Background
Allergic disease is caused by the inappropriate activation of Th2 cells in response to harmless or non-infectious stimuli such as pollens, foods or insect stings. In allergic individuals, Th2 recognition of allergens in the context of antigen presenting cells triggers the release of cytokines, including Interleukin (IL)-3, IL-4, IL-5 and IL-13, which underlie the typical allergic pathology with recruitment of mast cells and eosinophils, mucus production, IgE and tissue remodeling [
1].
In the airway, allergic disease is elicited by allergen inhalation and manifests itself with the typical asthma symptoms of tightness of chest and reduced lung function. Broncho-alveolar lavage of segmentally challenged allergic asthma patients has revealed the presence of eosinophils and Th2 cells in the airway with increased mucin, Th2 cytokines and increased smooth muscle mass [
2]. Animal models of allergic airway inflammation faithfully replicate several of these hallmarks of the late asthmatic response: type 2-associated cytokines such as IL-4, IL-5 and IL-13 can be found in the broncho-alveolar lavage (BAL) and are responsible for the observed inflammatory pathologies [
3,
4].
Chicken ovalbumin (OVA) has been used extensively as a model allergen, proving valuable in elucidating many features of airway disease. OVA is inexpensive and readily available, with well-characterized MHCI and MHCII epitopes. In addition, the availability of the OVA-specific OTI and OTII T cell receptor transgenic mice greatly facilitates the monitoring of OVA-specific immune responses in airways and local lymph nodes [
5]. These features have made OVA a reagent of choice when studying the cellular mechanisms underlying airway inflammatory responses. However, although OVA is not a completely irrelevant allergen, as food allergies to egg are relatively common in humans [
6], it is not clinically relevant as an airway allergen. In addition, OVA is not naturally immunogenic upon inhalation, and can induce tolerance [
7] unless supplemented with low doses of LPS [
8]. In order to sensitise to allergic airway inflammation, OVA is normally used adsorbed to the adjuvant aluminium hydroxide and must be administered via the non-physiological i.p. route, thus bypassing the airway innate immune environment during sensitisation [
9]. Due to these drawbacks, the clinical relevance of information obtained using the OVA model has been questioned by some investigators.
Other allergens of higher clinical relevance being used in experimental airway allergy models include the house dust mite (HDM)
Dermatophagoides pteronyssinus, cockroach, and the fungus
Alternaria alternata. These allergens can sensitise mice when given via i.n. instillation without adjuvant, and indeed asthma patients are commonly atopic against one or more of them. One of their common features is the harbouring of innate properties, such as protease activity that can directly affect lung epithelium and other cell populations [
10,
11], or the ability to engage Toll-like receptors (TLR) through protein mimicry [
12], thereby eliciting epithelial cell production of alarmins and cytokines such as IL-33, thymic stromal lymphopoietin (TSLP) and IL-25 [
13]. These cytokines act on multiple cell types including dendritic cells (DC) and pathogenic Th2 cells [
14], and can also induce the production of IL-5 and IL-13 from local innate immune cell populations such as type-2 innate lymphoid cells (ILC2) independently of conventional Th2 cells [
15‐
17]. Therefore, the innate properties of allergens may have a substantial impact on the cell populations involved in allergic responses in different models.
In this study, we wished to assess to what degree the choice of allergen can affect the mechanism of allergic airway inflammation in different experimental models. We compared four models, which use the allergens OVA or HDM and employ different protocols of allergen sensitisation and challenge, focusing on airway eosinophilia and goblet cell hyperplasia as these responses can be elicited by either ILC2 or CD4+ Th2 cells. We compared these models in different strains of knockout (KO) mice that are defective in selected components of the adaptive or innate immune response, and found that, in most cases, allergic inflammation was comparably reduced in each KO strain across all four models, with the degree of reduction ranging from mild to strong. Our results reveal notable similarities, but also some subtle differences, in the molecules and cell types driving each model, thus suggesting the involvement of similar immunological mechanisms in each case.
