The nasal microbiome, nasal transcriptome, and pet sensitization

Abstract

Background:

Pet allergies are common in children with asthma. Microbiota and host responses may mediate allergen sensitization.

Objective:

To uncover host-microbe relationships in pet allergen sensitization via joint examination of the nasal microbiome and nasal transcriptome.

Methods:

We collected nasal samples from 132 children with asthma for parallel 16S rRNA and RNA sequencing. Specific IgE levels for cat and dog dander were measured. Analyses of the nasal microbiome, nasal transcriptome, and their correlations were performed with respect to pet sensitization status.

Results:

Among the 132 children, 91 (68.9%) were cat sensitized and 96 (72.7%) were dog sensitized. Cat sensitization was associated with lower nasal microbial diversity by Shannon Index (P=0.021) and differential nasal bacterial composition by weighted UniFrac distance (PERMANOVA P=0.035). Corynebacterium sp. and Staphylococcus epidermidis were significantly less abundant, and the metabolic process “fatty acid elongation in mitochondria” was lower in pet sensitized vs. unsensitized children. Correlation networks revealed that the nasal expression levels of 47 genes representing inflammatory processes were negatively correlated with the relative abundances of Corynebacterium sp. and Staphylococcus epidermidis. Thus, these species were not only directly associated with the absence of pet sensitization, but also with the under-expression of host gene expression of inflammatory processes that contribute to allergen sensitization. Causal mediation analyses revealed that the associations between these nasal species and pet sensitization were mediated by nasal gene expression.

Conclusion:

Higher abundances of nasal Corynebacterium sp. and Staphylococcus epidermidis are associated with absence of pet sensitization and correlate with lower expression of inflammatory genes.

Capsule Summary:

Nasal microbiota and nasal gene expression vary by pet sensitization status. Higher abundances of nasal Corynebacterium sp. and Staphylococcus epidermidis are associated with absence of pet sensitization and correlate with lower host expression of inflammatory genes.

Introduction

Allergies to cat and dog dander are common, and exposure to these pet allergens often triggers asthma exacerbations in individuals with asthma.^{1, 2} Pet exposure is associated with risk of allergen sensitization, and this is thought to be partly mediated by microbiota.^{2, 3} While many studies have reported associations between gut microbiota and allergen sensitization^{4, 5}, including to pet allergens ^{6, 7}, fewer studies have specifically examined interactions between airway microbiota and allergen sensitization.⁸ In particular, given the nasal passages are the first point of contact between inhalant allergens and the airway, examination of nasal microbiota with respect to allergen sensitization and nasal gene expression could reveal novel insight into host-microbe relationships in allergen sensitization. We hypothesized that multi-omic examination of subjects with and without pet sensitization could reveal novel associations and mediation relationships between nasal microbiota, nasal gene expression, and allergen sensitization to cat and/or dog.

Results and Discussion

Detailed methods are provided in the Online Supplement.

Study population and design

Children with persistent asthma based on National Heart Lung Blood Institute Expert Panel Report 3 guidelines⁹ were recruited from the Mount Sinai Health System, New York, NY, USA. The study was approved by the Mount Sinai Institutional Review Board, and written informed consent was obtained from participants’ parents. The 132 children (median age=12.0 years, interquartile range=5 years) included those with mild/moderate (74%) and severe (26%) persistent asthma (Table S1). Asthma control varied among the participants with mean Asthma Control Test (ACT) score 16.6 (SD 4.3). Among the children, asthma medication use included short-acting beta agonist (92%), inhaled corticosteroid (ICS) (23%), combined ICS/long-acting beta agonist (LABA) (21%), leukotriene receptor agonist (21%), and omalizumab (2%). Additional characteristics are shown in Table S1.

