Ozone activates airway nerves via the selective stimulation of TRPA1 ion channels

Abstract

Inhalation of ozone is a major health risk in industrialized nations. Ozone can impair lung function and induce respiratory symptoms through sensory neural-mediated pathways, yet the specific interaction of ozone with airway sensory nerves has yet to be elucidated. Here we demonstrate, using a vagally innervated ex vivo tracheal–lung mouse preparation, that ozone selectively and directly evokes action potential discharge in a subset of nociceptive bronchopulmonary nerves, namely slow conducting C-fibres. Sensitivity to ozone correlated with the transient receptor potential (TRP) A1 agonist, cinnamaldehyde, with ozone having no effect on cinnamaldehyde-insensitive fibres. C-fibre responses to ozone were abolished by ruthenium red (TRP inhibitor). Ozone also stimulated a subset of nociceptive sensory neurones isolated from vagal ganglia of wild-type mice, but failed to activate neurones isolated from transient receptor potential ankyrin 1 (TRPA1) knockout mice. Ozone activated HEK293 cells transfected with TRPA1, but failed to activate non-transfected HEK293 or HEK293 transfected with the capsaicin-sensitive transient receptor potential vanilloid 1 (TRPV1) channel. Thus, ozone is not an indiscriminate neuronal activator, but rather it potently and selectively activates a subset of airway C-fibres by directly stimulating TRPA1.

Introduction

Tropospheric ozone has been associated in epidemiological studies (reviewed in Mudway & Kelly, 2000) with decreased pulmonary function and increased hospitalization, particularly in subjects with pre-existing airway disease. Exposure studies with volunteers have shown that ozone acutely causes cough, increased airway-specific resistance, decreased inspiratory capacity, decreased forced vital capacity (FVC), decreased forced expiratory volume in 1 s (FEV₁), chest pain, inspiratory pain and shortness of breath at concentrations observed in the troposphere (Kerr et al. 1975; Hazucha et al. 1989; McDonnell et al. 1999; Mudway & Kelly, 2000; Uysal & Schapira, 2003). In addition, ozone causes non-specific hyperreactivity, lung inflammation (characterized by a pronounced neutrophilia), pulmonary oedema and airway epithelial damage over a 24 h period (reviewed in Mudway & Kelly, 2000). Clinical and animal studies suggest that the cough and noxious sensations and at least a component of the decrements in lung function caused by ozone exposure are mediated through sensory neural mechanisms (Lee et al. 1979; Hazucha et al. 1989; Schelegle et al. 1993, 2001; Tepper et al. 1993). There is, however, no mechanistic information on how ozone interacts with sensory nerves.

The airways are densely innervated by sensory nerves, the majority of which are afferent C-fibres adapted to detect noxious stimuli (Carr & Undem, 2003). Activation of vagal bronchopulmonary C-fibre afferent nerves can result in sensations and reflex-mediated decrements in lung function similar to those observed with ozone exposure (Coleridge & Coleridge, 1984). Ozone may indeed cause these effects through the direct activation of bronchopulmonary C-fibres but surprisingly little attention has been given to this hypothesis, probably due to methodological issues. In dogs, ozone (2–3 ppm) administered via a ventilator has been associated with an increase in action potential discharge in rapidly adapting receptors (RAR fibres), but this would appear to be an indirect consequence of lung compliance. Ozone also led to an increase in action potential discharge in bronchial C-fibres (Coleridge et al. 1993). However, in rat lungs ozone failed to overtly activate C-fibres, but increased their response to subsequent exposure to capsaicin (Ho & Lee, 1998). Several studies have reported cellular and physiological responses to ozone in the lungs that are sensitive to neurokinin receptor blockade (Graham et al. 2001; Fu et al. 2002; Oslund et al. 2008). These studies indirectly support the idea that ozone in some fashion leads to the stimulation of tachykinergic C-fibres in the lungs. Bronchopulmonary C-fibres selectively express ion channel proteins that are gated by noxious temperature, osmotic and chemical stimuli (Taylor-Clark & Undem, 2006). In particular, recent studies have found that transient receptor potential vanilloid 1 (TRPV1) and transient receptor potential ankyrin 1 (TRPA1) are ion channels responsible for the action potential initiation in nociceptive nerves in response to a wide range of inflammatory mediators and noxious irritants (Bessac & Jordt, 2008).

In the present study, we addressed the hypothesis that ozone can directly lead to the activation of mouse bronchopulmonary C-fibres. In addition, we evaluated the ionic mechanisms that may underlie this response. The data obtained support the conclusion that ozone evokes high-frequency action potential discharge in bronchopulmonary C-fibres, and this is a direct effect that occurs via the gating of TRPA1 channels within the C-fibre membranes.

Methods

All experiments were approved by the Johns Hopkins Animal Care and Use Committee and they comply with regulations and policies stipulated by The Journal of Physiology (Drummond, 2009). In total 30 mice were used in this study.

