**Increased Ratio of Matrix Metalloproteinase-9 (MMP-9)**/**Tissue Inhibitor Metalloproteinase-1 from Alveolar Macrophages in Chronic Asthma with a Fast Decline in FEV1 at 5-Year Follow-up**

**Fu-Tsai Chung 1,2,3,4, Hung-Yu Huang 1,3, Chun-Yu Lo 1,2, Yu-Chen Huang 1,2, Chang-Wei Lin 1,3, Chih-Chen He 1, Jung-Ru He 1,2, Te-Fang Sheng 1,2 and Chun-Hua Wang 1,2,\***


Received: 9 August 2019; Accepted: 9 September 2019; Published: 12 September 2019

**Abstract:** Chronic asthma is associated with progressive airway remodeling, which may contribute to declining lung function. An increase in matrix metalloproteinases-9 (MMP-9)/tissue inhibitor metalloproteinase-1 (TIMP-1) may indicate airway inflammation and bronchial injury. Bronchial biopsy specimens and alveolar macrophages (AMs) were obtained from patients with asthma under regular treatment with inhaled corticosteroids or combination therapy and normal subjects (*n* = 10). Asthmatics included those with a slow forced expiratory volume in one second (FEV1) decline (<30 mL/year, *n* = 13) and those with a fast FEV1 decline (≥30 mL/year, *n* = 8) in 5-year follow-up. Immunostaining expression of MMP-9 and TIMP-1 was detected in airway tissues. MMP-9 and TIMP-1 was measured from AMs cultured for 24 h. After the 5-year treatment, the methacholine airway hyperresponsiveness of the slow FEV1 decline group was decreased, but that of the fast FEV1 decline group was increased (PC20, provocative concentration causing a 20% decrease in FEV1, 3.12 ± 1.10 to 1.14 ± 0.34 mg/dL, *p* < 0.05). AMs of asthma with a fast FEV1 decline released a higher level of MMP-9 (8.52 ± 3.53 pg/mL, *p* < 0.05) than those of a slow FEV1 decline (0.99 ± 0.20 pg/mL). The MMP-9/TIMP ratio in the fast FEV1 decline group (0.089 ± 0.032) was higher than that of the slow FEV1 decline group (0.007 ± 0.001, *p* < 0.01). The annual FEV1 decline in 5 years was proportional to the level of MMP-9 (r = 57, *p* < 0.01) and MMP-9/TIMP-1 ratio (r = 0.58, *p* < 0.01). The airways of asthma with greater yearly decline in FEV1 showed an increased thickness of submucosa and strong expression of MMP-9. An increase in MMP-9 and MMP-9/TIMP-1 in airways or AMs could be indicators of chronic airway inflammation and contribute to a greater decline in lung function of patients with chronic asthma.

**Keywords:** asthma; airway remodeling; matrix metalloproteinases-9; tissue inhibitor of metalloproteinase-1; alveolar macrophages

#### **1. Introduction**

Asthma is a chronic respiratory disease of airway inflammation that manifests as variable airflow limitation [1]. Patients with asthmatic airway inflammation develop tissue injury with subsequent

structural changes, which is termed airway wall remodeling [2,3]. Elderly or longer duration of asthma is correlated to airway wall remodeling and contributes to a decline in lung function [4]. The features of airway remodeling include smooth muscle hypertrophy, goblet-cell hyperplasia, subepithelial fibrosis, inflammatory cell infiltration, as well as epithelial shedding [5].

Matrix metalloproteinases (MMPs) are a family of enzymes that can break down proteins of the extracellular matrix (ECM), thus contributing to pathological processes of inflammation, wound healing, and fibrosis [6]. The MMPs also play an important role in several lung or airway diseases, or even lung cancer [7–10]. Additionally, increased levels of MMP-9 in serum, sputum, or lavage fluid were observed in patients with asthma [11–13]. MMP-9-deficient animals could inhibit airway inflammation and the immunoreactivity of MMP-9 has also been reported to be correlated with asthma severity [14]. Nevertheless, a defensive function of MMP-9 in asthma was reported through a heightened inflammation in MMP9-deficient mice [15,16]. Taken together, MMP-9 may be an important factor involved in asthma, but the production of MMP-9 in chronic asthma with persistent airway obstruction is still undetermined.

The tissue inhibitors of matrix metalloproteinase (TIMPs) inhibit enzymatic activity of MMPs through binding to the MMPs [17,18]. The secretion of TIMP-1 is associated with MMP-9. TIMP-1 may possibly result in the thickening process of the basement membrane in asthma [19]. Therefore, MMP and TIMP imbalance may cause clinical differences in chronic airway diseases [20,21]. The ratio of MMP-9/TIMP-1 in the sputum of asthmatic patients has been shown to decrease after recovery from an acute exacerbation of asthma, which may imply that MMP9/TIMP has a negative correlation [22]. Through increasing the thickness of the airway wall by collagen deposition, a decreased ratio of MMP-9/TIMP-1 in chronic asthma may result in airway obstruction [4]. Vignola et al. [21] reported that the ratio of sputum MMP-9/TIMP-1 has a positive correlation with the forced expiratory volume in one second (FEV1) of asthmatic patients. Additionally, a previous report [23] showed that a low serum MMP-9/TIMP-1 ratio was found in asthmatic patients who show little FEV1 improvement with the treatment of corticosteroids. Despite some reports [22,23], which revealed the possibly negative correlation of MMP9, TIMP, and lung function in asthmatic patients, the relation among these parameters remains controversial. Some reports also presented no correlation between MMP9, TIMP, and lung functions [24], or positive correlations among MMP9, TIMP, and lung functions [25,26].

Our aim was to evaluate whether the ratio of MMP-9/TIMP-1 released from cultured alveolar macrophages was higher in chronic asthmatic patients who had a fast, yearly decline in FEV1. We also investigate whether the ratio of MMP-9/TIMP-1 was correlated to the magnitude of yearly FEV1 decline.

#### **2. Materials and Methods**

#### *2.1. Patient Population*

Non-smoking asthmatic patients, aged 18 to 65 years, were recruited from outpatient clinics of the Chang Gung Memorial Hospital. Asthma was defined according to the American Thoracic Society criteria [27]. All asthmatic subjects had a >12% improvement in forced expiratory volume in one second (FEV1) with inhaled albuterol (400 μg) and bronchial hyperreactivity to methacholine (provocative concentration (PC) causing a 20% decrease in FEV1, PC20 < 8 mg/L). These patients received anti-asthma medications, which included inhaled and/or oral corticosteroids, inhaled long-acting β2-agonist, or a combination of these, and had been followed up at our clinics for more than 5 years. All of the patients with asthma had used inhaled corticosteroids all the time. All patients had been stable for at least 3 months and were taking their usual medications before entry into the study. Inhaled β2-agonists were withheld for 12 hours before methacholine testing.

The asthmatic subjects were divided into 2 groups: 13 asthmatic subjects (7 women and 6 men, aged 49.2 ± 3.4 years) with a slow FEV1 decline (< 30 mL/year) in the 5-year follow-up, and 8 asthmatic subjects (3 women and 5 men, aged 49.4 ± 3.7 years) with a fast FEV1 (≥ 30 mL/year) in the 5-year follow-up, as previously described [28]. During the following years, one of the asthmatics with a slow FEV1 decline

and 4 asthmatics with a fast FEV1 decline experienced difficulty in controlling their asthma symptoms despite taking the maximum recommended dose of inhaled corticosteroids and inhaled long-acting β2-agonist. These patients needed additional therapy with long-term oral corticosteroids or a long-acting muscarinic antagonist. The calculation of annual FEV1 decline (mL/year) was determined by subtracting FEV1 at the first year by the second measurement 5 years later, then dividing this by 5 [28].

Ten normal non-smoking subjects (6 women and 4 men, aged 47.3 ± 2.8 years), who had normal pulmonary function and no evidence of bronchial hyperresponsiveness, allergic rhinitis, or asthma, were recruited. Their total IgE levels and eosinophil counts were normal and their serum-specific IgE was negative.

#### *2.2. Study Protocol*

All patients who satisfied the enrollment criteria performed routine pulmonary function tests and methacholine provocation testing after abstaining from short-acting oral or metered-dose inhaler bronchodilator use for 6 h, and long-acting β<sup>2</sup> agonist for 24–48 hours. Asthmatic subjects underwent the procedure of fiberoptic bronchoscopy with bronchial biopsy and bronchoalveolar lavage. Informed consent was obtained from all patients before entry into the study. The study was approved by the Chang Gung Memorial Hospital Ethics Committee (IRB number: 90-13).

#### *2.3. Fiberoptic Bronchoscopy*

All study patients received fiberoptic bronchoscopy under sedation. The procedure of bronchoscopy under sedation and administering of local anesthesia was performed in the study institution. Sedation with intravenous midazolam (5–10 mg), and local anesthesia with 2% xylocaine solution was performed during bronchoscopy. Oxygen saturation, blood pressure, and electrocardiography (ECG) were monitored during bronchoscopy. The bronchoscope was advanced through the nose and larynx, and then into the tracheal and bronchial lumen; bronchial biopsies were taken from the 4th or 5th subsegmental bronchus. The specimens were fixed by 4% paraformaldehyde for immunocytochemistry.

#### *2.4. Preparation of BAL Cells*

Bronchoalveolar lavage (BAL) was completed in subjects using 300 mL of 0.9% saline solution [29]. Sterile saline solution was instilled into the right fourth or fifth subsegmental bronchus. The lavage fluid was retrieved by gentle aspiration, collected, and filtered through two layers of sterile gauze. BAL fluid was kept on ice throughout processing. The collected BAL fluid was centrifuged at 600× *g* for 20 min at 4 ◦C. The cell pellet was obtained after centrifugation followed by consecutive washes and finally resuspended at 10<sup>6</sup> cells per mL in RPMI-1640 (GIBCO, Grand Island, New York, NY, USA) containing 5% heat-inactivated fetal calf serum (FCS, Flow Laboratories, Paisley, Scotland, UK). Trypan blue exclusion was used to determine cell viability. Differential cell counts were determined by counting 500 cells on cytocentrifuge preparations using a modified Wright–Giemsa stain. BAL fluids were stored at −70 ◦C until analysis. The purified alveolar macrophages were placed in 6-well plates at <sup>10</sup><sup>6</sup> cells/mL for 24 h at 37 ◦C and 5% CO2. The culture supernatant was collected and frozen at <sup>−</sup><sup>70</sup> ◦<sup>C</sup> until use.

#### *2.5. Immunocytochemistry*

Immunoreactivity for tissues was performed with the use of the avidin–biotin peroxidase complex method. Tissue sections (5 μm) from asthmatic subjects were incubated overnight at 4 ◦C with a variety of primary antibodies, including anti-human MMP-9 (Oncogen Science Inc, Cambridge, MA, USA) and TIMP-1 antibodies (Fuji Pharmaceutical Co, Toyama, Japan) [30]. Mouse immunoglobulin G1 (Dako, Kyoto, Japan) was used for negative controls. After washing in PBS/Tween 20 twice, the slides were counterstained by hematoxylin. Positive immunostaining was visualized as brown granules contained in the cytoplasm. The scores corresponding to MMP-9 and TIMP-1 immunostaining

expression were evaluated by a semi-quantitative assessment using the intensity and percentage of positively stained cells, such as epithelial cells, inflammatory cells, and mucus gland or smooth muscle. The intensity of MMP-9 and TIMP-1 staining was scored as follows: 1, weak; 2, moderate; and 3, strong (Figures 1 and 2). The percentage scores were determined by the following definition: 1, ≤25%; 2, 26–50%; 3, 51–75%; and 4, >75%. These scores were multiplied by the intensity and the percentage score. The range was between 1 and 12.

**Figure 1.** Immunohistochemical expression levels of MMP-9. An avidin–biotin complex immunohistochemical study for MMP-9 labeling was performed in airway tissues obtained from asthmatic patients with a slow FEV1 decline (A) and a fast FEV1 decline (B and C). Positive staining was defined as brown–yellow particles or tan–brown particles in the cytoplasm (magnification, 200×). (**A**) Airway tissue with weak staining (score 1); (**B**) airway tissue with moderate staining (score 6); (**C**) airway tissue with strong staining (score 12); (**D**) the immunohistochemistry score of MMP-9 was determined through a semi-quantitative assessment by calculating the intensity and percentage of positive cells. The central horizontal lines indicate the median, and the error bars (upper and lower horizontal lines) are the 75th percentile and 25th percentile, respectively, \*\*\* *p* < 0.0001. IHC, immunohistochemistry; MMP-9, matrix metalloproteinase-9.

**Figure 2.** Immunohistochemical expression levels of TIMP-1. An avidin–biotin complex immunohistochemical study for TIMP-1 expression was performed in airway tissues derived from asthmatic patients with a slow FEV1 decline (A and B) and a rapid FEV1 decline (C). Positive staining was defined as brown–yellow particles or tan–brown particles in the cytoplasm (magnification, 200×). (**A**) airway tissue with weak staining (score 1); (**B**) airway tissue with moderate staining (score 6); (**C**) airway tissue with strong staining (score 12); (**D**) immunostaining of TIMP-1 was scored through a semi-quantitative assessment by calculating the intensity and percentage of positive cells. The central horizontal lines indicate the median, and the error bars (upper and lower horizontal lines) are the 75th percentile and 25th percentile, respectively. IHC, immunohistochemistry; TIMP-1, tissue inhibitor of matrix metalloproteinase-1.

#### *2.6. Hematoxylin and Eosin Staining (H and E)*

The thickness of the epithelium, basement membrane, and subepithelial layer was investigated by H and E staining for 20 min at room temperature. The results were visualized using an Olympus BX51 microscope (Olympus Corporation, Tokyo, Japan).

#### *2.7. MMP-9 and TIMP-1 ELISA*

Quantitative sandwich-type enzyme-linked immunoassay techniques (ELISA) were used to assay the secretory products in macrophage supernatants [26]. MMP-9 and TIMP-1 kits were used (Amersham Life Sciences, Arlington Heights, IL, USA) according to the manufacturer's instructions. The optical density was measured with a spectrophotometer set to 450 mM for all assays. Quantification was performed by interpolation from a standard curve.

#### *2.8. Statistical Analysis*

Data were presented as mean ± SEM. The data were analyzed using a Student's *t*-test for paired or unpaired data. For data with even or uneven or variation, a Mann–Whitney U test or Wilcoxon signed rank test was used for unpaired or paired data, respectively. ANOVA with post hoc analysis was used when comparing data from three groups; *p* < 0.05 was considered significant.

#### **3. Results**

#### *3.1. Demographic Features of Patients*

Table 1 summarizes pulmonary function tests on asthmatic patients and normal subjects. The FVC, FEV1, and levels of PC20 in methacholine tests demonstrated no difference between the two groups of asthmatics. Initially, the PC20 was similar in both asthmatic groups. After 5 years, airway hyperresponsiveness of patients with a slow FEV1 decline was significantly decreased (PC20 from 1.87 ± 0.52 to 3.52 ± 0.81 mg/dL, *n* = 13, *p* < 0.05), while that of asthmatics with a fast FEV1 decline demonstrated an increased airway hyperresponsiveness (PC20 from 3.46 ± 1.02 to 1.14 ± 0.34 mg/dL, *n* = 8, *p* < 0.05) (Figure 3).

**Table 1.** Baseline characteristics of the normal subjects and asthmatics.


Values are mean ± SEM; F/M, female/male; FEV1: forced expiratory volume in one second; FVC: forced vital capacity; % pred., percent of predicted value; L: liter; PC20: value of methacholine provocative concentration causing a 20% decrease in FEV1. \*\*\* *p* < 0.0001 compared with group A.

**Figure 3.** The individual concentration of methacholine provocation in asthmatic patients who had a slow FEV1 decline (open circle, *n* = 13) or a fast FEV1 decline (solid circle, *n* = 8) after receiving 5 years of inhaled anti-asthma medicine, compared with the initial values (initial). The significance is indicated.

#### *3.2. Cellular Profile Analysis of Bronchoalveolar Lavage*

The cellularity of lavage fluid was similar between the asthmatic groups but was significantly lower in normal subjects. Compared to normal subjects, there was an increase in the percentage of eosinophils and lymphocytes and a corresponding decrease in the proportion of alveolar macrophages in asthmatics with either a fast or a slow decline in FEV1 (Table 2). The cellularity of different cell types in normal subjects and asthmatics is presented in Figure 4 and had the same result as the percentage of cell types. The asthmatic patients with a fast FEV1 decline over 5 years had a marked rise in the proportion of neutrophils compared with those who exhibited a slow decline in FEV1 or normal subjects (Table 2). In addition, the magnitude of the annual FEV1 decline in asthmatics was highly correlated with the cellularity of neutrophils in BAL (r = 0.718, *n* = 21, *p* = 0.0002).


**Table 2.** Characteristics of bronchoalveolar lavage.

Abbreviation: AMs = alveolar macrophages; Values present as mean ± SEM; \* *p* < 0.05, \*\* *p* < 0.01 compared with normal subjects. # *p* < 0.05 compared with normal subjects or asthma with a slow FEV1 decline.

**Figure 4.** The concentration of different cell types in bronchoalveolar lavage fluid derived from normal subjects (Normal, *n* = 10), and asthmatic patients with a slow FEV1 decline (Slow FEV1 decline, *n* = 13) or a fast FEV1 decline (Fast FEV1 decline, *n* = 8); \*\* *p* < 0.01 is indicated.

#### *3.3. Expression of MMP-9 and TIMP-1 in Bronchial Biopsies*

MMP-9 was expressed in airway tissue obtained from asthmatics, especially in epithelial cells, inflammatory cells, or gland cells (Figure 1A–C). Furthermore, asthmatic patients with a rapid FEV1 decline had a significantly higher immunohistochemistry (IHC) score of MMP-9 (median, 8.5; IQR, 6.5–9.0), when compared with asthmatic patients with a slow FEV1 decline (median, 2.0; IQR, 1.0–4.0) (*p* < 0.0001, Figure 1D). The IHC score of TIMP-1 expression in the airway showed no significant difference between the two groups (Figure 2D). The thickness of the basement membrane (15.5 ± 2.2 μm, *n* = 8, *p* = 0.0002) and subepithelial layer (138.3 ± 12.2 μm, *n* = 8, *p* < 0.0001) in asthmatics with a fast FEV1 decline was significantly increased (Table 3). However, the asthmatics with a slow FEV1 decline had a thinner basement membrane and subepithelial layer.

**Table 3.** The thickness of airway structure in patients with asthma.


Data expressed as mean ± SEM.

#### *3.4. Generation of MMP-9 and TIMP-1 From Macrophages*

Most importantly, alveolar macrophages (AMs) from chronic asthma with a fast FEV1 decline spontaneously released a higher amount of MMP-9 (8.52 ± 3.53 ng/mL, *n* = 8, *p* < 0.05) than those of asthma with a slow FEV1 decline (0.99 ± 0.20 ng/mL, *n* = 13) or normal subjects (0.47 ± 0.20 ng/mL, *n* = 10). The level of MMP-9 was significantly higher in asthma with a slow FEV1 decline compared to normal subjects (Figure 5A). Compared with MMP-9, concentrations of the inhibitor, TIMP-1, were higher in supernatants from normal subjects and asthmatic groups (Figure 5B). A significantly higher level of TIMP-1 released from AMs into the culture medium in asthmatics with a slow FEV1 decline over 5 years was observed when compared to that of asthmatics with a fast decline in FEV1 or normal subjects (Figure 5B). When data are expressed as the molar ratio of enzyme to inhibitor, chronic asthmatics with a fast FEV1 decline in 5 years had a significant increase in this ratio in AMs (Figure 6). The ratio of MMP-9/TIMP-1 showed no difference between the normal subjects and those suffering from chronic asthma with a slow FEV1 decline at the 5-year follow-up (Figure 6).

**Figure 5.** Individual concentration of (**A**) MMP-9 and (**B**) TIMP-1 spontaneously released from alveolar macrophages after 24 h culture in normal subjects (open square, *n* = 10), chronic asthma with a slow FEV1 decline (open circle, *n* = 13), and chronic asthma with a fast FEV1 decline in (solid circle, *n* = 8). The significance is indicated.

