Airway Microbiota Dynamics Uncover a Critical Window for Interplay of Pathogenic Bacteria and Allergy in Childhood Respiratory Disease

Abstract

Repeated cycles of infection-associated lower airway inflammation drive the pathogenesis of persistent wheezing disease in children. In this study, the occurrence of acute respiratory tract illnesses (ARIs) and the nasopharyngeal microbiome (NPM) were characterized in 244 infants through their first five years of life. Through this analysis, we demonstrate that >80% of infectious events involve viral pathogens, but are accompanied by a shift in the NPM toward dominance by a small range of pathogenic bacterial genera. Unexpectedly, this change frequently precedes the detection of viral pathogens and acute symptoms. Colonization of illness-associated bacteria coupled with early allergic sensitization is associated with persistent wheeze in school-aged children, which is the hallmark of the asthma phenotype. In contrast, these bacterial genera are associated with "transient wheeze" that resolves after age 3 years in non-sensitized children. Thus, to complement early allergic sensitization, monitoring NPM composition may enable early detection and intervention in high-risk children.

Keywords: airway microbiota; allergic sensitization; asthma; lower respiratory infection.

Graphical abstract

Figure 1

Definition and Distribution of Microbiome Profile Groups (A) Heatmap shows relative abundances of 20 common operational taxonomic units (OTUs), aggregated values for other OTUs from the common genera, and aggregated values for all other rare OTUs within each sample. Tree to the left shows phylogenetic relationships between the sequenced V4 region of the 21 common OTU sequences. Dendrogram at the top indicates complete linkage clustering of Bray-Curtis distances between samples; colored bars indicate assignment to microbiome profile groups (MPGs) based on this clustering. Bar plot to the right shows the total abundance of each OTU or group of OTUs within the whole dataset; OTUs that dominate a common MPG are colored to match that MPG. (B) Distribution of MPGs within each time period, shown separately for healthy and acute respiratory illness (ARI) samples. See also Figures S1 and S2; Table S2.

Figure 2

Within-Sample Diversity Is Associated with Age and Acute Respiratory Illness Symptoms (A) Shannon diversity index (SDI) per sample over time, colored by symptom status as indicated (URI, upper respiratory illness; LRI, lower respiratory illness). Solid lines, loess smoothed curves; dashed lines, 95% confidence intervals. (B) SDI distributions within common MPGs. Asterisk indicates FDR adjusted p value of <0.05 in GEE linear regression of SDI against healthy versus LRI, adjusted for age at collection (as in Table S3). (C) Relative abundances of the dominant OTU within each MPG (as specified in Table S1). See also Figure S3 and Table S3.

Figure 3

Time-Varying Associations of Bacterial Taxa with Acute Respiratory Illness Symptoms (A) Log₂ fold change (solid lines) and 95% confidence intervals (dashed lines) comparing symptomatic versus healthy samples, estimated using smoothing splines ANOVA. Non-significant segments are colored gray. (B) Same as (A) but including further adjustment for Moraxella OTU 4398454 abundance (dark-green curve) and vice versa (dark-red curve). See also Figure S4.

Figure 4

Microbial Interaction Networks and Stability (A) Pairwise correlations among eight characteristic OTUs, calculated separately for samples collected up to and including 2 years of age (triangle of the heatmap below the diagonal), and samples collected after 2 years of age (triangle of the heatmap above the diagonal). Cell colors indicate correlation coefficients; non-significant correlations (p > 0.001) are white. ^∗Bonferroni-corrected p < 0.05/28, testing for change in correlation before and after 2 years of age using Fisher's z test. (B) Correlations between Alloiococcus or Corynebacterium and Moraxella or Streptococcus or Haemophilus OTUs (bold gray box in A) over half-yearly time periods (filled circles, significant correlations, p = 0.001; empty circles, non-significant correlations, p > 0.001). (C) Transitions between microbiome profile groups (MPGs) for consecutive pairs of healthy samples collected from the same individuals 6–12 months apart. OTU key: Haemophilus: A = 240051, B = 4469627, C = 956702; Moraxellaceae: A = 1057260, B = 854899; Corynebacterium: A = 4474764, B = 1049188, C = 4376867. (D) Proportion of healthy samples collected at each time point, for which the same MPG was detected in the next healthy sample from each individual. Colors indicate the specific MPGs involved, colored as in (C). See also Figure S6 and Table S4.

Figure 5

NPM Associations with Symptoms of Acute Respiratory Illness and Wheeze (A) Frequency of symptoms (URI, upper respiratory illness; LRI, lower respiratory illness) among samples stratified by the presence or absence (+/−) of known respiratory viruses and presence or absence (+/−) of bacterial communities assigned to Moraxella, Streptococcus, or Haemophilus microbiome profile groups (MPGs). (B) Association of acute respiratory illness (ARI) symptoms with specific MPGs, stratified by the presence or absence (+/−) of common respiratory viruses (RV, rhinovirus; RSV, respiratory syncytial virus; Vir, any virus). Odds ratios (OR) and 95% confidence intervals were estimated using generalized estimating equations with unstructured correlation and robust standard errors, adjusting for age, gender, and season. (C) Proportion of healthy samples assigned to Moraxella, Streptococcus, or Haemophilus MPGs, stratified by time relative to a recorded LRI episode. SE bars are given for the Moraxella MPG. We regressed assignment to Moraxella MPG against time to LRI (separate models for each time category versus all other healthy samples) (^∗p < 0.05). (D) Frequency of pre-school wheeze phenotypes (y axis), stratified by frequency of Moraxella, Streptococcus, or Haemophilus MPGs among healthy samples collected from 6 months to 2 years of age (x axis, in tertiles). Data are shown separately for 73 children who were allergic sensitized by 2 years of age, and 64 who were not. See also Figures S5 and S7.