2.1 Subjects

Patients aged 16–60 were recruited from across 18 UK centers with physician-diagnosed asthma of a minimum of 1 year and on current prescription for asthma (see methods section in the supplement for further detail on the inclusion and exclusion criteria and information on the Genetics of Asthma and Severe Phenotypes [GASP] cohort). 708 asthma patients were used for the serum biomarker analysis, 4,205 patients were used for the clinical analysis with 3037 used for the clinical and genetic analysis (Table 1).

2.2 Collection of Clinical and Immunological Patient Data

Demographics including age, sex, age of onset, height, and weight, along with clinical data baseline lung function (FEV₁, FVC and PEF) after 4 h withholding reliever medication, smoking status, smoking pack years, current asthma medication, Global Initiative for Asthma (GINA) classification 2018 [19] and ACQ score (see supplement methods section for further details) were recorded. Total serum IgE, number of hospital admissions within the last year, comorbidities such as allergic rhinitis and allergic dermatitis and blood eosinophil count were also recorded.

2.3 Collection and Processing of Serum Samples

708 asthma patients, representative of the larger cohort (Table 1) had blood for serum collected when their asthma was stable. Serum separating tubes were centrifuged at 3000 rpm for 10 min at recruitment site and aliquots were made and stored at −80°C. Aliquots of serum samples were sent on dry ice and stored at −80°C at the University of Nottingham Respiratory Medicine laboratories.

2.4 Luminex Assay

R&D systems Magnetic Luminex was used to measure the concentration of Th1 (IFN-y), Th2 (IL-4, 5, 13) and Th17 (IL-17) cytokines. In addition, potential upstream regulators (TSLP, IL-33, IL-33 receptor [ST2]) and downstream inflammatory markers of inflammation (IL-6, CCL11 and Periostin) were quantified in the sera according to manufactures instructions (see Supporting Information S1: Table S1 for the range of each analyte measured in all 708 patients and Supporting Information S1: Table S2 for the maximum detection limit). When protein levels were below the level of detection for specific individuals, these individuals did not contribute to these analyses.

2.5 Statistical Analysis

All statistical analyses were carried out on IBM SPSS Statistics 24. See methods in the supplement section for further details on the statistical tests used in this study.

3.1 Demographics and Clinical Features of the Serum Cohort

A total of 708 asthma patients from 18 UK centers were available for analysis as part of the GASP initiative (Table 1). These subjects were enriched with patients meeting GINA 3-5 (GINA 2018) criteria. Basic demographics, clinical features including spirometry and immunological data were collected, however not all data was available for all 708 patients. Overall patients had poorly controlled asthma with a female predominance (Table 1).

3.2 Serum Cytokines Associated With Th2 Inflammation Showed High Coordinated Expression

Distribution of the serum cytokine data is displayed in Supporting Information S1: Figure S1 and Supporting Information S1: Table S1. Most of the cytokines demonstrated a non-normal distribution. TSLP showed high correlation with IL-4, IL-5 and IFN-ƴ (R² = 0.68, R² = 0.59, R² = 0.58 and R² = 0.55; respectively) (Figure 1B,F–H). IL-4 showed high correlation with IL-33, IL-5 and IFN-y (R² = 0.54, R² = 0.513 and R² = 0.52; respectively) (Figure 1D–I). Periostin and ST2 (soluble IL-33 receptor) showed no correlation with any of the biomarkers.

3.3 Cytokines in the Serum Identify a Heterogeneous Inflammatory Profile in Asthma and Suggests Overlapping Inflammatory Profiles

Patients in the top quartile of the three cytokines that reflect current biological targets that are currently in clinic, that is, IL-5, TSLP, IL-4/13 (Supporting Information S1: Table S2) demonstrate 46.6% of the asthma patients were high for all three targets (Figure 1J). Whereas Figure 1K demonstrates those patients in the top quartile for the Th2 cytokines (Supporting Information S1: Table S3) with 43.7% of patients being high for all 3 Th2 cytokines.