Discussion
We used four models of allergic airway inflammation that employ different allergens (OVA and HDM) priming protocols (systemic vs. local) and airway challenges (acute vs. repeated), to compare the contribution of four key immune response genes; MHCII, TSLPR, CD1d and TLR4, to various parameters of airway inflammation in each of the models. Differences between C57BL/6 males and females were also assessed. We found that inactivation of MHCII, TSLPR and TLR4 had an overall similar impact across all four models used, suggesting similarities in the immune mechanisms underlying each of them. The impacts of sex and CD1d inactivation showed some variation among models, preferentially affecting the responses to HDM or OVA, respectively, regardless of the protocols of sensitisation and challenge used in each case.
The clearest-cut results were observed in MHCII-KO mice. These mice lack conventional CD4+ T cells, whereas CD1d-restricted CD4+ T cells are reported to be present in normal numbers [
25]. ILC2 expression of MHCII is also expected to be defective in these mice [
26]. We observed that eosinophil accumulation and PAS-positive staining in each of the models used were strictly dependent on the presence of conventional CD4+ T cells. This observation clearly points to an essential role of CD4+ T cells, regardless of allergen or immunization protocol, in each of our models, and is consistent with published studies reporting a key role of CD4+ T cells in several models of airway inflammation [
27‐
29]. Our data also suggest that in these models cytokine-producing ILC2 or NKT cells are not sufficient for a response but may cooperate with CD4+ T cells for their function [
30,
31]. In contrast to eosinophil infiltration and PAS staining, cellular infiltration in the peribronchial and perivascular areas of the lung were variably affected, suggesting that these responses also rely, at least in part, on innate components [
32].
The inactivation of TSLPR also had a considerable impact on airway inflammation, and substantially reduced airway eosinophils in all models. Reduced eosinophil responses could be fully rescued by adoptive transfer of in vitro-primed Th2 cells into TSLPR-KO hosts, suggesting that, at the challenge phase, the role of TSLP was predominantly on the Th2 population [
33]. In contrast, the effects of TSLPR inactivation on PAS-positive staining in bronchioles were less marked compared to the effects on eosinophils, or the effects observed in MHCII-KO mice, except for the local HDM model. This observation differs from previous studies [
33‐
35] which used BALB/c TSLPR-KO mice (vs. our C57BL/6) and OVA preparations of unknown endotoxin content (vs. our low-endotoxin OVA) to report that goblet cell hyperplasia was also decreased, perhaps suggesting that genetic background and/or allergen composition can affect the impact of TSLPR deficiency on specific responses.
TSLP is known to have an essential function in allergic airway inflammation. TSLPR-KO mice developed reduced allergic airway inflammation in an OVA model [
34], and lung-specific overexpression of TSLP can induce spontaneous allergic airway inflammation in mice [
35]. TSLP is produced by airway epithelium exposed to LPS or protease-containing allergens [
35,
36], and acts on multiple immune populations. In addition to activating DC [
23,
37,
38], TSLP is necessary for the survival and effector function of memory Th2 cells [
33,
39,
40]. Together with IL-33, TSLP can also induce cytokine production by ILC2 [
13], with Th2 cells and ILC2 both contributing to cytokine production during allergic airway inflammation [
17].
