Research Article

Pulmonary neuroendocrine cells amplify allergic asthma responses

See allHide authors and affiliations

Science  08 Jun 2018:
Vol. 360, Issue 6393, eaan8546
DOI: 10.1126/science.aan8546

Finding a role for PNECs in asthma

Pulmonary neuroendocrine cells (PNECs) are a rare cell type located in airway and alveolar epithelia and are often in contact with sensory nerve fibers. They have a wide phylogenic distribution and are found even in the relatively primitive lungs of amphibia and reptiles, suggesting a critical function. Sui et al. found that mice lacking PNECs have suppressed type 2 (allergic) immune responses. PNECs were observed in close proximity to group 2 innate lymphoid cells (ILC2s) around airway branch points. The PNECs enhanced ILC2 activity by secreting CGRP (calcitonin gene-related peptide). They also induced goblet-cell hyperplasia via the neurotransmitter GABA (γ-aminobutyric acid). Interestingly, human asthma patients were found to have increased PNEC numbers, suggesting a potential therapeutic target for the treatment of asthma.

Science, this issue p. eaan8546

Structured Abstract


The lung, with its vast surface area, senses and responds to signals in inhaled air. Aberrant interactions between the lung and the environment underlie many diseases, including asthma. In vitro data show that pulmonary neuroendocrine cells (PNECs), a rare airway epithelial cell population, can act as chemosensors. Once stimulated in culture, they release dense core vesicles rich in neuropeptides, amines, and neurotransmitters. These bioactive molecules are capable of eliciting immune and physiological responses. A recent in vivo study by our group revealed that the proper development of PNECs into self-clustering units called neuroepithelial bodies is essential for restricting the number of immune cells in the naïve lung. However, whether PNECs can function in vivo to translate exogenous airway signals such as allergens into the cascade of downstream responses is unknown.


To test the hypothesis that PNECs act as sensors in the lung, we generated mouse mutants that lack PNECs by inactivating Ascl1 in the airway epithelium—i.e., mutants that were depleted of PNECs starting at development. We exposed these mutants to either ovalbumin or house dust mites, following regimes of existing asthma models. We determined whether the mutants showed different asthmatic responses than controls. We elucidated the underlying mechanisms by identifying molecular effectors and cellular targets of PNECs. To complement the functional tests in mice, we investigated whether human asthma patients showed pathological changes in their PNECs.


Although normal at baseline, Ascl1-mutant mice exhibited severely reduced goblet cell hyperplasia and immune cell numbers compared with controls after allergen challenge. In investigating possible molecular effectors, we found that several PNEC products were decreased in mutants relative to controls after allergen challenge, including calcitonin gene-related peptide (CGRP) and γ-aminobutyric acid (GABA). In exploring possible cellular tar­gets, we found that innate lymphoid group 2 cells (ILC2s) were enriched at airway branch points, similar to PNECs. The PNEC product CGRP stimulated ILC2 production of interleukin-5 in culture. Conversely, inactivation of the CGRP recep­tor gene Calcrl in ILC2s led to dampened immune responses to allergens. In contrast to CGRP, GABA did not increase ILC2 cytokine secretion. Rather, inactivation of GABA biogenesis led to defective goblet cell hyperplasia after allergen challenge, suggesting that GABA is required for this response in the airway epithelium. The instillation of a mixture of CGRP and GABA in Ascl1 mutants restored both immune cell increases and goblet cell hyperplasia after allergen challenge, indicating that these products are the primary molecular effectors of PNECs in vivo. Consistent with these results from mice, we found increased PNEC numbers and cluster sizes in human asthma patients, which may underlie the heightened response to allergens in these individuals.


Our results demonstrate that PNECs, despite being a rare population of cells in the airway, are critical for amplifying the airway allergen signal into mucosal type 2 responses. Specifically, PNECs act through their product GABA to stimulate airway epithelial mucus production. In parallel, PNECs act through another product, CGRP, to stimulate ILC2 production of cytokines, which in turn recruit downstream immune cells. PNECs and ILC2s form neuroimmunological modules at the airway branch points, which are also the sites where airway particles are enriched. Our findings indicate that the PNEC-ILC2 axis functions to sense inhaled inputs, such as allergens, and amplify them into lung outputs, such as the allergic asthma response.

PNECs are preferentially localized at branch points.

A mouse airway stained by antibody against CGRP, to label PNECs (magenta) and antibody against SCGB1A1 to label club cells (green) (200× magnification). PNECs often cluster into neuroepithelial bodies and are preferentially localized at branch points.


Pulmonary neuroendocrine cells (PNECs) are rare airway epithelial cells whose function is poorly understood. Here we show that Ascl1-mutant mice that have no PNECs exhibit severely blunted mucosal type 2 response in models of allergic asthma. PNECs reside in close proximity to group 2 innate lymphoid cells (ILC2s) near airway branch points. PNECs act through calcitonin gene-related peptide (CGRP) to stimulate ILC2s and elicit downstream immune responses. In addition, PNECs act through the neurotransmitter γ-aminobutyric acid (GABA) to induce goblet cell hyperplasia. The instillation of a mixture of CGRP and GABA in Ascl1-mutant airways restores both immune and goblet cell responses. In accordance, lungs from human asthmatics show increased PNECs. These findings demonstrate that the PNEC-ILC2 neuroimmunological modules function at airway branch points to amplify allergic asthma responses.