**Figure 6.** Ratio of MMP-9 to TIMP-1. Data are expressed as the individual value of MMP-9 to TIMP-1 ratio in normal subjects (open square, *n* = 10), chronic asthma with a slow FEV1 decline (open circle, *n* = 13), and chronic asthma with a fast FEV1 decline (solid circle, *n* = 8). The *p* values are presented.

The generation of MMP-9 released from AMs was positively associated with the annual decline in FEV1 (r = 570, *n* = 21, *p* < 0.01) (Figure 7A). However, there was no correlation between the level of TIMP-1 and the annual decline in FEV1 (Figure 6B). When the MMP-9/TIMP-1 ratios were plotted against the magnitude of FEV1 change, the annual decline in FEV1 was significantly proportional to the ratio of MMP-9/TIMP-1 (r = 0.584, *n* = 21, *p* < 0.01) (Figure 7C).

**Figure 7.** Correlation of the level of (**A**) MMP-9 and (**B**) TIMP-1 released from alveolar macrophages (AMs) cultured for 24 h with the magnitude of FEV1 decline per year (mL/year) from patients with chronic asthma receiving inhaled corticosteroids. (**C**) Correlation of the MMP-9/TIMP-1 ratio in 24 h AM culture with the magnitude of FEV1 decline per year over 5 years from patients with chronic asthma receiving inhaled corticosteroids.

#### **4. Discussion**

We demonstrated that the levels of MMP-9 expression in the epithelium, inflammatory cells, and submucosa as determined from immunostaining showed upregulation in asthmatic patients who regularly received inhaled corticosteroids and who had a rapid decline in pulmonary function at 5-year follow-up. The alveolar macrophages from these unstable asthmatics also spontaneously released higher amounts of MMP-9. Most importantly, the MMP-9 level and MMP-9/TIMP-1 ratio produced from AMs were significantly decreased in chronic asthma with a slow FEV1 decline. The higher levels of TIMP-1 released from cultured AM were observed in clinically stable asthmatics. The magnitude of annual FEV1 decline was proportional to the MMP-9 generation and the ratio of MMP-9/TIMP-1 released from alveolar macrophages, even though the patients had regularly received inhaled corticosteroids. An increased thickness of the basement membrane and subepithelial layer was observed in asthmatics with a fast FEV1 decline. Taken together, MMPs (mainly MMP-9) and TIMP-1 may contribute to progressive loss of lung function and increased cellular matrix fibrosis of the airway wall in chronic asthmatics who demonstrated a poor response to anti-asthma treatments.

We reported that the generation of MMP-9 from alveolar macrophages and the ratio of MMP-9/TIMP-1 are strongly associated with the magnitude of FEV1 decline in chronic asthma, which is in agreement with the data of Vignola et al. [21]. The same authors have stressed the potential importance of MMP-9 and TIMP-1 imbalance in asthma by showing that the basal airway caliber was related to the ratio of MMP-9 to TIMP-1. It was reported that concentrations of MMP-9 in airway neutrophils and BAL fluid revealed a significant correlation after allergen challenge [31]. We also observed that the BAL fluid of asthmatics with a rapid decline in FEV1 exhibited a higher amount of neutrophils, while that of stable asthmatics did not show an increased amount of neutrophils. The cellularity of neutrophils in BAL was highly correlated to the annual FEV1 decline in asthmatics. These observations raise the possibility that neutrophils may cause injury to the airway and result in further remodeling of chronic unstable asthma, as has been suggested by studies showing persistent bronchial neutrophilia in severe asthma and status asthmatics [32]. Our results strongly suggest that alveolar macrophages and neutrophils may be important sources of MMP-9 release, leading some asthmatic patients—who are poorly responsive to inhaled corticosteroid treatment—to develop progressive irreversible airway scarring and fibrosis. Accordingly, an excess of MMP-9 release in the airway was associated with impairment in the lung function of FEV1.

It is thought that the imbalance between MMPs and TIMPs may play an important role in the process of the degradation and synthesis of the extracellular matrix of the airway. Hoshino et al. [33] reported that deposition of the basement membrane matrix components—including collagen III, collagen V, and tenascin—in asthma correlated to the upregulated expression of MMP-9 and, therefore, airway remodeling in asthma could cause airflow obstruction and airway hyperresponsiveness. As well as the consuming mechanisms of the extracellular matrix, MMP-9 may modulate cytokines and other proteases [34]. MMP-9 may degrade alpha1-antitrypsin and preserve neutrophil elastase activity [35], thus activating the function of fibroblasts [36]. In addition, the binding of MMP-9 to CD44 can result in the release of TGF-β1 and, therefore, regulate extracellular matrix remodeling via fibroblast activation [37]. MMP-9 may also potentiate angiogenesis by vascular endothelial growth factor activation [34,38] and increase the production of angiostatin [39]. Our results showed increased thickness in the submucosa and basement membrane, as well as higher expression of MMP-9 in chronic asthma with rapid pulmonary function decline. The asthmatics with a fast FEV1 decline presented with higher airway hyperresponsiveness. Therefore, macrophages from these unstable asthmatics may release more MMP-9, leading to collagen deposition or neovascularization in the basement membrane of airways and contributing to airway hyperresponsiveness.

The TIMPs could bind to MMPs and inhibit their enzymatic activity. In our study, alveolar macrophages from chronic stable asthma patients responsive to inhaled corticosteroids released higher amounts of TIMP-1 than those of chronic unstable asthma patients who do not respond to inhaled corticosteroids or healthy subjects. Hoshino and colleagues [27] found that corticosteroid treatment can decrease the deposition of subepithelial collagen through downregulation of MMP-9 and upregulation of TIMP-1 in asthma. In addition, a higher level of TIMP in stable asthma potentially helps the airways to mitigate the degrading activities of MMPs [40]. Nevertheless, this process may restrict cell trafficking and tissue repair, and may cause increased deposition of the extracellular matrix through in vivo inhibition of MMP-9 or other MMPs. Russell et al. reported that the release and activity of MMP-9 and TIMP-1 by alveolar macrophages from patients with chronic obstructive pulmonary disease might be important in the development of COPD because these cells exhibit increased levels of MMP-9 elastolytic activity [41]. Meanwhile, Russell et al. also reported that dexamethasone prevented the increase in MMP-9 release, and increased TIMP-1 release. Similarly, alveolar macrophages released a higher amount of MMP-9 in chronic unstable asthma patients than those with stable asthma or normal subjects in our study (Figures 5 and 6). Asthmatic patients had higher cell counts of BAL, including total cell counts, macrophage, lymphocytes, neutrophils, and eosinophils than those observed in normal subjects in Table 2 and Figure 4. MMP-9 is known to be produced by several inflammatory or structural cells, including bronchial epithelial cells, eosinophils, mast cells, and alveolar macrophages, and these may have a greater contribution relative to neutrophils in chronic asthma. This means that persistent airway or lung inflammation of chronic unstable asthmatics may be less responsive to anti-asthma treatment, which may contribute to higher cellularity of inflammatory cells, thus leading to the increased amount of MMP-9.

Although our patients have received inhaled corticosteroids for more than 5 years, some of these patients still demonstrated a progressive decline in pulmonary function and persistent airway structure change. A report showed that a high dose of inhaled corticosteroids did not change the levels of MMP-1, MMP-9, and TIMP-1 in induced sputum [42], although other authors have described a significant decrease in MMP-9 in the bronchial biopsies of asthmatics treated with inhaled corticosteroids [27]. Our results revealed that inhaled corticosteroids do not modify the expression of MMP-9 and TIMP-1 in AMs in patients with chronic asthma exhibiting a rapid decline in FEV1. Thus, inhaled corticosteroids cannot inhibit the airway remodeling of chronic asthma. Several reports have shown that neutrophil levels may increase in the airways in severe asthmatics or in patients with chronic asthma who exhibit a poor response to inhaled steroids [32,43]. Persistent generation of inflammatory products and their inhibitors in chronic asthma may cause airway injury and remodeling. The mechanisms should be further studied.

Previous reports revealed that MMP-9 increases in severe asthma [40,44], and that it may be the cause of airflow obstruction through the induction of airway structural changes [45]. The persistently high levels of MMP-9 indicate that the remodeling of airways may be initiated at the beginning of asthma development and may be expressed as severe asthma.

Extensive studies have revealed the role of MMPs in the pathogenesis of asthma, airway hyperresponsiveness (AHR), and asthma-associated airway remodeling [45,46]. MMP-9 was the first MMP to be investigated in depth for its role in the development of asthmatic pathology and was also the most highly expressed MMP in the BAL fluid and sputum of asthma patients [21,47]. Moreover, the high level of MMP-9 expression in bronchial biopsies from asthmatic patients [48] was associated with asthma severity [49] as well as the number of macrophages and neutrophils [50]. MMP-9 can degrade elastin and type IV collagen [51], an important component of the basement membrane and, thus, MMP-9 plays a role in the disruption of the basement membrane and contributes to increased ECM deposition, subepithelial fibrosis, and airway wall thickness [52]. Airway remodeling related to the thickness of the sub-basement membrane [45] has been linked to airway AHR and leads to the subsequent development of fixed airflow limitation and to a long-term decline in lung function in asthmatics [53]. Our recent study revealed that AMs produced excessive MMP-9 over TIMP-1, which was associated with increased AHR and was a predictor of the development of accelerated lung function decline [20]. Other MMPs, including MMP-1, -2, -3, or -12, are also studied as being involved in asthma [6]. MMP-1 is activated by mast cell tryptase, leading to a proproliferative extracellular matrix. The interactions of airway smooth muscle/mast cell contribute to asthma severity by transiently increasing MMP activation, thus inducing airway smooth muscle hyperplasia and AHR [54]. The potentially pro-remodeling roles for MMP-1 are involved in the promotion of airway smooth muscle proliferation. In stable mild-to-moderate asthma, MMP-2 in association with MMP-3 is released from bronchial fibroblasts and may have a negative effect on lung function and AHR [55]. However, studies on MMP-9 and MMP-2 double-knockout mice revealed that MMP-9, and not MMP-2, is the dominant airway MMP controlling inflammatory cell egression [56]. Although MMP-12 is suggested to be involved in asthma, this conclusion is based on studies in which mostly animal models are used [6]. In humans, the MMP-12-mediated pathological degradation of the ECM is associated with COPD patients [57]. Our study aimed to investigate whether the MMP released from AM was associated with AHR and increased expression in airways, thus contributing to subepithelial thickness and accelerated lung function decline. Therefore, our study focused on MMP-9. Nevertheless, a greater understanding of the involvement of MMPs in airway remodeling in asthma is indispensable and, thus, further studies are needed to examine the expression of other MMPs in airways and BAL fluid as well as their release from AMs.

#### *Limitations*

Alveolar macrophages are the predominant immune cells in the lung and are represented by the classically activated (or M1) and the alternatively activated (or M2) phenotypes according to their function [58]. Increased polarization and activation of M2 macrophages, which was induced by interleukin (IL)-4 and IL-13, are found in asthma and are suggested to be involved in asthma pathogenesis [59,60]. A previous study has shown that activation of M1 or M2 macrophages may upregulate a distinct group of MMPs and TIMP-3 [61]. A group of matrix metalloproteinases also influences M1/M2 polarization of macrophages [62]. However, our study did not investigate the association of MMP-9 with M1/M2 polarization in the asthma with accelerated lung function decline. Therefore, to understand the causal relationship of MMP-9 and M1/M2 macrophage activation in the development of airway remodeling in asthma, the macrophage polarization in BAL and airway tissue should be addressed in future studies.

#### **5. Conclusions**

We concluded that there was an increase in MMP-9 and MMP-9/TIMP-1 in airways and alveolar macrophages from chronic asthmatics with a rapid FEV1 decline, despite having been regularly treated with inhaled corticosteroids. An increase in MMP-9 and MMP-9/TIMP-1 in airways or AMs could contribute to a greater decline in lung function in cases of chronic asthma and could therefore be used as an indicator of chronic airway inflammation.

**Author Contributions:** Conceptualization, F.-T.C., C.-Y.L., and C.-H.W.; Data curation, F.-T.C., C.-Y.L., C.-C.H., and J.-R.H.; Funding acquisition, C.-H.W.; Investigation, F.-T.C. and C.-H.W.; Methodology, C.-W.L., C.-C.H., J.-R.H., and T.-F.S.; Project administration, F.-T.C. and H.-Y.H.; Resources, C.-Y.L., C.-W.L., Y.-C.H., and C.-H.W.; Supervision, C.-H.W.; Validation, F.T.C. and C.-Y.L.; Writing—original draft, F.-T.C. and C.-H.W.; Review and editing, C.-H.W.

**Funding:** This research was funded by Chang Gung Memorial Hospital Research Project Grant (CMRPG3F1492, CMRPG3F1493, CMRPG3B1321, CMRPG3B1322, CMRPG3B1323, CMRPG3F1501, and CMRPG3F1502) and Taiwan National Science Research Project (NMRP) grant (NSC90-2314-B182A-036).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Lung Microbiome in Asthma: Current Perspectives**

**Konstantinos Loverdos 1, Georgios Bellos 2, Louiza Kokolatou 2, Ioannis Vasileiadis 1, Evangelos Giamarellos 3, Matteo Pecchiari 4, Nikolaos Koulouris 1, Antonia Koutsoukou <sup>1</sup> and Nikoletta Rovina 1,\***


Received: 29 October 2019; Accepted: 12 November 2019; Published: 14 November 2019

**Abstract:** A growing body of evidence implicates the human microbiome as a potentially influential player actively engaged in shaping the pathogenetic processes underlying the endotypes and phenotypes of chronic respiratory diseases, particularly of the airways. In this article, we specifically review current evidence on the characteristics of lung microbiome, and specifically the bacteriome, the modes of interaction between lung microbiota and host immune system, the role of the "lung–gut axis", and the functional effects thereof on asthma pathogenesis. We also attempt to explore the possibilities of therapeutic manipulation of the microbiome, aiming at the establishment of asthma prevention strategies and the optimization of asthma treatment.

**Keywords:** microbiome; pathogenesis; inflammation; immune responses; asthma

#### **1. Introduction**

Asthma is the most common chronic respiratory disease, affecting more than 300 million people of all ages worldwide and killing about 250,000 of them each year [1], posing a substantial socioeconomic burden, especially in low- and middle-income countries. Asthma is a multifactorial and heterogeneous disease, comprising several different disease "phenotypes" and "endotypes" [2–4]. The current approach acknowledging that different phenotypes may share a common endotype and vice versa and, more important, that the disease phenotype may change over time has boosted our understanding on asthma pathogenesis and facilitated the development of novel targeted biological therapies, especially where they are most needed—that is severe corticosteroid-insensitive asthma [5]. However, while highly effective biologics are now available, and several more are in the pipeline for severe uncontrolled asthmatics with the Th2-high endotype [6], an almost-empty therapeutic arsenal is the case for those with the Th2-low endotype. Moreover, given the fact that Th2 inflammatory markers are absent in up to 50% of asthmatic patients (although an even higher corresponding percentage of 76% was reported in a very recent randomized controlled trial of patients with mild persistent asthma) [7,8], it is clear that further research is urgently needed to shed light on the biological pathways leading to non-eosinophilic asthma and, thus, promote the discovery of new treatment strategies for this large group of patients.

A growing body of evidence implicates the human microbiome as a potentially influential player that is actively engaged in shaping the pathogenetic processes underlying the aforementioned and other unresolved issues both in asthma [9,10] and in the other chronic respiratory diseases, particularly of the airways [11–13]. In contrast to earlier beliefs, a well-developed metabolically active microbial community, termed lung microbiota, resides in the lower respiratory tract of healthy humans. The entire habitat, including the microorganisms (bacteria, archaea, lower and higher eukaryotes, and viruses), their genomes (i.e., genes), and the surrounding environmental conditions are defined as the lung microbiome [14]. The latter is involved in a constant cross-talk with the host, and the same applies for the other bacterial communities residing in and on the human body, as well, with the most prominent being the gut microbiota, forming the "gut–lung axis" [15]. The diverse mechanisms mediating this cross-talk are now being gradually recognized. Under normal conditions; the interaction between the microbiota and the host confers mutual benefits for both ("symbiosis"). However, we are becoming increasingly cognizant of the fact that lung microbiota composition and diversity are affected in disease (including asthma) and that these changes can be translated in altered host immune responses, influencing asthma susceptibility, phenotype, exacerbation pattern, and treatment responsiveness ("dysbiosis" instead of "symbiosis").

In this article, we review current evidence on the characteristics of lung microbiome, the modes of interaction between lung microbiota and host immune system, the role of the "lung–gut axis" and the functional effects thereof on asthma pathogenesis. We also attempt to explore the possibilities of therapeutic manipulation of the microbiome, aiming at the establishment of asthma prevention strategies and the optimization of asthma treatment.

#### **2. Microbiome**

#### *2.1. Historical Perspectives*

The perception of the existence of bacteria inhabiting certain parts of the human body without causing disease is not new. For decades, diverse microbial genera were isolated after in vitro cultivation of biological samples collected from healthy individuals. However, the sensitivity of these traditional culture-dependent microbiological methods was severely limited. For instance, in an early study, it was estimated that only 24% of the entire microbiota present in an adult male fecal sample could be recovered by cultivation [16]. Using a variety of media and incubation methods, the corresponding sensitivity of in vitro cultivation for bacterial species identification in bronchoalveolar lavage samples of healthy individuals was substantially higher (61%), but it was still limited [17]. However, the exact role and teleology of this "normal (micro) flora" remained largely unknown, although it was postulated that changes in its composition could be detrimental, as in the case of antibiotic-induced *Clostridium di*ffi*cile* overgrowth in the large intestine of patients with pseudomembranous colitis [18]. The advent of novel molecular techniques for microbiological profiling, with the most important being the implementation of polymerase chain reaction (PCR) amplification and sequencing of the highly specific and ubiquitous among bacteria 16S-rRNA gene, led to a tremendous progress in our understanding of the extent, diversity, composition, and location of the sum of microbial communities living inside or on the human organism (now termed microbiota instead of normal flora) [19]. Large-scale research collaborations, such as the two phases of the Human Microbiome Program (HMP), funded by the US National Institutes of Health (NIH), and the Metagenomics of the Human Intestinal Tract (MetaHIT) project, funded by the European Community, established enormous reference databases of human microbiota genomes and metagenomes, after analyzing dozens of thousands of samples derived from 48 primary sites (mostly feces) in hundreds of healthy individuals and patients with specific conditions or disorders [20–24]. Based on these advances, it is now estimated that human microbiome consists of approximately 3.8 <sup>×</sup> 10<sup>13</sup> bacteria, probably marginally outnumbering human cells (3 <sup>×</sup> 10<sup>13</sup> for the standard age and somatotype) [25,26]. As expected, the microbial community residing in the gut is the most abundant, comprising slightly more than 1000 bacterial species [24,27]. These commensal bacteria harbor about 3.3 million genes, surpassing in number the genes contained in the host genome by approximately 150 times [24]. It soon became evident that this incredibly rich microbial ecosystem could not be uninvolved in the biological processes underlying health and disease. Further evolution

in molecular biology, namely the emergence of "-omics" technologies (genomics, transcriptomics, proteomics, and metabolomics), have now enabled the investigation of the functional effects of human microbiome by detecting and studying the functional genes encoded by the microbial community and their products (proteins, metabolites etc.) [19].

Amplicon-based sequencing of marker genes, such as 16S rRNA, is a powerful tool for assessing and comparing the structure of microbial communities at a high phylogenetic resolution. Because 16S rRNA sequencing is more cost-effective than whole-metagenome shotgun sequencing, marker gene analysis is frequently used for broad studies that involve a large number of different samples. With the expanded use of 16S rRNA sequencing for resident microbiota recognition on different human surfaces, organs such as the lungs, the stomach, and the uterus, previously considered sterile based on culture-dependent studies, were shown to host a substantial microbial burden under normal conditions. These findings gave birth to the notion of lung microbiome and primed tenacious research endeavors for its characterization.