The differential impact of TSLPR deficiency we observed on eosinophils vs. PAS-positive staining in mice sensitised systemically with OVA or HDM was very unexpected, but consistent across the three models. This observation raises interesting questions about a differential TSLPR requirement for increased eosinophils vs. mucus production in lung and airway, which are dependent on IL-5 and IL-13, respectively. As both of these responses are strictly dependent on conventional CD4+ T cells, as indicated by experiments in MHCII-KO mice, this result may suggest a heterogeneity in TSLP requirement by either CD4+ Th2 subsets producing IL-5 vs. IL-13, or the CD4+ Th2 vs. ILC2 populations that can produce these cytokines. Results in this paper showing that repeated OVA challenges tend to decrease eosinophil numbers while increasing the percent of PAS-positive staining in bronchiole epithelium might also suggest a similar possibility. Studies in mice where the TSLPR is conditionally inactivated in selected immune populations in lung will be necessary in order to address these questions. Interestingly, decreased eosinophils and PAS-positive cells were both observed after i.n. HDM immunisation, which is consistent with a stronger dependence of this model on local innate immune mechanisms [
9]. Overall, the results of these experiments are consistent with the recognized role of TSLP in supporting both innate and adaptive immune responses to allergens. Importantly, anti-TSLP has proven effective in ameliorating asthma symptoms in patients [
41,
42], a result that is consistent with the essential role of TSLP in maintaining memory Th2 populations in vivo [
33,
39].
NKT cells are glycolipid-reactive, innate-like T cells that can mediate allergic airway inflammation in the absence of conventional CD4+ T cells [
43] and are necessary for airway inflammation and hyper-reactivity in an OVA model [
44]. In contrast, the role of NKT cells in HDM models has not been examined. Unlike OVA, which does not contain NKT ligands, allergens such as HDM [
16] and pollens [
16,
45] are reported to contain glycolipids that induce NKT cell activation and production of IL-4 and IL-13. Consistent with previous reports, we observed that CD1d-KO mice, which lack NKT cells, generated impaired airway eosinophilia and PAS staining in OVA models, which were partially compensated by multiple airway challenges. In contrast, lack of NKT cells had no significant impact on the two models of HDM response, one of which used the same i.p. route of sensitisation and alum adjuvant also employed in the OVA models. This observation may suggest a role of NKT cells in amplifying responses to antigens that do not effectively engage innate immune mechanisms, such as the low-endotoxin OVA preparations used in our studies. However, the mechanism by which NKT cells might contribute to the OVA response measured here, the endogenous or exogenous source of potential NKT cell ligands in this model, as well as the role of NKT cells in clinical disease, remain poorly understood. Initial reports of a major contribution of NKT cells to allergic airway inflammation in patients [
46] have been questioned [
47]. The subsequent description of multiple NKT subsets with varying cytokine profiles [
48] further adds to the complexity of this question.
In contrast to the clear impact of defective MHCII or TSLPR expression on airway eosinophilia, we found that responses in TLR4-KO mice were, for the most part, similar to the responses in wild-type C57BL/6 mice. The response to OVA was weakly to moderately stronger in TLR4-KO mice than in C57BL/6 mice, which might be due to the reported capacity of endotoxin to suppress priming to allergic airway inflammation [
49]. In contrast to OVA, the response to HDM was not affected by TLR4 inactivation. Since the HDM protein Der p 2 is a known MD-2 mimic that facilitates signalling through TLR4 [
12], and as such it is reported to exacerbate airway allergic responses via interaction with airway epithelial cells [
36], this observation might suggest a low Der p 2 content in our HDM preparations [
50]. Alternatively, other immunologically active components of HDM such as proteases, β-glucans and chitins [
51] may be compensating for the loss of TLR4-dependent signalling in our model. While the impact of TLR4 inactivation on the models used here was unremarkable, it is important to note that TLR4 ligands have been shown to play important functions in other models of allergic response. Allergic conjunctivitis to Short Ragweed pollen [
52] required TLR4 to drive the production of TSLP/OX40L and the priming of a productive Th2 response. Low endotoxin content in OVA and Cockroach extract preparations increased airway inflammatory responses [
8,
53], and co-operated with Proteinase-activated receptor-2 (PAR
2) signalling in inducing allergic sensitisation [
53]. These studies highlight the multiple mechanisms through which allergens can engage with the immune system.
