Host–microbial interactions differ with age of asthma onset

Extract Asthma is a heterogenous disease [1] and dichotomisation between childhood/early-onset (EO) and adult/late-onset (LO) disease [2] identified differences in lung function decline and response to anti-inflammatory therapies, including biologics [3]. This suggests distinct inflammatory mechanisms underpin EO and LO asthma. In parallel, a relationship exists between airway neutrophilia and the airway microbiome [4, 5]. We postulate that differences in host–microbial interactions are associated with the age of asthma onset and would be maintained over time. Here, we applied a recently described machine learning framework, sparse canonical correlation analysis (Sparse-CCA) [6], to identify differences in host–microbial interactions in the airways of EO and LO asthma.

Sparse-CCA incorporates a lasso (Least Absolute Shrinkage and Selection Operator) penalty for feature selection and a linear projection of two sets of observations into a shared latent space [6] which identifies a smaller subset of paired host genes and bacterial species, known as components that are most highly correlated for age-of-onset-group.The analysis was conducted using R-4.1.3with the PMA package (version 1.2.1).Hyperparameter tuning was performed using a grid search approach to identify parameters for sparsity penalties, as previously described [6].Sparse CCA components were computed for each group with no components being correlated with each other.
Pathway enrichment analysis of the Kyoto Encyclopaedia of Genes and Genomes (KEGG) and pathway interaction database (PID) gene sets from the MsigDB canonical pathways collection (https://www.gsea-msigdb.org/gsea/msigdb/human/collections.jsp)across all components identified six pathways that were shared between phenotypes.In addition, 10 pathways were only enriched in LO asthma and 13 pathways only enriched in EO asthma according to absolute weighting using a Fisher's exact test ( p<0.05).
Compared to LO asthma or pathways shared between LO and EO asthma, there was greater enrichment of pathways associated with adhesion molecules in the components of EO asthma.This is consistent with genetic studies that identify barrier function as a contributor to EO disease [10] as well as its greater association with atopic conditions [11] (figure 1b).In addition, PID pathways specific for EOC1 were associated with cell adhesion, migration and proliferation, whilst KEGG pathways in EOC2 are linked to pathways involved in immune signalling in response to microbial infection.Finally, EOC3 is associated with cell proliferation/death pathways and insulin resistance (figure 1b).
Sparse-CCA identified tumour necrosis factor (TNF) signalling to be more prominent in LOC1 (figure 1b).TNF is a pro-inflammatory cytokine associated with neutrophilic asthma [12].While direct therapeutic targeting of TNF in asthma has not been successful, azithromycin therapy modulates the TNF axis [13].Integrin-and other cell surface receptor-mediated intracellular signalling (FAK pathway), ribosome and gene expression, proliferative pathways and calcium signalling pathways were associated with LOC2-5, respectively (figure 1b).
Several pathways were shared between LO and EO phenotypes, particularly EOC1 and LOC2 and 4 (figure 1b).However, the genes and species constituting those components were not identical; for example, LOC4 and EOC1 were both enriched for leukocyte transendothelial migration (figure 1b and c), but LOC4 was characterised by Streptococcus species and Wnt5a while EOC1 was characterised by Moraxella catarrhalis, Haemophilus influenzae and CFLAR (figure 1c).SQSTM1, present in both components, modulates microbe-induced inflammatory pathways in an autophagy-dependent and -independent manner [14].The combination of Wnt5a and CFLAR (figure 1c) with SQSTM1, for example, on microbial growth and on host-microbe immune interactions should be the target of further studies.
Using gene set variation analysis (GSVA) to calculate a sample-wise enrichment score (ES) from the top 10 genes in a component by absolute weights from the Sparse-CCA, genes in LOC4 correlated with sputum neutrophilia in LO patients but not in EO asthma.Conversely, genes in EOC1 correlated with sputum neutrophilia in EO but not LO asthma (figure 1d).These findings indicate that these host-microbial interactions are unique to each neutrophilic asthma phenotype.
Haemophilus influenzae was associated with pathways enriched in both EO and LO components (15 in EO and 10 in LO), Moraxella catarrhalis was only associated with pathways enriched in EO (19 pathways) and Tropheryma whipplei with pathways enriched in LO (6 pathways).LOC3 had a geneset, whose ES was correlated with sputum eosinophils and was dominated by Neisseria and Haemophilus influenza.
In summary, sparse-CCA identified several host gene-microbiome associations; however, longitudinal/ dynamic conclusions regarding causality cannot be inferred in this cross-sectional/static analysis.Furthermore, exposures cannot be accounted for and will clearly influence disease evolution, and a temporal microbiomics approach may be required to identify underlying endotypes [15].Moreover, U-BIOPRED is an adult cohort and so age of disease onset is confounded by disease duration, number of exacerbations and treatment including corticosteroids and macrolides.Nevertheless, microbial analysis of https://doi.org/10.1183/13993003.00428-2024the U-BIOPRED data has previously identified differences in microbial profiles and age of onset [4].This analysis extends those findings by identifying shared and unique host-microbial interactions between EO and LO phenotypes.
This study demonstrates the utility of integrating the sputum microbiome and host gene expression together to obtain insight into their to the disease process, which is superior to single-dataset omics alone.While the composition of the airway microbiome changes throughout life, it is particularly dynamic in the early years of life, when perturbations are thought to be critical to lower airway immune maturation [16].Our findings demonstrate that the heterogeneity of asthma immunopathophysiology may be better understood though host-microbial interactions.Faculty of Medicine, Southampton University, Southampton, UK. 3 Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore. 4Amsterdam University Medical Centers, Department of Pulmonary Medicine, University of Amsterdam, Amsterdam, The Netherlands.

FIGURE 1 a
FIGURE 1 a) Flowchart showing the overall method of sample collection, cohort grouping and the application of sparse canonical correlation analysis (Sparse-CCA).b) Component genes pathway analysis showing the enrichment of pathways for each component.c) Visualisation of leukocyte transendothelial migration for both adult and child onset showing the top 10 genes and species by absolute weights.d) Gene set variation analysis (GSVA) sample-wise enrichment score (ES) correlation with clinical characteristics showing the correlation of the ES of each gene set from the components in the leukocyte transendothelial migration pathway with sputum neutrophils and eosinophils.KEGG: Kyoto Encyclopaedia of Genes and Genomes; PID: pathway interaction database; TNF: tumour necrosis factor.Significance of Spearman coefficient correlation: not significant ( p>0.05) indicated by clear box; **: p⩽0.01; ***: p⩽0.001.