Airway Sensory Nerve Density Is Increased in Chronic Cough

Abstract

Rationale: Chronic cough is characterized by frequent urges to cough and a heightened sensitivity to inhaled irritants. Airway sensory nerves trigger cough. We hypothesized that sensory nerve density is increased in chronic cough, which may contribute to excessive and persistent coughing.

Objectives: To measure airway nerve density (axonal length) and complexity (nerve branching, neuropeptide expression) in humans with and without chronic cough.

Methods: Bronchoscopic human airway biopsies were immunolabeled for nerves and the sensory neuropeptide substance P. Eosinophil peroxidase was also quantified given previous reports showing associations between eosinophils and nerve density. Three-dimensional image z-stacks of epithelium and subepithelium were generated using confocal microscopy, and from these z-stacks, total nerve length, the number of nerve branch points, substance P expression, and eosinophil peroxidase were quantified within each airway compartment.

Measurements and Main Results: Nerve length and the number of branch points were significantly increased in epithelium, but not subepithelium, in chronic cough compared with healthy airways. Substance P expression was scarce and was similar in chronic cough and healthy airways. Nerve length and branching were not associated with eosinophil peroxidase nor with demographics such as age and sex in either group.

Conclusions: Airway epithelial sensory nerve density is increased in chronic cough, suggesting sensory neuroplasticity contributes to cough hypersensitivity.

At a Glance Commentary

Scientific Knowledge on the Subject

Chronic cough is a common and difficult-to-treat condition. Mechanisms underlying the development and persistence of cough are incompletely understood.

What This Study Adds to the Field

Airway sensory nerves trigger cough. In this study, we show that airway epithelial sensory nerve density is increased in chronic cough, which may contribute to excessive and persistent cough triggering.

Cough is a protective response that expels inhaled particulates, microorganisms, and mucus from the airways and is essential for maintaining healthy lungs (1, 2). However, when cough becomes chronic, it is a debilitating condition characterized by frequent urges to cough and excessive bouts of coughing that may persist for months or years. Chronic cough significantly impairs quality of life (3, 4) and is a major global problem, estimated to affect nearly 10% of the world’s population (5). Despite this burden, our understanding of why chronic cough develops is surprisingly limited, and few effective treatments exist for this bothersome disease (6, 7).

Cough is triggered by vagal afferent sensory nerves that innervate airway epithelium and respond to a variety of chemical and mechanical stimuli (8). Patients with chronic cough frequently report increased sensitivity to inhaled stimuli, which has been termed cough hypersensitivity syndrome (6). Cough hypersensitivity is increasingly recognized as a neuropathic process in which sensitization of neuronal reflex pathways contributes to persistent coughing (9, 10). Inflammatory mediators, such as IFNs and tumor necrosis factor ⍺, “sensitize” nerves by increasing neuronal irritant nociceptor expression and by altering sensory neuropeptides such as substance P (11–13). Accordingly, neuromodulatory drugs such as gabapentin and purinergic P2X₃ antagonists suppress cough in clinical trials (14–17).

Whether airway nerves undergo structural remodeling in chronic cough is unknown. Historically, airway nerve structure has been difficult to study because of nerves’ complex, three-dimensional arrangement that spans tens to hundreds of histologic tissue sections (18). Consequently, features such as nerve length and the number of nerve branch points could not be accurately measured within individual slices. However, advances in confocal microscopy and tissue optical clearing have enabled detailed analysis of nerve structure in intact, whole mount airway samples (19, 20). Using these techniques, we recently demonstrated that airway epithelial sensory nerve density is increased in patients with eosinophilic asthma (21), highlighting an unexpected degree of airway nerve structural plasticity and suggesting that structural remodeling may contribute to manifestations of disease.