Finally, we report that C57BL/6 females develop stronger airway eosinophilia compared to male mice, but only in selected models. Intriguingly, the difference was strongest in the systemic HDM model, where eosinophil numbers are high, compared to the local model in which the weaker response might have been expected to be more dependent on cooperation with ILC2 [
24]. Our results may also suggest that the impact of androgens and ILC2 in airway inflammation models may depend in part on the mouse colony and/or strain, as well as other properties of the allergen used.
A perhaps unexpected result from our study is the overall similarity of the results obtained using the i.n. HDM sensitisation model to results in other models using i.p. sensitisation with allergen in alum adjuvant. Whereas i.p. sensitisation with adjuvant is clearly not a physiological model of allergen exposure, the impact of KO mutations in MHCII, TSLPR, CD1d, and TLR4 on these artificial models was not dissimilar from the impact observed after i.n. sensitisation, suggesting that, regardless of the priming route, all these models essentially measure the activity of a population of effector Th2 cells with similar activation requirements. In this respect, it is also important to note that the natural route of airway allergen sensitisation in humans remains unknown, and may not necessarily be via the airway, with sensitisation via the skin remaining a realistic possibility [
54].
While the innate properties of allergens clearly have an impact on their immunogenicity, and may have a stronger influence on inflammation than described here if used in settings of low-level or chronic allergen exposure, in our experiments they appeared mostly insufficient to directly drive inflammation independently of conventional CD4+ T cells. The study of such innate responses mostly requires tailored models in which specific allergens are used in high amounts and/or after careful purification to preserve their innate properties [
50]. A relevant example is the powerful fungal airway allergen
Alternaria alternata [
15], which is itself a trigger of innate allergic responses [
15] but, similar to IL-33, can also prime adaptive T cell responses in mice [
55,
56] and humans [
57]. It is also of interest that IL-33, which is essential for the innate function of
Alternaria and many other environmental allergens [
10], is also induced in macrophages by treatment with alum [
58], and is rapidly produced in the peritoneum after i.p. injection of alum adjuvant [
59]. Studies to examine lung ILC2 after i.p. alum injection may help establish whether the systemic and local priming of allergic immune responses in models of allergic airway inflammation such as those used here may involve common innate mechanisms.
Methods
Mice
The following mouse strains were used: C57BL/6 J (originally from Jackson Laboratory, Bar Harbor, Maine, USA), TSLP-receptor (R) KO [
60], MHCII-KO [
22], CD1d-KO [
61], TLR4-KO [
62] and OTII [
63]. Mice were bred by brother x sister mating and maintained in specific pathogen-free conditions at the Malaghan Institute of Medical Research, Wellington, NZ, with water and food ad libitum. Mice were age and sex-matched within experiments and used when 6–12 weeks old. Mice were euthanised for sample collection at the end of experiment by intraperitoneal injection of a high dose of Ketamine+Xylazine (300 and 9 mg/Kg, respectively). All experimental procedures were approved by the Victoria University of Wellington Animal Ethics Committee and carried out according to Institutional guidelines. No adverse effects were observed during this study.
Allergic airway inflammation
Naïve mice were randomly assigned to control or experimental groups and sensitised i.p. with 200 μl Alu-S-Gel (1.3%) (Serva, Heidelberg, Germany) containing 2 μg OVA Grade V, (Sigma-Aldrich, Saint Louis, MO, USA), 40 μg
D. pteronyssinus soluble extract (Greer Labs, Lenoir, NC, USA) or PBS (Gibco, Carlsbad, CA, USA). One hundred micrograms
D. pteronyssinus crushed bodies (Greer Labs) were administered in 50 μl PBS i.n. to anaesthetised mice; HDM extract was also used in this model but gave very low eosinophil responses (not shown) and was not used further. HDM models were carried out using only female mice except for Fig.