#### *2.2. The Lung Microbiome in Health*

#### 2.2.1. The Early Life Shaping

Although not specifically studied in humans, the development of the lung microbiome probably adheres to that of the rest of the human body microbial ecosystem. The exact starting time point for the bacterial colonization of the human body cannot be accurately determined. Until recently, amniotic fluid, which fills fetal lungs prenatally, was considered sterile. This historical belief was challenged by the discovery of bacterial DNA in amniotic fluid and placental samples [28], which may be suggestive of a prenatal initiation of lung microbial colonization and development, although the actual existence of an amniotic fluid microbiome remains controversial [29] and its potential significance vastly unknown. Detectable microbial communities in multiple body sites have been identified in newborns as early as <5 min after delivery, and their synthesis initially resembles the maternal vagina or skin microbiota composition, depending on the mode of delivery (vaginal or caesarian section) [30]. This premature microbiome has been shown to change in composition and diversity and mature functionally during the first two to three years of life, after which it gradually stabilizes to a pattern closely matching that of adults [31–33]. This early life microbiota instability, probably in parallel with the concurrent immune system immaturity, is believed to render microbiome particularly susceptible to the influence of diverse environmental factors, including diet (e.g., breastfeeding), day care, crowding, and antibiotic use [34,35], which ultimately shape the structure of the adult human microbiome and, presumably, confer predisposition or resistance to disease ("window of opportunity" theory) [36] (see Figure 1). A similar trajectory of diversity and composition changes with increasing age has recently been described in the lung microbiota of mice [37].

#### 2.2.2. The Immigration/Elimination Balance

The synthesis of any bacterial (or other living organism) community at any given time is dictated by the interaction between three factors: (1) immigration, (2) elimination, and (3) local reproduction of community members (see Figure 2).

**Figure 2.** Factors determining the immigration/elimination balance.

With regard to lung microbiome, immigration mostly originates from subclinical micro-aspiration from the upper respiratory tract (URT), although other sources, including the inspired air (which carries approximately 105–106 bacteria/m3) [38], and the upper gastrointestinal tract via aspiration of gastric contents [39], may also make minor contributions. The elimination of lung microbial community members can be assumed to be mediated by the complete armory of lung defense mechanisms, including cough reflex, mucociliary clearance, and innate and adaptive immunity [40], and, thus, to depend on its effectiveness. In health, the balance between dispersal of microbes from the URT and eradication of lung microbial community members via local defense mechanisms is considered the major determinant of lung microbiome characteristics, whereas local bacterial reproduction most probably plays a rather minor role. The URT as the major source of lung microbiome is strongly supported by data showing a close resemblance between upper and lower respiratory tract (LRT) microbiome composition in healthy individuals [41–43]. Further corroboration of these results comes from the study of Venkataraman et al., demonstrating that lung microbiome composition in healthy individuals could be best attributed to neutral dispersal of microbes from the oropharynx rather than active local bacterial selection in the lungs [16]. Even more so, in a study investigating the possibility of spatially determined intrapulmonary discrepancies in the characteristics of lung microbiome, Dickson et al. showed that microbial richness and Firmicutes phylum abundance in the right upper lobe was more similar to those of the supraglottic region compared with the other lung lobes sampled, in which microbiome is practically identical [44]. Given the closer vicinity of the right upper lobe to the URT, these findings suggest that not only is the LRT microbiome directly related to that of the URT, but also this association is probably inversely proportional to the distance from the oropharynx (i.e., the more proximal to the oropharynx, the closer the microbiome resemblance). Conversely, if intrapulmonary bacterial growth was a major determinant of lung microbiome synthesis, significant intra-subject variations in the microbiome characteristics of different lobes should have been expected, given the well-established between-lobes disparities in local growth conditions (e.g., oxygen tension, pH, temperature) [45]. This was not the case in the study of Dickson et al, in which all parts of the lung located distally from the URT, irrespective of exact lobe, had practically identical microbiome [44].

#### 2.2.3. Composition and Structural Features

In contrast with the microbial communities residing in other parts of the human body, our knowledge on normal lung microbiome features, in terms of its development, composition (particularly at the genus and species levels), functional effects, and their determinants remains largely incomplete. To some extent, this reflects sampling difficulties, resulting in most relevant studies suffering from small-size limitations and lack of longitudinal data [13].

Lung microbiota is a relatively small bacterial community. Based on the findings of studies applying 16S rRNA sequencing in endobronchial brushing (EB) samples from healthy and diseased individuals, it is estimated that there are on average 103–105 bacterial genomes (or 16S copies) per cm2 of bronchial tissue sampled, although with significant inter-subject variability [41,46]. Comparatively, colon microbiota, which is the most abundant microbial ecosystem in the human body, comprises up to 10<sup>11</sup> CFU/gr of luminal content [47]. In a study evaluating the associations between the diverse microbial communities of the aero-digestive tract, bacterial density in bronchoalveolar lavage (BAL) fluid was found to be 100- to 1000-fold and 10- to 100-fold lower than in oral washes and gastric aspirates, respectively, of the same healthy subjects [43]. Sequencing of these 16S rRNA genes and comparison with established microbial genomic databases have led to the identification of 38 bacteria phyla, with 303 [48], or even more [49], genera residing in the human lung. However, these are far from equally represented with the top six phyla and the top 25 genera accounting for 86% and 65%, respectively, of all sequences identified [48]. Specifically, *Bacteroidetes* and *Firmicutes* are the most abundant phyla in the lung microbiota of healthy humans, followed by *Proteobacteria* and, to a lesser extent, *Actinobacteria* and *Fusobacteria* [41,43,46,48,50]. At the genus level, *Prevotella*, *Veillonella*, and *Streptococcus* are generally considered the most dominant taxa [41,42,46,51], although there is substantial variation in the relevant abundance of lung commensal microbes between studies and *Neisseria, Haemophilus, Fusobacterium*, or other genera (e.g., *Actinomyces, Porphyromonas,* and *Lactobacillus*) are occasionally found in comparable counts [42,46–49,52]. Although these discrepancies are probably, at least in part, due to the small sizes of the relevant studies, the presence of considerable inter-subject variability in lung microbiota composition has been documented in healthy individuals [44,52], so that, to date, it is not possible to define a typical ("normal") lung microbiome.

#### *2.3. Potential E*ff*ects of Sampling Methods on the Assessment of Lung Microbiome Structure*

The study of lung microbiota parameters requires bronchoscopy for BAL or EB samples acquisition. Induced sputum analysis may be a less invasive alternative, although differences in bacterial composition of sputum samples compared with those retrieved by bronchoscopic techniques, as a result of contamination by the rich oral/oropharyngeal microbiota, have been well documented [49]. Moreover, there is a certain degree of uncertainty surrounding putative effects of the specific bronchoscopic technique used for sampling on the characteristics of the derived microbiota. Denner et al. reported significant discrepancies in the extent, diversity, and relative affluence of the retrieved microbial communities between BAL and EB bronchoscopic sampling, the latter being associated with a denser and more diverse microbiome [48]. An earlier study comparing the microbial communities sampled by multiple respiratory tract sites of healthy individuals, demonstrated that the bacterial population of the left lower lobar bronchus retrieved by EB was generally larger than the populations of the right middle lobe segmental bronchi, which were sampled by consecutive BALs, although no differences were found in the corresponding bacterial communities' composition [41]. Other researchers have ruled out bronchoscopic-technique-specific effects on the lung microbiota synthesis after applying both BAL and EB in contralateral lobes of healthy volunteers [44]. Although the reported discrepancies could be merely the result of differences in the nature of the samples (and in the modes of their acquisition), along with small study sizes, they might also represent a shift in the microbiota characteristics of the small peripheral airways and the lung parenchyma (sampled by BAL) compared with the more proximal bronchi (sampled by EB). Until further research addresses these issues and establishes a standardized technique for lung microbiota derivation, sampling methods should be taken into account when designing or evaluating the results of human lung microbiome studies.

#### *2.4. Spatial Discrepancies in the Structure of Lung Microbiome*

Some degree of spatially dependent intra-subject variability in lung microbiota features has been demonstrated, although this is substantially more limited than the aforementioned between-subjects variability [44]. Specifically, it has been shown that the right upper lobe microbial community richness, composition, and variation are all significantly different from those of more distal parts of the lungs (left upper lobe, right middle lobe, and lingula) of the same healthy individuals and more closely resemble the upper respiratory tract microbiota [44]. Finally, it should be mentioned that, in contrast with the gut microbiome which has been shown to present significant geographical variation possibly associated with the regional lifestyle [32], there are no data suggesting a similar trend in lung microbiome. Despite the lack of direct comparisons between populations from different parts of the world, no dissimilarities have been found in the lung microbiome of HMP initiative participants from eight different US cities [42] and, to date, the relatively few studies conducted in non-Western populations do not seem to yield different results from the majority of lung microbiome studies, which have generally been confined to the Western world [53,54].

#### Cross-Talk between Lung Microbiota and the Host

The host immune system is primarily responsible for the conduct of most of the host–microbiome interplay, and there is now a growing body of evidence establishing the presence of an active and multiform cross-talk between the lung microbiome and the host immune system [11]. Invading pathogens activate the inflammasomes (multi-meric protein complexes) both directly and indirectly [55], to produce inflammasome associated pro-cytokines (IL-18, IL-1β), after the recognition of the pathogens by a family of receptors through pathogen-associated molecular patterns (PAMPs) [56]. Structural components of the bacterial cells and LPS (a ubiquitous structural component of Gram-negative

bacteria outer membrane) are ligands for the pattern recognition receptors (PRRs) expressed by the host antigen-presenting cells. Upon stimulation, PRRs (with Toll-like receptors (TLRs)-2 and -4 being the principle representatives) may trigger diverse cellular processes regulating immune responses in the lung. Importantly, these responses appear to be bacterial species- or genus-specific, underscoring the potentially significant effects of lung microbiome composition alterations on the host immune regulation (see Figure 3).

**Figure 3.** Respiratory microbiota produce metabolites (SCFAs, NO, or nitrite, aromatic amino acids), which influence host immune activity. Inflammasome-associated pro-cytokines can be produced and activated by a family of receptors which detect the presence of pathogens through PAMPs. SCFA: short chain fatty acids; NO: nitric oxide; AAs: aromatic amino acids; TLR: Toll-like receptors; ROS: reactive oxygen species; PAMP: pathogen associated molecular patterns; MMP: matrix metalloproteinase; MDC: macrophage-derived chemokine; MIP1α: macrophage inflammatory protein 1α.

Interspecies differences in LPS structure are thought to account for the corresponding variations in TLR subtype specificity and lung inflammatory capacity [57,58]. The Bacteroides *Prevotella*, one of the most abundant genera in the healthy lung microbiome, appears to exhibit a TLR2-dependent low inflammatory potential, whereas the Proteobacteria *Haemophilus influenzae* and *Moraxella catarrhalis,* linked with lung microbiome alterations in asthma and COPD, induce severe TLR2 independent (and probably TLR4-dependent) lung inflammation and injury in mice [57]. Other potentially pathogenic Proteobacteria residing in the lung, such as *Pseudomonas aeruginosa*, *Stenotrophomonas maltophilia*, and *Burkholderia* spp, possess flagella, the major structural component of which flagellin is recognized by host TLR5, leading to the induction of pro-inflammatory mediators' secretion [59]. Bacterial DNA may also stimulate host immune responses. This effect is mediated by the abundant in bacterial DNA sequences unmethylated CpG dinucleotides, which bind to the host TLR9, inducing an inflammatory response of the T-helper-1 (Th1) type [60]. Furthermore, all four major phyla of the lung microbiome (Bacteroidetes, Firmicutes, Proteobacteria, and Actinobacteria) have been shown to stimulate NOD2 receptors in vitro, and this effect is mediated by the bacterial cell wall component peptidoglycan [61].

Beyond structural microbial components, resident microbes-derived metabolites are also believed to be actively involved in the interplay between lung microbiome and the host immune system. Short-chain fatty acids (SCFAs) and amino acids metabolism products are the most extensively studied metabolites generated by human microbiota. SCFAs, such as butyrate, propionate, and acetate, are the main end-product of dietary fiber fermentation undertaken by the gastrointestinal tract microbiota. However, there are also data implying SCFAs production from lung microbiota, since active expression of bacterial genes associated with SCFAs metabolism has been described in bronchial brushings [62]. SCFAs have been implicated in numerous mechanisms promoting maintenance of homeostasis in health, including preservation of gut barrier integrity, control of appetite and energy intake, protection against autoimmunity and tumorigenesis in colon, and regulation of blood–brain barrier permeability, among others [63,64]. Importantly, they have also been shown to possess immunomodulatory properties [64,65]. The latter are mediated by G protein coupled receptors, particularly GPR41, GPR43, and GPR109a, which are all expressed on the surface of most types of inflammatory cells (macrophages/monocytes, dendritic cells, and neutrophils), as well as by direct inhibition of histone deacetylases (HDACs) [63–65], i.e., enzymes actively participating in post-translational modifications of the histones-DNA interaction inside the chromatin structure that regulates cell transcriptional activity and gene expression [66]. The principal immunomodulating effect of SCFAs appears to be the induction of differentiation and proliferation of extra-thymic regulatory T cells (Tregs) [67–69], a lymphocyte subset with established anti-inflammatory and anti-allergic effects [70,71]. Apart from the SCFAs, indole-3-acetate, a metabolite of the amino acid tryptophan produced by gut microbiota, was recently shown to attenuate LPS-induced pro-inflammatory cytokine and chemokine secretion in alveolar macrophages derived from smokers, and this observed anti-inflammatory action of indole-3-acetate was hypothesized to potentially mediate the effect of azithromycin on lowering exacerbation rates in COPD patients [72]. Indole-3-acetate and the other tryptophan metabolites are activators of the aryl hydrocarbon receptor [73], another well-known modulator of inflammation and immunity, with a proposed role in Treg generation [74].

We previously focused on microbiota-derived mediators engaged in shaping diverse aspects of host immune response. It must be realized that the opposite is most probably also valid. This means that numerous signaling molecules originating from host cells can be sensed by resident microbes and are capable of modifying the composition and diversity of lung microbiome. Indeed, there are data from several in vitro studies demonstrating potential influences of catecholamines and cytokines on the growth and virulence of various bacterial strains, perhaps with a species- or genus-specific manner [75–79]. This host-to-microbiome signaling pathway may be implicated in respiratory disease pathogenesis by contributing to lung microbiome alterations favoring the dominance of specific potentially pathogenic species. In this context, it has been shown that increased intra-alveolar levels of the catecholamines epinephrine and norepinephrine in BAL samples from lung-transplant recipients are significantly associated with both indices of acute infection and reduced lung microbiome diversity, characterized by the predominance of a single bacterial species, notably *Pseudomonas aeruginosa* [80].

#### *2.5. Microbiome and Asthma Susceptibility*

#### The "Gut–Lung Axis" and the Hygiene Hypothesis

Accumulated evidence arising from both human and animal studies suggests that the development of allergic diseases, including asthma, may, in fact, be dependent on the bacterial communities residing in the gut. Gut microbiota actively interact with the host immune system via bacterial structural components and secreted metabolites, and these interactions possess the ability to modulate immune responses both locally in the gastrointestinal tract and systemically, influencing various distal sites, including the lungs. The term "gut–lung axis" has been coined to account for these effects of gut microbiota on lung immunity, both in health and in disease [81] (see Figure 4).

**Figure 4.** The "lung–gut" axis interactions with host immunity. TLR: Toll-like receptors; Treg: T regulatory cell; MDC: macrophage-derived chemokine; MIP1α: macrophage inflammatory protein 1α.

The presumable involvement of resident microbiota in allergic asthma development is primarily supported by data showing an exaggerated Th2 immunity-driven airway inflammation in germ-free (GF) mice sensitized with ovalbumin compared with specific pathogen-free (SPF) counterparts, which can be abrogated when GF mice are recolonized by the commensal flora of the SPF animals prior to sensitization [82]. Likewise, gut microbiota disruption as a result of antibiotics administration promotes allergic airway disease in experimental murine asthma [83,84]. Importantly, antibiotic-induced long-term alterations in gut microbiota diversity and composition have also been observed in humans [85–87] and both early life and maternal antibiotic use have been associated with an increased risk for recurrent wheeze and asthma development in childhood [88,89]. Aside from antibiotic exposure, formula feeding [90,91] and Caesarian-section delivery [92,93] have been correlated both with differences in infant gut microbiota composition and with a heightened childhood asthma susceptibility compared with breastfeeding and vaginal delivery. Other exposures operating in early life, when the microbiome and the host immune system are both in the process of maturation, might also be relevant (see Figures 1 and 5). For instance, living with dogs or cats as pets during the first year of life has been linked to a decreased prevalence of atopy at age six to seven [94], and dust from households with dogs has been shown to enrich cecal microbiota, together with downregulating Th2-mediated airway inflammation in a murine model of allergic sensitization [95]. Fujimura et al. [95] elegantly showed that exposure to pets may be associated with distinct gut microbiota composition characteristics conferring protection against airway allergen challenge [95]. Specifically, the cecal microbiome of mice exposed to dust derived from households with dogs was significantly enriched in Firmicutes compared with mice exposed to dust from residencies without pets, with *Lactobacillus johnsonii* being the most dominant taxon.

**Figure 5.** Early life exposures that may affect the symbiosis/dysbiosis balance and predispose to asthma.

The relative microbial diversity of the environment and the level of exposure during the first years of life have also been highlighted as important factors influencing subsequent allergy and asthma susceptibility. Several studies from around the world have consistently demonstrated that children growing up in rural settings with high-level exposure to bacterially enriched farming environments presented significantly lower rates of atopic asthma compared with their non-farm peers [96]. In a landmark study published in 2016, 60 schoolchildren from two reproductively isolated US agricultural communities, the Amish and the Hutterites, sharing strong similarities in terms of genetic background and lifestyle, but with diametrically opposite farming practices, were examined [97]. The children of the Amish, who practice a traditional type of farming, presented a significantly lower rate of allergic sensitization compared to those of Hutterites, who use highly industrialized farming infrastructure, and no asthma cases were identified among Amish, in contrast to the six asthmatic Hutterite children.

Short-chain fatty acids (SCFAs) have been proposed as pivotal mediators of the gut–lung interplay. Knockout mice lacking the *gpr43* receptor gene have been shown to present an exaggerated allergic airway inflammatory response upon ovalbumin challenge [98]. In a landmark study, Trompette et al. demonstrated that the dietary fiber content may influence the extent of allergic inflammatory changes in the lungs via SCFAs-mediated defects in DC activation, leading to an impaired Th2 cell differentiation [99]. Interestingly, they showed that high-fiber diet attenuated Th2-driven lung inflammation, as assessed by total and differential cell count in BAL fluid, Th2 cytokines mRNA levels in lung tissue, serum total IgE, metacholine challenge, and lung histological analysis. On the contrary, low-fiber diet aggravated allergic inflammation in the lungs. Furthermore, the administration of SCFAs to mice through drinking water led to an acceleration of Th2 inflammation resolution in propionate-supplemented wild-type mice compared with un-supplemented controls, which was abrogated in a separate group of GPR41-dedicient mice. These results indicate that circulating SCFAs (propionate in particular) produced by gut resident bacteria in direct proportion to the dietary fiber content may reduce susceptibility to allergic airway disease through the activation of the GRP41 receptor.

#### *2.6. The Lung Microbiome in Asthma*

During the last decade, an ever-growing number of studies have attempted to shed light on the features of lung microbiome in patients with asthma [46,48,50,62,100–108]; however, the small sample sizes, the lack of uniformity and standardization in the selection and processing of respiratory samples used for microbiome characterization in asthma, the scarcity of data on potential longitudinal changes in lung microbiome composition during the course of the disease and in association with treatment implementation, and finally, the absence of a clearly described 'normal' lung microbiome with which comparisons can be safely made pose limitations to the complete delineation and interpretation of lung microbiome in asthmatic patients. Despite these caveats, studies investigating the lung microbiome in asthma have been relatively successful in capturing microbiome alterations in a large part of the spectrum of disease severity and phenotypes, thus providing an initial premature understanding of a potential association between features of lung microbiome and specific disease characteristics.