In this study, we tested the hypothesis that airway sensory nerve density is increased in chronic cough. We generated three-dimensional models from confocal z-stack images of airway nerves in whole mount biopsies collected from humans with chronic cough and from healthy control subjects. Total nerve length and the number of nerve branch points were quantified within epithelium and subepithelium. Neuronal substance P expression was also assessed as a marker of a neuropeptide-expressing sensory nerve subpopulation, and eosinophil peroxidase was measured to test eosinophilic inflammation in airway tissue.

Methods

Study Population

Study subjects were recruited from a respiratory specialty practice at a regional academic referral center. Eligible subjects were 18 years or older with a chief complaint of daily cough lasting 12 weeks or longer and an increased sensitivity to cough triggers. For inclusion, subjects were also required to have both normal spirometry and chest radiography, performed contemporaneously to enrollment. Subjects with underlying lung diseases (e.g., asthma, chronic obstructive pulmonary disease, and interstitial lung disease) were ineligible. Exclusion criteria also included a history of 10 or more pack-years of smoking tobacco or current smoking regardless of duration. Control subjects were healthy nonsmokers without existing lung disease. See online supplement for additional eligibility criteria. All subjects provided written informed consent. The protocol was approved by the Institutional Review Board of Beaumont Hospital.

Airway Sampling

Three forceps biopsies were obtained by flexible bronchoscopy from the third through fifth generation of airways in the right middle lobe. Biopsy samples were immediately fixed in formalin overnight and then washed and stored in phosphate-buffered saline.

Immunohistochemistry

Airways were immunostained as previously described (21). Briefly, biopsy samples were washed with tris-buffered saline (TBS) and blocked overnight at 4°C with a solution of 1% Triton X-100, 4% normal goat serum, and 5% powdered milk in TBS. Airway nerves were then labeled using rabbit polyclonal antibody against the pan-neuronal marker protein gene product 9.5 (PGP9.5) (Millipore) and a rat polyclonal antibody for substance P (BD Pharmingen), followed by secondary anti-rabbit 488 and anti-rat 555 antibodies (Life Technologies). Eosinophils were labeled with a mouse polyclonal antibody against eosinophil peroxidase followed by Alexa Fluor anti-mouse 647 secondary antibody (Life Technologies). Isotype controls were performed concurrently using rat IgG, rabbit IgG, and mouse IgG in place of the primary antibodies. After washing with TBS, airway samples were labeled with the nuclear counterstain DAPI and dehydrated with methanol and cleared using 2:1 benzyl benzoate:benzyl alcohol. Tissues were mounted in 2:1 benzyl benzoate:benzyl alcohol on well slides (1-mm thick), covered with a glass coverslip, and sealed with Permount (Thermo Fisher Scientific).

Image Acquisition and Processing

Images were acquired at nine nonoverlapping locations for each specimen. Z-stack images of airway epithelium and subepithelium were obtained for each location using a Zeiss LSM780 confocal microscope with a 63X/0.45 PlanApo objective and a 2-mm working distance (Carl Zeiss). In-plane (x and y) and out-of-plane (z) resolution was 0.264 μm × 0.264 μm × 1.00 μm. The total in-plane field of view and scan depth for a single human biopsy z-stack image acquisition was approximately 135 μm × 135 μm × 80 μm.

Nerve Modeling and Morphologic Quantification

Nerve models were created from each z-stack image by applying a computer-generated intensity-based filament over PGP9.5-positive voxels (Imaris software) (19). From these models, total nerve length (cumulative length of all nerves within a z-stack) and the number of branch points were quantified within airway epithelium and subepithelium by a blinded reviewer. Epithelium was distinguished from subepithelium by identifying the basement membrane, and subepithelial nerve quantification was performed to a depth of 20 μm from the basement membrane, as biopsy depth varied between samples. In some samples, biopsy depth was insufficient for subepithelial nerve quantification, and these samples were excluded from the final analysis. Substance P and eosinophil peroxidase were also quantified within each airway compartment by applying a computer-generated surface around positive voxels and were expressed as a volume in cubic microns. Image z-stacks from each biopsy were averaged, and then biopsy means were averaged to generate a single value for each subject.

