive control). CA12 levels were approximately twofold higher in freshly isolated cells compared with cultured cells and higher in non-CF HBE cells compared with CF HBE cells.

These results contrast with a previous report using piglet proximal large airways, which concluded that surface liquid pH would not change during breathing (Thornell et al., 2018). In an attempt to reconcile our results, we compared CA12 mRNA transcript levels in human and pig airways. qPCR revealed extremely low CA12 mRNA expression in fresh pig tracheal scrapings and well-differentiated pig tracheal epithelial (PTE) primary cultures compared with CF bronchial tissue, well-differentiated non-CF and CF HBE cell cultures, and HEK cells expressing endogenous CA12 (Fig. 4 B). Different primers were used to amplify human and pig CA12; therefore, we also analyzed pig skeletal muscle tissue and kidney cortex as negative and positive controls. CA12 mRNA expression was negligible in pig muscle and very high in pig renal cortex, as expected. Similar results were obtained at the protein level using immunoblots (Fig. 4, C and D). Strong CA12 protein expression was observed in freshly isolated human bronchus but was also abundant in primary cultures of non-CF and CF HBE cells and in HEK cells expressing endogenous CA12. CA12 protein was expressed at high levels in pig kidney cortex, as expected, confirming that the antibody recognizes pig CA12. By contrast, CA12 protein expression was negligible in pig trachea, PTE primary cultures, and pig skeletal muscle (Fig. 4, C and D), consistent with the mRNA data in Fig. 4 B and with previous studies (Lee et al., 2016; Tarun et al., 2003). Fig. 4 D also shows very high CA12 protein expression in HEK cells that had been transiently transfected with CA12 cDNA as a positive control. These results indicate there is much higher epithelial CA12 expression in human airways than in pig airways, and this species difference may explain the previous conclusion regarding the effect of breathing on ASL pH.

Airway epithelial cells are heterogeneous; therefore, it was of interest to examine which cell types express CA12. CA12 immunofluorescence was detected at the apical pole of ciliated cells, which were readily identified by coimmunostaining with antibody against acetyl-tubulin (Fig. 4, E and F; and Fig. S2). CA12 was not detected in MUC5AC-positive goblet cells, either in well-differentiated HBE primary cultures (Fig. S2) or in tissue sections (Fig. S3 B). CA12 immunofluorescence was weaker in primary cell cultures than in freshly isolated cells obtained by nasal or bronchoscopic brushing (Fig. S4, A and B) and was also less intense in CF primary HBE cultures compared with non-CF cultures when imaged using the same microscope settings (Fig. S4, C and D), consistent with the qPCR and immunoblotting results shown in Fig. 4, B–D. The reasons for these differences in CA12 expression level remain to be elucidated and may involve such environmental and hormonal factors as oxygen tension and estrogen receptor activation (Gruvberger et al., 2001; Ivanov et al., 2001). In summary, CA12 is highly expressed in human but not pig airway epithelium. It is expressed at the apical pole of ciliated HBE cells and has an extracellular catalytic domain; thus, it may catalyze oscillations in ASL pH (Fig. 4 G).

We compared breathing-induced pH changes on the nasal mucosae of 10 non-CF and seven CF subjects. During inspiration, ASL reached pH 9.0 in non-CF subjects (Fig. 5, A and C) but only pH 7.7 in those with CF (Fig. 5, B and C). pH during expiration was also 0.5 pH unit lower in CF subjects compared with non-CF subjects (Fig. 5 D). This shift to a lower baseline pH may reflect reduced ASL HCO₃⁻ concentration, since spraying the nasal turbinate with 50 mM sodium bicarbonate in saline increased the baseline pH by >0.5 pH unit and enabled large pH excursions during breathing (compare Fig. 5, E and F, with Fig. 5, G and H). Applying HCO₃⁻ onto the mucosal surface increased the maximum pH during exhalation by >1 pH unit. The increase was significant (P = 0.0051 paired t test with four CF patients, P = 0.0017 using pooled results from all CF patients; Fig. 5, I–K), indicating that exogenous HCO₃⁻ can restore large, breathing-induced alkalinizations of the ASL on CF nasal mucosa.