Allergen sensitization to cat and dog dander was determined by serum specific IgE (sIgE) ≥ 0.10 kU_A/L (ImmunoCAP, ThermoScientific). Among the 132 children, 91 (68.9%) were cat sensitized and 96 (72.7%) were dog sensitized. Compared to their unsensitized counterparts, there were no differences in age, sex, race, asthma severity, ACT score, last upper respiratory infection, last antibiotic use, or asthma medication use except for ICS/LABA in subjects sensitized to cat or dog (Table S1). Pet ownership was not associated with sensitization to the respective pet, although dog ownership was associated with cat sensitization.

All subjects underwent nasal swab of one nare and nasal brushing of the contralateral nare. The swab and brush samples respectively underwent DNA and RNA isolation, 16S rRNA and RNA sequencing, sequence data processing, quality control, and analysis using previously described methods.¹⁰ Briefly, nasal microbiome profiling of isolated DNA was performed with dual indices and Illumina sequencing adapters attached to the V3V4 amplicon using Nextera XT Index Kit v2 (Illumina, San Diego, CA), with sequencing of amplified 16S rRNA on the Illumina MiSeq platform with 2X250bp paired-end reads; except where indicated, analyses with amplicon sequence variants (ASVs)¹¹ were performed using Qiime 2¹² with Silva reference data version 138¹³. Nasal RNA sequencing involved library preparation with the TruSeq RNA Sample Prep Kit v2 protocol (Illumina, San Diego, CA) and sequencing on the Illumina HiSeq 2500 platform with a per-sample target of 40–50 million 100bp paired-end reads. Following data processing and quality control of the above data, there were 19,269 ASVs and 17,337 genes for analysis.

Cat sensitization is associated with lower nasal microbial diversity and differential nasal bacterial composition

Alpha diversity measured by Shannon index, a measure of the richness and evenness of species at a site, was significantly lower in children sensitized (n=91) vs. unsensitized (n=41) to cat (P=0.021, Figure 1A). Although children sensitized to dog (n=96) showed a similar trend of lower nasal bacterial diversity relative to those without dog sensitization (n=36), this difference was not significant (Figure 1B). Comparison of nasal microbial composition showed compositional differences associated with cat sensitization (weighted UniFrac PERMANOVA P=0.035, pseudo-F = 2.8; Jaccard PERMANOVA P=0.016, pseudo-F=1.3) but not dog sensitization (weighted UniFrac PERMANOVA P=0.23, pseudo-F=1.3; Jaccard PERMANOVA P=0.46, pseudo-F=0.98) (Figure 1C, 1D, Figure S1).

Figure 1: Alpha diversity and relative abundances of nasal bacterial species in children sensitized and unsensitized to cat and dog.

Alpha diversity of nasal bacterial species was measured by Shannon index at a rarefaction depth of 4,980. Alpha diversity comparison by Kruskal-Wallis test between (A) children sensitized (Cat+, n=91) and unsensitized (Cat-, n=41) to cat, and (B) children sensitized (Dog+, n=96) and unsensitized (Dog-, n=36) to dog. Relative abundance of nasal bacterial species in (C) children sensitized (Cat+, n=91) and unsensitized (Cat-, n=41) to cat, and (D) children sensitized (Dog+, n=96) and unsensitized (Dog-, n=36) to dog. Unnamed species are indicated with “sp.”. Species with relative abundance less than 1% were combined as “Other”.

Linear discriminant effect size (LEfSe) analysis¹⁴, a method for biomarker discovery, was then used to identify species with relative abundance ≥1% that best characterize cat and dog sensitization. LEfSe revealed that an unnamed Corynebacterium sp. was significantly associated with absence of sensitization to cat (linear discriminant analysis (LDA) score=5.4, P=0.011) and dog (LDA score=5.0, P=0.043). Staphylococcus epidermidis was also associated with absence of sensitization to cat (LDA score=5.1, P=0.0045). No other species showed significant results. Corynebacterium spp. have been previously associated with reduced risk of airway diseases.^{10, 15} Staphylococcus epidermidis is a commensal bacterium associated with healthy maturation of the nasal microbiome and the production of beneficial antimicrobial peptides.16

Sensitivity analyses with LEfSE adjusted for age, sex, ICS/LABA use, and cat or dog ownership revealed no significant changes to these findings (Online Supplement). To test the specificity of our findings to pet sensitization, we used the same metrics to compare children in the cohort with and without sensitization to pollen (tree, grass, or weed), finding no significant differences by Shannon Index, weighted UniFrac, or LEfSe.