3.4 Childhood Onset Asthma Patients Are More Atopic Than Adult Onset, However, Have a Similar Serum Cytokine Profile

Table 2 shows the demographic, clinical and immunological data stratified by childhood onset < 16 years, early adulthood, 16–34 years, and late adulthood, 35–60 years. The childhood onset group had more atopic comorbidities including atopic dermatitis 32.6% (p = 0.001) and allergic rhinitis 63.9% (p = 0.001) as well as higher total blood IgE levels 150 KU/L (p = 7.8 × 10⁻⁸) compared to both later onset groups (Table 2). Both later age of onset groups demonstrated a higher weight 83.31 kg and 86.12 kg respectively compared to the < 16 group 81.53 kg (p = 0.002). Later age of onset groups also demonstrated higher smoking pack years 9 and 10 years respectively (p = 1.0 × 10⁻⁶) suggesting early onset patients are less likely to take up smoking in adulthood. FEV₁ (%Pred) was lower in the < 16 and 16–34 age groups, 80.17% and 79.11% respectively compared to the 35–60 group 83.92% (p = 0.003). There was a higher female predominance in the 16–34 group, supporting previous studies that asthma development in middle age has a female bias [20]. No serum cytokine level was associated with age of asthma onset (Supporting Information S1: Table S4).

3.5 Patients Prone to Exacerbation Have a Smoking History, Higher BMI, Worse Lung Function and Poor Asthma Symptom Control But Do Not Show Differences in Their Serum Cytokine Profile

Data were first analysed based on patients that did or did not have an exacerbation within the last year (Table 3), then the clinical features that met multiple testing were further analysed based on hospital admission frequency (Supporting Information S1: Table S5). BMI was higher in patients who exacerbated once (29.97 kg/m²) or more than once (30.22 kg/m²) compared to those who never exacerbated in the previous 12 months (28.61 kg/m²) (p = 1.0×10⁻⁶) (Table 3). The data showed smoking frequency is highly associated with the risk of an exacerbation and admission to a high dependency/ICU. Smoking pack years was the highest in patients admitted to high dependency/ICU (9 pack years) and high in those who exacerbated once (4.35 pack years) and more than once (4.75 pack years) compared to those that never exacerbated (3.5 pack years) (p = 0.0002). FEV₁ (%Pred) was inversely related to hospital admission frequency (p = 8.1 × 10^-9). Similarly, the more frequent the hospital admissions the higher the ACQ6 score demonstrating that patients who frequently exacerbate have poor symptom control. Patients with one or more hospital admission or in the high dependency/ICU groups had more severe asthma, GINA 3/4 30.7% and GINA 5 49.2% respectively (p = 5.07 × 10⁻²⁰), (Supporting Information S1: Figure S2). No serum cytokine level was related to exacerbation prone asthma (Supporting Information S1: Table S6).

3.6 Asthma Patient Subgroups Defined by Elevated Blood Eosinophil Counts Have Lower Smoking Pack Years and Elevated Total IgE

Asthma patients were stratified based on blood eosinophil count ≥ 300 cells/μL. Patients with elevated eosinophil counts had lower smoking pack years compared to the eosinophil low group (p = 0.0003) (Supporting Information S1: Table S7). Total blood IgE levels were higher in the eosinophil high group, 250 KU/L compared to 109 KU/L (p = 2.0 × 10⁻⁶).

3.7 Two Genetic Signals for Moderate-Severe Asthma Were Associated With Childhood Onset Asthma

We explored whether carrying 0, 1 or 2 risk alleles for the 25 genetic signals associated with moderate–severe asthma [9] were related with childhood onset, exacerbation prone, eosinophilic high asthma or any of the Th1, 2, 17 biomarkers (Figure 2). Carrying 1 or 2 risk alleles for SNPs rs9303277 (IKZF3) and rs61816764 (FLG) was associated with childhood onset asthma and met statistical significance after Bonferroni correction. Other genetic signals did show suggestive association with other outcomes but did not meet correction (Supporting Information S1: Tables S8–S11). The genetic risk factors for moderate-severe asthma did not drive the serum biomarker levels in the blood associated with Th1, Th2 and Th17 inflammation (Supporting Information S1: Tables S12–S14).