The gram-negative bacterium Pseudomonas aeruginosa is the most common lung pathogen in CF adults and is considered a neutrophile because its growth is optimal in the range pH 6.6–7.0 (Booth, 1985; Todar, 2019). We examined the pH sensitivity of P. aeruginosa using the mucoid strain PA508 from a CF patient. Bacteria were incubated for 24 h in artificial ASL at pH values ranging from 7.4 to 9.1. Aliquots were then plated on agar to determine the number of viable bacteria (colony-forming units [CFUs]; Fig. 6 A). There was little killing in the range of pH 7.4–8.0, as expected; however, CFUs declined at more alkaline pH values. The time course of the decline in air-equilibrated saline suggested that >50% of the CFUs are killed within ~3 h under these conditions (Fig. 6 B).

We also examined the antimicrobial effect of CO₂ oscillations in fluid secreted by the airway epithelial cell line Calu-3. The number of CFUs was unchanged after bacteria were incubated for 6 h in fluid from unstimulated Calu-3 airway epithelial cells that had been equilibrated with air. However, when incubated in fluid from Calu-3 cultures that had been stimulated with forskolin (FSK) to secrete HCO₃⁻-rich fluid, 90% killing was observed (Fig. 6 C, i). By contrast, the number of CFUs increased exponentially regardless of FSK stimulation when bacteria were incubated in Calu-3 secretions that had been equilibrated with 5% CO₂ (Fig. 6 C, ii). Thus, elevating PCO₂ in the physiological range to prevent ASL alkalinization inhibits bacterial killing, and the increase in CFUs suggests that glycoproteins and other macromolecules in the secretions are nutrients that support bacterial growth. The secretions were bacteriostatic when CO₂ was maintained constant at 2.5% (i.e., there was no bacterial growth; Fig. 6 C, iii). However, PCO₂ oscillations that yielded the same time-averaged PCO₂ (i.e., 2.5%) were strongly antimicrobial (Fig. 6 C, iv), indicating that fluctuations in pH are inherently antimicrobial (Fig. 6 D). Similar results were obtained with artificial ASL containing a range of HCO₃⁻ concentrations (3.2–50 mM; Fig. 6 E). At each HCO₃⁻ concentration, more killing was observed when PCO₂ oscillated than when it remained constant at the same mean value (i.e., 2.5%). Further evidence that the antimicrobial effect is due to pH rather than HCO₃⁻ was obtained using a bioluminescence assay. Increasing pH to 8.6 or 9.3 caused similar killing in solutions buffered with phosphate or a mixture of phosphate and HCO₃⁻ (Fig. 6 F). Taken together, the results indicate that breathing-induced oscillations in PCO₂ produce large, transient alkalinizations that are antimicrobial in both artificial ASL and epithelial secretions.

Rapid, oscillating changes in ASL pH are an important feature of the airway microenvironment that has not been reported previously and has broad implications for airway physiology and host defense mechanisms. It is distinct from the current paradigm in which there is a slight reduction in steady-state pH in CF airways (Pezzulo et al., 2012). pH measurements using antimony and gold pH probes have not detected a lower steady-state pH in CF compared with non-CF subjects (McShane et al., 2003), and bronchial pH was not altered in CF versus non-CF children when measured using a luminescent dye–based fiber-optic probe having dimensions similar to the capillary electrodes used here (Schultz et al., 2017). Although we observed a slightly lower baseline pH on CF nasal epithelium in vivo, a more striking finding was the presence of large excursions to high pH during breathing that were absent or greatly reduced in CF subjects.

Previous modeling suggested that ASL pH oscillations would not occur during breathing due to low carbonic anhydrase activity (Thornell et al., 2018); however, the conclusion was based on data from pig airways. We found that airway CA12 expression is up to 100-fold higher in humans, and this species difference may account for different conclusions in the previous study. CA12 is a type I membrane protein with an extracellular catalytic domain and is expressed on the apical surface of many epithelia (Lee et al., 2016). Its physiological role remains uncertain in most tissues; however, in salivary glands it is essential for HCO₃⁻ secretion (Hong et al., 2015). In the renal proximal tubule, membrane-bound carbonic anhydrases CA4 and CA14 facilitate HCO₃⁻ reabsorption by accelerating its conversion to CO₂ in the lumen (Purkerson and Schwartz, 2007). The present results suggest that in the airways, membrane-bound CA12 accelerates CO₂ hydration and dehydration reactions so that they approach equilibrium during the breathing cycle, generating large pH oscillations.