Statistical Analysis

Cough and control subject demographics, nerve length, nerve branch points, and eosinophil peroxidase were compared using a Wilcoxon rank-sum test. Nerve length was modeled in terms of branching and age using linear regression. Nerve length was modeled in terms of eosinophil peroxidase using nonlinear regression. All statistics were performed using GraphPad Prism 8 (GraphPad). All P values were unadjusted and considered significant if less than 0.05.

Results

Subject Demographics

A total of 22 subjects with chronic cough and 21 healthy control subjects were included in the final analysis (Table 1). Age, body mass index, blood eosinophil count, and prebronchodilator spirometry were similar between the groups. More females were present in both groups. Four subjects in each group were former smokers, and all had less than 10 pack-years of cigarette exposure. At the time of airway sampling, three subjects in the cough cohort were receiving treatment with inhaled corticosteroids, and one was treated with systemic corticosteroids, compared to none in the control cohort.

Table 1.

Characteristics of Study Subjects

	Control (n = 21)	Cough (n = 22)
Age, yr, median (range)	57 (21–76)	56 (31–76)
Female, n (%)	15 (71)	13 (59)
Body mass index*	27 ± 4.8	27 ± 4.4
Blood eosinophil count, cells/μl	0.159 ± 0.1	0.200 ± 0.1
Prebronchodilator spirometry
FEV1, L	2.73 ± 0.6	2.93 ± 0.8
Percent predicted	96.3 ± 12	99.0 ± 12
FVC, L	3.51 ± 0.8	3.84 ± 1.2
Percent predicted	97.8 ± 12	102.4 ± 15
FEV1/FVC ratio	78 ± 7	77 ± 7
Use of inhaled corticosteroid, n (%)	0	3 (14)
Use of systemic corticosteroid, n (%)	0	1 (5)
Former smoker, n (%)	4 (19)	4 (18)

Values are means ± SD unless otherwise stated.

^*The body mass index is weight in kilograms divided by height squared.

Airway Epithelial Sensory Nerve Density Is Increased in Chronic Cough

Airway epithelial sensory nerves were immunolabeled with antibody against the neuronal marker PGP9.5 in bronchoscopic biopsies obtained from the right middle lobe. Nerve length and the number of branch points were quantified from three-dimensional nerve models superimposed onto PGP-positive voxels (Figures 1A–1F and 2A and 2B). Airway epithelial sensory nerve length (Figure 2C) and the number of branch points (Figure 2D) were significantly increased in subjects with chronic cough compared with healthy control subjects (length and branch points, mean [95% confidence interval]: control 549.2 μm [341.3–757.1 μm] and 11.3 [3.2–19.4] vs. cough 993.8 μm [726.5–1261.0 μm] and 23.2 [14.4–31.9]). Longer nerve length correlated with increased branch points within individuals (Figure 2E; R² = 0.72; P < 0.001).

Figure 2.

Airway epithelial sensory innervation is increased in chronic cough. Representative images of airway epithelial nerves in bronchoscopic biopsies from (A) healthy subjects (control) and (B) subjects with chronic cough (cough). Images are three-dimensional z-stacks of ×63 confocal images comprising the full thickness of airway epithelium. (C) Total epithelial nerve length and (D) the number of branch points in cough (n = 22) and healthy control subjects (n = 21). Box plots represent median values, bounds of the upper and lower quartiles, and error bars indicating minimum and maximum values. (E) Correlation between nerve length and number of branch points in healthy (white circles) and chronic cough (blue circles) groups. Data points represent an average value for each subject derived from nonoverlapping z-stacks. Scale bars, 15 μm. See online supplement for movie of three-dimensional modeling.

Figure 1.