4.1 Characterization of Asthma Subgroups

Childhood onset is a reproducible subgroup in asthma and highly associated with atopy, high total IgE, prevalence of atopic dermatitis and rhinitis and eosinophilia [20, 25-30] with the current data further supporting these observations. Late onset patients had a higher BMI, lower lung function, a higher prevalence of smokers and less allergic comorbidities [31-33], which again was replicated in the current study. Furthermore, signals rs9303277 (IKZF3) and rs61816764 (FLG) showed the strongest effect size with age of onset when carrying both risk alleles for moderate-severe asthma compared with all the genetic signals and clinical features. Signals around the IKZF3 gene have also previously been reported in childhood age of onset GWAS studies [34-36] suggesting that these signals may drive childhood onset of asthma. A recent study provided evidence for a novel disease-driving mechanism induced by the 17q21risk allele, which is the strongest genetic risk factor for childhood-onset asthma. They revealed that increased mucosal GSDMB expression was associated with a cell-lytic immune response linked with compromised airway immunocompetence. Their findings suggested GSDMB-related airway cell death and disturbances in the mucosal IFN signature explain the increased vulnerability of 17q21 risk allele carriers to respiratory viral infections in early life. This potentially provides potential novel therapeutic biological targets for interventions in early life with IFN [37]. Overall, these data support distinct clinical and immunological features of these different onset asthma groups [20, 25-29] and confirm that genetics is important for the risk of developing childhood disease. The systemic inflammatory profile was not different between these groups, again limiting the utility and insight provided.

Exacerbation prone asthma is another important subgroup. We found that higher BMI is associated with more frequent hospital admissions extending data suggesting that BMI is a marker of disease severity in asthma [38]. Poor symptom control and severe asthma defined were also associated with frequent hospital admissions in line with other studies [39-41]. Patients with frequent exacerbations in the UK severe asthma register also had the strongest association with a high ACQ6 symptom score despite treated with or without OCS [42]. In the current study ACQ6/7 showed a clear association with exacerbation including frequency. We did not identify a specific serum inflammatory profile in these groups and none of the genetic variants associated with the risk of developing moderate-severe asthma are risk factors for exacerbation in this adult cohort.

We also examined a subgroup characterized by eosinophilia. In our study blood eosinophil count had a weak relationship with IgE driven disease but not with any other clinical distinct subgroups. However, these analyses may have been confounded by the use of corticosteroids, and this may also suggest why elevated eosinophils were specific in identifying distinct asthma subsets. We also demonstrated smoking pack years influenced low blood eosinophil levels. Studies have demonstrated cigarette smoke to influence the immune inflammatory response and can regulate inflammatory mechanisms of eosinophil migration therefore reducing the number of eosinophils [43]. Eosinophil levels in COPD patients have been shown to be lower in the blood suggesting smoking influences eosinophil levels [44]. Again, we did not identify a relationship between systemic inflammation and this asthma subgroup. While not meeting Bonferroni correction, the two nominal genetic associations with rs1342326 (IL33), rs36045143 (CLEC16A) and blood eosinophil numbers have been described previously [45].

Our study has many strengths; it included well characterized asthma patients with combined used of clinical, genetics and serum biomarker data. Limitations included OCS and ICS use was not well recorded for each patient and so could not be taken into account in the analyses as a covariate. However, in a small subset of patients where we had OCS and ICS data, we looked at whether corticosteroids could influence the levels of these biomarkers with statistically significant effects after Bonferroni correction. A second limitation was some of the biomarkers were difficult to measure in the serum as they fell below the standard curve particularly IL-5 and IL-13 this could potentially be because some cytokines have short half-lives and so a more sensitive assay may be need. In addition, serum is not a direct measurement of airway inflammation and so cytokine levels in sera may not reflect airway inflammation therefore particular cytokines could be lower in the sera than the airways. The population sample was also relatively small for genetic association studies and there was no exhale nitric oxide or sputum eosinophil data recorded as well as limited blood eosinophil data to stratify type-2 high asthma patients.