Potential functional implications of nasal microbiota associated with pet sensitization

To explore the functional implications of our findings, we pursued two paths. In the first, we applied metagenomic inference using PICRUSt¹⁷ and STAMP¹⁸ to infer microbial functions associated with the compositional differences in nasal bacteria associated with pet sensitization. Metagenomic inference showed that compared to their unsensitized counterparts, the metabolic process “fatty acid elongation in mitochondria” was lower in children sensitized to cat (FDR = 7.5E-03) (Figure 2A) and dog (FDR = 1.6E-02) (Figure 2B). Others have shown that altered lipid metabolism in mitochondria modulates regulatory T cell (Treg) suppressive capacity.¹⁹ Experimental blockade of fatty acid binding proteins leads to decreased fatty acid elongation in mitochondria and perturbation of Treg function.¹⁹ Given the role of Tregs in mediating tolerance against airborne antigens²⁰, our finding of reduced Treg-associated mitochondrial fatty acid processes in pet sensitized children raises the possibility that nasal microbiota could potentiate allergen sensitization via downregulation of this metabolic process. These metagenomic inference findings merit direct validation.

Figure 2: Fatty acid elongation in mitochondria of children sensitized and unsensitized to cat and dog.

Comparison of fatty acid elongation in mitochondria, a metagenomically inferred metabolic process, between (A) children sensitized (Cat+, n=91) and unsensitized (Cat-, n=41) to cat, and (B) children sensitized (Dog+, n=96) and unsensitized (Dog-, n=36) to dog. FDR values from White’s non-parametric t-tests are shown.

The second path of our functional investigation aimed to explore the potential effects of Corynebacterium sp. and Staphylococcus epidermidis on host from the lens of host gene expression and as related to pet sensitization. We first identified nasal genes differentially expressed based on cat sensitization, finding 58 differentially expressed genes (DEGs) at FDR≤0.05 and |log2 fold change| >1 (Figure 3A); DEG analysis was performed using R packages voom²¹ and limma¹⁰. We then analogously identified nasal genes differentially expressed based on dog sensitization, finding 20 dog DEGs (Figure 3B). Weighted gene coexpression network analysis²² showed that most of these DEGs belonged to two gene coexpression modules representing inflammatory response and allergy (Figure 3A, 3B) (Online supplement). Fourteen of the cat and dog DEGs overlapped with one another (Figure 3C). There was no significant change in DEGs when analyses were adjusted for age, sex, ICS/LABA use, and cat or dog ownership (Online Supplement).

Figure 3: Differentially expressed nasal genes of children sensitized and unsensitized to cat and dog.

Differentially expressed nasal genes (rows) associated with (A) cat sensitization and (B) dog sensitization at FDR ≤ 0.05 and |log2 fold change| >1 across subjects (columns). Green = higher expression in sensitized subjects. Cyan = higher expression in unsensitized subjects. Gene expression values were centered and scaled by gene. Hierarchical clustering was performed based on complete linkage. Genes in magenta and purple fonts are members of the inflammatory response and allergy coexpression modules, respectively, from weighted gene coexpression network analysis. (C) Venn diagram for differentially expressed genes associated with cat sensitization and dog sensitization.