Three-dimensional modeling of human airway nerves in bronchoscopic biopsies. (A) Human airway biopsy immunostained for nerves with antibody against the pan-neuronal protein PGP9.5 (cyan) and for cell nuclei with DAPI (blue). The image comprises full thickness of the biopsy specimen from airway epithelium to subepithelium moving proximal to distal in the z plane. (B) Airway nerve models are generated in three-dimensions along PGP9.5-positive voxels using Imaris software filament generator. (C) Nerve branch points (yellow globes) and total nerve length are quantified from filament-based models (cyan lines) for each image z-stack. (D) Volumetric airway nerve model based on PGP9.5+ voxels. (E) Lateral view of airway biopsy demonstrating epithelial and subepithelial airway compartments separated by the basement membrane. (F) Epithelial and subepithelial nerves are quantified and analyzed separately. Scale bar, 30 μm.

Subepithelial Nerve Density Is Similar in Chronic Cough and Healthy Airways

Subepithelial nerve morphology was quantified in airway parenchyma below the basement membrane (Figures 3A and 3B). Subepithelial nerve length (Figure 3C) and the number of branch points (Figure 3D) were similar between chronic cough and healthy control subjects (length and branch points, mean [95% confidence interval]: control 379.5 μm [271.7–487.3 μm] and 5.3 [3.1–7.3] vs. cough 479.8 μm [274.2–685.4 μm] and 10.9 [2.8–14.8]). Within individual subjects, longer subepithelial nerve length was associated with increased branch points (Figure 3E; R² = 0.40; P < 0.001).

Figure 3.

Subepithelial nerve length and branching are similar in chronic cough and healthy airways. Representative images of three-dimensional z-stacks showing airway subepithelial nerves (cyan) and nerve branch points (yellow globes) in bronchoscopic biopsies from (A) healthy subjects (control) and (B) subjects with chronic cough (cough). (C) Nerve length and (D) number of branch points in healthy subjects (n = 16) and subjects with chronic cough (n = 15) were quantified from three-dimensional models. Box plots represent median values, bound of the upper and lower quartiles, and error bars indicating minimum and maximum values. (E) Correlation between nerve length and number of branch points in subepithelium of control subjects (white circles) and subjects with chronic cough (pink circles). Scale bars, 15 μm. Data points represent an averaged value for each subject derived from nonoverlapping z-stacks.

Substance P-Expressing Airway Nerves Are Rare in Human Airways

Substance P expression was quantified on PGP9.5-labeled nerves in the airway epithelium and subepithelium and expressed as total volume in cubic microns (Figures 4A and 4B). Substance P volume was similar between cough and control subjects in epithelium (Figure 4C, left panel) and subepithelium (Figure 4D, left panel). Overall, substance P-expressing nerves were rare in both cough and control subjects, accounting for an average of 5.9% and 4.5% of total epithelial nerve volume in cough and control subjects (Figure 4C, right panel) and 2.6% and 6.2% of total subepithelial nerve volume in cough and control subjects (Figure 4D, right panel), respectively.

Figure 4.

Neuronal substance P expression is not increased in chronic cough. (A) Representative image of a three-dimensional z-stack showing substance P expression in a bronchoscopic biopsy from a subject with chronic cough. (B) Computer-generated mask overlying substance P–positive voxels shown in relation to airway nerve filaments. (C and D) Neuronal substance P expression and the ratio of substance P volume to total nerve volume were similar in subjects with chronic cough and control subjects in epithelium (C) and subepithelium (D). Box plots represent median values, bound of the upper and lower quartiles, and error bars indicating minimum and maximum values. Data points represent an averaged value for each subject derived from nonoverlapping z-stacks. Scale bar, 30 μm. PGP9.5 = protein gene product 9.5.

Epithelial Nerve Density Is Not Associated with Age or Sex

Epithelial nerve length was similar between men and women (Figure 5A) and was not associated with age in either group (Figure 5B control R² = 0.04, P = 0.38; and Figure 5C cough R² = 0.05; P = 0.29).