#### Alterations of Lung Microbiome Structure in Asthma

The most constant finding of lung microbiome studies in asthma is probably an observed increase in the relative abundance of the Proteobacteria phylum in asthmatic lungs [9,46,62,106–109]. At the genus level, this change is seemingly driven by a corresponding increase in the prevalence of *Haemophilus* and/or *Neisseria* [46,62,103,105–107], although other potentially pathogenic genera belonging to the Proteobacteria phylum, such as *Moraxella*, *Pseudomonas*, and members of the *Enterobacteriaceae* family, might also be involved [48,103,106,108]. It appears, though, that this Proteobacteria expansion is rather specific to non-severe asthma, as studies directly comparing non-severe with severe asthmatics have demonstrated distinct patterns of lung microbiome composition in these two groups, with Proteobacteria dominating in the lung microbiome of non-severe asthmatics, while other phyla, possibly Actinobacteria [108] or Firmicutes (mainly *Streptococci*) [106], are more prevalent in severe asthma. However, an increased relative abundance of certain Proteobacteria, such as the *Pseudomonadaceae* and *Enterobacteriaceae* (most notably *Klebsiella* spp), has also been reported in severe asthma patients [104,108].

Other taxa, e.g., the common respiratory commensals *Prevotella* and *Veillonella*, have generally been shown to be less common in the lung microbiome of patients with both severe and non-severe asthma [46,48,50,106], although these findings are not universal [62]. Overall, asthma-associated alterations in lung microbiome composition are significantly less firmly determined at the genus level compared with the phylum level, possibly reflecting, at least in part, the corresponding variations described in the lung microbiome composition of healthy individuals.

The relationship between asthma and lung microbial community diversity is even more controversial. Earlier studies in patients with mild asthma have reported an increased bacterial burden and diversity in induced sputa and EBs compared with healthy individuals [101,102]. In line with these observations, Durack et al. found a marginally increased phylogenetic diversity in EB samples retrieved from 42 asthmatic patients compared with 21 healthy controls [62]. Likewise, Sverrild et al. showed bacterial diversity augmentation in BAL fluid specimens from 23 patients with non-eosinophilic asthma [107]. On the contrary, other studies having included patients with more severe or corticosteroids refractory disease failed to replicate these findings and occasionally came up with opposite results [50,105,106,108].

#### *2.7. The Potential Role of Lung Microbiome in Shaping Asthma Phenotypes and Endotypes*

Several clinical, physiological, and inflammatory characteristics involved in the definition of asthma phenotypes (and, to a lesser extent, endotypes) [3] have been linked with features of lung microbiome (see Figure 6).

**Figure 6.** Implication of lung microbiome in asthma. TNF: Tumor necrosis factor; IL: Interleukin; IFN: Interferon; MAPK: Mitogen-activated protein kinase; SCFA: short chain fatty acid; BHR: bronchial hyperresponsiveness.

#### 2.7.1. Inflammatory Profile

Eosinophilic Inflammation

Lung microbiota diversity and composition have been investigated in patients with both severe and non-severe eosinophilic asthma and compared with the corresponding features in non-eosinophilic asthma. In a cohort of mild corticosteroid (CS)-naïve asthma, Sverrild et al. reported increased α-diversity (a measure of bacterial taxa richness and evenness) and decreased β-diversity (an index of heterogeneity between bacterial communities) in the BAL microbiome of eosinophil-high asthmatic patients compared with eosinophil-low ones [107]. At the same time, various genera were significantly enriched (e.g., *Aeribacillus, Halomonas,* and *Sphingomonas*) or depleted (e.g., *Neisseria, Bacteroides,* and *Actinomyces*) in eosinophil-high as opposed to eosinophil-low asthma. On the contrary, in another cohort encompassing patients with both severe and non-severe asthma, the *Actinomycataceae* family members were significantly more abundant in eosinophilic compared with non-eosinophilic, asthma and their relative abundance was positively associated with the eosinophil count in sputa [108]. *Tropheryma whipplei* has also been identified as a prevalent member of the lung bacterial community in severe asthmatic patients of the eosinophilic inflammatory phenotype [105].

Instead of directly comparing lung microbiome composition in patients with eosinophilic and non-eosinophilic asthma, other groups have focused on the investigation of potential associations between features of microbiome and markers of eosinophilic airway inflammation. Eosinophil infiltration of bronchial tissue and eosinophil count in BAL fluid have been associated with lower total bacterial loads and diversity in EB samples from patients with mostly severe asthma [48,104]. Particularly, *Rickettsia* and certain Actinobacteria have been positively correlated with lung eosinophilia, whereas taxa negatively correlated mostly belong to Proteobacteria (including *Moraxellaceae*) and Firmicutes [48,104]. The bronchial epithelial cell expression of specific genes known to be involved in the Th2-mediated immune response (*CLCA1*, *SERPINB2* and *POSTN*) has also been examined in relation to lung microbiome composition [62,104]. These three gene expressions correlated negatively with total bacterial burden and relative abundance of certain taxa, including the previously mentioned *Moraxellaceae* [62,104].

Despite significant variability and, to some extent, rather contradictory results between studies, these data strongly support the existence of distinct features of lung microbiota burden, diversity, and composition in patients with the eosinophilic asthma subtype. Overall, it appears that both the eosinophilic inflammatory phenotype and the underlying Th2-high endotype correspond to lung microbial communities with comparatively low bacterial load, substantial diversity but limited heterogeneity, and possibly increased representation of Actinobacteria (perhaps most notably of the *Actinomyces* genus) and decreased representation of certain Proteobacteria (potentially including *Moraxella* sp) [110].

#### Neutrophilic Inflammation

Neutrophilic asthma, along with the paucigranulocytic inflammatory phenotype, belongs to the non-Th2 (or Th2-low) asthma endotype and Th1 and/or Th17 immune processes have been implicated in its pathogenesis [109,111], although therapeutic targeting of neither Th1- nor Th17-related cytokines, namely anti-TNFa [112] and anti-IL17A receptor agents [113], have proved effective. On the other hand, certain bacterial pathogens are established triggers of Th17-driven immune responses [114], and there are some data showing reduction of neutrophilic inflammatory markers (primarily IL-8) and quality-of-life improvement following treatment with clarithromycin in patients with severe refractory non-eosinophilic asthma [115]. These observations have drawn attention to the investigation of potential associations between lung microbiome and neutrophilic asthma in particular. In the most recent and largest to-date study of lung microbiome in asthma, Taylor et al. classified 167 patients with moderate-to-severe asthma in eosinophilic, neutrophilic, paucigranulocytic, and mixed granulocytic inflammatory phenotypes based on induced sputum differential cell count percentages and searched for interphenotype dissimilarities in sputum bacterial diversity and composition [110]. Although limited by the uneven distribution of participants across the different phenotypes (only 14 had neutrophilic asthma), the study provided evidence for a significantly decreased bacterial diversity, along with increased heterogeneity in patients with neutrophilic compared with both eosinophilic and paucigranulocytic asthma. This was accompanied by a strong inverse correlation between phylogenetic diversity and neutrophil percentage in sputum. *Haemophilus* and *Moraxella* genera were found enriched in neutrophilic asthma samples and significantly correlated with asthma inflammatory profile, with the opposite being the case for *Streptococcus*. In line with these results, Simpson et al. also demonstrated a decreased bacterial diversity in induced sputa collected from patients with severe neutrophilic compared with non-neutrophilic asthma in an earlier study [105]. Again, Proteobacteria, particularly *Haemophilus influenzae*, was the phylum dominating in the neutrophilic airway microbiota, with a relative depletion of Actinobacteria and Firmicutes. Similarly, others have shown positive correlations between *Moraxella catarrhalis*, *Haemophilus* sp, and *Streptococcus* sp total abundance and sputum neutrophils percentage and IL-8 concentration in severe asthma [103].

Although observational, and thus not establishing causality, these studies may support a case for lung microbiota dysbiosis in the pathogenesis of neutrophilic asthma. It appears that the LRT of asthmatic patients with predominantly neutrophilic airway inflammation harbors a relatively uniform microbiome, in which Proteobacteria, most prominently the potentially pathogenic *Haemophilus* and *Moraxella* genera, have outgrown the taxa normally over-distributed in the lung microbiota (e.g., Firmicutes). These gradually expanding new colonizers of the asthmatic airways may, in fact, constitute the triggering factor for the aberrant Th17 immune response, along with other inflammatory pathways activation, observed in neutrophilic asthma. Indeed, non-typeable *Haemophilus influenzae* intranasally administered at a sublethal dose has been shown to produce a robust Th17 response in the lungs of experimental mice [116] and, even more relevantly, infection with non-typeable *Haemophilus influenzae* may lead to reduced eosinophilic inflammation and increased neutrophilic infiltration of the airways in an IL-17-dependent manner in a murine model of allergic asthma [117]. Although the cause of Proteobacteria overgrowth in the LRT of certain asthmatics remains elusive and may be multifactorial (e.g., selective bacterial growth favored by structural changes and/or inflammatory mediators in the airways as a result of the underlying disease process), it must be noted that, as previously discussed, inhaled corticosteroid (ICS) treatment itself may promote Proteobacteria enrichment in the lung microbiota and, thus, contribute to the development of neutrophilic asthma [2]. Apparently, if this hypothetical association between respiratory dysbiosis and neutrophilic asthma pathogenesis stands true, strategies aiming at the manipulation of lung microbiome, possibly through Proteobacteria suppression and normal diversity restoration, may open up new avenues for the treatment of this, until now, untargeted asthma phenotype. Macrolides may be part of such strategies.

#### 2.7.2. Corticosteroid Responsiveness

Several data support the correlation between lung microbiota burden and diversity and corticosteroid responsiveness. Goleva et al, in a study on CS sensitive and CS resistant asthmatics, demonstrated that, although bacterial composition both at the phylum and at the genus level was quite similar between the two groups overall, unique patterns of bacterial expansions were observed in the majority of patients with both CS-resistant and CS-sensitive asthma [50]. In CS-resistant asthmatics, the genera *Neisseria, Haemophilus,* and *Tropheryma* were abundant, while the genera *Pasteurella* and *Fusobacterium* were predominant in CS-sensitive asthma. Of note, CS-resistant patients had significantly higher levels of interleukin (IL)-8 mRNA in their BAL cells, implying that distinct lung microbiota profiles may be responsible for CS resistance in the neutrophilic asthma phenotype. To further validate their findings, the authors subsequently incubated alveolar macrophages isolated from BAL samples of asthmatic patients, with either *Haemophilus parainfluenzae* (a genus solely expanded in CS resistant patients) or *Prevotella melaninogenica*, and found a significantly reduced CS responsiveness in vitro following treatment with dexamethasone in cells cultured with *H. parainfluenzae,* but not *P*. *melaninogenica*. Durack et al. assessed bacterial composition in ten ICS responders and an equal number of ICS non-responders, both with mild ICS-naive asthma [62]. They reported significantly different microbiota profiles in the two groups, with responder's lung microbiome synthesis sharing more resemblance with that of healthy controls. In particular, the *Microbacteriaceae* and *Pasteurellaceae* (including *Haemophilus*) families were found enriched in ICS non-responders and expansions of the *Streptococcaceae, Fusobacteriaceae,* and *Sphingomonodaceae* families were shown in responders. Interestingly, in the same study, lack of CS responsiveness was associated functionally with increased expression of bacterial genes involved in biodegradation pathways, which may explain resistance to CS treatment.

Finally, Huang et al. demonstrated a significant positive correlation between lung microbiota diversity and *FKBP5* gene expression, a marker of steroid response [104]. In terms of bacterial composition, increased *FKBP5* gene expression was associated mainly with Actinobacteria and Proteobacteria phyla enrichment. These results suggesting an association between lung-microbial community diversity and CS responsiveness in severe asthma are further supported by more recent data showing a significant inverse correlation linking phylogenetic diversity in induced sputum specimens collected from patients with moderate-to-severe asthma and ICS dose with the use of both univariate and multivariate regression analyses [110].

#### 2.7.3. Effect of Treatment

Data regarding the effect of asthma treatment on lung microbiome are scarce in literature. In their study, after initial bronchoscopic sampling (EBs) for lung microbiome assessment at enrollment, Durack et al. randomized (2:1 ratio) 42 patients with mild well-controlled asthma, who had not received ICS previously, to receive a six-week course of a medium dose of ICS (250mcg fluticasone propionate twice daily) or placebo and repeated EBs post-treatment [62]. Despite limitations related to sample quantity insufficiency, the authors did not discern changes in bacterial burden and diversity after ICS treatment. However, they did find fluticasone-induced alterations in the relative abundance of certain taxa, namely an expansion of the *Microbacteriaceae* family and the *Moraxella* and *Neisseria* genera and a depletion of *Fusobacterium*, in those who responded to ICS treatment.

In another study, patients with both mild and more severe asthma were stratified according to the use of corticosteroids (inhaled and oral) [48]. Both those on ICS only and those on combined ICS and OCS treatment regimens exhibited significant alterations in EBs bacterial composition at the phylum, as well as the genus, level compared with corticosteroid-naïve asthma patients. These alterations comprised Proteobacteria enrichment and Bacteroidetes (specifically *Prevotella*) and Fusobacteria depletion in all corticosteroid groups, with decreased *Veillonella* and increased *Pseudomonas* abundance in those on ICS only and OCS, respectively. Although limited and centered exclusively on corticosteroids, these data clearly imply that asthma treatment may modify important compositional characteristics of the lung microbiome, including selection of potentially pathogenic species, with as yet unknown potential consequences.

Interestingly, most recent studies have attempted to provide insights into the possible functional properties of lung microbiome in asthma. Durack et al. employed an algorithmic prediction model (PICRUSt) to infer lung bacterial metagenomic characteristics based on 16S-rRNA sequencing in patients with mild asthma and showed potentially increased expression of genes involved in pathways mediating the metabolism of amino acids and carbohydrates, particularly SCFAs, in these patients compared with healthy controls [62]. On the contrary, a relative reduction was predicted in the activation of LPS synthesis-specific processes. Using the same analytical method in severe asthmatic patients, Huang et al. managed to associate distinct metabolic (e.g., carbohydrate digestion, indole alkaloid biosynthesis) and immune (e.g., NOD-like and RIG-I-like receptor signaling) pathways with taxa (positively or negatively) correlated with specific phenotypical features of asthma, including body-mass index (BMI), asthma control assessed by Asthma Control Questionnaire (ACQ), corticosteroid responsiveness, and Th17-driven inflammation [104]. The authors concluded that concrete members of the asthmatic lung microbiota may be actively engaged in the pathogenetic mechanisms leading to the acquisition of different disease characteristics between asthmatics (i.e., phenotypes). Furthermore, Sverrild et al., again with the use of PICRUSt, showed different predicted functional profiles of lung microbiota associated with mild eosinophilic compared with non-eosinophilic asthma [107]. Clearly, further research taking advantage of the novel technologies in metagenomic analysis of lung microbiota is urgently needed to better describe the potential mechanistic role of lung microbiome alterations in asthma pathogenesis and disease phenotype/endotype determination.

#### 2.7.4. Bronchial Hyperresponsiveness (BHR)

BHR to direct or indirect stimuli, in addition to airway inflammation, has long been considered inherent to asthma. However, neither its presence nor its severity is uniform or stable in all asthmatic individuals, and, thus, BHR may potentially constitute a contributing factor in asthma phenotyping [118]. BHR (indicated by metacholine PC20 concentrations) has been shown to correlate significantly with lung microbiota diversity in asthma, with greater bacterial diversity associated with lower metacholine PC20 concentrations [101]. In the same study, Proteobacteria comprised the majority of resident taxa correlated with greater BHR. Likewise, lung microbiota diversity was greater in those patients with asthma that exhibited the larger reductions in BHR following a course with clarithromycin [101].

#### 2.7.5. Lung Function

The potential impact of lung microbiome composition on lung function in patients with asthma has mainly been addressed by the study of Denner et al. [48]. The authors stratified their cohort of asthmatic patients (from across the range of disease severity) according to FEV1 and noticed a significantly decreased relative abundance of various bacterial phyla (Firmicutes, Bacteroidetes, and Actinobacteria) and genera (*Prevotella, Veillonella,* and *Gemella*) in the BAL fluid of those with the lowest FEV1 values (FEV1 < 60%). In contrast, *Lactobacillus* was found to be enriched in patients with the most severely impaired lung function. Similar positive correlations between the abundance of the *Bacteroidaceae* family or lung microbiota phylogenetic diversity and FEV1 have also been reported elsewhere [108,110]. Others have reached different conclusions with respect to taxa-specific associations with lung function in severe asthma, showing significantly reduced FEV1 levels in those patients with *M. catarrhalis, Haemophilus* sp, or *Streptococcus* sp predominance in induced sputa microbiota compared with others who had different taxa dominating the bacterial communities of their sputa [103].

#### 2.7.6. Obesity

Obesity has repeatedly been identified in cluster analyses of asthmatic populations as a key clinical feature discriminating a distinct subset of patients (mostly female) with adult onset, non-Th2 mediated, highly symptomatic and difficult-to-treat asthma from other asthma phenotypes [119–121]. Although the exact pathogenetic mechanisms linking obesity with asthma remain unknown, studies in obese asthmatics have shown improvements in BHR, asthma control, lung function, and inflammatory indices following bariatric surgery [122,123] or diet-induced [124,125] weight loss.

In a cohort of patients with severe asthma, high BMI was significantly associated with a distinctive lung microbiota composition, mainly consisting of Bacteroidetes (including *Prevotella* species) and Firmicutes (e.g., *Clostridium* species) [104]. As expected, these obese asthmatic individuals presented less bronchial eosinophilic infiltration in EB specimens (non-eosinophilic asthma). It is noteworthy that bacterial taxa associated with obesity in this study exhibited distinct predicted functions, including activation of pathways involved in host immune response, such as NOD-like receptor signaling. These data suggest a potential role for lung microbiome in shaping the obesity-related asthma phenotype.

#### 2.7.7. Smoking Asthma

Although no significant difference has been reported in BAL bacterial communities composition between healthy smokers and non-smokers [42], an analysis of induced sputum microbiota profiles in patients with severe asthma demonstrated increased total bacterial diversity and relative abundance of Fusobacteria in ex-smokers compared with never smokers, while smoking intensity (assessed by pack–years of smoking) was directly associated with the prevalence of Actinobacteria [105]. These findings raise the possibility that smoking may be involved in shaping lung microbiome alterations observed in asthma.

#### *2.8. Lung Microbiome and Asthma Exacerbations*

The most common precipitating factor triggering asthma exacerbations is viral respiratory tract infections, particularly from rhinoviruses [126]. On the contrary, bacterial lung infections have not been shown to trigger asthma exacerbations but for in a minority of cases, with the possible exception of the self-limiting atypical bacterial infections caused by *Mycoplasma pneumoniae* and *Chlamydophila pneumoniae* for both of which a relative prevalence of 18% has been serologically detected during asthma attacks in adults and children, respectively [127]. On the other hand, considerably high rates of the potentially pathogenic Proteobacteria *Haemophilus influenzae* (the most prevalent), *Streptococcus pneumoniae*, and *Moraxella catarrhalis* have been detected by both culture-based and molecular techniques from upper- and lower-respiratory-tract samples of asthmatic children during

exacerbation periods [128,129]. Moreover, a cluster analysis based on sputum cellular and mediator profiles in adult asthmatic patients during exacerbations has linked the predominance of different bacterial phyla in lung microbiota with specific clinical features and inflammatory markers [130] (see Figure 7).

**Figure 7.** Exacerbation biologic clusters in asthmatic and COPD patients using the subjects' discriminant scores. Adapted from ref. [130], Creative Commons Attribution License (CC BY).