Solutions

Ozone gas was generated at a rate of 200 mg h⁻¹ by passing dry oxygen through a silent arc discharge (Model 200, Sanders Ozonizer). The resultant gas was bubbled (using a glass fritted gas dispersion tube, Pyrex) into a small glass vial (28 mm diameter) containing 4–5 ml of 1 mm H₃PO₄ at a flow rate of approximately 1 ml s⁻¹. All ozone solutions were generated and used in a laboratory with a stabilized atmospheric environment at 24°C. Preliminary tests indicated that the concentration of solubilized ozone was maximal when the meniscus level of the H₃PO₄ was barely above the fritted portion of the inserted dispersion tube and the flow rate was not more than 5 ml s⁻¹. Solubilized ozone concentration was determined by analysing the absorption spectra at 250 nm (Abs_250nm) of the solution in a 1 cm² quartz cuvette using a Jenway 6405 UV Spectrophotometer (Bibby Scientific, Burlington, NJ, USA) and applying the formula [ozone]= Abs_250nm/ɛ(ozone), where ɛ(ozone) is the molar absorptivity of ozone at 250 nm, i.e. 2900 m⁻¹ cm⁻¹ (Hart et al. 1983). Ozone concentrations were measured 2–3 s after removal from the ozone gas source. Maximal [ozone] was attained after 10 min of ozone gas bubbling through the H₃PO₄ solution and, although acidic pH is known to increase ozone solubility/stability in solution, increasing the H₃PO₄ from 1 mm to 10 mm and 100 mm had no effect on the maximal [ozone] or the time taken to attain it. Mean maximal [ozone] in 1 mm H₃PO₄ was calculated to be 327 μm, which is consistent with other reports (Hart et al. 1983; Lotriet et al. 2007). Standard deviation for these solutions over the course of the entire study was 25 μm, although on any give day the standard deviation was consistently around 10 μm. Ozone in acidic solution is known to have a half-life of approximately 50 min and we observed similar stability with our solutions (data not shown).

Next, we investigated the stability of ozone in biological buffer solutions. Ozone is known to react with a wide variety of molecules, in particular organic compounds. For in vitro fura-2 AM calcium assays we have previously employed Locke solution: 34°C; composition (mm): 136 NaCl, 5.6 KCl, 1.2 MgCl₂, 2.2 CaCl₂, 1.2 NaH₂PO₄, 14.3 NaHCO₃ and 10 d-glucose (gassed with 95% O₂–5% CO₂, pH 7.3–7.4). In these preliminary experiments we investigated the stability of ozone in Locke solution with 10 mm, 1 mm and 0 mm glucose. To minimize decay before measurement, 2 ml of Locke solution (34°C) were placed in the quartz cuvette in the Jenway spectrophotometer and a kinetic analysis was started (recording every 250 ms). Ozone (0.2 ml of in 1 mm H₃PO₄) was added directly to the cuvette and pipette-mixed rapidly with the Locke solution. Absorbance was recorded continuously and reliable data were attained <1.5 s after solubilized ozone was added. Absorbance data were converted to [ozone] and normalized to the [ozone] at the first data point after solubilized ozone was added to the cuvette (t= 0). The effect of glucose on ozone stability in Locke solution was dramatic: [ozone] decay followed ‘first order’-like kinetics with a half-life of 5.4 ± 0.6 s for 10 mm glucose (n= 4), 11.0 ± 1.3 s for 1 mm glucose (n= 4) and 121.0 ± 38.3 s for 0 mm glucose (n= 16) (Fig. 1A). In addition, we investigated the reproducibility and predictability of [ozone] in glucose-free Locke solutions (generated from the addition of 1 mm H₃PO₄ solutions of ozone to glucose-free Locke solution). The correlation of recorded [ozone]_{glucose-free Locke} to the predicted [ozone] (calculated from the original recorded [ozone] in the H₃PO₄ solution) was extremely strong (Fig. 1B, fitted with linear regression to yield an equation of y= 1.03x+ 0.33 (n= 13)), even at low micromolar concentrations. Finally, we briefly investigated the effect of H₃PO₄ on the pH of glucose-free Locke solution. Using an AB15 accumet pH meter (Fisher Scientific, Pittsburgh, PA, USA), the pH of a 10 ml Locke solution (pH 7.3–7.4) did not decrease with addition of up to 10 ml of 1 mm H₃PO₄. Overall, the data demonstrate that glucose-free Locke solution is a suitable medium for studying solubilized ozone. Identical experiments using glucose-free Krebs solution showed similar results (data not shown).

Figure 1. Ozone stability in Locke solution.

A, effect of glucose concentration (0, 1 and 10 mm) on ozone stability as measured by the absorption at 250 nm. Data normalized to the [ozone] at t= 0. B, correlation between recorded [ozone] and the predicted [ozone] based on calculations from the stock ozone/H₃PO₄ solution.