A report by the National Center for Health Statistics of the U.S. Centers for Disease Control and Prevention estimates that 24.6 million people in the United States have asthma (1). Although much attention has been placed on immune cells as the source of symptoms, in recent years, the concept that nonhematopoietic tissues are a critical source of cytokines has gained support (2, 3). For example, lung epithelial cells such as alveolar type 2 cells produce interleukin-33 (IL-33), which in turn activates immune cells, including residential group 2 innate lymphoid cells (ILC2s) (4, 5). ILC2s secrete type 2 cytokines, including IL-5 and IL-13, which promote smooth muscle contraction, eosinophil infiltration, and goblet cell hyperplasia—key features of an asthmatic response (6, 7).

PNECs are a distinctive and poorly understood cell population in the lung. They are rare, endoderm-derived epithelial cells that constitute ~1% of the airway cell population (8, 9). They are the earliest specified lung epithelial cells, expressing Ascl1, their defining marker, starting at embryonic day 12.5. In rodents, a majority of PNECs form clusters, which are called neuroepithelial bodies (NEBs). NEBs are highly innervated by both afferent and efferent neurons (10). PNECs contain dense core vesicles filled with neuropeptides, amines, and neurotransmitters. In vitro, PNECs can be stimulated by oxygen, mechanical stretch, and chemical stimuli and release their vesicular content (11). Excess PNECs have been reported in lungs with various diseases, including chronic obstructive pulmonary disease, sudden infant death syndrome, bronchopulmonary dysplasia, and small cell lung cancer (1216). Recently, we showed that proper clustering of PNECs is essential for restricting immune cell number in the naïve neonatal lung (17). However, the precise in vivo role of PNECs in lung pathogenesis remains unknown.

PNECs are required for goblet cell hyperplasia in asthma models

To generate a mouse mutant lacking PNECs, we inactivated Ascl1, which encodes a transcription factor that is essential for their formation, as shown in Ascl1 global null mutants (18, 19). To bypass the perinatal lethality of the global nulls, we inactivated Ascl1 in precursors of PNECs in the airway epithelium by using Shhcre (a cre insertion into the sonic hedgehog gene) at the onset of lung development (hereafter, Ascl1CKO for conditional knockout) (fig. S1) (20). This early inactivation led to a complete absence of PNECs, as indicated by the lack of calcitonin gene-related peptide (CGRP)– or synaptophysin-positive cells in the airway epithelium (Fig. 1). Synaptophysin also labels the nerves that normally innervate PNECs. In the mutant, these nerves remained present subjacent to the airway, even though they no longer intercalated into the epithelium (Fig. 1, C, D, C', and D'). In contrast to the absent PNECs, the intrinsic neurons, which are also dependent on Ascl1, remained, as indicated by TUJ1 (tubulin β3) expression—presumably because Shhcre is not active in these cells to delete Ascl1 (fig. S2).

Fig. 1 Inactivation of Ascl1 by Shhcre in lung epithelium prevents PNEC formation, whereas nerves remain subjacent to airways.

(A to D and A' to D') At embryonic day 18.5 (E18.5), CGRP staining outlines PNECs in E-cadherin–positive epithelium. Synaptophysin staining outlines PNECs and nerves, including those that innervate PNECs. Arrows indicate PNECs in the control airway. Boxed areas in the upper panels are magnified in the corresponding lower panels. No CGRP or synaptophysin staining is evident in the epithelium in Ascl1CKO mice, indicating a complete absence of PNECs. In the mutant, although the nerves no longer innervate epithelial cells [compare (C') and (D')], the nerve tracks are still apparent nearby. Data are representative of sections from n = 3 mice of each genotype.

Ascl1CKO mutants were viable at birth, which is counter to the postulation that PNECs are required for the transition from the intrauterine environment to breathing air. The mutant lungs showed normal branching morphogenesis and alveolar morphology (fig. S3). Using immunofluorescent staining and quantitative reverse transcription polymerase chain reaction (qRT-PCR) of key cell type markers, we found no apparent defects in airway club and ciliated cells or in type I and II alveolar cells. Like in the airways of control mice, no goblet cells were detected in the naïve mutant airways. These data indicate that, despite PNECs being the first differentiated cell type in the airway, their absence has no gross effects on the development of other cell types in the lung at baseline.

To test whether disruption of PNECs affects lung responses to environmental cues, we used an established model to induce allergic asthma-like responses by sensitizing early postnatal mice to ovalbumin (OVA) injected in combination with alum. Mice were then challenged with OVA inhalation (Fig. 2A) (21). As predicted, control mice showed robust goblet cell hyperplasia along the airways (Fig. 2B). All MUC5AC-positive goblet cells were also positive for SCGB1A1, a club cell marker, suggesting that club cells were partially converted into mucus-producing cells (fig. S4). In contrast, Ascl1CKO mice showed substantially reduced goblet cell hyperplasia, which registered at about 10% of the control response on the basis of a histological mucin index (Fig. 2, C to F). This observation was corroborated by dramatically reduced expression of Muc5ac and goblet cell–promoting factors including Spdef and Foxa3 (Fig. 2G). Markers for other airway epithelial cell types—club cells and ciliated cells—remained normal in Ascl1CKO mice after OVA challenge (fig. S4).

Fig. 2 PNECs are required for goblet cell hyperplasia and immune cell infiltration in a model of asthma.