In order to provide an explanation for the above observations, a model of respiratory dysbiosis involving an altered lung microbiome and a dysregulated host immune response has been proposed to delineate the occurrence of asthma (and other chronic respiratory diseases) exacerbations [131]. According to this model, an acute stimulus, most commonly a viral infection, induces an immune-mediated inflammatory response, leading to a modification of local conditions in the airways, which may promote alteration of lung microbiota composition by favoring the predominance of specific previously underrepresented species and/or by predisposing to invasion by new potentially pathogenic strains. In turn, this deranged microbial growth may result in a further amplification of local airway inflammation, presumably via bacterial metabolite production and activation of signaling pathways mediated by the interaction between PAMPs and PRRs. Ultimately, a self-perpetuating vicious cycle arises, leading to the overwhelming inflammatory response that underlies asthma exacerbations.

Indeed, there is considerable evidence suggesting that viral infections can increase susceptibility to bacterial growth and invasion. Viral infections may impair mucociliary clearance through diverse mechanisms, including increased mucous production, ciliary dyskinesia, and cilia depletion [132,133]. Moreover, rhinoviral and possibly other viral infections have been reported to facilitate bacterial transmigration by increasing airway epithelial barrier permeability in vitro, and this effect has been attributed to the loss of occludins, a major component of tight junctions connecting adjacent epithelial cells [134,135]. Respiratory viruses alter the nasopharyngeal microbiome and may be associated with a distinct microbial signature. In the study by Rosas-Salazar et al. [136], which mostly included children <6 months of age, nasopharyngeal microbiome of infants during HRV and RSV acute respiratory tract infections (ARIs) was largely dominated by Moraxella, Streptococcus, Corynebacterium, Haemophilus, and Dolosigranulum. In addition, there was a higher abundance of Staphylococcus and a trend toward a higher abundance of Haemophilus in RSV-positive infants, and in the overall bacterial composition

between infants with HRV and RSV ARIs. Furthermore, Santee et al. [137] showed that previous history of acute sinusitis influences the composition of the nasopharyngeal microbiota, characterized by a depletion in relative abundance of specific taxa. Diminished diversity was associated with more frequent upper-respiratory infections. These, among many other published data, highlight the possibility that virus-specific compositional shifts in the nasopharyngeal microbiome contribute to worse outcomes after early-life ARIs.

Various viruses, including human rhinovirus (HRV), respiratory syncytial virus (RSV), and influenza virus, can augment the expression of adhesion molecules, such as ICAM-1, CEACAM-1, CEACAM-6, and PAFR, on the surface of epithelial cells, thus promoting invasion by certain bacterial species, including *H influenza, S pneumoniae*, and *M catarrhalis* [138–140]. Apart from the upregulation of host receptors, some viral structures appear to directly promote bacterial binding to host tissues [141,142].

Furthermore, viral infections may interfere with host immune system integrity and function and, consequently, increase susceptibility to bacterial colonization and infection. For instance, HRV infection has been shown to attenuate the inflammatory response to *H. influenzae* and delay bacterial elimination by decreasing TLR2 responsiveness to bacterial insult [143]. An almost complete disappearance of alveolar macrophages has been observed after influenza infection in a murine model [144] and HRV and RSV have been found to compromise phagocytosis and diminish pro-inflammatory cytokine release from alveolar macrophages in response to bacterial products [145,146]. Dendritic cell (DC) as well as CD4 and CD8 T-cell numbers and/or function may also be impaired after influenza or other viral infections with detrimental effects on host defense against bacteria [147]. Finally, viral respiratory infections induce changes in the lung microenvironment, such as increased temperature, nutrient availability, and cytokine and catecholamine levels, all of which are potential enhancers of bacterial virulence and immunogenicity [78,79,148]. The above mechanisms may act synergistically to elicit potentially significant alterations in the composition and diversity of lung microbiota, thus resulting through the dynamic microbiome–host immune system interplay, in the amplified and deregulated inflammatory response characterizing asthma exacerbations and even predispose to recurrent exacerbations.

Conversely, there are data, albeit rather limited, suggesting that bacteria can influence susceptibility to viral infections. In this context, several experimental studies have provided evidence for a protective role of a healthy intact microbiome against influenza. Ichinohe et al. showed that mice with impaired microbiome as a result of pretreatment with antibiotics presented attenuated CD4 and CD8 T-cell responses and reduced antibody production following respiratory influenza virus infection [149]. Likewise, increased mortality has been reported in SPF mice intranasally infected with a lethal dose of influenza virus compared with non-SPF mice, and this finding could be replicated by priming mice with *Staphylococcus aureus*, used as a surrogate of upper-respiratory-tract commensal flora, before influenza virus inoculation [150]. TLR2 signaling and an M2 alveolar macrophage phenotype were essential for the relative insensitivity against influenza-induced lethal lung injury observed in this S aureus-primed murine model. Other studies focusing on the nasopharyngeal microbiota in children with RSV bronchiolitis have highlighted the potentially deleterious effects of a deranged Proteobacteria (*H influenzae*, *M catarrhalis*)-dominated microbiome on the risk of viral infection [151,152]. RSV infection occurrence and severity and pro-inflammatory cytokine concentrations were all found to be positively correlated with nasopharyngeal colonization with the potentially pathogenic *H. influenzae, M. catarrhalis,* and *Streptococcus species.* Similarly, *S. pneumoniae* colonization was associated with increased rates of seroconversion to human metapneumovirus in infants in an older study [153]. This viral infection predisposition conferred by the coexistence of potentially pathogenic bacteria could be the result of upregulation of viral entry receptors and/or impaired antiviral immune response induced by these bacteria. Such examples are the upregulation of the major HRV entry receptor ICAM-1 by *H. influenza* in airway epithelial cells [154] or the reduction of TLR3 expression and IFN secretion in bronchial epithelial cells infected with *M catarrhalis* [155].

Collectively, these data strengthen the hypothesis that lung microbiome alterations observed in asthma patients are actively involved in the pathogenesis of disease exacerbations and may represent a potential therapeutic target in asthma exacerbations management.

#### *2.9. Therapeutic Implications of Microbiome Manipulation*

Based on the above, the proposed role of human microbiome in shaping both asthma susceptibility and phenotypes renders exogenous manipulation of microbiome a potentially attractive therapeutic strategy for both asthma prevention and treatment. Interventions in microbiome composition have previously been recommended as putatively effective approaches in three areas of asthma management: (1) prevention of asthma development during early life by favoring factors that promote immune tolerance to allergens and minimizing those that predispose to the emergence of atopy; (2) management of Th2-low asthma phenotypes/endotype, particularly neutrophilic asthma with frequent exacerbations; and (3) reverse of CS resistance or prevention of its emergence [156]. As yet, potential interventions applied for these purposes comprise lifestyle measures, vaccinations, and pharmacological treatment, including (1) probiotics, (2) prebiotics, and (3) antibiotics. It must be emphasized that, until now, only scarce data are available for the exact impact of these interventions on human microbiome composition and function, and even less is known about their presumed ability to ameliorate clinically meaningful outcomes in asthmatic patients. Furthermore, the effects of medications currently used in asthma treatment, including CS and bronchodilators, on microbiome remain mostly undetermined, and microbiome manipulations could ideally enhance beneficial actions and/or minimize side effects possibly induced by these agents.

Theoretically, any environmental exposure capable of modifying the composition of human microbiome during the dynamic period of its acquisition and maturation starting in utero and extending to the first few years of life may influence the likelihood of subsequently developing allergic asthma. This means that there is a so-called 'window of opportunity' estimated to operate within the first 100 days of life, during which lifestyle interventions, as well as exogenously administered microbial agents, could promote the formation of a tolerogenic microbiome, thus limiting the risk of allergic asthma [157]. For instance, raw cow milk consumption and vitamin D and omega-3 fatty acids supplementation during pregnancy have all been shown to diminish the risk of asthma in the offspring, presumably by affecting the microbiome synthesis of the child [158]. Avoidance of unnecessary maternal antibiotic exposure is another potentially beneficial pre-birth measure linked with microbiome [89]. As previously mentioned, vaginal delivery, breastfeeding, reasonable antibiotic use, pet ownership, exposure to 'unhygienic' traditional farming environments, and high-fiber dietary content in infancy have, more or less consistently, been associated with favorable effects on human body microbiota composition and function (especially in the gut) and a reduced risk of allergic asthma development later in life. However, it must be stressed that available evidence is generally inconclusive and probably only a few of these associations can be considered definite. Moreover, implementation of some or even all of these measures may be practically unattainable for various reasons.

To overcome these limitations, more "interventional" approaches have been proposed. Probiotics are live microorganisms (bacteria or fungi) that may favorably affect the host health, whereas prebiotics are defined as non-digestible compounds that, after processing by intestinal microbiota, promote the expansion and/or activation of beneficial commensals. Symbiotics refer to preparations containing both probiotics and prebiotics [159]. A multitude of orally administered probiotics have been shown to attenuate allergic airway inflammation in mice [160–163]. As described above, high-fiber diets and SCFAs administration (which are essentially prebiotics) have also been found to mitigate inflammatory changes in experimental models of allergic airway disease [99,100]. The potential effect of these biologic agents with microbiome-modifying capacity on allergic asthma has further been studied in several randomized placebo-controlled trials (RCTs). Daily supplementation of asthmatic school children with the probiotic *Lactobacillus gasseri* A5 or a mixture of *Lactobacillus acidophilus*, *Bifidobacterium bifidum*, and *Lactobacillus delbrueckii* for two to three months have both been shown to improve

asthma symptoms and lung function [164–168]. Similarly, a symbiotic preparation consisting of short-chain galacto-oligosaccharides, long-chain fructo-oligosaccharides, and the *Bifidobacterium breve* M-16V increased PEF and reduced serum IL-5 in an adult cohort of allergic asthma [168]. However, most [167–169], albeit not all [170], meta-analyses performed to date have failed to demonstrate a beneficial effect of preventive probiotic supplementation to mothers during pregnancy and/or infants during the first year of life on asthma risk. Further adequately powered and well-designed RCTs are warranted to clarify the role of probiotics/prebiotics on asthma prevention and treatment.

While the possible propitious effects of orally administered probiotics/prebiotics on asthma are thought to be derived by favorable alterations in the gut microbiota and the gut–lung axis, a more targeted approach involving direct interventions to lung microbiome via inhalational delivery of such agents remains relatively unexplored. In mice, intranasal administration of *Acinetobacter Iwo*ffi*i, Lactococcus lactis,* or *Staphylococcus sciuri*, all of which had previously been isolated from cowsheds in the context of farming asthma studies, have been shown to hinder eosinophilic airway inflammation induced by ovalbumin sensitization and challenge [2,171]. Similar findings have also been reported by another group for *Escherichia coli* administered via inhalation in the same murine model of allergic airway inflammation [172]. In all cases, allergy protection conferred by inhaled bacteria was associated with modifications in DC activation, leading to altered T-cell responses [172–174], in line with the previously presented observations on SCFAs administration. However, the effects of inhaled probiotic treatments on lung microbiota structure and function have not been addressed. In another experimental study assessing the potential influence of the administration route on the protective role of probiotics, *Lactobacillus paracasei* more efficiently suppressed eosinophilic inflammation when administered intranasally rather than via a feeding tube [175]. These results might suggest superior efficacy of microbiome-modifying therapeutic interventions, probiotics, and probably others (e.g., antibiotics) [176], when applied locally to the respiratory tract as compared to oral administration, which additionally may affect other organs. As yet, no human data are available for inhaled probiotic treatment in asthma.

The proposed role of human microbiome on shaping asthma phenotypes and exacerbation susceptibility, as analyzed above, may form the conceptual basis for the use of antibiotics as a means for therapeutic manipulation of the microbiome, aiming at the restoration of a symbiotic state between the host and the resident microbiota through the suppression of overgrown detrimental taxa and the foster of beneficial ones. This holds especially true for the Th2-low, CS-insensitive, exacerbation prone, severe neutrophilic asthma phenotype, the treatment options for which are particularly limited. As Proteobacteria-dominated changes in lung microbiome have quite consistently been identified in relation with this phenotype, with *Haemophilus influenzae* and *Moraxella catarrhalis* being the most prominent species involved, these taxa could probably be considered the primary targets of microbiome-modifying antibiotic treatment in neutrophilic asthma. The anticipated gradual advent of non-culture-based methods for lung microbiota identification in clinical practice may allow more personalized approaches in antibiotic targeting of dysbiotic lung bacteria in the future.

Both in asthma and in a growing list of other respiratory diseases, macrolides have become the subject of much attention due to the fact that they combine well-known antimicrobial activity along with multiform immunomodulatory properties, including attenuation of neutrophil chemotaxis, inhibition of biofilm formation, reduction of mucus hypersecretion, and even downregulation of viral entry receptors with amplification of viral infection-induced IFN production [177–179]. Overall, clinical trials of macrolides in asthma, most of which in patients with severe uncontrolled asthma, have yielded conflicting results. However, in a pre-specified subgroup analysis of the AZISAST multicenter randomized placebo-controlled trial, it was shown that a six-month course of azithromycin, given at a dose of 250 mg, three times a week, in patients with severe asthma and frequent exacerbation, significantly reduced a composite primary outcome comprising the rates of severe exacerbations and lower respiratory-tract infections requiring antibiotics in the non-eosinophilic asthma subgroup of patients [180]. More recently, the larger AMAZES randomized controlled trial (RCT) reported

significantly decreased rates of exacerbations, along with improved asthma-related quality of life in patients with uncontrolled asthma despite therapy with medium-to-high doses of ICS plus LABA who received a higher dose of azithromycin (500 mg three times per week) for 12 months [181]. Interestingly, in this trial, azithromycin appeared equally effective in all studied subgroups, including both those of eosinophilic and non-eosinophilic asthma. Although these beneficial effects of azithromycin (and perhaps other macrolides) on asthma exacerbations could be attributed to the aforementioned immunomodulatory anti-inflammatory properties of macrolides, there are data showing alterations of microbiota composition following treatment with azithromycin. In a research letter published in 2014, Slater et al. described the longitudinal effects of six weeks of daily therapy with 250 mg azithromycin on lung microbiota characteristics, using bronchoscopic washings for sampling and DNA pyrosequencing analysis in five patients with moderate-to-severe asthma [182]. The study reported a reduction of lung microbiota diversity post-treatment, accompanied by an increased relative abundance of *Anaerococcus* and a decreased recognition of the potentially pathogenic genera of *Pseudomonas*, *Staphylococcus,* and *Haemophilus*. Furthermore, in a bacteriological sub-study of the AZISAST trial, azithromycin administration was associated with post-treatment changes in the microbiota composition of the oropharynx compared with both pretreatment status and control patients receiving placebo [183]. At the phylum level, Firmicutes increased and Fusobacteria decreased in oropharyngeal swabs derived from eight severe asthma patients after a six-month course of azithromycin. This finding corresponded to an increased abundance of *Streptococcus salivarius*, with a parallel reduction of *Leptotrichia wadei* at the species level. Of notice, in half of the washout samples collected one month following completion of azithromycin treatment, the microbiota composition was almost identical to the pretreatment oropharyngeal flora of the same patient, indicating a possibility for original microbiome recovery after cessation of antibiotics administration and a consequent requirement for long-term treatment.

Admittedly, there are considerable limitations in the long-term use of antibiotics as a therapeutic tool for favorable manipulations of the host microbiome. Apart from well-founded concerns with respect to potential adverse effects (sometimes irreversible or life threatening) and emergence of antimicrobial resistance arising from long-term antibiotic regimens, the lack of specificity in the elimination of resident bacteria does not allow currently available antimicrobial agents to eradicate exclusively dysbiotic members of the lung bacterial community while sparing beneficial commensals, and may, in fact, produce the opposite results. As previously noted, maternal and early life antibiotic usage has been associated with an increased risk of asthma development during childhood. The detrimental effects of antibiotic use (and not necessarily long-term) on gut microbiome are firmly established and involve the depletion of normal biodiversity and the overgrowth of potentially harmful bacterial strains, leading to dysbiosis. Occasionally, this can be expressed clinically in the form of antibiotic-associated diarrhea and pseudomembranous colitis. Such antibiotics-associated hazards (i.e., selection of potentially pathogenic microbiota) are also pertinent to the lung microbiome, and broad-spectrum antibiotic use may lead from a state of respiratory dysbiosis to another, each time dominated by different and probably more-resistant strains. Despite potentially abolishing systemic adverse effects, inhaled antibiotics are probably subject to these same limitations and cannot be considered as an ideal alternative to systemically administered agents.

Novel approaches to microbiome-modifying therapies have focused on the manipulation of metabolites known to be produced or processed by the (gut) microbiota and to act on the host [184,185]. Metabolite-targeting molecules are under development. Fecal transplantation has successfully been used as a means of gut microbiota manipulation in several intestinal disorders [186]. Its potential role in allergic and respiratory diseases remains unexplored.

The endeavor to therapeutically manipulate human microbiome is clearly still in its infancy, and, by all means, intensive innovative research applying to all stages of the treatment discovery and development pipeline is mandatory before any of these interventions could be incorporated into clinical practice guideline recommendations.

#### **3. Conclusions**

Studies conducted over the last two decades have dramatically changed our perspective on LRT microbiology, with the use of culture-independent molecular techniques. It is now universally accepted that the lung is far from a sterile organ. It has convincingly been shown that, in asthma, lung microbiome undergoes significant alterations in terms of diversity and composition, with certain species outgrowing others, functionally leading to a presumable state of respiratory dysbiosis. Many of these alterations have been associated with specific phenotypic features of asthma, including disease exacerbations, and aspects of this dysbiosis may represent the missing link in the pathogenesis of some expressions of the asthmatic disease, perhaps, in particular, the neutrophilic inflammatory phenotype. A considerably large body of evidence arising from both experimental and epidemiological studies is now available, suggesting that early life environmental exposures affecting the gut microbiome structure may be involved in shaping susceptibility for asthma development later in life by modifying microbiota-derived factors thought to be actively involved in the configuration of the gut–lung axis, such as the SCFAs. Given its proposed roles in influencing the risk for asthma development, as well as the phenotypic expression of an established disease, microbiome has emerged as a potential therapeutic target in asthma. As yet, interventions potentially modifying the microbiome have not clearly been shown to be efficacious in preventing and/or treating asthma, with few exceptions.

More accurate delineation of lung microbiota structural and functional characteristics, both in health and in asthma, remains an unmet need in the microbiome research. Larger longitudinal studies applying standardized sampling methods in well-characterized asthmatic cohorts are indispensable in this regard. Unquestionably, the impending widespread application of the novel "-omics" technologies can be expected to provide invaluable insights into the functional effects of lung microbiome and their presumable role in shaping asthma predisposition and phenotypes. Advancing our understanding on the mechanistic links between human microbiome and asthma will hopefully culminate in the discovery of novel therapeutic interventions targeting specific aspects of respiratory dysbiosis.

**Author Contributions:** Conceptualization, N.R.; literature search and data extraction, K.L., G.B., L.K., and N.R.; writing—original draft preparation, K.L., G.B., and L.K; writing, review, and editing, E.G., M.P., I.V., and N.R.; supervision, A.K., N.K., and N.R.

**Acknowledgments:** In this section you can acknowledge any support given which is not covered by the author contribution or funding sections. This may include administrative and technical support, or donations in kind (e.g., materials used for experiments).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Targeting NLRP3 Inflammasome Activation in Severe Asthma**

**Efthymia Theofani 1, Maria Semitekolou 1, Ioannis Morianos 1, Konstantinos Samitas <sup>2</sup> and Georgina Xanthou 1,\***


Received: 31 July 2019; Accepted: 26 September 2019; Published: 4 October 2019

**Abstract:** Severe asthma (SA) is a chronic lung disease characterized by recurring symptoms of reversible airflow obstruction, airway hyper-responsiveness (AHR), and inflammation that is resistant to currently employed treatments. The nucleotide-binding oligomerization domain-like Receptor Family Pyrin Domain Containing 3 (NLRP3) inflammasome is an intracellular sensor that detects microbial motifs and endogenous danger signals and represents a key component of innate immune responses in the airways. Assembly of the NLRP3 inflammasome leads to caspase 1-dependent release of the pro-inflammatory cytokines IL-1β and IL-18 as well as pyroptosis. Accumulating evidence proposes that NLRP3 activation is critically involved in asthma pathogenesis. In fact, although NLRP3 facilitates the clearance of pathogens in the airways, persistent NLRP3 activation by inhaled irritants and/or innocuous environmental allergens can lead to overt pulmonary inflammation and exacerbation of asthma manifestations. Notably, administration of NLRP3 inhibitors in asthma models restrains AHR and pulmonary inflammation. Here, we provide an overview of the pathophysiology of SA, present molecular mechanisms underlying aberrant inflammatory responses in the airways, summarize recent studies pertinent to the biology and functions of NLRP3, and discuss the role of NLRP3 in the pathogenesis of asthma. Finally, we contemplate the potential of targeting NLRP3 as a novel therapeutic approach for the management of SA.