C-fibre extracellular recordings

Mice were killed by CO₂ asphyxiation followed by exsanguination. The innervated isolated trachea/bronchus preparation was prepared as previously described (Nassenstein et al. 2008). Briefly, the airways and lungs with their intact extrinsic innervation (vagus nerve including vagal ganglia) were taken and placed in a dissecting dish containing Krebs bicarbonate buffer solution composed of (mm): 118 NaCl, 5.4 KCl, 1.0 NaH₂PO₄, 1.2 MgSO₄, 1.9 CaCl₂, 25.0 NaHCO₃ and 11.1 d-glucose, and equilibrated with 95% O₂ and 5% CO₂ (pH 7.2–7.4). Connective tissue was trimmed away leaving the trachea and lungs with their intact nerves. The airways were then pinned to the larger compartment of a custom-built two-compartment recording chamber which was lined with silicone elastomer (Sylgard). A vagal ganglion was gently pulled into the adjacent compartment of the chamber through a small hole and pinned. Both compartments were separately superfused with Krebs bicarbonate buffer containing d-glucose (37°C). A sharp glass electrode was pulled by a Flaming Brown micropipette puller (P-87; Sutter Instruments, Novato, CA, USA) and filled with 3 m NaCl solution. The electrode was gently inserted into the vagal ganglion so as to be placed near the cell bodies. The recorded action potentials were amplified (Microelectrode AC amplifier 1800; A-M Systems, Everett, WA, USA), filtered (0.3 kHz of low cut-off and 1 kHz of high cut-off), and monitored on an oscilloscope (TDS340; Tektronix, Beaverton, OR, USA) and a chart recorder (TA240; Gould, Valley View, OH, USA). The scaled output from the amplifier was captured and analysed by a Macintosh computer using NerveOfIt software (Phocis, Baltimore, MD, USA). To measure conduction velocity, an electrical stimulation (S44; Grass Instruments, Quincy, MA, USA) was applied to the centre of the receptive field. NerveOfIt software analysis was also able to discriminate individual nerve fibre responses on the rare occasion that more than one bronchopulmonary afferent was recorded from during stimulation (electrical, mechanical or chemical) of the lung tissue. The conduction velocity of the individual bronchopulmonary afferents was calculated by dividing the distance along the nerve pathway by the time delay between the shock artifact and the action potential evoked by electrical stimulation. Drugs were intratracheally applied as a 1 ml bolus over 10 s. Ozone and its vehicle were given as a 1 ml bolus of glucose-free Krebs solution. A 1 mm H₃PO₄ solution containing a known [ozone] was added to 1 mm H₃PO₄ to give an [ozone] of 6.667 times the final desired concentration. A 150 μl sample of this ozone and H₃PO₄ solution was mixed with 850 μl of glucose-free Krebs solution immediately prior to bolus injection to give the desired [ozone]_{glucose-free Krebs}.

In the extracellular recording studies, the action potential discharge was quantified off-line and recorded in 1 s bins. A response was considered positive if the number of action potentials in any 1 s bin was >2 times the average baseline response. The baseline activity was usually either absent or less than 2 Hz. The peak frequency evoked by a stimulus was quantified as the maximum number of action potentials that occurred within any 1 s bin less the mean baseline frequency. The total number of action potentials (AP_total) was quantified as the total action potential count during the response (AP_recorded) minus the mean baseline frequency (Freq_BL, in s⁻¹) multiplied by the duration of the response (t, in s) (AP_total= AP_recorded− (Freq_BL×t)).

HEK293 cell culture

Wild-type HEK293 cells, cells stably expressing human TRPA1 (hTRPA1-HEK) or human TRPV1 (hTRPV1-HEK) were used in this study, as previously described (Taylor-Clark et al. 2008b). Cells were maintained in an incubator (37˚C, 5% CO₂) in DMEM (containing 110 mg l⁻¹ pyruvate and 564 mg l⁻¹l-glutamine) supplemented with 10% FBS and 500 μg ml⁻¹ Geneticin as a selection agent. Cells were removed from their culture flasks by treatment with Accutase (Sigma), then plated onto poly-d-lysine-coated coverslips and incubated at 37˚C for >1 h before experimentation.

Dissociation of mouse vagal ganglia

Mouse vagal ganglia were isolated and enzymatically dissociated from wild-type C57BL/6J mice and TRPA1^−/− mice using previously described methods (Taylor-Clark et al. 2008a). Isolated neurones were plated onto poly-d-lysine-coated and laminin-coated coverslips and used within 24 h.