(A) Regime of the early postnatal OVA model, indicating postnatal day (P) time points for sensitization, challenge, and analysis. (B and C) Representative PAS staining of P24 longitudinal airway sections of OVA-challenged control and Ascl1CKO mice. Arrows from insets point to the areas that are magnified. (D to F) Representative PAS staining of transverse sections, from which histological mucus index was calculated by the linear percent of epithelium positive for PAS-stained mucus (n = 3 for each). Student’s t test. Arrows in (E) indicate residual PAS staining. (G) Goblet cell–related gene expression as assayed by qRT-PCR of P24 whole lungs after OVA challenge (n = 3 for each). Student’s t test. (H to J) Flow cytometry analysis of immune cells from whole lungs, after either PBS (control) or OVA challenge as labeled. Mann-Whitney U test (n = 4 for PBS groups and 7 for OVA groups). (K) Cytokine gene expression as assayed by qRT-PCR of P24 whole lungs (n = 3 each). Student’s t test. N.S. (not significant), P ≥ 0.05; *P < 0.05; **P < 0.01. Data are representative of three experiments. Error bars represent means ± SEM.

To determine whether PNECs are also required for allergen-induced goblet cell hyperplasia in the adult lung, we exposed mutant and control mice to house dust mites (HDMs), a common allergen that also causes asthma in humans (22). By periodic acid–Schiff (PAS) staining and quantification of Muc5ac transcripts, we found that goblet cell hyperplasia response was reduced in the Ascl1CKO mutants compared with controls in the HDM model (fig. S5). Thus, PNECs amplify airway goblet cell hyperplasia in the asthma models tested.

PNECs are required for type 2 immune responses in asthma models

A cardinal feature of asthma is eosinophilic infiltration and type 2 T helper (TH2) cell priming in the lung (7). To further explore the role of PNECs in this aspect of the asthmatic response, we compared immune cell frequencies in neonatal OVA and alum–treated Ascl1CKO and control mice. The Ascl1CKO mice exhibited lower numbers of ILC2s, eosinophils, and TH2 cells than control mice after OVA challenge (Fig. 2, H to J, and fig. S6). Further, Ascl1CKO mice showed reduced Il5 and Il13 expression (Fig. 2K). These differences were not due to baseline immune cell differences, given that phosphate-buffered saline (PBS)–treated Ascl1CKO and control mice had similar levels of immune cells (Fig. 2, H to J, and fig. S6).

To determine whether PNECs are required for allergen-induced immune response in the adult lung, we returned to the adult HDM model. In this adult model, Ascl1CKO mice showed decreased immune cell infiltration compared with controls (fig. S7). We also depleted PNECs in the postnatal stage by using a different genetic model. We used Shhcre to activate diphtheria toxin receptor (DTR) expression under the transcriptional control of the PNEC product CGRP (encoded by Calca) in a CalcaloxP-GFP-loxP-DTR strain (GFP, green fluorescent protein) (23). The introduction of diphtheria toxin into the lung resulted in a ~70% reduction in PNECs. When challenged with HDMs, these mice also showed an immune response that was dampened, although to a lesser extent than in the Ascl1CKO mutant, likely owing to residual PNECs in the DTR model (fig. S8). Thus, PNECs amplify allergen-induced type 2 immune responses in the asthma models tested.

Absence of PNECs leads to the reduction of key neuropeptides and the neurotransmitter GABA

In a normal lung, type 2 immune responses are led by changes in cytokines such as IL-25, IL-33, and TSLP (thymic stromal lymphopoietin) (2426). In Ascl1CKO lungs, these cytokines were detected at similar levels to those in controls after each of the OVA challenges (fig. S9). These results suggest that these epithelium-derived cytokines may not explain why type 2 immune responses are different when PNECs are absent.

A key feature of PNECs is their production and secretion of neuropeptides and neurotransmitters (11). To test the hypothesis that PNECs control type 2 responses through some of these molecules, we assayed the expression of genes encoding several neuropeptides that have been implicated in inflammatory diseases. These included Vip (encoding vasoactive intestinal peptide), Calca (CGRP), Chga (chromogranin A), and Npy (neuropeptide Y) (11, 2729). In control lungs, Calca, Chga, Npy, and Vip levels were greater after OVA challenge than after administration of PBS, with Calca showing the largest fold increase. In Ascl1CKO lungs, Calca, Chga, and Npy, but not Vip, were significantly lower than in lungs of control mice after OVA challenge (Fig. 3A). The relative fold changes in Calca and Chga were larger than that in Npy, possibly owing to the concentrated expression of Calca and Chga, but not Npy, in PNECs (11, 27).

Fig. 3 PNECs reside in proximity to ILC2s and stimulate their cytokine production through CGRP.

(A) Neuropeptide gene expression as assayed by qRT-PCR in P24 whole lungs after PBS or OVA challenge (n = 3 each). Student’s t test. (B) GABA level as assayed by ELISA in P24 whole lungs after OVA challenge (n = 3 each). Student’s t test. (C to E) From naïve mice, representative images of PNEC (green, anti-CGRP antibody, indicated by arrows) and ILC2 (magenta, Il5-tdTomato reporter, indicated by arrowheads) localization near branch points in longitudinal [(C) and (D)] or transverse (E) vibratome sections of airways. The boxed area in (C) is magnified in (D). Anti-CGRP antibody also labels sensory nerve (E). (F) Of the 217 total ILC2s analyzed, 178 (82%) were localized within 70 μm of airway branch (nodal) points. (G) Of the 217 total ILC2s analyzed, 132 (61%) of them were localized within 70 μm of PNECs. (H to J) Effects of CGRP or GABA on IL-5 production by sorted primary ILC2s in culture, in the presence of cytokines indicated at the top left. CGRP, but not GABA, can increase IL-5 secretion as assayed by ELISA (n = 3 each). Student’s t test. (K to M) Data from flow cytometry analysis of immune cells from whole lungs of HDM-treated adult control and mutant mice as labeled. Mann-Whitney U test (n = 7 for each group). N.S., P ≥ 0.05; *P < 0.05; **P < 0.01. Error bars represent means ± SEM.