**Keywords:** severe asthma; innate immunity; immune regulation; NLRP3; IL-1β; allergic airway inflammation

#### **1. Introduction**

Asthma represents a serious global health problem that affects 1%–18% of the population of all age groups. Its prevalence has increased in the last decades, especially among children [1]. Asthma is characterized by variable symptoms of wheezing, dyspnea, chest tightness, coughing, and reversible airflow obstruction, and is usually associated with airway hyperresponsiveness (AHR) to innocuous environmental allergens and chronic airway inflammation. Factors, such as allergen or irritant exposure, respiratory infections, exercise, climate changes, and stress, are responsible for the disparities and severity of asthma symptoms [1]. Asthma has been long considered as a heterogeneous chronic lung disease that encompasses multiple groups of patients characterized by varying features or phenotypes [2,3]. A small percentage of asthmatics exhibit severe disease exacerbations despite the fact that they are already under treatment with high doses of inhaled and/or systemic corticosteroids [2,3]. These patients suffer from severe asthma (SA) that is poorly controlled and, in some cases, life-threatening [4,5]. Although patients with SA comprise a small percentage of the total asthma population (5%–10%), they denote 50% of total healthcare costs, rendering SA a substantial health and socioeconomic burden [6,7]. SA is characterized by marked thickening and

structural changes of the airway wall, excessive airway narrowing, and fixed airflow obstruction [6,7]. An in-depth understanding of the heterogeneity of SA and the immunological mechanisms underlying its pathophysiology is critical for the identification of novel biomarkers and molecular pathways that can be targeted in novel treatment modalities.

In the lung, innate immune responses provide the first line of defense against environmental signals, including pathogens, allergens, and other irritants, and act through downstream signaling by numerous extracellular and intracellular receptors, termed pattern recognition receptors (PRRs) [8–10]. NOD-like Receptor Family Pyrin Domain Containing 3 (NLRP3) is an intracellular PRR that detects microbial motifs, endogenous danger signals, and environmental irritants, and induces the formation and activation of the NLRP3 inflammasome. Although the NLRP3 inflammasome is essential for providing protective immunity, overactivation of inflammasome-mediated responses can cause excessive inflammation, tissue damage, and lead to chronic inflammatory diseases, including asthma [10,11]. In this review, we describe the immunological mechanisms underlying aberrant inflammatory responses in the airways and their link to SA pathogenesis. We also present the biology and functions of NLRP3 and discuss its role in the initiation and propagation of SA features. Finally, we present recent findings pertinent to targeting NLRP3 functions as a novel therapeutic approach for the control of inflammatory responses in the airways.

#### **2. Severe Asthma Pathogenesis**

#### *2.1. Type 2 Asthma*

To address SA complexity, the concept of asthma endotyping has emerged [12–14]. Depending on the type of immune cell responses implicated in disease pathogenesis, asthma endotypes are categorized as (a) type 2 asthma, characterized predominantly by T helper type 2 (Th2) cell-mediated inflammation and (b) nontype 2 asthma, associated with Th1 and/or Th17 cell inflammation [15–17].

Upon allergen exposure, dendritic cells (DCs) in the lung mucosa take up allergens, reach the mediastinal lymph nodes, and present allergen components to naive T cells in the context of major histocompatibility complex class II [18]. Allergens with proteolytic activity, such as those derived from house dust mites (HDM), pollen grains, fungi, and occupational sensitizers activate protease activated receptors expressed on DCs, disrupt epithelial tight junctions, and initiate inflammatory responses [18]. Moreover, certain allergens and airborne particulates contain microbial components which interact with PRRs, including Toll-like receptors (TLRs) and NOD-like receptors (NLRs), on DCs and airway epithelial cells, and serve as "danger signals" for the host immune response [18]. Upon interaction with allergen-loaded DCs, naive Th cells differentiate into Th1, Th2, Th9, or Th17 cells, depending on the type and dose of allergen and the local cytokine milieu [18]. Allergen-specific Th2 cells, generated in the presence of type 2 cytokines, migrate into the airways wherein upon allergen re-exposure, secrete cytokines and promote mucus secretion, subepithelial fibrosis, bronchial remodeling, and AHR [19]. The production of Th2 cytokines also leads to the recruitment of innate effector cells, including mast cells, basophils, and eosinophils, as well as to isotype switching of B cell-secreted IgG to allergen-specific IgE [19]. Additionally, Th9 cells, a recently identified Th cell subset characterized by high levels of IL-9, exacerbate allergic airway inflammation (AAI), predominantly through activation of mast cell functions [20,21]. More specifically, experimental studies have shown that IL-9 production by Th9 cells and by type 2 innate lymphoid cells (ILC2s) enhances the production of IL-2 by mast cells, leading to further expansion of ILC2s, which activate Th9 cells, in a positive feedback loop [22]. Of clinical relevance, increased numbers of Th9 cells were demonstrated in peripheral blood mononuclear cells (PBMCs) isolated from HDM or pollen allergic subjects and correlated with IgE levels [23]. Moreover, elevated IL-9-secreting T lymphocytes were observed in the bronchoalveolar lavage (BAL) of asthmatics [24].

Apart from DCs, the asthmatic airway epithelium represents a major source of cytokines termed "alarmins", such as, IL-25, IL-33, and thymic stromal lymphopoietin (TSLP), and chemokines, including RANTES (Regulated on Activation, Normal T cell Expressed and Secrete or CCL5), TARC (Thymusand Activation-Regulated Chemokine or CCL17), eotaxins (CCL11, CCL24 and CCL26), and MCP-3 (Monocyte Chemotactic Protein-3 or CCL7) that trigger Th2 cell polarization upon exposure to allergens, pollutants, viral, fungal, and bacterial components [18]. As mentioned above, recent studies have highlighted a key role for ILC2s in asthma immunopathogenesis [25]. ILC2s are activated in response to TSLP, IL-25, and IL-33 signaling [25], and produce IL-5, IL-13, and prostaglandin [26], further propagating Th2-cell mediated responses in the airways.

Type 2 asthma is characterized by any combination of the following processes: eosinophilia in the sputum or blood, atopy and a high level of fractional exhaled nitric oxide (FeNO) [16]. Several biomarkers of type 2 inflammation, such as FeNO, serum IgE, blood or sputum eosinophils, and serum periostin distinguish type 2-high and type 2-low asthma phenotypes and also predict the responsiveness to type 2 cytokine-targeted therapy [27]. Eosinophils play a crucial role in the initiation and propagation of inflammatory responses in asthma [16]. Asthmatic patients with increased numbers of eosinophils in the periphery suffer from more severe disease exacerbations and poorer disease control [16]. FeNO represents an indicator of IL-13-mediated and corticosteroid-responsive airway inflammation, as the presence of IL-13 activates inducible nitric oxide synthase (iNOS), leading to increased FeNO production in the airways [16]. Approximately 70% of patients with asthma have an allergic phenotype, characterized by allergen-specific IgE and elevated total IgE levels [16]. Periostin, an extracellular matrix protein, induced by IL-4 and IL-13 signaling, is secreted by bronchial epithelial cells and represents another important biomarker for severe eosinophilic type 2 asthma [28]. Periostin is involved in airway remodeling, sub-epithelial fibrosis, eosinophil recruitment, and mucus production. Notably, a high serum concentration of periostin denotes one of the most significant indicators of eosinophilic inflammation in asthma [29].

Clustering studies revealed that one of the major reasons SA patients remain unresponsive to corticosteroid treatment is that apart from Th2 inflammation, other mediators are also implicated in disease pathogenesis [12,30]. Indeed, enhanced Interferon gamma *IFNG* expression was detected in BAL cells, accompanied by increased secretion in SA patients compared to mild-to-moderate asthmatics (MMA) [31]. The same study also showed elevated percentages of CD4+IFN-γ<sup>+</sup> T cells in the BAL [31]. In line with above, increased IFN-γ mRNA levels were observed in lung tissue and sputum of SA patients [32,33]. It must be asked what triggers these IFN-γ-mediated responses in the airways of SA patients? It is known that persistent viral (especially rhinoviruses) and bacterial (*Chlamydia pneumoniae*, *Streptococcus pneumoniae*, *Mycoplasma pneumoniae*, *Haemophilus influenzae*, *Moraxella catarrhalis* and *Staphylococcus aureus*) infections augment IFN-γ secretion in SA patients, highlighting a key role for respiratory tract infections in disease severity and asthma exacerbations [34].

#### *2.2. Nontype 2 Asthma*

The pathophysiology of nontype 2 asthma remains less well characterized. Nontype 2 asthma is denoted by the absence of type 2 biomarkers, such as eosinophils and IgE, and the predominance of Th17 cells and neutrophils in the airways [35]. In fact, growing evidence has highlighted a key role for IL-17 in the pathogenesis of nontype 2 asthma [35]. Tissue-infiltrating CD4<sup>+</sup> T cells isolated from bronchial biopsies of SA patients produce copious amounts of IL-17 and IL-22 upon ex vivo polyclonal stimulation [36,37]. Another study revealed that in vitro administration of recombinant human IL-17 enhanced the secretion of IL-8 by human bronchial epithelial cells (HBECs) and venous endothelial cells [38]. Moreover, conditioned medium from IL-17-treated HBECs increased neutrophil migration in vitro [38]. Increased IL-17 mRNA levels are also observed in the sputum of SA patients [39]. In addition, IL-17 levels in the peripheral blood (PB) of SA patients positively correlate with disease severity [40]. A very interesting recent study revealed that an IL-4Rα polymorphism associated with SA leads to the conversion of regulatory T cells (Tregs) to Th17 cells in vitro [41]. In brief, the authors isolated from the PB of asthmatics and healthy controls that were either wild type (WT) (IL4RQ576/Q576), heterozygous (IL4RQ576/R576), or homozygous (IL4RR576/R576) for the mutated allele, naive T cells, and differentiated them into Treg cells. Interestingly, naive CD4<sup>+</sup> T cells from

asthmatics bearing the IL4RR576 mutation exhibited defective induction of Treg cells and were skewed towards a Th17-like phenotype, exemplified by increased secretion of IL-17 [41]. Still, targeting IL-17 has failed to improve disease symptoms in SA patients as opposed to anti-type 2 cytokine therapy, suggesting that other mechanisms compensate for its pathogenic effects or that targeting pathogenic Th17 cells specifically would be more appropriate [42,43].

To date, biomarkers of type 2-low or neutrophilic asthma have not been identified. Several studies have shown that there is no inverse correlation between sputum eosinophil and neutrophil numbers in SA patients and that eosinophils are present in excess, additionally to neutrophilic accumulation, in the airways [44]. Moreover, although measuring eosinophil numbers in the periphery correlates with the percentages of eosinophils in induced sputum, blood neutrophil parameters represent poor surrogates for the proportion of neutrophils in the sputum [45,46]. Recent studies showed that the levels of the chitinase-like protein YKL-40 in the blood of SA patients could potentially be used as a marker for nontype 2 neutrophilic asthma [47]. In fact, combining the measurement of YKL-40 with other clinical parameters of the disease may provide a more reliable strategy for defining nontype 2 asthma. Tumor necrosis factor (TNF)-α has been also shown to have a critical role in nontype 2 asthma through acting on smooth muscle cells or by modifying the release of the cysteinyl leukotrienes, LTC4 and LTD4 [48]. TNF-α is increased in the BAL and bronchial biopsies in SA patients compared to MMAs [49]. Importantly, the administration of inhaled recombinant TNF-α to normal subjects led to the development of AHR and airway neutrophilia [50,51]. Subsequent clinical trials utilizing antiTNF-α therapeutic administration contributed to the elucidation of the role of this cytokine in vivo [52]. Indeed, improvement in patient quality of life, lung function, and a reduction in AHR and exacerbation frequency was observed in patients treated with antiTNF-α [52]. Still, considering the significant heterogeneity observed in SA patients, benefit from antiTNF-α therapy is likely to occur only in a small group of patients.

Apart from type 2 and nontype 2, asthma can be categorized into four endotypes (eosinophilic, neutrophilic, mixed granulocytic, and paucigranulocytic) based on the type of airway infiltrating immune cells [53]. Among these endotypes, paucigranulocytic asthma (PGA) presents no evidence of increased eosinophils or neutrophils in induced sputum, and instead is characterized by low-grade airway inflammation associated with airway smooth muscle (ASM) dysfunction, persistent airflow limitation, and AHR [54,55]. Molecules involved in oxidative stress, matrix metalloproteinases, neutrophil elastase, and galectin-3, commonly used for the discrimination between eosinophilic and neutrophlic asthma, also remain unaltered in patients with PGA [56–59]. Moreover, patients with PGA display lower FeNO levels compared to those with eosinophilic asthma [60]. Notably, Wang and colleagues documented that PGA was the most common endotype observed in children with stable asthma [61]. Corticosteroids do not seem to exert beneficial effects in this cohort of asthmatics, irrespective of the dose used [62]. Considering that symptoms dominating in this cohort are mostly due to ASM phenotypic changes and/or neuronal dysfunction, therapies directed toward ASM responses, including mast cell-targeted therapies and/or bronchial thermoplasty, might benefit PGA patients [63]. Although the precise mechanisms of action remain incompletely understood, bronchial thermoplasty is considered to diminish ASM mass through the delivery of localized thermal energy [63]. In addition, factors associated with dysregulated ASM functions, as well as mediators involved in the thickening of the subepithelial basement membrane, could be used as disease biomarkers and guide the design of effective therapeutic approaches for PGA [56].

In sharp contrast to adult asthma, wherein the type of airway inflammation has been extensively investigated due to the use of invasive and semi-invasive techniques, such as bronchial biopsies, BAL, and induced sputum, the majority of analyses in childhood asthma have been performed only in severe forms of disease [64–67]. The most common endotype of childhood asthma is the eosinophilic, characterized by unrestrained symptoms, atopy, impaired lung function, enhanced AHR, increased numbers of exacerbations, and steroid responsiveness [68]. The percentages of eosinophils in induced sputum significantly vary over time in children with SA, and these variations are not associated with asthma therapy [69]. Importantly, persistent airway eosinophilia has been detected in

a small cohort of children with SA, even after a high dose of systemic corticosteroids [70]. Regarding neutrophilic pediatric asthma, the functional role of airway neutrophils in asthma pathophysiology remains elusive. Infiltration of neutrophils in the airways in pediatric SA represents a feature of airway inflammation at all ages and is triggered by viral and/or bacterial infections, exacerbating asthma symptoms [71]. Increased numbers of intraepithelial neutrophils, along with enhanced epithelial and submucosal expression of IL-17R, have been observed in lung biopsies of a subgroup of children with therapy-resistant SA. These findings correlate with improved lung function suggesting a potential beneficial rather than adverse role for neutrophils in pediatric SA pathophysiology [72]. Other studies show that neutrophils coexist with eosinophils in the airways of a group of children with SA, highlighting the complexity of defining the distinct endotypes in children [73]. Overall, it becomes evident that a more detailed and careful assessment of all the inflammatory endotypes is essential for the characterization and management of severe pediatric asthma.

#### **3. Targeted Therapies For Severe Asthma**

In-depth investigation of the molecular and cellular mechanisms underlying SA pathophysiology has significantly contributed to the development of novel therapeutic strategies for disease management. Indeed, antibodies targeting factors involved in SA pathology are already being used as a first line medication. An excellent example is omalizumab, a monoclonal antibody directed against IgE that has become an established add-on therapy for patients with uncontrolled allergic asthma [74]. In addition, monoclonal antibodies against IL-5 (reslizumab, mepolizumab), IL-5R (benralizumab), and IL-4R (dupilumab) have been approved as add-on treatments for uncontrolled type 2 eosinophilic asthma [74]. Although these therapies have proven effective in certain asthma cohorts, some patients that suffer from severe allergic and/or eosinophilic asthma, as well as most patients with severe non type 2 asthma, experience weakly-regulated disease manifestations [74]. Notably, the reported adverse effects of these monoclonal antibody therapies should be also considered. For example, administration of omalizumab has been associated with an anaphylaxis rate of 0.09% that most frequently occurs within 2 hours after the first dose and 30 min after subsequent doses, necessitating the close monitoring of patients [75,76]. Moreover, a higher incidence of cardiovascular and cerebrovascular adverse effects (AEs) has been observed upon omalizumab administration [77]. Mepolizumab, the first anti-IL-5 monoclonal antibody approved for eosinophilc asthma, has been associated with headaches, injection site reactions, back pain, and fatigue [78]. The most common AEs of reslizumab, another FDA-approved anti-IL-5 antibody, are a 0.3% anaphylaxis rate, elevated serum creatinine kinase, and musculoskeletal and oropharyngeal pain [79]. Regarding benralizumab, an anti-IL-5R antibody that is currently undergoing phase 3 trials, there have been no documented AEs apart from nasopharyngitis and injection site reactions [80]. AEs in patients receiving dupilumab, a monoclonal antibody that targets the common receptor for IL-4 and IL-13, include nasopharyngitis, injection site reactions, and headaches [81]. Monoclonal antibodies targeting TSLP, IL-33 and its receptor ST2, the receptor for stem cell factors on mast cells, and a DNA enzyme directed at GATA3 are currently being evaluated for their efficacy in SA. It is worth mentioning that a number of antagonists directed against other potential targets, such as, IL-25, IL-6, TNF-like ligand 1A, CD6, and activated cell adhesion molecules are under consideration for future clinical trials [74]. Results from these clinical trials will be of great importance as they may introduce novel treatment modalities that will successfully replace the existing ones and lead to the efficient management of SA.

Taken together, it becomes evident that as the airway lumen is continually exposed to external and endogenous stimuli, its ability to distinguish between innocuous environmental allergens and pathogenic agents is crucial for the maintenance of immune tolerance and lung homeostasis. In fact, the complex interactions between innate and adaptive immune responses in the lung micromilieu represent a major determinant of the development of tolerance or allergic inflammation. Innate immune responses have considerable bearing on ensuing adaptive responses, and if left uncontrolled, can lead to detrimental pathological consequences. Hence, delineation of the precise mechanisms involved in

the regulation of innate immune reponses in the airways is essential for the design of more efficient treatment modalities for SA patients.

#### **4. Inflammasomes: A Key Component of Innate Immunity**

The innate immune system acts as the first line of defense during exposure to environmental pathogens. In the lung, innate immune responses act through downstream signaling by numerous extracellular and intracellular receptors termed pattern recognition receptors (PRRs). PRRs recognize pathogen-associated molecular patterns (PAMPs), such as lipopolysaccharide (LPS), bacterial and viral RNA, danger-associated molecular patterns (DAMPs) in damaged and/or dying cells, including reactive oxygen species (ROS), ATP and mitochondrial DNA, and homeostasis-altering molecular processes (HAMPs) that detect alterations in cell homeostasis and elicit inflammatory responses in the host [8,9]. PRRs are expressed on macrophages, monocytes, DCs, and on tissue-resident cells, including airway epithelial cells, and upon ligand binding, induce the secretion of inflammatory cytokines and chemokines [82,83]. PRRs consist of the Toll-like receptors (TLRs), the RIG-I-like receptors (RLRs), the nucleotide-binding oligomerization domain-like receptors (NLRs), the Scavenger receptors, the C-type lectin receptors (CLRs), and the absent-in-melanoma (AIM)-like receptors (ALRs). Among PPRs, TLRs and NLRs represent the most well-known and studied receptors [83,84]. TLRs are located at the cell membrane and the intracellular compartment, while NLRs are located solely in the cytosol [84]. Signaling through these receptors enables the innate immune system to monitor and respond to infectious agents and other damage-inducing stimuli, eliciting protective immunity.