Several factors are expected to influence the amplitude of pH oscillations. They require luminal HCO₃⁻, which is reduced in CF (Smith and Welsh, 1992) due to diminished CF transmembrane conductance regulator (CFTR)–mediated HCO₃⁻ efflux (Linsdell et al., 1997; Poulsen et al., 1994) and also reduced HCO₃⁻ secretion by pendrin (SLC26A4), an anion exchanger that is modulated by CFTR (Gorrieri et al., 2016; Kim et al., 2019). The steady-state HCO₃⁻ concentration in ASL will also be reduced by concurrent H⁺ secretion via mechanisms that continue unabated when CFTR-dependent HCO₃⁻ secretion is defective (Coakley et al., 2003; Li et al., 2016; Scudieri et al., 2018). Abnormal mucus is a hallmark of CF (Quinton, 2008), and we found that buffering by endogenous mucous secretions and exogenous, semi-purified mucins reduces the pH excursions. Mucus accumulation and airway obstruction would reduce the amplitude of pH excursions further, since airflow is required for variations in PCO₂. Loss of alkaline transients may impact other respiratory diseases in which there is excess mucus and Pseudomonas infection, such as chronic obstructive pulmonary disease and non-CF bronchiectasis.

We found that a constant, low level of CO₂ is less antimicrobial than PCO₂ oscillations that have the same mean PCO₂ value. This may reflect greater cell stress when bacteria must maintain intracellular pH homeostasis in a fluctuating environment. In addition to direct bacterial killing, excursions to high pH in vivo may exert antimicrobial effects through other mechanisms, for example, by increasing the efficacy of cationic antimicrobial peptides (Abou Alaiwa et al., 2014) or through enhanced oxidative killing by the dual oxidase–lactoperoxidase system (Iovannisci et al., 2010). HBE mucus properties vary little in the range of pH 6.0–8 (Hill et al., 2018); however, the present results encourage studies of rheological properties and mucus plugging at higher pH (Tang et al., 2016). We observed similar pH oscillations on the nasal mucosae of male and female subjects and on cultured cells from male and female subjects; however, future studies should also examine sex differences systematically, as CF females acquire P. aeruginosa infections earlier in life (Demko et al., 1995) and have more frequent pulmonary exacerbations (Block et al., 2006). Estrogen may suppress pH oscillations by inhibiting HCO₃⁻ secretion (Sheridan et al., 2013) or by reducing the sensitivity of bacteria to pH by inducing mucoidy (i.e., bacterial secretion of a protective, alginate-containing matrix; Chotirmall et al., 2012).

Using alkalinization when air is inhaled to kill pathogens contained in that air requires rapid CO₂ hydration and dehydration reactions catalyzed by carbonic anhydrase. This may explain why mutations in CA12 produce lung disease with symptoms resembling those in CF (Lee et al., 2016). An American adult misdiagnosed with CF and a boy from a consanguineous Bedouin family who displayed classic CF symptoms without mutations in the cftr gene were both found to be homozygous for loss-of-function mutations in CA12. It would be interesting to measure ASL pH in these individuals to determine if an inability to alkalinize ASL pH during inhalation might explain their susceptibility to respiratory infections.

In vivo measurements of pH on the nasal mucosa were performed with informed written consent following protocol #MM-CODIM-16-296, which was approved by the West-Central-Montreal-Health’s Medical/Biomedical Research Ethics Committee. Nasal pH was measured in 10 adult control subjects: mean age, 30.8 yr (range, 24–38 yr); six males. Nasal pH was also measured in 11 subjects with CF carrying a range of CF mutations: mean age was 30.2 yr (range, 22–47 yr); eight subjects were males, and all genotypes are indicated in the figure legends.

For in vitro studies of pH on nasal and bronchial epithelial cell cultures, non-CF lung tissue was obtained from the International Institute for the Advancement of Medicine (Edison, NJ) and National Disease Research Interchange (Philadelphia, PA). CF bronchial tissue from patients undergoing lung transplantation was obtained with informed written consent from the Biobank of respiratory tissues at the Centre Hospitalier de l’Université de Montréal and Institut de recherche cliniques de Montréal. Tissue handling followed protocols approved by the institutional review boards of the Centre Hospitalier de l’Université de Montréal and McGill University (#A08-M70-14B). HBE cells were isolated in the Primary Airway Cell Biobank at the McGill Cystic Fibrosis Translational Research center as described previously (Fulcher et al., 2005; Matthes et al., 2018) and cultured for 1 mo under an air–liquid interface. All CF cells used for in vitro studies were homozygous for F508del-CFTR.