To further investigate the source of the neuropeptide changes, we focused on CGRP as an example. CGRP is expressed in the PNECs and in lung sensory nerves with cell bodies located in the vagal ganglia. By qRT-PCR, we found that Calca expression in the vagal ganglia was similar in wild-type mice treated with OVA and PBS controls under the regime that we used (fig. S10A). Calca expression was also similar in the vagal ganglia of OVA-treated Ascl1CKO mice and genotype controls (fig. S10B). Within the lung epithelium, we found by immunofluorescent staining that CGRP expression remained restricted to PNECs after challenge (fig. S11). PNEC proliferation did not increase (fig. S11). Rather, the CGRP signal was detected at a notably higher intensity, suggesting that this PNEC-specific increase contributes to the overall increase of CGRP levels in the lung (fig. S11).

In addition to neuropeptides, we also assayed for the levels of γ-aminobutyric acid (GABA), a PNEC-produced neurotransmitter that can stimulate goblet cell hyperplasia (30, 31). GABA levels were reduced in Ascl1CKO mutants, possibly owing to loss of GAD1, a rate-limiting enzyme in GABA biosynthesis, and VGAT, the vesicular GABA transporter, which are both specifically expressed in PNECs in the lung (Fig. 3B) (31). Together, these differences in neuropeptide and neurotransmitter levels suggest that they may be responsible for the reduced immune cell infiltration and goblet cell hyperplasia in the Ascl1CKO mutant.

PNECs reside in proximity to ILC2s and can stimulate ILC2 cytokine production

To identify possible direct targets of PNEC signaling, we made use of a distinct feature of these cells: their preferential localization to airway branch points (8, 9). We searched for immune cells residing in close proximity to these junctions. ILC2s are a population of tissue-resident immune cells that have been shown to play a central role in type 2 immune responses (32, 33). We investigated the localization of ILC2s by using a mouse reporter line with tdTomato knocked into the Il5 locus (28). In naïve mice, in addition to confirming the previous finding that ILC2s are localized near the basement membrane subjacent to the airway epithelium (11, 2729), we found that a majority of ILC2s (178 of 217 total cells counted, or 82%) reside within 70 μm of airway branch points , also called nodal points (Fig. 3, C to F) (9). By labeling PNECs with antibody against CGRP (anti-CGRP antibody) in the Il5-tdTomato background, we found that ILC2s were preferentially localized within 70 μm of PNECs (132 of 217 total cells counted, or 61%) (Fig. 3G). This proximity suggests ILC2s as a candidate direct target of PNEC signaling.

To test whether ILC2s can perceive PNEC signals, we first addressed whether they express the appropriate receptors. Existing RNA-sequencing data indicate that ILC2s express the CGRP co-receptor genes Calcrl and Ramp1 and the GABA receptor gene Gabrr1, but not the CHGA receptor gene (34). We confirmed the expression of CGRP and GABA receptor genes by qPCR assay in primary ILC2s sorted from lungs of naïve wild-type mice (fig. S12).

To test whether ILC2s could be stimulated by PNEC signals, we cultured primary lung ILC2s from naïve lungs and addressed their response to CGRP or GABA. ILC2s respond to cytokines such as IL-33 (35). Consistent with this, after in vivo OVA challenge, there was an increase in IL-33 receptor gene St2, but no change in the more general IL-1 family signal mediator gene Il1/Rap, in sorted primary ILC2s (fig. S13). CGRP increased IL-5 production by ILC2s when cultured in the presence of IL-33. However, GABA had a negligible effect (Fig. 3H). In addition, CGRP enhanced IL-5 production in the presence of both IL-25 and IL-33 (Fig. 3I) (35). CGRP stimulation did not occur in the absence of IL-25 and IL-33 (Fig. 3J). To complement data from IL-5 enzyme-linked immunosorbent assays (ELISAs), we also assayed an array of candidate cytokines using the LEGENDplex TH2 system (Biolegend). We confirmed that IL-5 and IL-6 were increased when ILC2s were cultured in the presence of CGRP (fig. S14).

Using CFSE (carboxyfluorescein succinimidyl ester) staining, we observed little change in ILC2 division, indicating that CGRP promotes ILC2 cytokine production but not proliferation (fig. S15). IL-33 did not appear to alter CGRP receptor gene expression level in ILC2s, suggesting that CGRP and IL-33 may work in parallel to stimulate ILC2s (table S1). We also tested CGRP and GABA function on sorted TH2 cells but observed no change in IL-5 production (fig. S16). Furthermore, the addition of CGRP to sorted eosinophils led to only a slight increase in leukotriene C4 production, a minor change compared with the effect of CGRP on ILC2s at the same concentration (fig. S17). Thus, CGRP can directly act on ILC2s to promote their maturation and production of cytokines such as IL-5. Activated ILC2s, in turn, have been shown to recruit eosinophils and trigger a cascade of TH2 responses (32, 33).

To further validate the PNEC-CGRP-ILC2 axis in vivo, we deleted Calcrl in ILC2 cells by using Il5cre. When treated with HDMs, mutants showed reduced immune cell infiltration compared with controls, consistent with the notion that CGRP signaling through ILC2s is required for a full TH2 immune response (Fig. 3, K to M, and fig. S18).