Calcium imaging

HEK293-covered coverslips were loaded with fura-2 acetyoxymethyl ester (fura-2 AM; 8 μm) (Molecular Probes, Carlsbad, CA, USA) in DMEM (containing 110 mg l⁻¹ pyruvate and 564 mg l⁻¹l-glutamine) supplemented with 10% FBS and incubated (40 min, 37°C, 5% CO₂). Neurone-covered coverslips were loaded with fura-2 AM (8 μm) in L-15 media containing 10% FBS and incubated (40 min, 37°C). For imaging, the coverslip was placed in a custom-built chamber (bath volume of 600 μl) and superfused at 4 ml min⁻¹ with glucose-free Locke solution (34°C; composition (mm): 136 NaCl, 5.6 KCl, 1.2 MgCl₂, 2.2 CaCl₂, 1.2 NaH₂PO₄, 14.3 NaHCO₃ (gassed with 95% O₂–5% CO₂, pH 7.3–7.4)) for 15 min before and throughout each experiment by an infusion pump. Changes in intracellular free calcium concentration (intracellular [Ca²⁺]_free) were measured by digital microscopy (Universal; Carl Zeiss, Inc., Thornwood, NY, USA) equipped with in-house equipment for ratiometric recording of single cells. The field of cells was monitored by sequential dual excitation, 352 and 380 nm, and the analysis of the image ratios used methods previously described to calculate changes in intracellular [Ca²⁺]_free (Taylor-Clark et al. 2008b). The ratio images were acquired every 6 or 12 s. Superfused buffer was stopped 30 s before each drug application, when 300 μl buffer was removed from the bath and replaced by 300 μl of 2× test agent solution added between image acquisitions. At the end of the dissociated vagal neuronal studies, neurones were exposed to KCl (30 s, 75 mm) to confirm voltage sensitivity. At the end of all calcium imaging experiments, both neurones and HEK cells were exposed to ionomycin (30 s, 1 μm) to obtain a maximal response.

Solutions of ozone were generated by the addition of 1 mm H₃PO₄ solution containing a known [ozone] to glucose-free Locke solution. Therefore, we first performed preliminary studies to determine the effect of vehicle on hTRPA1-HEK cells. Serial dilution of glucose-free Locke solution with 1 mm H₃PO₄, ranging from (H₃PO₄ : Locke) 1 : 40 to 1 : 2, was applied to hTRPA1-HEK cells followed by the selective TRPA1 agonist allyl isothiocyanate (AITC, 100 μm). Vehicle had no effect on hTRPA1-HEK cells as measured in the fura-2 AM assay, although the 1 : 2 1 mm H₃PO₄ demonstrated minor activation together with visible deformation of the cells (Table 1 and data not shown). As expected, AITC activated the hTRPA1-HEK cells (maximum response 60.5 ± 3.0% of ionomycin, n= 64). From these data, we chose a 1 : 6 dilution as being a suitable vehicle for these studies. Thus glucose-free Locke solutions of ozone were generated by the addition of 1 mm H₃PO₄ solution containing a known [ozone] to 1 mm H₃PO₄ to give an [ozone] of 6 times the final desired concentration. A 100 μl aliquot of this ozone and H₃PO₄ solution was mixed with 200 μl of glucose-free Locke solution immediately prior to the 300 μl of glucose-free Locke solution in the chamber to give the desired [ozone]_{glucose-free Locke}. In this way the glucose-free Locke solution was diluted by the same amount of 1 mm H₃PO₄ for all [ozone]_{glucose-free Locke}.

Table 1.

Response of hTRPA1-HEK to dilution of Locke_glucose-free with H₃PO₄

Treatment	Molarity (mosmol l−1)	pH	Change in [Ca2+] (%)
1 : 40	332.6	7.4	0.35 ± 0.2
1 : 20	324.0	7.4	−0.49 ± 0.2
1 : 10	307.0	7.4	−1.25 ± 0.2
1 : 4	255.8	7.4	−0.76 ± 0.2
1 : 2	170.6	7.4	2.15 ± 1.7
AITC	341.1	7.4	60.5 ± 3.0

Data represent mean ±s.e.m. Ca²⁺ responses of hTRPA1-HEK cells (n= 64). Osmolarity of the solution (mosmol l⁻¹) was calculated from dilution of Locke_glucose-free.

For the analysis of fura-2 AM-loaded cells, the measurement software converted ratiometric information to intracellular [Ca²⁺]_free using Tsien parameters ([Ca²⁺]=K_d((R−R_min)/(R_max−R))(b)) particular to this instrumentation and the HEK cells and dissociated mouse vagal neurones (Taylor-Clark et al. 2008a). Preliminary calibration studies yielded an R_min (352/380 nm ratio under calcium-free conditions) of 0.3 for both HEK cells and mouse sensory neurones and an R_max (352/380 nm ratio under calcium-saturating conditions) of 18 and 14 for HEK cells and neurones, respectively. These values were not appreciably different between cells exposed to glucose-free Locke solution and normal Locke solution (containing 10 mm d-glucose). b (380 nm in calcium-free conditions/380 nm in calcium-saturating conditions) was estimated as being 10 and the K_d was estimated as being 224 nm. In the following experimental studies we did not specifically calibrate the relationship between ratiometric data and absolute calcium concentration for each specific cell, choosing instead to use the parameters provided from the calibration studies and relate all measurements to the peak ionomycin response in each viable cell. This effectively provided the needed cell-to-cell calibration for enumerating individual cellular responses. Only cells that had a robust response to ionomycin were included in analyses. At each time point for each cell, data were presented as the percentage change in intracellular [Ca²⁺]_free, normalized to ionomycin:

where [Ca²⁺]_x was the apparent [Ca²⁺]_free of the cell at a given time point, [Ca²⁺]_bl was the cell's mean baseline apparent [Ca²⁺]_free measured over 120 s, and [Ca²⁺]_max was the cell's peak apparent [Ca²⁺]_free during ionomycin treatment. For the neuronal experiments, neurones were defined as ‘responders’ to a given compound if the mean response was greater than the mean baseline plus 2× the standard deviation. Only neurones that responded to KCl were included in analyses. Given that vagal ganglia are probably composed of heterogeneous neuronal populations, it is important to emphasize the point that results are presented in two distinct ways. Firstly, the number of neurones responding (based on the criteria described above) to a given stimulus compared to the total number of neurones is reported. Secondly, the mean percentage change in intracellular [Ca²⁺]_free normalized to ionomycin of those neurones that (based on the above criteria) were defined as ‘responders’ is reported. Perfusion with glucose-free Locke solution for up to 40 min had no effect on the baseline [Ca²⁺]_{intracellular}.

Chemicals

Stock solutions (200×+) of all agonists were dissolved in 100% ethanol (final concentration of 0.5% ethanol or less). Fura-2 AM was purchased from Molecular Probes. All other chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA).

Results

To investigate whether ozone could directly activate bronchopulmonary C-fibres, we applied ozone to C-fibre receptive fields in a vagally innervated ex vivo tracheal–lung mouse preparation (Kollarik et al. 2003). Using this technique we were able to record action potential discharge in individual bronchopulmonary C-fibres in response to ozone, independently of the effects of blood pressure, or changes in lung compliance and resistance. Bronchopulmonary C-fibres that conduct action potentials below 0.7 m s⁻¹ are typically activated by the selective TRPV1 agonist capsaicin and are considered nociceptive (Kollarik et al. 2003). In seven bronchopulmonary C-fibres, the vehicle for ozone (150 μl of 1 mm H₃PO₄ in 1 ml bolus of glucose-free Krebs solution) failed to elicit action potential discharge. By contrast, the C-fibres innervating the airway responded strongly to exposure to ozone (10–150 μm) with action potential discharge (Table 2).

Table 2.

Effect of cinnamaldehyde sensitivity on the activation of bronchopulmonary C-fibres by ozone

	Cinnamaldehyde sensitive	Cinnamaldehyde insensitive
Vehicle	0/5	0/2
10 μm O3	3/6	0/4
30 μm O3	10/12	0/7
150 μm O3	2/2	2/3

Data represent number of ozone-sensitive fibres out of total fibre number.

We have previously shown that TRPV1-expressing slow conducting mouse bronchopulmonary C-fibres also selectively express TRPA1 channels and evoke action potentials in response to the selective TRPA1 agonist cinnamaldehyde (300 μm) in a ruthenium red-sensitive manner (Nassenstein et al. 2008). In the present study, cinnamaldehyde (300 μm) activated 12 out of 19 fibres, with a peak frequency of 10.5 ± 1.0 Hz. The conduction velocity of cinnamaldehyde-sensitive fibres (mean of 0.51 m s⁻¹, range 0.4–0.6 m s⁻¹) was slower than the cinnamaldehyde-insensitive fibres (mean of 0.68 m s⁻¹, range 0.5–1.0 m s⁻¹).

We noted that ozone selectively stimulated the C-fibres that responded to the TRPA1 agonist cinnamaldehyde. At 30 μm, ozone stimulated 10 of 12 cinnamaldehyde-responsive C-fibres, but 0 of 7 C-fibres that were cinnamaldehyde-insensitive responded to ozone. Quantitatively the ozone response correlated well with the cinnamaldehyde response for each given fibre (r= 0.61, P < 0.01, data not shown). At 150 μm ozone activated C-fibres irrespective of cinnamaldehyde sensitivity. The ozone-induced C-fibre activation was typically characterized by a delay in action potential initiation of approximately 10–40 s, with the response lasting up to 3 min before returning to baseline (<0.4 Hz) (Fig. 2A). This compares with a quicker but shorter-lived response to cinnamaldehyde (data not shown). When the C-fibre was exposed to a second ozone treatment 15 min after the first response had returned to baseline, there was no evidence of tachyphylaxis. The first ozone response had a peak discharge of 6.5 ± 1.5 Hz, total number of action potentials 116 ± 31, and the second ozone response had a peak discharge of 7.3 ± 1.3 Hz, total number of action potentials 210 ± 47. The slight increase in the second response was statistically significant (P < 0.05).

Figure 2. Ozone elicits action potential discharge in bronchopulmonary C-fibres in a ruthenium red-sensitive manner.