Figure 5.

Nerve length is not related to age or sex. (A) Airway epithelial nerve length was similar between women and men within each study group. ♂ = male; ♀ = female. (B and C) Epithelial nerve length was not associated to the age of study subjects in either group.

Airway Tissue Eosinophilia Does Not Predict Increased Nerve Density in Cough

Airway eosinophils were measured in epithelium and subepithelium by immunostaining for eosinophil peroxidase. The amount of eosinophil peroxidase was similar in chronic cough and healthy control subjects within both epithelium (Figure E1A in the online supplement) and subepithelium (Figure E1D). There was no association between airway eosinophil peroxidase and nerve length in individual subjects in either group (Figure E1B control epithelium R² = 0.14 and Figure E1C cough epithelium R² = 0.01; Figure E1E control subepithelium R² = 0.02 and Figure E1F cough subepithelium R² = 0.11).

Discussion

Here, we show that airway epithelial sensory nerve density is increased in humans with chronic cough. These structural changes support the concept that neuronal plasticity contributes to cough hypersensitivity syndrome, in which both noxious and innocuous stimuli inappropriately trigger cough reflexes leading to excessive and persistent coughing (6). Once activated, sensory nerve input is integrated into a complex neuronal cough reflex involving ascending and descending vagal nerve pathways and central nervous processes that can both trigger and suppress coughing. Changes in any portion of this reflex have potential to result in cough hypersensitivity. As evidence of this system’s complexity, cough responses to inhaled capsaicin and prostaglandin D2 vary greatly between individuals (22), as do responses between individuals with different airway diseases such as chronic cough and asthma (9). Accordingly, neuromodulatory drugs such as gabapentin (14), opioids (23), P2X₃ antagonists (15–17), and neurokinin 1 receptor antagonists (24, 25) suppress cough, providing further indication that neuropathic processes have a central role in chronic cough.

Our results are in line with previous studies showing neuronal substance P expression is relatively sparse in human airways and not increased in chronic cough (26, 27). Substance P’s scarcity in human airways, coupled with the ineffectiveness of antagonists of substance P receptors (also termed neurokinin receptors) in early clinical trials (28), raised doubts about its role in chronic cough in humans. Nonetheless, humans with chronic cough have increased substance P in blood (29) and nasal lavage (30), and in animal studies, substance P potentiates citric acid–induced coughing (31–33). More recent data may shed light on these discrepancies. The brain-penetrant neurokinin 1 receptor–selective antagonist Orvepitant improved a number of secondary endpoints in patients with chronic cough, including cough severity, urge to cough, and quality of life, although its primary outcome, cough frequency, failed to meet statistical significance because of a prominent placebo effect in the control group (24, 25). Unlike prior antagonists that were peripherally restricted, Orvepitant crosses the blood-brain barrier and blocks both central and peripheral neurokinin receptors. Thus, substance P may have a central mode of action in humans. Improvements in cough frequency and quality of life were also seen with another centrally acting neurokin 1 receptor antagonist, Aprepitant, in patients with lung cancer (34). On the basis of these results, investigations of substance P antagonists in cough are ongoing.

Sensory nerve purinergic P2X₃ receptors are another target with promising clinical potential. P2X₃ receptors are neuronal cation channels activated by ATP, an important endogenous mediator released during cellular stress that provokes coughing in healthy subjects and in chronic cough (35, 36). Recent trials demonstrated that a purinergic P2X₃ antagonist Gefapixant reduced cough frequency and improved quality of life in subjects with chronic cough (15–17). Interestingly, cough suppression was observed up to 4 weeks after discontinuation of the study drug, which raises the intriguing possibility that these antagonists may “reset” neuronal cough sensitivity to restore normal nerve function.