Inactivation of GABA synthesis or transport disrupts goblet cell hyperplasia without affecting immune responses

Increased expression of GABA receptors in lung epithelial cells from asthmatic patients has been reported, and intranasal administration of GABA inhibitor has been shown to suppress OVA-induced goblet cell hyperplasia in mice (30). Thus, even though GABA had no effect on ILC2 stimulation, we tested the hypothesis that GABA signaling from PNECs is required for goblet cell hyperplasia in vivo. We disrupted GABA signaling by inactivating Gad1, which encodes a GABA synthesis protein, or Vgat, which encodes GABA transporter. Because PNECs are the only cells in lung that express these genes, inactivation by Shhcre in these animals (hereafter, Gad1CKO or VgatCKO mice) led to the specific disruption of GABA production from PNECs (31). Airway epithelial cell development remained normal in these mutants (fig. S19). However, when treated with OVA as neonates, both mutants showed a striking deficit in goblet cell hyperplasia (Fig. 4, A to D), confirming recent findings (36). To test whether GABA is also required in adults for the goblet cell hyperplasia response, we subjected adult Gad1CKO mice to the adult OVA model and found dampened goblet cell production of mucus (fig. S20). To further validate that disrupting GABA production in the lung was responsible for reduced goblet cell hyperplasia, we inactivated Gad1 in the lung epithelium by using Nkx2-1creERT2 [generating Nkx2-1creERT2 Gad1fl/fl (Nkx2-1creER;Gad1) mutant]. Nkx2-1creER;Gad1 mutants also showed a deficit in goblet cell hyperplasia (fig. S21). These results suggest that GABA production from the PNECs is essential for the goblet cell hyperplasia response.

Fig. 4 GABA from PNECs is essential for goblet cell hyperplasia.

(A to D) Representative PAS staining of OVA-challenged airways at P24. Arrows from insets point to the areas that are magnified. (E to G) Percentage of immune cells in whole lungs after OVA challenge (n = 7 for control and 6 for mutant). Mann-Whitney U test. (H) Cytokine gene expression as assayed by qRT-PCR in whole lungs after OVA challenge. Student’s t test (n = 3 for each group). N.S., P ≥ 0.05. Error bars represent means ± SEM.

Next, we addressed whether the goblet cell phenotype in the GABA-signaling mutants was associated with a disruption of immune cell infiltration and cytokine production in vivo. We found that the Gad1CKO mice had similar levels of ILC2s, TH2 cells, and eosinophils compared with controls upon OVA challenge, unlike in the case of the Ascl1CKO mutant (Fig. 4, E to G, and fig. S22). Il5 and Il13 levels were not significantly different. Thus, consistent with in vitro data (Fig. 3H), GABA signaling from PNECs is essential for goblet cell hyperplasia, but not type 2 immune cell recruitment, in the OVA asthma model.

Instillation of CGRP and GABA in the Ascl1CKO mutant reverses defective goblet cell hyperplasia and immune cell infiltration

To address whether CGRP and GABA are sufficient to reverse the defective asthma-like response in the absence of PNECs, we introduced a mixture of CGRP and GABA via intratracheal administration immediately after each OVA challenge to mimic the in vivo role of PNECs. On the basis of previously published doses, we used a 40-μl mixture of 100 nM CGRP and 10 mM GABA (37, 38). We found that the exogenous administration of CGRP and GABA was sufficient to restore asthma-like phenotypes in Ascl1CKO mice, including goblet cell hyperplasia, immune cell infiltration, and type 2 cytokine expression (Fig. 5 and fig. S23). Other airway cell types were not perturbed (fig. S24). As a control, we also administered the same mixture to wild-type mice that were sensitized but not challenged with OVA. GABA and CGRP did not induce goblet cell hyperplasia in these mice, indicating that these PNEC products collaborate with immune cells to mount goblet cell hyperplasia triggered by allergen challenge (fig. S25). Together, these findings offer an in vivo demonstration that CGRP and GABA are key bioactive products from PNECs that can drive lung allergic responses in the absence of these cells.

Fig. 5 In vivo restoration of OVA responses in Ascl1CKO mutants with intratracheal instillation of CGRP and GABA.

(A to C) PAS staining after OVA challenge. Arrows from insets point to the areas that are magnified. (D to F) Immune cell numbers from whole lungs after OVA challenge (n = 13 for control, 6 for Ascl1CKO, and 11 for Ascl1CKO + CGRP + GABA). Mann-Whitney U test. (G) Gene expression as assayed by qRT-PCR in whole lungs after OVA challenge. Student’s t test (n = 3 for each group). N.S., P ≥ 0.05; *P < 0.05; **P < 0.01. Error bars represent means ± SEM.

PNECs are increased in the airways of human asthma patients

Although PNEC pathology has been documented in a number of lung diseases, whether they change in human asthma has not been directly addressed. To investigate this, we stained lung sections from asthmatics and age-matched controls by using anti-bombesin antibody, which is commonly used to outline all human PNECs. We also used anti-CGRP antibody, which, like in mice, outlines the subset of CGRP-producing PNECs (Fig. 6, A to D, and table S1 for patient characteristics, treatment regime, and primary autopsy findings). We quantified total PNECs, PNECs in the proximal bronchiole and distal respiratory bronchiole, and PNEC cluster size, and we found that all of these were increased in asthmatic samples compared with controls (Fig. 6, E to L). The increase in the CGRP-positive subset of total PNECs was more significant than that in the total bombesin-positive PNECs (P = 0.0052 and 0.0485, respectively). These findings in humans are consistent with our findings in mice and suggest that an increase in PNECs, particularly CGRP-expressing PNECs, may contribute to allergic asthma.

Fig. 6 The number of PNECs is higher in asthmatic lungs.