A, representative trace of action potential discharge from a single cinnamaldehyde-sensitive bronchopulmonary C-fibre evoked by repeated doses of ozone (30 μm) in the absence (1st and 2nd controls) and presence of ruthenium red (30 μm). B and C, mean ±s.e.m. Action potential discharge from individual identified bronchopulmonary C-fibres to cinnamaldehyde (CA, 300 μm) and ozone (30 μm) in the absence and presence of ruthenium red (30 μm). B, peak action potential discharge. C, total number of action potentials. *Significant inhibition of ozone response by ruthenium red (P < 0.05, n= 5).

The correlation of ozone-induced responses with cinnamaldehyde suggested that the presence of TRPA1 channels on the peripheral terminals of bronchopulmonary C-fibres was required for ozone-induced activation at 30 μm. We tested this hypothesis by evaluating the effect of the non-selective TRPA1/TRPV1 channel blocker ruthenium red on ozone-sensitive C-fibres. We used a concentration of ruthenium red, 30 μm for 15 min, that we previously found was effective at blocking the response of the C-fibres to TRPA1 agonists (Nassenstein et al. 2008). In five fibres, ruthenium red virtually abolished the ozone response (P < 0.05). After treatment with ruthenium red, ozone (30 μm) only activated 1 out of 5 (peak discharge of 1.4 ± 0.8 Hz, total number of action potentials 16.2 ± 7.6; P < 0.01) (Fig. 2A–C).

The evidence suggests that ozone (30 μm) may selectively activate bronchopulmonary C-fibres through the activation of the ion channel TRPA1. To more specifically address this hypothesis we examined the effect of ozone in HEK293 cells stably transfected with human TRPA1 (hTRPA1-HEK) or human TRPV1 (hTRPV1-HEK) using fura-2 AM calcium imaging. For all experiments a 1 : 6 bolus of 1 mm H₃PO₄ was used as the vehicle for solubilized ozone (see Methods). This vehicle failed to activate hTRPA1-HEK, hTRPV1-HEK and non-transfected HEK293 (ntHEK) cells (maximum response 1.0 ± 0.1% of ionomycin, 1.2 ± 0.2% of ionomycin and 0.8 ± 0.1% of ionomycin, respectively) (Fig. 3A).

Figure 3. Activation of hTRPA1-HEK cells by ozone.

A, mean ±s.e.m. Ca²⁺ responses of hTRPA1-HEK (black line, n= 447), hTRPV1-HEK (grey line, n= 188) and ntHEK cells (broken line, n= 389) to vehicle and ozone (3 μm). Drugs were applied for 60 s (black line). B, dose–response relationships of Ca²⁺ responses of hTRPA1-HEK (black squares), hTRPV1-HEK (grey circles) and ntHEK cells (open triangles) to ozone (300 nm to 30 μm) (data comprise >188 cells). Data represent the maximum response during the 60 s agonist treatment taken from mean cell response versus time curves (note that the s.e.m. is contained within symbol).

Ozone (3 μm, 60 s treatment) failed to stimulate a substantive calcium response in hTRPV1-HEK or in non-transfected HEK cells (maximum response 0.5 ± 0.1% of ionomycin and 0.3 ± 0.1% of ionomycin, respectively) (Fig. 3A). As a positive control, capsaicin (500 nm) robustly activated hTRPV1-HEK cells (maximum response 51.2 ± 1.5% of ionomycin, data not shown). By contrast, ozone (3 μm) caused a robust increase in intracellular calcium in hTRPA1-HEK cells (maximum response 26.6 ± 1.2% of ionomycin) (Fig. 3A). The response to ozone was detectable at the first time point after ozone exposure (recording every 12 s) and reached a maximum within 40 s.

The concentration–response analysis for ozone-induced activation of hTRPA1-HEK revealed an EC₅₀ of ∼3 μm and a maximum response of 46.2 ± 1.0% of ionomycin (n= 447) (Fig. 3B). Only at the largest concentration tested, 30 μm, were non-selective effects observed (i.e. minor effects in ntHEK cells that averaged 10.8 ± 0.9% of ionomycin (n= 389), respectively). At this non-selective concentration there was as expected a similar small response noted in hTRPV1-HEK cells.

We next evaluated whether ozone could activate vagal C-fibre neurones, isolated from all other tissue elements, via TRPA1 channel activation. The neurones in the vagal sensory ganglia were dissociated and isolated, and evaluated using fura-2 AM calcium imaging. Our previous studies using this experimental design have shown that TRPA1 channels are expressed by 30–65% of vagal neurones (Nassenstein et al. 2008; Taylor-Clark et al. 2008a, 2009a). Vagal neurones from wild-type mice were treated with vehicle (a 1 : 6 bolus of 1 mm H₃PO₄ in glucose-free Locke solution, 60 s treatment), ozone (10 μm, 60 s), the selective TRPA1 agonist AITC (100 μm, 30 s) and the selective TRPV1 agonist capsaicin (1 μm, 30 s). Vehicle failed to activate wild-type vagal neurones, whereas ozone activated 75 out of 130 neurones (maximum response 22.8 ± 2.6% of ionomycin) (Fig. 4). The response to ozone was detectable at the first time point after ozone exposure (recording every 6 s) and reached a maximum within 90 s. The response was not reversible within the limits of the study (4 min of washout). Subsequent AITC treatment had little effect on top of the ozone response, consistent with the lack of additivity we have observed with other TRPA1 agonists (Taylor-Clark et al. 2008a, 2009a,b). Among the 55 neurones that failed to respond to ozone, all but 3 also failed to respond to the TRPA1 agonist AITC. Capsaicin, however, activated 16 of the remaining 55 ozone-insensitive neurones (data not shown).