NLRs consist of a conserved nucleotide binding and oligomerization domain (NACHT), a carboxy-terminal ligand-binding region, composed of leucine-rich repeats (LRRs), involved in ligand binding or activator sensing, and an amino-terminal effector domain required for protein–protein interactions [85–87]. The human NLR gene family consists of 22 members, classified into four subfamilies depending on their N-terminal regions: NLRA, NLRB, NLRC, and NLRP. The NLRA region contains an acidic transactivation domain, the NLRB a baculoviral inhibitory repeat-like domain, NLRC contains a caspase activation and recruitment domain (CARD), and NLRP contains a pyrin domain (PYD) [85–87]. The detection of PAMPs, DAMPs, and HAMPs by NLRs leads to the formation of a large multimolecular signaling platform called the inflammasome. Inflammasomes respond to a constellation of endogenous and pathogenic signals and are critical inducers of host defense. Five major inflammasomes have been identified so far: NLRP1, NLRC4, RIG-I, AIM2, and NLRP3. These consist of an active NLRP receptor, the inflammasome adaptor protein, Apoptosis-associated Speck-like protein Containing CARD (ASC), and caspase-1 [88]. NLRP6, NLRP7, NLRP12, and IFI16, can also form inflammasomes, but their composition remains unclear. Apart from the physiological role of inflammasomes in providing protective immunity, inflammasomes also regulate cell proliferation and tissue repair processes [11]. Still, overactivation of inflammasome-mediated responses can cause excessive inflammation, tissue damage, and lead to chronic inflammatory diseases and metabolic disorders.

#### **5. Nlrp3 Biology and Functions**

NLRP3 is expressed in granulocytes, monocytes, DCs, T cells, and epithelial cells [89]. The NLPR3 inflammasome consists of the NLRP3 receptor, the adaptor protein ASC, also known as Pycard, and caspase-1 that acts as an effector protein [89]. The NLRP3 receptor is a tripartite protein that contains an amino-terminal PYD, a nucleotide-binding NACHT, and a carboxy-terminal LRR domain. ASC has two domains, an amino-terminal PYD and a carboxy-terminal CARD domain. NLRP3 activation leads to protein–protein interactions between the NLRP3 and ASC via PYD domains. This facilitates ASC polymerization to form long helical filaments that are condensed into an intracellular macromolecular aggregate, known as ASC speck [90]. Subsequently, the ASC CARD domain associates with the CARD domain of caspase-1, inducing caspase-1 self-cleavage and activation [91]. A serine–threonine kinase known as NIMA-related kinase 7 (NEK 7) binds to NLRP3 directly and oligomerizes with NLRP3 into a complex that is essential for ASC speck formation and caspase-1 activation [92]. Activated caspase-1 cleaves the inactive pro-IL-1β and pro-IL-18 forms into bioactive cytokines that activate downstream inflammatory pathways [93] (Figure 1). Recent studies have shown that ASC specks can be exocytosed and accumulated in the extracellular space, retaining their ability to produce IL-1β. Extracellular ASC specks can also be internalized by macrophages, further activating IL-1β production [94,95]. Moreover, ASC specks isolated from cells can induce the aggregation of other ASC specks located intracellularly, a feature shared with prionoid proteins. These findings suggest that extracellular ASC can propagate inflammatory responses even at distinct sites, promoting systemic inflammation. Of note, increased extracellular ASC specks were documented in bone marrow from patients with myelodysplastic syndrome [96]. Pertinent to lung diseases, increased extracellularly-assembled ASC specks were observed in the BAL of patients with chronic obstructive pulmonary disease and pneumonia compared to patients with pulmonary hypertension and healthy controls [94]. Considering the documented effects of extracellular ASC specks on the activation of distant cell populations, their presence in SA patients may have important implications, and warrants further investigation.

**Figure 1. Mechanisms involved in activation and regulation of NLRP3 canonical pathway.** NLRP3 inflammasome activation requires two signals. The "priming" signal is triggered by PAMP/DAMP recognition by PPRs (e.g. TLRs) and certain cytokines (e.g. TNF-α, IL-1β) and activates NF-κB in the cell nucleus. This leads to NLRP3, pro-IL1-β and pro-IL-18 gene transcription. The second signal induces the assembly of NLRP3, ASC, and caspase-1 to form an active NLRP3 inflammasome and ultimately leads to the release of mature IL-1β and IL-18. Gasdermin D is cleaved and becomes inserted into the cell membrane, forming pores and inducing pyroptosis. The mechanisms proposed for the second NLRP3 activating signals are shown and include: a) changes in cytosolic levels of ions, such as K<sup>+</sup>, Cl- and Ca<sup>+</sup>2, b) lysosomal destabilization and the release of cathepsins, c) mitochondrial dysfunction-derived signals such as mtROS, mtDNA and d) metabolic changes. PtdIns4P on dTGN drive NLRP3 activation. Aerobic glycolysis pathways and the TCA cycle also activate NLRP3. Autophagy and mitophagy inhibit NLRP3 inflammasome activation. IFNs also inhibit NLRP3 activation through NO production. IL-1R, IL-1β receptor; TLR, Toll-like receptor; TNFR, tumor necrosis factor receptor; IFNAR, IFNα/β receptor; NEK7, NIMA- related kinase 7; NF- κB, nuclear factor- κB; P2X7, P2X purinoceptor 7; PtdIns4P, phosphatidylinositol-4-phosphate; PYD, pyrin domain; ROS, reactive oxygen species; HK1, hexokinase; mTORC1, rapamycin complex 1; SDH, succinate dehydrogenase; EIF2AK2, eukaryotic translation initiation factor 2-alpha kinase 2.

IL-1β secretion upon NLRP3 inflammasome activation initiates acute phase reactions, including the recruitment of inflammatory cells at the site of infection and expression of proinflammatory cytokines, such as, IL-6, TNF-α, and chemokines [97]. Briefly, active IL-1β binds to the extracellular domain of the IL-1 type 1 receptor (IL-1R) that recruits the second receptor chain, termed IL-1R accessory protein (IL-1RAcP) [98]. This leads to the activation of intracellular signaling molecules, such as the myeloid differentiation primary response 88 (MYD88), TNF receptor-associated factor 6 (TRAF6) and IL-1R-associated kinases (IRAK), which, in turn, activate the nuclear factor-κB (NF-κB) transcription factor, eliciting cytokine gene expression [99]. Active IL-1β exerts its functions through both autocrine and paracrine mechanisms, and therefore its regulation is under tight control to circumvent pathologic implications. IL-1β signaling is also involved in Th2 and Th17 cell differentiation and is implicated in the pathogenicity of allergic airway inflammation [100]. The biological role of IL-18 is different from that of IL-1β. IL-18 induces Th1 cell responses and the production of IFN-γ by CD4<sup>+</sup> and CD8<sup>+</sup> T cells, natural killer cells, and macrophages, and is essential for antiviral immunity. IL-18 also plays a role in Th2 cell differentiation and is involved in AHR in experimental models, as well as in asthmatic patients [101,102].

Apart from proinflammatory cytokine release, a key outcome of NLRP3 inflammasome activation is pyroptosis, a form of lytic cell death characterized by cell swelling, membrane rupture, and release of proinflammatory cellular contents [103,104]. The formation of plasma membrane pores during pyroptosis drives ion changes, inducing increased osmotic pressure, water influx, and cell swelling. Pyroptosis is triggered by caspase-1-driven cleavage of the pore-forming protein gasdermin D (GSDMD) that leads to the formation of the GSDM N-terminal fragment, that, in turn, introduces pore formation upon insertion into the plasma membrane, thus killing cells from within [103,104]. Pyroptosis is a highly inflammatory process that is accompanied by the release of IL-1β, IL-18, IL-1α, high mobility group box 1 protein, and lactate dehydrogenase to the extracellular milieu [103,105] (Figure 1). NLRP3 inflammasome activation occurs through three distinct molecular pathways: the canonical, the noncanonical, and the alternative pathways [11,86,89,92,105,106] (Table 1).


**Table 1.** Key characteristics of canonical, noncanonical and alternative NLRP3 activation.

#### **6. Activation of the Canonical Nlrp3 Pathway**

Activation of the NLRP3 inflammasome through the canonical pathway requires 2 steps: a priming signal and a second activating signal (Figure 1). The priming signal is crucial for the transcription of the inflammasome components, *NLRP3*, *CASP1*, and the *IL1B* and *IL18* genes. The priming signal is usually a PAMP, such as, LPS, lipoproteins, carbohydrates, and flagellin [107]. IL-1β and TNF-α signaling can also induce NLRP3 priming, known as sterile priming. Gene transcription upon priming is mediated predominantly through NF-κB activation and nuclear translocation [108,109]. Interestingly, recent studies have discovered a nontranscriptional priming process that relies on post-translational modifications (PTMs), such as ubiquitylation, phosphorylation, and sumoylation of NLRP3 components [110].

A wide variety of secondary stimuli activate the NLRP3 inflammasome, including bacteria, viruses, environmental nanoparticles such as alum and silica, and endogenous molecules, including ATP, monosodium urate (MSU), and cholesterol crystals. The main mechanisms through which these

PAMPs and DAMPs trigger NLRP3 inflammasome activation are associated with: a) changes in cytosolic levels of ions, such as, K+, Ca<sup>+</sup>2, and Cl- , b) lysosomal destabilization, c) ROS production, and d) mitochondrial dysfunction [92,111]. The second activation signal leads to the assembly of the NLRP3 complex, the activation of caspase-1, and the release of the mature forms of IL-1β and IL-18 (Figure 1).

Decreased extracellular K<sup>+</sup> levels trigger NLRP3 activation, while high extracellular K<sup>+</sup> concentrations block NLRP3 signaling [112]. For example, extracellular ATP, upon binding to its receptor, the purine-dependent phenoxin-1 channel P2X7, induces K<sup>+</sup> efflux and initiates NLRP3 assembly and downstream signaling [113,114]. The bacterial toxin nigericin also promotes activation of NLRP3 by inducing K<sup>+</sup> efflux in a pannexin-1-dependent manner [115]. In addition to ATP and pore-forming toxins, alum, silica, and calcium pyrophosphate crystals also induce K<sup>+</sup> efflux. Moreover, the complement cascade component membrane attack complex (MAC) activates NLRP3 [116]. Apart from K<sup>+</sup> efflux, mobilization of Ca+<sup>2</sup> in the cytosol through the opening of plasma membrane channels or the release of endoplasmic reticulum (ER)-linked Ca+<sup>2</sup> stores represents another upstream event in NLRP3 activation [117]. It should be emphasized that K<sup>+</sup> efflux regulates Ca+<sup>2</sup> flux and these two channels act cooperatively to activate NLRP3. In fact, NLRP3 activation induced by nigericin, alum, MSU crystals, and the MAC depends on Ca+<sup>2</sup> flux along with K<sup>+</sup> efflux [118]. Cl- efflux through chloride intracellular channel proteins (CLICs) enhances NLRP3 activation via the polymerization of ASC [119]. Translocation of CLIC1, CLIC4, and CLIC5 to the plasma membrane depends on the release of mitochondrial ROS (mtROS), whereas Cl- efflux occurs downstream of K<sup>+</sup> efflux [119]. Lysosomal swelling and damage by phagocytosed but resistant to degradation crystals, such as silica, β-amyloid, liposomes, and asbestos, represents another mechanism of NLRP3 activation [120,121]. Briefly, the accumulation of crystals intracellulary destabilizes the lysophagosome and leads to the release of its components, including proteases, lipases, cathepsins, and Ca+<sup>2</sup> in the cytosol, which, in turn, drive NLRP3 assembly and activation in a K<sup>+</sup> efflux-dependent manner [120,121] (Figure 1).

Numerous sources of ROS, such as, NADPH-oxidases, xanthine oxidase, cytochrome P450, cyclooxygenases, and lipoxygenases, induce NLRP3 activation [122] (Figure 1). The production of mtROS and mitochondrial DNA (mtDNA) also activate NLRP3. In fact, chemical inhibitors preventing ROS production inhibit NLRP3 inflammasome activation in response to several activators. Furthermore, factors that cause mitochondrial dysfunction increase the oxidation of mtDNA, which activates NLRP3 inflammasome. Increased mtROS production oxidizes thioredoxin (TRX), leading to its dissociation from the thioredoxin (TRX)-interacting protein (TXNIP). The dissociated TXNIP directly binds to NLRP3, leading to its activation [123]. Mitochondria also act as docking sites for NLRP3 inflammasome assembly. For example, the mitochondrial antiviral signalling protein (MAVS), an adaptor protein in RNA sensing, is critical for NLRP3 inflammasome activation during infections with RNA viruses and stimulation with the synthetic RNA polyinosinic-polycytidylic acid. MAVS recruits NLRP3, directing its localization to the mitochondria [124]. Still, MAVS is not essential for NLRP3 inflammation induced by other NLRP3 stimuli. Recently, it was shown that trans-Golgi network disassembly into vesicles, known as dispersed trans-Golgi network (dTGN), is another process that leads to NLRP3 inflammasome activation. More specifically, the phospholipid phosphatidylinositol-4-phosphate on dTGN drives NLRP3 aggregation, ASC oligomerization and caspase-1 activation, and downstream signaling [125].

Interestingly, recent studies have implicated cellular metabolic events in the activation of the NLRP3 inflammasome. Indeed, aerobic glycolysis and the mitochondrial electron transport chain (ETC) enhance NLRP3-driven responses [126–129] (Figure 1). In LPS-stimulated macrophages, activation of mammalian target of rapamycin complex 1 (mTORC1) promotes hexokinase (HK1)-dependent glycolysis which, in turn, induces NLRP3 activation [126–129]. Consequently, inhibition of mTORC1 or deficiency of Raptor, an mTORC1-binding partner, decreases HK1-dependent glycolysis and suppresses NLRP3 signaling [126–129]. Moreover, glucose deprivation, 2-deoxyglucose (2-DG) treatment, or HK-1 knockdown suppresses ATP-driven NLRP3 activation and inhibits IL-1β secretion

by macrophages [126–129]. Additionally, saturated fatty acids, such as palmitate, suppress the activation of the anti-inflammatory AMP-activated kinase, leading to increased ROS production and NLRP3 activation [130].

Altogether, these secondary activating signals lead to IL-1β and IL-18 secretion downstream of NLRP3 activation and also to pyroptosis (Figure 1). Still, despite extensive research in understanding the upstream events during NLRP3 activation, there is still no single unifying model, and further studies using genetic approaches, rather than pharmacological inhibition that could lead to indirect and off-target effects, need to be performed.

#### **7. Role of the Noncanonical and Alternative Nlrp3 Activation Pathways**

The noncanonical pathway of NLRP3 activation is associated with the detection of intracellular LPS generated following infection by Gram-negative bacteria, such as *Escherichia coli*, *Salmonella typhimurium*, *Shigella flexneri*, and *Burkholderia thailandensis* [131–133]. As such, the noncanonical NLRP3 pathway induces pyroptotic cell death and restricts the growth of intracellular bacteria in myeloid and nonmyeloid cells. The noncanonical NLRP3 pathway requires signaling through caspase-11 in mice and caspases 4 and 5 in humans. The binding of LPS to caspases 11, 4, and 5 results in their autoactivation and cleavage of GSDMD, triggering pyroptosis and the secretion of IL-1α [134]. Pyroptosis enhances K<sup>+</sup> efflux which activates the canonical NLRP3 pathway and the release of IL-1β and IL-18. Hence, caspases 4, 5, and 11 do not cleave IL-1β and IL-18, but only lead to pyroptosis, and the subsequent canonical NLRP3 inflammasome activation pathway is responsible for caspase-1 activation and cytokine secretion. Activated caspase-11 also cleaves pannexin-1, a membrane ATP channel, which induces K<sup>+</sup> efflux and activates the canonical NLRP3 pathway [131–133].

The cytosolic accessibility of LPS is driven by Guanylate-Binding Proteins (GBPs) and the Immunity Related GTPase family member 10 (IRGB10) which lyse the Gram-negative bacterium-containing vacuoles, releasing bacterial LPS into the cytoplasm [134,135]. Bacterial outer membrane vesicles (OMVs) also deliver LPS into the cytoplasm through endocytosis. Another process through which extracellular LPS activates the noncononical NLRP3 pathway is the binding and activating of its receptor, TLR4. TLR4 induces TRIF and MyD88 signaling, and drives the production of type I IFNs which, in turn, enhance the expression of the noncanonical inflammasome components, caspase-11, GBPs and IRGB10 [135,136]. In fact, type I IFNs, along with the complement C3-C3aR axis, upregulate caspase-11 expression. In neutrophils, the activation of the noncanonical NLRP3 pathway through detection of cytosolic LPS induces the release of neutrophil extracellular traps (NETs) that, in turn, activate the canonical NLRP3 pathway [137]. Reciprocally, IL-18 released upon NLRP3 inflammasome assembly induces NETosis [137].

An alternative NLRP3 inflammasome pathway was recently discovered in human and porcine monocytes and does not require a secondary signal [106]. LPS recognition by TLR4 induces the intracellular activation of the TIR-domain-containing adapter-inducing interferon-β - Receptorinteracting serine/threonine-protein kinase 1 - Fas-associated protein with death domain - Caspase-8 (TRIF-RIPK1-FADD-CASP8) cascade. Cleavage of caspase-8 induces NLRP3 activation and the maturation of IL-1β and IL-18, through an as yet unknown mechanism. This alternative pathway of NLRP3 activation occurs independently of K<sup>+</sup> efflux and ASC speck formation. An additional unique feature is that it does not trigger pyroptosis, and the secretion of IL-1β is independent of GSDM [106]. Notably, in mouse bone marrow-derived macrophages (BMDM), simultaneous TLR and NLRP3 stimulation leads to rapid inflammasome activation independent of de novo gene transcription [138]. This type of NLRP3 activation does not promote IL-1β secretion and pyroptosis, but enhances IL-18 production and provides a fast protective response against intracellular pathogen burden.

#### **8. Regulation of Nlrp3 Functions**

The activation of the NLRP3 inflammasome is associated with a diverse range of human diseases. Mutations in the *NLRP3* gene are associated with the dominantly inherited autoinflammatory diseases

known as cryopyrin-associated periodic syndromes (CAPS), including familial cold autoinflammatory syndrome, Muckle–Wells syndrome, and chronic infantile neurological cutaneous and articular syndrome [139–141]. Single nucleotide polymorphisms in the genes encoding NLRP3 inflammasome components have been also associated with the pathophysiology of Crohn's disease and rheumatoid arthritis. Other studies have revealed that excessive NLRP3 activation is implicated in diseases driven by metabolic dysfunction such as type 2 diabetes and nonalcoholic steatohepatitis, in neurodegenerative diseases including Alzheimer's disease and Parkinson's disease, and in cancer [142–145]. Notably, NLRP3 activation through exposure to crystals and protein aggregates is associated with silicosis and fibrosis in the lung, atherosclerosis, gout flares, and kidney dysfunction [142]. Hence, stringent regulation of NLRP3 responses is essential for the control of overactive inflammatory processes and the prevention of tissue damage.

NLRP3 regulation takes place both at the transcriptional and the post-transcriptional levels. Type I-IFNs repress the expression of pro-IL-1β through the secretion of IL-10 [146] (Figure 1). Type-I IFNs also induce the expression of iNOS, which inhibits NLRP3 activation through the production of Nitric oxide (NO ) and TRIM that reduces ROS release [147]. Post-transcriptional regulation of NLRP3 involves signaling through the microRNA, mir-223, which binds to the 3' untranslated region UTR of NLRP3 and inhibits its expression [148] (Figure 1). PTMs can also negatively regulate NLRP3 responses. Indeed, the ubiquitylation of the LLR domain of NLRP3 by the membrane-associated Ring finger protein 7 (MARCH-7), and the phosphorylation of its Ser291 residue, negatively regulates NLRP3 activation [149,150].