(A to D) Representative images of control and asthma autopsy lung sections stained with anti-bombesin antibody or anti-CGRP antibody to delineate PNECs expressing each marker. Arrows indicate labeled PNECs. (E to J) Quantification of the percentage of bombesin- and CGRP-positive epithelial area [total, in the proximal bronchiole (PB), and in the distal respiratory bronchiole (RB)]. Asthmatic samples have higher percentages in all cases. Bombesin: total, P = 0.0485; PB, P = 0.0274; and RB, P = 0.0037. CGRP: total, P = 0.0052; PB, P = 0.0076; and RB, P = 0.0274. (K and L) Quantification of the mean large bombesin- or CGRP-positive cluster size. Asthmatic samples have larger cluster sizes in both cases. Bombesin, P = 0.010; CGRP, P = 0.0075. Mann-Whitney U test (n = 7 controls and 7 asthmatics). Ages, 16 months to 13 years (table S2 shows age-match and other clinical data). *P < 0.05; **P < 0.01. Error bars represent means ± SEM.


The precise in vivo function of PNECs has been a long-standing question in lung biology. Here we show that PNECs are critical for mounting type 2 immune responses in mouse models of asthma. They act through secreted neuropeptides and neurotransmitters, including CGRP and GABA (Fig. 7A). PNECs signal directly to ILC2s, and together they form a neuroimmunological module to sense and respond to environmental stimuli that enter the airway (Fig. 7B).

Fig. 7 The PNEC-ILC2 axis functions at airway branch points to amplify the allergic asthma response.

(A) A model for PNEC function in asthmatic responses. PNECs act through CGRP to stimulate ILC2s and through GABA to stimulate goblet cell hyperplasia. (B) A diagram illustrating PNEC-ILC2 interactions in neuroimmunological modules (boxed areas) at airway branch points, where particles carrying allergens or other antigens congregate.

The finding that the collaboration between PNECs and ILCs occurs preferentially at airway branch points is intriguing because computational modeling of particle dynamics shows that over time, particles that enter the airway concentrate at branch points (39). Thus, cells that reside at these sites are at a prime location to sample diverse inputs, including chemosignals (Fig. 7B). This preferential positioning of PNECs is established in utero, poising these cells to respond to the environment at first breath. In line with growing evidence supporting the impact of early life exposures on the immune system (40), we propose that in the early postnatal environment, with every stimulation, the PNEC-ILC2 neuroimmunological modules mature into signaling centers that are critical in airway defense against pathogens and tissue damage.

Our results show that PNECs act through distinct secreted products to control different aspects of the responses to allergens (Fig. 7A). Although GABA synthesis and secretion are essential for goblet cell hyperplasia (36), GABA regulates neither cytokine levels nor immune cell infiltration. In contrast, CGRP can stimulate residential ILC2s at close range, inducing them to release IL-5, which can in turn recruit eosinophils and elicit a cascade of immune responses (32, 33). In the absence of PNECs and consequently of the CGRP from these cells, the ILC2 response is severely dampened even though IL-33 is still present. Consistent with this finding, global inactivation of either Calcrl or Ramp1—genes encoding co-receptors for CGRP—results in diminished lung inflammation upon OVA challenge (29). Furthermore, our data show that Il5cre-mediated inactivation of Calcrl also results in a deficit in the immune response. These results demonstrate that signals from PNECs are important triggers of ILC2 activation in the two allergic asthma models that we used.

It remains unclear how PNECs are activated after allergen challenge. Our data suggest that the increase in CGRP production in the lung is not due to an increase in PNEC cells, but rather to an increase in the levels of CGRP per PNEC. Recent data indicate that innervation of PNECs is required for the increased production of GABA, suggesting that stimulating efferent neurons that innervate PNECs may be a mechanism that can activate PNECs (36). It is also possible that allergen-induced upstream immune cells and signals may activate PNECs.

In addition to PNECs, neurons are also a source of neuropeptides in the lung. Although genetic ablation of TRPV1-expressing sensory neurons did not disrupt an OVA-induced immune response, ablation of a larger pool of NAV1.8 nociceptor–expressing sensory neurons blunted eosinophil and T cell responses (41, 42). Several recent studies also showed that neuron-derived neuromedin can directly interact with ILC2s and regulate immune responses in both the intestine and lung (4345). Another study showed that TRPV1 neurons, acting through CGRP, regulate protective immunity against lethal Staphylococcus aureus pneumonia (46). With the increasing awareness that neuropeptides are important regulators of lung immune responses, our findings here establish PNECs as a crucial source of signals.

PNECs are the first specialized cell type that forms in the lung epithelium, suggesting that early establishment of this sensory population is of prime importance to building a functional organ that is responsive to environmental inputs. Our finding that PNECs are increased in asthmatic patient lungs raises the possibility that this increase may contribute to disease. Given the evolutionary conservation of these cells, it is unlikely that PNECs act as the lung “appendix,” a site of pathogenesis with no apparent role in homeostasis. Rather, PNECs are likely essential for lung defense, in addition to conferring responses to allergens as demonstrated here. Elucidating the full capacity of PNECs is important to understanding how blocking their function, or the function of their secreted products, can be used safely and effectively to treat allergic lung diseases such as asthma. Cells similar to PNECs are found in other tissues—for example, enteric endocrine cells in the intestine. These neuroendocrine cells may prove to play critical roles in translating organ-specific environmental inputs into appropriate tissue responses.

Materials and methods


All mice were housed and all experimental procedures were carried out in American Association for Accreditation of Laboratory Animal Care accredited facilities and labs at either the University of Wisconsin–Madison or University of California, San Diego. This study followed the Guide for the Care and Use of Laboratory Animals. Shhcre, Ascl1fl, Vgatfl, Gad1fl, Calcrlfl, Il5tdTomato-cre Nkx2-1creERT2, and Calcalox-Gfp-lox-Dta mice have all been described previously (20, 23, 28, 4749). Il5tdTomato-cre, Gad1fl, Calcrlfl, and Ascl1fl mice were on the B6 background. Shhcre, Nkx2-1creERT2, Calcafl-Gfp-fl-Dta, and Vgatfl mice were back-crossed to the B6 background for at least three generations starting from a mixed background. Littermates were used as controls to minimize potential genetic background effects.