Several nonpharmacologic approaches for ablating airway nerves have also recently garnered interest in airway diseases. For example, in asthma, bronchial thermoplasty, which applies heat to the airway wall, decreases airway epithelial innervation as well as airway smooth muscle mass and reduces asthma exacerbations (37–39). Patients with hypersensitivity to capsaicin had a particularly favorable response to thermoplasty, underscoring the importance of sensory nerves in asthma as well as cough (40). Similarly, targeted lung denervation applies a radiofrequency to airway walls to ablate airway nerves (41, 42). Our findings suggest nerve ablation may also benefit chronic cough patients, although this approach has yet to be tested.

In our previous study of patients with asthma, airway eosinophils increased sensory nerve density (21). Despite excluding patients with asthma from our current analysis, we quantified airway tissue eosinophils to test for the presence of eosinophilic bronchitis, which is a cause of chronic cough, and to establish whether eosinophils were similarly associated with nerve plasticity in this cohort. Overall, eosinophils were similar in subjects with cough and control subjects, and neither airway nor peripheral blood eosinophils were associated with increased innervation. These data suggest that increased airway sensory nerve density in cough and asthma are the result of distinct mechanisms. Both mechanisms may ultimately involve mediators derived from airway epithelium, as increased innervation was isolated to epithelium in both diseases. Although the specific mediator of nerve growth in chronic cough is unknown, epithelial cells are a rich source of NGF (nerve growth factor) and other neurotrophins (43) that promote airway nerve growth (44–47). NGF also induces neuronal TRPV1 nociceptor expression in guinea pigs (48) and potentiates TRPA1-induced inward currents in isolated cells (49), which are expected to increase the likelihood of nerve activation leading to cough. In humans, levels of sputum and serum NGF, brain-derived neurotrophic factor, and neurotrophin-3 were not increased in those with chronic cough compared with healthy control subjects (50, 51). However, these results may not reflect local effects of neurotrophins within specific airway compartments like the epithelium, where neuronal remodeling occurs.

Nerve density was not related to the age of study participants, suggesting that airway nerve growth, or conversely regression, is not a product of normal aging in adults. Our cross-sectional study design could not determine the timing of onset or pace of pathologic nerve growth. Thus, although increased nerve density may promote development of chronic cough, it is equally possible that nerve growth is a consequence of chronic coughing that contributes to its persistence. In animal models, exposures during critical periods of lung development provoke increased sensory innervation, such as prenatal and early postnatal second-hand smoke and allergen exposure in mice (46, 47, 52, 53) and early postnatal allergen and ozone exposure in nonhuman primates (54). Humans may be particularly vulnerable during specific developmental periods as well.

Our study has some limitations to note. First, our analysis of airway epithelial innervation assumes that remodeling is specific to sensory nerves because parasympathetic and sympathetic nerves are not present in this layer in humans. However, modeling was based on PGP9.5 expression, which is not sensory afferent specific. Regarding modeling, because PGP9.5 expression is noncontiguous along some nerve axons, nerve endings in close proximity that followed the same course were considered to represent a single nerve axon. Since reviewers were blinded to study group, measurement error related to these assumptions is likely to be randomly distributed between groups. Finally, it is possible that chronic cough in some subjects was due to cough-variant asthma, as a clinical history of cough hypersensitivity, an absence of wheezing, and normal spirometry cannot entirely exclude this diagnosis. Methacholine challenge and fractional exhaled nitric oxide were not measured in our study, and though all patients failed to respond to inhaled corticosteroids, medication adherence was not measured. That said, airway eosinophilia was similar between subjects with cough and control subjects, arguing against inclusion of patients with type 2 high asthma in the cough cohort.

In summary, we have shown that airway sensory nerves undergo substantial remodeling in chronic cough, which may contribute to cough hypersensitivity and persistence. Sensory nerve growth was limited to airway epithelium, suggesting epithelial-derived factors have a role in the pathogenesis of sensory hyperinnervation. Our results reinforce the prominent yet underappreciated role of airway nerves in healthy and diseased airways.