One of the most important mechanisms that restrain NLRP3 functions is autophagy (Figure 1). Autophagy is an endogenous recycling process utilized by the host to maintain cell homeostasis in response to stress [151]. In autophagy, dysfunctional or unnecessary cellular components become degraded. In addition, mitophagy promotes the clearance of damaged mitochondria from the cytoplasm and reduces mtROS [151]. Recent studies have shown that activation of NLRP3 in macrophages deficient in autophagy components, such as *Beclin* or *Lc3b*, leads to increased secretion of mtDNA and ROS, and increased activation of caspase-1 and release of IL-1β and IL-18 [152,153]. Furthermore, in vitro treatment of monocytes with rapamycin, an autophagy inducer, reduces IL-1β secretion in response to LPS [154]. In addition, induction of mitophagy through activation of the receptor-interacting serine/threonine-protein kinase 2 (RIP2) limits virus-induced NLRP3 activation [155]. Moreover, infection with influenza A virus in mice deficient in *Nod2* and *Ripk2* results in defective mitophagy, leading to excessive activation of NLRP3 and increased IL-18 production [155].

Signaling through metabolic regulators can also inhibit NLRP3 activation. Dimethyl fumarate (DMF) activates the transcription factor NF-E2-related factor 2 (NRF2) that represses IL-1β and NLRP3 gene expression in LPS-treated microglia and in the human acute monocytic leukemia cell line (THP-1) [156]. NRF2 also regulates levels of antioxidant genes to support cell survival during oxidative stress, and through limiting ROS levels, it inhibits NLRP3 activation [157]. DMF also decreases mtROS release and suppresses NLRP3 assembly [156]. In addition, nicotinamide adenine dinucleotide (NAD+) activates sirtuins (Sirt2) which inhibit NLRP3 inflammasome activation and decrease pro-IL-1β production [158] (Figure 1). Some metabolites of fatty acids, including β-hydroxybutyrate and α-linolenic acid, suppress NLRP3 activation via inhibiting K<sup>+</sup> efflux and the oligomerization of ASC, and by reducing ROS levels [159,160]. Interestingly, cholesterol metabolism is also associated with the regulation of NLRP3 responses. Of note, a recent report documented that bile acids through the TGR5-cAMP-protein kinase A axis inhibit NLRP3 activation and prevent LPS-induced systemic inflammation, alum-mediated peritoneal inflammation, and type 2 diabetes in mouse models [161]. Other studies in human primary monocyte-derived macrophages showed that PGE2 treatment following LPS stimulation inhibits NLRP3 activation through increasing intracellular cAMP levels [162]. In support, inhibition of cyclooxygenase 2, resulting in PGE2 blockade, enhances NLRP3 activation in human macrophages. In response to PGE2, protein kinase A is also upregulated, phosphorylates Ser295 of NLRP3, and attenuates its ATPase function [150]. Conversely, treatment of

BMDM with PGE2 prior stimulation with LPS and ATP increased IL-1β and active caspase-1 release in culture supernatants, highligting species-specific differences as well as differences dependent on the timing of PGE2 administration [163].

Several small-molecule compounds have been described as potent inhibitors of NLRP3 activation, including MCC950 [164], β-hydroxybutyrate (BHB) [159], Bay 11-7082 [165], dimethyl sulfoxide (DMSO) [166], and type I IFN [167]. Seminal studies by Coll et al. have documented that MCC950 inhibits both the canonical and the noncanonical pathways of NLRP3 activation, while it does not affect other inflammasomes. MCC950 prevents NLRP3-induced ASC oligomerization without affecting NLRP3 priming, and inhibits IL-1β secretion by human and mouse macrophages [164]. Subsequent studies demonstrated that MCC950 blocks nigericin-induced NLRP3 activation via inhibition of Clefflux in LPS-primed BMDMs [168]. Moreover, MCC950 inhibits NLRP3 inflammasome formation through blocking ATP hydrolysis [169]. Of relevance, preclinical studies have shown that MCC950 alleviates the severity of experimental autoimmune encephalomyelitis and prevents neonatal lethality in a model of CAPS [164]. Pharmacological inhibition of NLRP3 with MCC950 also protected against dopaminergic degeneration in a mouse model of Parkinson's disease (PD), and reduced total leukocytes and inflammatory macrophages in the BAL of mice infected with influenza A virus [170,171].

The aforementioned paragraphs highlight crucial functions for NLRP3 inflammasome activation in the eradication of pathogens, the protection against damaged and/or dying cells, and the maintenance of tissue homeostasis. However, several gaps in our understanding of the precise molecular pathways involved in the initiation and regulation of NLRP3-driven inflammatory responses remain, and represent promising avenues for future research. In the next section, we describe current knowledge on the role of NLRP3 activation in the initiation and propagation of allergic airway inflammation and human asthma, and discuss the implications of excessive NLRP3 responses in the pathogenesis of allergic diseases.

#### **9. Role of Nlrp3 Signaling in Allergic Airway Inflammation**

NLRP3 inflammasome activation is involved in the initiation and the propagation of allergendriven inflammatory responses in the airways. Studies using mouse models of allergic asthma induced by adjuvant-free ovalbumin (OVA) and adjuvant (aluminum hydroxide)-coupled OVA, demonstrate enhanced protein levels of NLRP3 and caspase-1, along with elevated IL-1β and TNF-α release by epithelial cells and macrophages in the airways, compared to Phosphate-buffered saline (PBS)-treated mice [172]. Moreover, in mice sensitized with OVA and LPS and challenged with OVA (OVALPS-OVA), as well as in HDM-instilled mice, increased ROS and mtROS production in BAL cells and in primary tracheal epithelial cells was detected, inducing augmented NLRP3, caspase-1, and IL-1β protein expression in the airways [173]. Interestingly, recent studies using three distinct HDM-induced mouse models of allergic airway inflammation (AAI), corresponding to eosiniphilic, mixed granulocytic, and neutrophilic asthma subtypes, documented increased expression of Nlrp3, Nlrc4, Nlrc5, Pycard, Casp-1 genes, and pro-IL-1β protein levels in the lungs, especially in the neutrophilic asthma model, while mature IL-1β was not shown, suggesting that although inflammasome molecules are upregulated, they do not form functional complexes without an additional trigger [174]. Moreover, increasing inflammasome sensor, caspase-1, and pro-IL-1β expression was documented from eosinophilic to neutrophilic asthma, illuminating an association of inflammasome signaling pathways with the type of airway inflammation [174]. Notably, induction of neutrophilic airway inflammation in mice challenged with HDM and polyinosinic-polycytidylic acid increased the concentration of Apolipoprotein E (APOE) in the epithelial lining fluid and enhanced IL-1β levels in the BAL, pointing to a potential role of APOE in IL-1β production [175].

Pertinent to the role of NLRP3 activation in allergic responses, studies using mice deficient in *NLRP3* and *ASC* in a model of OVA-induced AAI demonstrated decreased eosinophil influx, dampened AHR and reduced airway inflammation, and goblet cell accumulation, accompanied by decreased IL-1β expression in the airway, compared to wild type (WT) littermates [176]. Other studies also in a model of OVA-AAI, revealed that *NLRP3-*/*-* mice exhibit decreased pulmonary inflammation with suppressed mucus secretion, Th2 cytokine and chemokine production, and IgE levels [177]. Interestingly, in a model of OVA-AAI, production of IL-1α and IL-1β in the airways propagated Th2 cell-driven allergic responses and exacerbated pulmonary eosinophilia in a process mediated by caspase-8 activation [178]. These studies highlighted a novel role for caspase-8 in NLRP3 activation in the allergic airway, which was independent of caspase-1 and 11 signaling [178]. Other reports using a model of serum amyloid A (SAA)-induced AAI documented that the secretion of IL-1β in response to SAA is dependent on NLRP3 activation [179]. In fact, using mice deficient in *NLRP3* and *caspase-1*, the authors demonstrated a reduction in infiltrating neutrophils in the lung and decreased inflammatory cytokine/chemokine release (IL-1β, IL-6, and MCP-1 or CCL2) in the BAL compared to WT controls [179]. Another study using a complete Freund's adjuvant (CFA)/HDM-induced mouse model of AAI showed that administration of CRID3, an NLRP3 inhibitor, reduced IL1-β and Th2 cytokine production in the BAL and inhibited AHR [180]. Interestingly, a recent report demonstrated that NLRP3, along with IRF4, transactivates the *Il4* promoter, enhances Th2 cell differentiation, and exacerbates asthma symptoms in a mouse model [181]. Moreover, studies by Kim et al., in an experimental model of high fat diet-induced obesity, demonstrated that obese mice had increased AHR driven by aberrant NLRP3 inflammasome-dependent responses in the adipose tissue, which contributed to the activation of ILCs and increased IL-17 responses in the lung, in the absence of allergic sensitisation [182]. These studies highlighted a novel NLRP3-IL-1β-Th17 link in AHR development in obesity-associated allergic airway disease.

Notably, sensitization with OVA-alum or *Aspergillus fumigates* followed by challenge with OVA or *A. fumigates*, respectively, increased ROS production in lung homogenates mediated by mitochondrial Ca<sup>2</sup>+/calmodium-dependent protein kinase II (CaMK II), which induced NLRP3, active caspase-1 and mature IL-18 at the mRNA, and protein levels in allergic lungs, as well as exacerbated OVA-AAI [183]. In contrast, inhibition of mitochondrial CaMK II reduced AHR, Th2 cytokine production and NLRP3 activation in the lungs [183]. Further studies in a mouse model of OVA-AAI, accompanied by infection with *Chlamydia muridarum* or *Haemofilus influenza*, demonstrated an increased expression of NLRP3 in airway epithelial and infiltrating immune cells, as well as enhanced IL-1β and caspase-1 mRNA levels in the lungs of infected mice [184]. Notably, in vivo administration of MCC950 decreased AHR and neutrophil accumulation in the airways of infected mice [184]. Furthermore, using mice deficient in *NLRP3* and *IL-1*β, another study documented reduced airway inflammation and cytokine release following rhinovirus (RV) infection in HDM-challenged mice, highlighting the implication of NLRP3 activation in RV-induced disease exacerbations [185]. Notably, subepithelial macrophages were the major source of IL-1β in response to RV, with low IL-1β expression by the airway epithelium, suggesting that NLRP3 activation mainly occurred in macrophages. Moreover, no induction of IL-18, either at the mRNA or the protein levels, was observed in airway epithelial cells upon RV infection. It should be emhpasized that, in contrast to the airway epithelium, the gut epithelium is characterized by increased IL-18 and low IL-1β production in response to NLRP3 signaling. Indeed, IL-1β levels remained unaltered in colonic intestinal epithelial cells from mice treated with a high-fibre diet, while active caspase-1 and IL-18 were increased [186]. Notably, using *NLRP3-*/*-* mice, the authors showed that the protective effects of a high-fibre diet against colitis were mediated through NLRP3 activation in colonic epithelial cells. Altogether, these studies suggest that epithelial cells at distinct sites respond differently to inflammasome activation, not only in terms of cytokine secretion but also in the inflammasome sensor utilized, and propose a key role of the tissue micromilieu in governing epithelial cell-induced inflammatory responses.

Contradictory data have been generated by studies showing that NLRP3 activation is not essential for allergic disease development in OVA and HDM induced AAI. Allen et al. showed that the adjuvant effects of aluminum hydroxide in the OVA-AAI model were not affected in *NLRP3-*/*-* , *Casp-*/*- ,* or *PYCARD-*/*-* mice [187]. In another study, *NLRP3-*/*-* mice did not exhibit significant differences in airway eosinophilia, mucus production, AHR, and Th2 cell responses upon exposure to uric acid

crystals compared to their WT counterparts [188]. Similar findings were observed in a combined particular matter (PM)/Ova-induced mouse model of experimental asthma [189]. Finally, Madouri et al. demonstrated that mice deficient in *NLRP3-*/*-* , *Casp-*/*- ,* or *PYCARD-*/*-* exhibited enhanced lung inflammation and pathology, including eosinophilic inflitration and Th2 cytokine release, upon exposure to HDM, supporting the notion that NLRP3 activation exerts protective functions against HDM-induced allergic lung inflammation [190].

#### **10. Nlrp3 Signaling in Human Asthma**

A series of recent studies suggest that NLRP3 activation is involved in human asthma pathogenesis. Hirota et al. was among the first to describe NLRP3 and caspase-1 protein expression in human lung sections and in primary airway epithelial cells from healthy volunteers following exposure to PM10. In fact, they demonstrated that NLRP3 silencing using short hairpin (sh) NLRP3 attenuated PM10-induced release of IL-1β by human airway epithelial cells [191]. In other studies, in vitro RV infection of HBECs upregulated NLRP3, NLRC5, and caspase-1 protein levels that triggered IL-1β secretion [192]. Knock down of NLRP3 or NLRC5, using shRNA, decreased IL-1β secretion by HBECs, while simultaneous knock down of NLRP3 and NLRC5 abrogated IL-1β secretion. Moreover, HBEC from asthmatics exhibited enhanced co-localization of caspase-1 and ASC and increased mRNA expression of caspase-4 after IAV infection compared to healthy controls [193] (Figure 2). Still, it should be emphasized that the previous studies were using human bronchial epithelial cells cultured in a monolayer, and NLRP3 activation in airway epithelial cells cultured in an air–liquid interface (ALI), which is a more physiologically relevant model, remains incompletely defined. Increased NLRP3 and IL-18 protein levels were observed in airway epithelial cells in lung biopsies from asthmatics compared to healthy individuals [183]. Notably, individuals with neutrophilic asthma had elevated mRNA levels of NLRP3, caspase-1, and IL-1β, as well as NLRP3 and caspase-1 protein expression in sputum macrophages and neutrophils, compared to eosinophilic and paucigranulocytic asthmatics [180,194] (Figure 2). Increased expression of NLRP3 and IL-1β was also detected in the sputum of patients with SA compared to MMA, and correlated with clinical parameters of disease, such as neutrophilic airway inflammation, Asthma Control Questionnaire (ACQ) score, and Forced Expiratory Volume in 1 second (FEV1)% [184]. In another study, increased extracellular DNA (eDNA) sputum levels in SA correlated with sputum neutrophilic inflammation, increased NETs formation, caspase-1 activity, and IL-1β levels [195]. NLRP3 gene expression and IL-1β protein levels were also increased in sputum inflammatory cells from obese asthmatics compared to non-obese asthmatics, and correlated with body mass index [196] (Figure 2). Kim et al. showed higher expression of NLRP3 and caspase-1 in the BAL of asthmatics compared to healthy subjects [173] (Figure 2). BAL macrophages from asthmatics treated ex vivo with HDM also upregulated *NLRP3* and pro-IL-1β expression, resulting in increased IL-1β secretion in an APOE-dependent manner [174]. Lui et al. also showed that macrophages isolated from the PB of patients with Th2/Th17-predominant asthma had higher mRNA and protein levels of NLRP3 components and IL-1β compared to healthy controls [197].

Pertinent to inflammasome related-cytokines, most studies have shown that IL-1β is increased in the sputum and BAL of patients with neutrophilic asthma [198], and in the serum of asthmatic patients with or without steroid treatment, compared to controls [199] (Figure 2). Contradictory findings were observed regarding IL-18 release, with some studies showing increased IL-18 in the serum [200,201] and sputum from SA patients [202,203], and others reporting decreased levels in induced sputum [204]. Considering the increased activation of the NLRP3 pathways, along with excessive IL-1β release, in asthmatic patients, the concept of IL-1β blocking as a therapeutic approach in SA appears promising. Therapeutic administration of canakinumab, a fully human anti-IL-1β monocloncal antibody, has been widely used in conditions ranging from CAPS to rheumatoid arthritis, atherosclerosis, and lung cancer [205–208]. Pertinent to asthma, there was only one randomized double-blind placebo-controlled study that evaluated the safety and tolerability of canakinumab in mild asthmatic patients, as well as its effects on the attenuation of the late asthmatic response (LAR) following allergen challenge [209]. Canakinumab appeared to be safe and attenuated LAR compared to pretreatment values. Despite these positive results, no further studies have been conducted since [209], and canakinumab is no longer under investigation as a treatment for asthma (searched on clinicaltrials.org on 11/09/2019).

**Figure 2. The role of NLRP3 inflammasome in the development of severe asthma**. Exposure to pathogens, allergens, cigarette smoke, and other noxious stimuli in the asthmatic airway triggers the production of ROS, cytokines, and NETs which, in turn, can activate the NLRP3 inflammasome in infiltrating eosinophils, neutrophils, and macrophages, as well as in airway epithelial cells. This results in the enhanced release of IL-1β and IL-18, which leads to increased Th1 Th2 and/or Th17 cell infiltration and associated pathological consequences, such as mucus hypersecretion, AHR, and airway remodelin. eDNA, extracellular DNA; NETs, neutrophil extracellular traps; AHR, airway hyperresponsiveness.

In summary, a growing body of evidence suggests that NLRP3-induced inflammasome responses are implicated in AAI both in experimental models and human asthma (Figure 2). Still, certain controversial results obtained in animal studies are mainly associated with variations in the experimental models utilized, including type and concentration of allergen, route and time of administration, as well as mouse strain differences. The observed differences could be also associated with the distinct housing conditions affecting the microbiota composition. Hence, further mechanistic studies are warranted to resolve these disparities. More importantly, the precise role of NLRP3 activation in experimental models of SA remains elusive and deserves investigation. Of note, certain findings observed in animal studies were not observed in human asthmatics. For example, in human airway epithelial cells, it seems that other than NLRP3 inflammasome sensors become activated and trigger IL-1β release. Moreover, the precise factors that initate inflammasome assembly and activation in asthmatics remain incompletely defined. Mechanistic studies using animal models that more closely resemble SA need to be performed to clarify the type of cells in which NLRP3 signaling is activated, as well as its upstream regulators. In addition, functional studies using human primary airway epithelial cells in ALI cultures are essential for the elucidation of the differences in NLRP3-induced signaling in humans and mice. Finally, elucidation of the role of NLRP3 activation in the functional crosstalk between airway epithelial cells and other lung structural cells, such as ASMs, may shed new light on the mechanisms underlying tissue remodelling in SA.

#### **11. Concluding Remarks**

The past decade has witnessed a burgeoning appreciation of the existence of a wide range of SA endotypes. Still, and particularly in the type 2 low asthma endotypes, there is a considerable gap in our understanding of the cellular and molecular mechanisms involved and a remarkable scarcity of relevant biomarkers. More importantly, no effective treatments targeted at these endotypes have emerged. Growing evidence has illuminated a key role for NLRP3 inflammasome activation in the development and exacerbation of allergic responses. Hence, targeting NLRP3 inflammasome pathways in the airways of allergen-challenged mice, particularly in the context of SA, will improve our understanding of how NLRP3 signaling contributes to the development of specific aspects of disease severity. In fact, the direct comparison of the expression and activation of NLRP3 in mouse models and patients with distinct asthma severities is essential for the identification of novel biomarkers pertinent to the diverse asthma endotypes.

Current treatment modalities of NLRP3-related inflammatory human diseases target IL-1β with IL-1β antibodies or recombinant IL-1βR antagonists, such as canakinumab and anakinra, respectively. Nevertheless, most of these inhibitors are relatively nonspecific and have low efficacy. Thus, the development of targeted NLRP3 inflammasome site-specific therapeutics may be more beneficial in suppressing inflammasome-associated disease whilst not predisposing to infection. However, a deeper understanding of the NLRP3 inflammasome assembly and activation is needed before translation of these findings into therapies in clinical practice, especially in the context of SA. To achieve this, we need the development and use of better in vivo models of SA, along with complementary human studies using physiologically relevant in vitro models. Ultimately, this will facilitate the development of personalized medicine for the growing numbers of patients with SA.

**Author Contributions:** E.T. and M.S. searched the literature and wrote the manuscript; E.T. and J.M. designed the figures; G.X. wrote the manuscript.

**Funding:** General Secretariat for Research and Technology: 5035; General Secretariat for Research and Technology: 09-12-1074; Hellenic Foundation for Research and Innovation: 1030.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