Figure 4. Ozone fails to activate TRPA1^−/− vagal neurones.

Mean ±s.e.m. Ca²⁺ responses of vagal neurones responding to ozone (10 μm). Response to vehicle, AITC (100 μm) and capsaicin (Caps, 1 μm) also shown. Data comprised of neurones from wild-type mice (black squares, 75 out of 130 neurones responding) and neurones from TRPA1^−/− mice (grey squares, 14/107). Black line denotes the application of agonist. All neurones responded to KCl (75 mm) applied immediately prior to ionomycin.

To further address the hypothesis regarding the molecular identity of the ozone-sensitive ion channel responsible for the activation of vagal sensory neurones, we compared the responses of wild-type neurones to vagal neurones isolated from TRPA1^−/− animals. Ozone had little effect on TRPA1^−/− vagal neurones. In vagal neurones derived from TRPA1^−/− mice, ozone exposure was associated with a minor elevation in only 14/107 neurones. The magnitude of the response in these neurones was <10% of that observed in wild-type neurones (maximum response 2.2 ± 0.8% of ionomycin) (Fig. 4). As expected, the selective TRPA1 agonist AITC also failed to activate TRPA1^−/− neurones, whereas capsaicin evoked robust responses in 60/107 neurones (data not shown), including 13 of the 14 ozone-sensitive neurones (maximum response of 52.7 ± 6.5% of ionomycin) (Fig. 4).

Discussion

Inhalation of ozone, one of the major air pollutants in the developed world, causes noxious respiratory sensations and decrements in lung function. Responses to ozone include cough, chest pain, sore throat, decreases in FEV₁ and increases in airway resistance – symptoms that are consistent with the activation of airway sensory neural pathways (Mudway & Kelly, 2000; Uysal & Schapira, 2003). The data presented here provide evidence that ozone potently, effectively and directly evokes action potential discharge in vagal bronchopulmonary C-fibres. The data also support the hypothesis that this occurs primarily by an effect of ozone on the C-fibre leading selectively to the opening of TRPA1 cation channels. These mechanistic studies obtained ex vivo, are consistent with studies in dogs and rats that show vagal cooling (which prevents action potential propagation) inhibited several respiratory responses evoked by ozone (Lee et al. 1979; Schelegle et al. 1993). They are also consistent with studies in human volunteers showing that local anaesthetic inhalation inhibits ozone-induced respiratory symptoms (Hazucha et al. 1989; Schelegle et al. 2001).

Ozone can have many effects in the lungs that may in turn lead to afferent nerve activation. For example, ozone causes a decrease in lung compliance in ventilated animals (although not in spontaneously breathing humans; Kerr et al. 1975; Hazucha et al. 1989), which is thought to explain the activation of rapidly adapting vagal afferent mechanonsensors (Coleridge et al. 1993), as this response was reversibly abolished by brief hyperinflations. Certain vagal C-fibres in the lungs are sensitive to haemodynamic effects such as venous congestion (Roberts et al. 1986). Therefore, to discern whether ozone can have a direct effect on afferent nerves, we used the isolated vagally innervated lung preparation. That ozone potently (∼10 μm) was capable of evoking action potential discharge in vagal C-fibres ex vivo shows that the neuronal activation can occur independently of effects on blood flow and respiration.

Stimulators of the capsaicin receptor TRPV1 evoke action potential discharge in the vast majority of bronchopulmonary C-fibres in all mammals studied to date. More recently, TRPA1 has been found to be co-localized with TRPV1 in mouse bronchopulmonary C-fibres (Nassenstein et al. 2008), and TRPA1 agonists, like TRPV1 agonists, evoke robust action potential discharge in these nerves (Nassenstein et al. 2008; Taylor-Clark et al. 2008a, 2009a). In the present study, ozone (10–30 μm) only evoked action potentials in cinnamaldehyde-sensitive bronchopulmonary C-fibres, and the response to ozone significantly correlated with that observed with cinnamaldehyde. In addition, the ozone-induced action potential discharge was blocked following pre-treatment with ruthenium red, a non-selective inhibitor of TRPA1 and various TRPV channels. Overall, these data are consistent with an involvement of TRPA1 in the ozone effect.