Electrodes ~1 m in length were constructed by heating and pulling 1-ml disposable polyethylene syringes (Becton Dickinson) into fine capillaries (O’Donnell, 1992). An internal reference solution (mM: 100 NaCl and 100 sodium citrate, pH 6) was injected into the back of the syringe. The tip was filled by capillarity with a 1–2-mm column of hydrogen ionophore I–cocktail B (Sigma-Aldrich) and coated with 15% (wt/vol) polyvinyl chloride in tetrahydrofuran to prevent leakage of the ion exchanger. A silver chloride wire was inserted into the internal reference solution and connected by a cable to a differential electrometer (HiZ-223; Warner Instruments) interfaced to a computer (PowerLab 2/26; AD Instruments). The reference electrode was filled with 3 M KCl in 3% agar, and the two flexible capillaries were glued together using 15% PVC (polyvinyl chloride). Capillary pH electrodes had response times <1 s. The voltage was digitized, displayed, and stored on a personal computer. The double-barreled ion-selective electrodes were calibrated in phosphate-buffered solutions adjusted to pH 6.5–9.5 before and immediately after recording nasal pH. The mean ± SE for slopes and r² values were −51.1 ± 3.0 mV and 0.9851 ± 0.0027, respectively, for n = 32 electrodes.

A new capillary electrode was used for each subject and was placed gently on the inferior nasal turbinate. Subjects were asked to hold their breath for 10 s, then take 5–10 slow breaths while pH was being recorded. Recordings were annotated to indicate inhalation and exhalation. In some experiments, the turbinate was lightly sprayed with solution containing 50 mM bicarbonate, 500 µM acetazolamide, or filter-sterilized pig gastric mucins (80 mg/ml). pH recordings were analyzed using LabChart 7 (AD Instruments). Statistics were performed using Prism 5 software (GraphPad).

Air–liquid interface cultures contained columnar cells with beating cilia and goblet cells with mucin granules. Mucus accumulated on the apical surface and was removed before most experiments by rinsing with PBS containing 5 mM N-acetylcysteine. Bronchial epithelial cultures were mounted in a closed chamber during CO₂ oscillations on the stage of an inverted fluorescence microscope (IX81; Olympus) equipped with an imaging system (Photon Technology International) as described (Kim et al., 2019). Cells were rinsed with artificial ASL (mM: 130 NaCl, 10 NaHCO₃, 5 KCl, 1 CaCl₂, 1 MgSO₄, 2.8 NaHepes, 2.2 Hepes acid, and 10 glucose) and covered with 30 µl artificial ASL for at least 30 min before measurements. The ratiometric fluorescent pH indicator BCECF conjugated to dextran (Molecular Probes; Thermo Fisher Scientific) was added to artificial ASL using the vehicle FC-72, a perfluorocarbon that evaporated rapidly from the surface. To mimic tidal breathing, cells were exposed to alternating flows of 0.035% and 5% CO₂ using a custom-designed gas-mixing pump equipped with a humidification system and interfaced to a computer (Okolab). CO₂ levels at the gas inlet and outlet ports of the perfusion chamber were measured using an infrared CO₂ analyzer (CD-3A; Applied Biosystem). In some experiments, purified carbonic anhydrase type 2 (1 unit; Sigma-Aldrich) or the carbonic anhydrase inhibitor acetazolamide (100 µM) or topiramate (200 µM) was added to the artificial ASL. The effects of mucins on pH oscillations were studied by adding porcine gastric mucins (20–80 mg/ml; Sigma-Aldrich) dissolved in artificial ASL.

Well-polarized primary cultures were scraped from Transwells and solubilized in RIPA lysis buffer (150 mmol/L NaCl, 1 mmol/L Tris/HCl, 1% deoxycholic acid, 1% Triton X-100, 0.1% SDS, and protease inhibitor) for 15–30 min at 22°C and centrifuged at 32,000 g for 15 min at 22°C. Pig tissues (trachea, skeletal muscle, and kidney cortex) were snap frozen in liquid nitrogen immediately after dissection and stored at −80°C for up to 2 wk, ground into a fine powder using mortar and pestle, then solubilized in RIPA buffer. HEK cells expressing endogenous CA12 or transfected with CA12 cDNA were used as positive controls in immunoblots. Cells at 50%–80% confluence were transfected in 100-mm Petri dishes with 4 µg CA12 cDNA (OriGene) using GeneJuice (Sigma-Aldrich) as per manufacturer’s instructions. Supernatants were collected and assayed for protein concentration (Bio-Rad). Samples containing 50 µg total protein were resolved using 10% SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 5% skimmed milk in TBST (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Tween 20) for 1 h at 22°C, then incubated with anti-CA12 (1:500; Proteintech) and anti-tubulin (clone KMX-1; 1:1,000; Millipore Sigma) overnight at 4°C. Cross-reactivity of the anti-CA12 antibody with the pig enzyme was established using skeletal muscle and renal cortex as negative and positive controls, respectively. Cross-reactivity of the anti-tubulin antibody with pig tubulin was confirmed by immunostaining cilia in pig trachea. The membranes were washed with TBST, and the bands were visualized using Amersham Bioscience ECL (enhanced chemiluminescence) detecting reagent (GE Healthcare) and a ChemiDoc Imaging System (Bio-Rad).