Allergen-induced asthma model

For the early postnatal OVA asthma model, pups were sensitized by intraperitoneal injections of 10 μg of ovalbumin in 10 μl of PBS (OVA, A5503; Sigma) with 10 μl Imject alum (ThermoScientific) on postnatal day (P) 5 and P10, followed by three 15 min challenges with 3% aerosolized OVA solution on P18, P19, and P20. For controls in this protocol, all pups were injected on P5 and P10 with OVA and alum, and challenged on P18, P19, and P20 with PBS. Mice were sacrificed on P24 for analysis.

For the adult OVA asthma model, both mutant and control mice were sensitized by intraperitoneal injections of 50 μg of ovalbumin (OVA, A5503; Sigma) with 10 μl Imject alum (ThermoScientific) at 6 weeks of age on day 0 and day 7, followed by five 15 min challenges with 3% aerosolized OVA solution on days 15, 16, 17, 18, and 19. Mice were sacrificed on day 23 for analysis.

For the adult HDM model, 50 μg of HDM extract (Dermatophagoides pteronyssinus, Greer Labs) was introduced intranasally at 4 weeks of age on days 0, 7, 17, and 21. For controls in this protocol, PBS was administered intranasally on the same schedule instead of HDM. Mice were sacrificed on day 24 for analysis.

In Shhcre;Calcalox-Gfp-lox-DTR mutant and corresponding controls, diphtheria toxin (List labs) at 10 ng/g body weight was administered intranasally at P18 and P22. Treated mice were rested for at least 4 days before HDM treatment.

In pregnant females used to generate Nkx2-1creERT2;Gad1fl/fl mutant and corresponding controls, tamoxifen was injected at embryonic days (E) 11 and 12 to induce cre for the inactivation of Gad1.

Histology, mean linear intercept, and histological mucus index analysis

Mice were euthanized using CO2. Lungs were inflated with 4% PFA (paraformaldehyde) at 35 cmH2O airway pressure, and then fixed overnight. Lungs were then prepared for paraffin (8 μm), cryo (10 μm), or vibratome (100 μm) sectioning. Goblet cells were stained using a periodic acid–Schiff (PAS) staining kit (Sigma). The histological mucus index (the percentage of PAS-positive cells in the bronchial epithelium) was determined by using Image-J software (NIH). To quantify mean linear intercept (MLI), 20× H & E images were used. For each genotype, three mice, three sections per mouse and three independent fields per section were analyzed. Samples were compared using Student’s t test, by MLI ± SEM with statistical significance set at P < 0.05.


The following primary antibodies were used at the indicated final concentrations for immunofluorescence or immunohistological staining: rabbit anti-Synaptophysin polyclonal antibody [5 μg/ml] (Fisher), rabbit anti-bombesin polyclonal antibody [5 μg/ml] (ImmunoStar), rabbit anti-Calcitonin gene-related peptide/CGRP polyclonal antibody [5 μg/ml] (LifeSpan BioSciences), rabbit anti-CGRP polyclonal antibody [2 μg/ml] (Sigma), rat anti-E-Cadherin polyclonal antibody [5 μg/ml] (Abcam), rabbit anti-SCGB1A1 polyclonal antibody [5 μg/ml] (Seven Hills Bioreagents), rabbit anti-SPC polyclonal antibody [5 μg/ml] (Seven Hills Bioreagents), mouse anti-MUC5AC monoclonal antibody [5 μg/ml] (MRQ-9, Sigma), and syrian hamster anti-T1alpha (PDPN) polyclonal antibody [5 μg/ml] (Developmental Studies Hybridoma Bank). The following secondary antibodies were used: Cy3-conjugated goat anti-rabbit IgG [2 μg/ml], FITC-conjugated goat anti-rabbit IgG [2 μg/ml], FITC-conjugated goat anti-rabbit IgG [2 μg/ml], goat anti-rat FITC [2 μg/ml] (all from Jackson Immunoresearch), HRP-conjugated goat anti-rabbit IgG [2 μg/ml] (Vector Laboratories), and ImmPACT DAB peroxidase (HRP) substrate following manufacturer’s protocol (Vector Laboratories). Images were acquired either by ZEISS LSM 880 with Airyscan (Fig. 3, C and D, or by ZEISS Axio Imager 2 (the rest of mouse images), or Nikon DSRi1 digital camera mounted on a Nikon Eclipse 80i microscope (human sample images).

EdU analysis of cell proliferation

For EdU analysis of cell proliferation, ~300 μl of 400 μM EdU solution (ThermoScientific) was intraperitoneally injected. Mice were sacrificed 1 hour after EdU injection. Lungs were fixed in 4% PFA overnight and prepared for cryo sectioning (20 μm). EdU was detected using the Click-iT EdU Kit with Alexa Fluor 488 (Invitrogen).

Tissue processing for cell sorting and flow cytometry

After euthanasia and before tissue harvest, mice were transcardially perfused with 5 ml of cold PBS. Whole lungs were digested by gentle shaking in 5 ml of HBSS with 0.1 Wünsch units (WU) ml of Liberase (Roche) and 25 mg DNase I (Roche) for 60 min at 37°C, and then mechanically dissociated using GentleMACS C tubes (Miltenyi Biotec) followed by straining through a 70-μm filter. The cells were then re-suspended in 40% Percoll, underlined with 66% Percoll and centrifuged. Hematopoietic cells were isolated from the interphase for sorting and analysis.