The hypothesis that ozone activates TRPA1 channels was more specifically addressed in fura-2 AM calcium studies of neurones and hTRPA1-HEK, hTRPV1-HEK and ntHEK cells. The data from transfected HEK cells provide direct evidence that ozone effectively stimulates TRPA1, but fails to stimulate TRPV1. Ozone proved potent (EC₅₀∼3 μm) and selective at concentrations from 300 nm to 30 μm. At larger concentrations, ozone, not surprisingly, led to minor increases in calcium by mechanisms independent of TRPA1. In primary cultures of dissociated vagal neurones, ozone (10 μm) activated approximately 60% of wild-type vagal neurones. These neurones also responded to capsaicin, thus identifying them as putative C-fibre nociceptive neurones (although it should be noted that these neurones were isolated from the entire vagal ganglion and, as such, would have previously projected peripheral terminations to a variety of visceral organs including but not limited to the airways). From previous studies, the number of TRPA1-expressing neurones would be predicted to be between 30–65% of vagal neurones (Nassenstein et al. 2008; Taylor-Clark et al. 2008a, 2009a). That TRPA1 was both necessary and sufficient for the ozone-induced increases in calcium in vagal sensory neurones is supported by the findings that ozone (10 μm) had virtually no effect on vagal neurones derived from TRPA1^−/− mice (capsaicin-induced responses (TRPV1) were unchanged). Therefore, despite the fact that ozone is a relatively indiscriminate reactant, our studies suggest that at relatively low concentrations, ozone selectively activates nociceptive nerves through a mechanism that is dependent on TRPA1 channels.

The data are consistent with a ‘direct effect’ of ozone on the TRPA1-expressing C-fibre neurones. This, however, should not be understood to be necessarily a direct effect of ozone on the TRPA1 ion channel. The precise mechanisms whereby ozone leads to TRPA1 gating remain unknown. The rapid onset of action in the in vitro fura-2 AM calcium imaging studies (<10 s, similar to the onset of action of direct TRPA1 activators in this system) would be consistent with a direct channel interaction. Modelling of the kinetics of ozone diffusion across lipid bilayers (Pryor, 1992), however, suggests that it is unlikely that meaningful concentrations of ozone would be able to participate with the direct intracellular modulation of the channel. An alternative mechanism is that ozone reacts with plasma membrane components to form secondary products that then initiate TRPA1 gating. TRPA1 is a non-selective cation channel (Story et al. 2003; Jordt et al. 2004) that is gated by a wide range of stimuli including products of lipid peroxidation such as acrolein, 4-hydroxy-2-nonenal and 4-oxo-2-nonenal (Bautista et al. 2006; Trevisani et al. 2007; Andersson et al. 2008; Taylor-Clark et al. 2008a). These electrophilic products of lipid peroxidation probably activate TRPA1 directly via a covalent modification of the channel (Hinman et al. 2006; Macpherson et al. 2007). Ozone may therefore have interacted with the lipid environment of neurones or transfected HEK cells to produce (via ozonolysis, peroxidation and autoxidation reactions (reviewed in Pryor, 1994)) an intermediate that in turn could activate TRPA1. It would be unproductive at this time to speculate on the nature of such a product.

Solubilized ozone was capable of stimulating bronchopulmonary C-fibres at a concentration of 10–30 μm. It is of course not possible to accurately extrapolate between intratracheal infusion of 10–30 μm solubilized ozone, and the concentration of atmospheric ozone that would need to be inhaled to reach these concentrations. Nevertheless, this is a relatively potent effect of ozone when compared to other studies that employed solubilized ozone (Ito et al. 2005; Lotriet et al. 2007), suggesting that the actions of ozone on TRPA1 may be directly relevant to airway responses in vivo.

In summary, ozone directly and selectively activates bronchopulmonary C-fibres and dissociated vagal neurones via the gating of TRPA1 channels. It should be kept in mind that many vagal C-fibres terminate in the epithelium juxtaposed to the airway lumen, ideally situated to sense the inhaled air environment. The initiation of action potentials in these nociceptive nerve fibres may therefore contribute to the cough and decrements in lung function (through central and peripheral reflexes) experienced following ozone exposure (Mudway & Kelly, 2000). Consistent with this is the recent report of TRPA1 agonists causing cough in humans (Birrell et al. 2009), as does ozone (Hazucha et al. 1989; Schelegle et al. 2001). It will be informative in future studies to determine if TRPA1 represents a therapeutic target for the inhibition of some of the deleterious health effects of respiratory exposure to ozone.

Glossary

Abbreviations

AITC

allyl isothiocyanate

Caps

capsaicin

FEV₁

decreased forced expiratory volume in 1 s

FVC

forced vital capacity

fura-2 AM

fura-2 acetyoxymethyl ester

HEK

human embryonic kidney 293 cells

RAR

rapidly adapting receptor

TRP

transient receptor potential

TRPA1

transient receptor potential ankyrin 1

TRPV1

transient receptor potential vanilloid 1

Author contributions

T.E.T.-C. and B.J.U. both contributed to the design of experiments, the collection and interpretation of results and the drafting and revising of the manuscript.