To predict airway surface pH during CO₂ oscillations, numerical simulations were performed using Matlab (The Mathworks). We assumed physiological dimensions for ASL thickness, L, so that diffusion effects would be negligible within the breathing period, T, which is justified since the effects of diffusion can be neglected when the ratio of the breathing period to the diffusion time is relatively large (DT/L² > 1), a condition that is satisfied for L < 65 µm and a period of 4 s for one breath. This enabled us to model bicarbonate-buffered ASL using a system of ordinary differential equations (ODEs).

This system of ODEs can be considered very stiff and was solved using ode15s with a very low absolute and relative tolerance (10⁻¹⁸ and 10⁻¹⁵, respectively) using the NonNegative option. To simulate the presence of carbonic anhydrase, the above system was modified by increasing kb₁ and kf₁ by a factor of 20,000. We modified kb₂ in Eq. 2 to be 2.14 × 10¹⁰/M/s to better match the pH values predicted by the Henderson-Hasselbalch equation for the bicarbonate buffer system. This modification yields a theoretical pK_a of 6.1 for the system defined by Eqs. 1 and 2, rather than 6.47 using the rate of of 5×10¹⁰/M/s used by Mitchell et al. (2010).

First-strand cDNA was prepared using 1 µg total RNA and SuperScript VILO MasterMix (Invitrogen). CA12 mRNA levels were determined using Taqman primers (Assay ID: CA12: Hs01080909_m1; Life Technologies). Reactions were analyzed using the Quant Studio 7 Flex Real-Time PCR system (Life Technologies) and “Fast” program: 95°C for 20 s, followed by 40 cycles of 95°C for 1 s and 60°C for 20 s. Data were normalized to GAPDH mRNA levels.

Tissue specimens were collected within 48 h of lung resection and fixed in 10% neutral buffered formalin at 4°C overnight for paraffin embedding and sectioning. Tissue sections (5 µm) were mounted on microscope slides, de-paraffinized in xylene (3 × 5 min), and rehydrated in alcohol baths (95%, 95%, 70%; 5 min each) followed by two 5-min rinses in double-distlled H₂O and equilibration in PBS. Sections were permeabilized using 1% Triton X-100 and blocked for 1 h with 2% BSA in PBS at room temperature, then incubated with primary antibody (rabbit anti-CA12 1:200; Proteintech Group Inc.; mouse anti-Tubulin 1:200; MilliporeSigma; or mouse anti-MUC5AC 1:200; Millipore Sigma) in 1% BSA/PBS for 1 h. They were then exposed to Alexa-488 goat anti-mouse and Alexa-594 goat anti-rabbit secondary antibody (1:200; Abcam) for 1 h and mounted in ProLong Diamond Antifade Mountant (Invitrogen). Immunofluorescence was imaged using an LSM-780 confocal microscope (Zeiss) with a multiline Argon laser (488 nm, 25 mW) and a 561-nm line laser (15 mW). For primary cells on filters, the apical side was incubated with PBS for 15 min at 37°C in 5% CO₂, rinsed three times to remove apical mucus, and immunostained as above. To image single cells, air–liquid interface cultures were washed with PBS at room temperature, gently scraped from the inserts into PBS, then centrifuged onto a glass slide using a Cytospin4 (500 rpm for 5 min; Thermo Fisher Scientific).