Cell sorting and flow cytometry

The single-cell suspensions from above were pelleted, aliquoted at ~1 × 106 cells per tube, and stained with Fc blocking antibody (5 μg/ml, BD) and Live/Dead Fixable Dead Cell Stain Kit (Invitrogen) at room temperature for 10 min. The cells were washed and then incubated with surface marker antibody cocktail for 45 min at 4°C. For sorting ILC2s and TH2 cells, the following antibodies were used (all antibodies are from Biolegend, used at 2 μg/ml, unless otherwise specified): BV785-conjugated anti-CD90.2 (30/H12), AF700-conjugated anti-B220 (RA36B2), AF700-conjugated anti-CD11b (M1/70), AF700-conjugated anti-CD11c (N418), AF700-conjugated anti-Nk1.1 (PK136), BV510-conjugated anti-CD4 (GK1.5), PE-Cy7-conjugated anti-TCR beta (H57-597), FITC-conjugated anti-TCR delta (GL3), and PE-conjugated anti-ST2 (DIH9). For sorting eosinophils, the following antibodies were used: PE-conjugated anti-CD64 (X54-5/7.1), PE-Cy7-conjugated anti-CD11c (N418), and BB515 anti-Siglec F (2 μg/ml, E50-2440, BD). Samples were analyzed on an LSR II (BD Biosciences) with four lasers (405 nm, 488 nm, 561 nm, and 635 nm). Data were analyzed with FlowJo software (Treestar).

Cell culture

Freshly sorted ILC2s were cultured in RPMI culture media (Sigma) with 10% (vol/vol) heat-inactivated FBS (from Japan BioSerum), 100 U/ml penicillin/streptomycin, 10 mM HEPES buffer solution, 1× MEM nonessential amino acids, 1 mM sodium pyruvate, 50 μM 2-mercaptoethanol and 50 μg/ml gentamycin sulfate. Cells were cultured with IL-7 (10 ng/ml) and additional indicated cytokines (each at 10 ng/ml) for 12 hours. Then either CGRP (Sigma, 100 nM) or GABA (Sigma, 10 mM) were added. Freshly sorted TH2 cells were cultured with immobilized anti-CD3 antibody (Biolegend, 145-2C11, 10 μg/mL) with IL-33 (R&D Systems, 10 ng/ml) and IL-7 (R&D Systems, 10 ng/ml), and either CGRP (100 nM) or GABA (100mM). Freshly sorted eosinophils were cultured with IL-5 (R&D Systems, 10 ng/ml), and with or without CGRP (100 nM). Cytokine production were determined by using IL-5 ELISA kit from eBioscience or the Biolegend LEGENDplex TH2 system. Leukotriene C4 production were measure by using the LTC4 ELISA kit (Cayman Chemical).

Human asthma study population

After Institutional Review Board approval, the pathology database at Seattle Children’s Hospital was queried from 1960–2016 to identify autopsies in which asthma was documented in the final report. An age-matched control was obtained from the same year as each asthma case.

Human sample morphometric analysis

All slides were stained, imaged, and quantified together. For each subject, immunostaining for CGRP and bombesin was quantified in up to 12 random peripheral (non–cartilage-containing) airways. Airways were classified as either proximal (comprising membranous and terminal bronchioles with complete muscular walls) or respiratory bronchioles (air passages composed of part muscular bronchial wall and part alveoli). Airways were visualized and captured with a Nikon DSRi1 digital camera mounted on a Nikon Eclipse 80i microscope and analyzed using NIS-Elements Advanced Research Software v4.13 (Nikon Instruments, Melville, NY). All cases were photographed and analyzed using the same magnification, exposure time, lamp intensity, and camera gain. No image processing was carried out prior to intensity analysis. In each airway, the immunostained area was expressed as a percentage of the total airway epithelial area (cell %). The largest immune-positive cluster size was documented for each airway and divided by total airways counted to obtain mean large cluster size.

Statistical methods

As indicated in the figure legends, Student’s t test was used for parametric data and the Mann-Whitney U test was used for nonparametric data. All statistical analyses were performed using Prism 6 (GraphPad).


All primers are listed in table S3.

Supplementary Materials

References and Notes

Acknowledgments: We thank Q. Wu, K. Caron, and S. Han for sharing mouse strains; J. Whitsett for providing antibody; V. Nizet for the use of a flow cytometer; D. Broide for assistance with the HDM model; L. Nantie for help with imaging; S. Gong and Y. Wang for consultation on the delivery of peptides; and H. Hoffman, G. Haddad, and M. Kronenberg for advice. We thank the University of Wisconsin Carbone Cancer Center Flow Cytometry Laboratory for assistance with the use of flow cytometers. Funding: This work was supported by NIH grants OT2OD023857, R01HL113870, R01HL119946, and R01HL122406 and by Wisconsin Partnership Program grant 2897 (to X.S.). Author contributions: P.S., D.L.W., J.X., Y.Z., J.L., S.V.D., A.L., C.Y., R.M.L., B.S.K., G.D., and X.S. designed and performed the research; P.S. and D.L.W. analyzed the data; and P.S. and X.S. wrote the manuscript with input from other authors. Competing interests: The authors declare no competing interests. Data and materials availability: Ascl1fl mice were obtained from F. Guillemot under a material transfer agreement with the Francis Crick Institute, Mill Hill Laboratory. All data are presented either in the body of the paper or in the supplementary materials.

Stay Connected to Science

Navigate This Article