6. For the adoptive transfer model, 5 million OTII Th2 cells were generated by co-culturing lymph node cells from OTII mice with LPS-activated C57BL/6 bone marrow-derived DC, generated as in [
64], at an 8:1 ratio in the presence of 60 ng/ml IL-4, 20 ng/ml IL-2, and OVA
323–339 peptide (ISQAVHAAHAEINEAGR) in 6 well plates. IL-4 and IL-2 were replenished on day 2 and 4. In vitro-activated CD4+ T cells were harvested on day 5 and their phenotype was checked by flow cytometry. Resulting cells were Vα2
+Vβ5
+CD62L
loCD69
hiCD44
hi as described [
5]. Five million cells were injected through the lateral tail vein into recipient mice (C57, CD1d-KO or TSLPR-KO), and mice were challenged i.n. 1 day after cell transfer.
For i.n. challenges, mice were anaesthetised with Ketamine+Xylazine at 100 and 3 mg/Kg, respectively, and 100 μg endograde OVA (Hyglos GmbH, Bernried, Germany), 100 μg HDM soluble extract or 25 μg HDM crushed bodies was administered dropwise into one nostril in 50 μl sterile PBS. The distribution of the i.n.-instilled solution was checked in preliminary experiments where mice were given coloured tracers. They were found to include lower airway and lung, although the lung was not uniformly involved. BAL was collected by flushing 1 ml of PBS through the lungs thrice. After red blood cell lysis, samples were processed for flow cytometry and counted using Accucount beads (Spherotech, Green Oaks, IL, USA).
Flow cytometry
Antibody staining was performed in FACS Buffer (PBS with 2% FCS, 2 mM EDTA and 0.01% NaN3) using the following antibodies: anti-(a) CD11c-BV650, aMHCII (I-A/I-E)-Pacific Blue, aCD3-Pe-Cy7, aGr-1-AF647, aCD69-PE and aVα2-APC were from Biolegend (San Diego, CA, USA); aCD62L-PE-Cy7 and aCD44-APC-eFluor780 were from eBioscience (San Diego, CA, USA); aCD40-PE, aCD86-FITC, aVβ5.1/5.2-FITC, aNK1.1-PE, aSiglecF-PE-CF594, aCD19-APC-H7 and aCD4-BV605 were from Becton Dickinson (Franklin Lakes, NJ, USA); while aCD8-FITC and aCD16/32 hybridoma supernatant were prepared in house. Data were acquired on a LSRII SORP (Becton Dickinson, San Jose, CA, USA) or LSR Fortessa SORP (Becton Dickinson), and analysed using FlowJo Software v 9.9 (FlowJo LLC, Ashland, OR, USA). 4’,6-Diamidino-2-Phenylindole, Dihydrochloride (Molecular Probes, Eugene, OR, USA) was used to exclude dead cells.
Histology
Lungs were harvested and fixed in 10% neutral buffered formalin (Sigma-Aldrich), cut in 5 μm sections on the coronal plane and stained with haematoxylin and eosin or Alcian Blue-Periodic Acid Schiff (PAS). Peribronchial and perivascular inflammation was scored by a blinded operator using the following criteria: (0), no peribronchial or perivascular infiltrates; (1), 1–2 centrally located microscopic foci of inflammatory infiltrates; (2), a dense inflammatory infiltrate in a perivascular or peribronchial distribution originating in the center of the lung and extending along the vessels or bronchi into the middle third of the lung parenchyma; (3), perivascular or peribronchial infiltrates extending to the periphery of lung and approaching the visceral pleura. To quantify AB-PAS staining, we used a quantitative and objective method in which 3 micrographs per lung were taken using an Olympus BX51 compound microscope at 20x magnification, and processed using an Image J macro to calculate the percent PAS-positive staining in the total bronchiole epithelial area excluding airspace (Additional file
1A).
Statistics
All bar graph data are shown as mean +/− Standard Error of the Mean (SEM), excepting inflammation scores, which are expressed as median +/− interquartile ranges. All statistical analyses used a Mann-Whitney non-parametric t-test; p values lower than 5% were considered significant. Prism 7 for MAC OS X (San Diego, CA, USA) was used for all analyses.