CFU assays of viable bacteria used the mucoid strain P. aeruginosa 508 kindly provided by Dr. Danuta Radzioch (Research Institute–McGill University Health Centre, Montréal, Canada). P. aeruginosa suspensions were prepared using fresh colonies from tryptic soy broth (TSB) agar plates as follows: bacteria were cultured overnight in 50 ml TSB (250 rpm, 37°C), then cultured in fresh TSB for 2–3 h until mid-log phase growth as determined by optical density at 600 nm. Bacteria were washed in sterile PBS (pH 7.2, 10 mM) and diluted to 10⁵–10⁶ cells/ml in epithelial secretions or artificial ASL with slight modifications as detailed in the legends for Fig. 6 and Fig. S5. Secretions were used immediately after collection from Calu-3 cells that had been cultured at an air–liquid interface for at least 10 d. Apical fluid was collected by pipette under control conditions or during stimulation with 10 µM FSK. TSB (1%) was added to artificial ASL as a nutrient source in some experiments (Fig. 6 E and Fig. S5). For the killing assays in Fig. 6, A, B, and F, bacteria and test fluids were placed in small polypropylene chambers with loose-fitting caps and incubated for up to 24 h. In the experiments shown in Fig. 6, C–E, a chamber was connected to the computer-controlled CO₂ gas-mixing pump shown in Fig. S1, and 500-µl aliquots containing bacteria were exposed to constant or oscillating PCO₂ for up to 6 h in a humidified incubator at 37°C. Bacteria were sampled from the chambers as indicated in Fig. 6 and plated in duplicate on agar plates. After incubating the plates overnight at 37°C, CFUs were counted using OpenCFU software (Geissmann, 2013).

Bioluminescence assays of bacterial killing used the PAO1-DN613 strain of P. aeruginosa in which the rpsL-lux reporter was integrated into the chromosome at the attB site using a pUC18-mini-Tn7T vector. Bacterial samples were prepared from fresh colonies on plates and grown to mid-logarithmic phase for 2.5 h after overnight culture. Bacteria were suspended in sterile PBS at 10⁹ CFUs/ml, then serially diluted to 10²/ml in bicarbonate or phosphate buffer (100 mM). To calibrate the bioluminescence signal, 100-µl aliquots of bacteria at each dilution were transferred into tubes, and their bioluminescence was measured using a luminometer (TD-20/20; Turner Designs). To assess viability under different conditions, the samples were placed in an incubator shaker at 36°C, incubated with orbital agitation at 200 rpm for 0, 1, and 2 h, and then vortexed briefly before measuring bioluminescence.

Requests for computer code should be addressed to Nathan B. Scales: [email protected].

Fig. S1 shows the programmable fast gas-mixing pump, which is used to mimic variations in PCO₂ during tidal breathing when measuring the surface pH on well-differentiated primary bronchial epithelial cells in vitro. Fig. S2 shows that CA12 protein is expressed in ciliated cells but not secretory cells. Fig. S3 shows the specificity of the CA12 antibody used for immunostaining human lung tissue. Fig. S4 shows that CA12 immunostaining is reduced in CF bronchial and nasal cell cultures compared with non-CF cell cultures. Fig. S5 shows that HCO₃⁻-dependent killing is enhanced by PCO₂ oscillations in the presence of nutrients.

We are grateful to the research subjects who volunteered to participate in the clinical studies. We thank Dr. Jacopo Mortola (McGill University, Montréal, Québec, Canada) for access to infrared CO₂ monitoring equipment, Dr. Danuta Radzioch (Research Institute–McGill University Health Centre, Montréal, Québec, Canada) for bacterial strains, Véronique Flynn and Olymel S.E.C. (Saint-Esprit, Québec, Canada) for providing pig tissue, and Hong-Seok Hur for preparing the graphical abstract and graphic in Fig. 1.

This work was supported by research grants from Cystic Fibrosis Canada (grant no. 3022) and the Canada Foundation for Innovation (grant no. 21375) to J.W. Hanrahan and Cystic Fibrosis Canada (grant no. 555681) to J.P. Ianowski.

Author contributions: D. Kim, S. Frenkiel, E. Matouk, R. Robert, and J.W. Hanrahan planned the project; D. Kim, C. Martini, M.A. Tewfik, C.D. Poirier, X. Luan, G.A. McKay, D. Nguyen, and J.P. Ianowski developed the tools; N.B. Scales wrote the software and performed the computer simulations; D. Kim, J. Liao, A. Abu-Arish, Y. Luo, and S. Frenkiel performed the experiments; D. Kim, J. Liao, and N.B. Scales analyzed the data; and D. Kim and J.W. Hanrahan wrote the manuscript. All authors helped to interpret the results and reviewed and approved the manuscript.