Of the 64 genes associated with cat and/or dog sensitization, the expression levels of 47 DEGs were each significantly correlated (Spearman correlation FDR≤0.05) with the relative abundance of Corynebacterium sp. and/or Staphylococcus epidermidis (Figure 4). Each of these DEGs has been previously associated with at least one outcome related to allergen sensitization, including IgE and IgE-mediated hypersensitivity (MS4A2, GATA2, CEBPE, PCSK6, GCNT4), Type 2 inflammation (POSTIN, B3GNT6, GCSAML), mast cell function (ADCYAP1, CPA3, TPSAB1, TPSB2, TPSD1, SIGLEC6, SIGLEC8, HDC), eosinophil activity (CAPN14, GATA1, GATA2, CCL24, CLC), lung function (SNTG2, DLC1, DPP4), and mucosal inflammation (SLC9A3, CDH26, SCGB3A1, ALOX15)^23–25. Except for SCGB3A1, Corynebacterium sp. and Staphylococcus epidermidis abundances were negatively correlated (FDR≤0.05) with the expression of each of these DEGs associated with pet sensitization (Figure 4); that is, higher abundances of these two species were associated with lower levels of pet sensitization gene expression.

Figure 4: Correlation network between nasal Corynebacterium sp., nasal Staphylococcus epidermidis, and nasal transcripts associated with cat and/or dog sensitization.

Circular and rectangular nodes represent differentially expressed genes (DEGs) associated with cat and/or dog sensitization at FDR ≤ 0.05 and |log2 fold change| > 1. Edges show significant Spearman correlations (FDR ≤ 0.05) between the relative abundance of nasal species and gene expression level. Circled genes are DEGs identified by causal mediation analysis to mediate associations between Corynebacterium sp., Staphylococcus epidermidis, and pet sensitization.

As LEfSe analysis had shown that Corynebacterium sp. and Staphylococcus epidermidis were directly associated with protection from sensitization to pet allergens, we found reassuring consistency from the correlation network that these nasal species also had near uniform negative correlation with the nasal expression of the sensitization-associated DEGs. In line with prior observations that Corynebacterium spp. are associated with reduced risk of airway disease^{10, 15} and Staphylococcus epidermidis promotes a healthy nasal microbiome,16 we found that both nasal species were not only associated with the absence of pet sensitization in children with asthma, but also with the under-expression of genes associated with inflammatory processes that contribute to allergen sensitization.

Nasal gene expression mediates associations between nasal microbiota and pet sensitization

To further examine relationships between Corynebacterium sp., Staphylococcus epidermidis, and DEGs associated with pet sensitization, we performed causal mediation analysis. The abundances of Corynebacterium sp. and Staphylococcus epidermidis were inversely associated with cat and/or dog sensitization, and these associations were causally mediated by reduced expression of 7 genes related to IgE-mediated hypersensitivity and mast cell function (FDR≤0.05, Figure 5). Interestingly, we did not find that associations between DEGs and pet sensitization were mediated by Corynebacterium sp. or Staphylococcus epidermidis.

Figure 5: Causal mediation analysis for Corynebacterium sp., Staphylococcus epidermidis, DEGs, and pet sensitization.

Associations between: (1) nasal species and pet sensitization are shown in blue; (2) nasal species and DEGs are shown in purple; and (3) DEGs and pet sensitization are shown in pink. Positive (+) and negative (−) associations are as indicated. Dashed arrows indicate mediation of the association between nasal species and pet sensitization by gene expression. FDR values from the causal mediation analysis are shown. Results are shown for Corynebacterium sp. (panels A,B) and Staphylococcus epidermidis (C).

Conclusions

To our knowledge, this is the first study to jointly examine the nasal microbiome, nasal transcriptome, and their relationship to pet allergen sensitization in children with asthma. The limitations of this study include its observational, cross-sectional design and the possibility of residual confounding. We found that the nasal microbiome and nasal transcriptome of children with asthma vary by pet sensitization. Higher abundance of nasal Corynebacterium sp. and Staphyloccocus epidermidis in particular were associated with the absence of pet sensitization and correlated with lower host expression of inflammatory genes, supporting that these nasal species are associated with allergy health. Causal mediation analysis identified several DEGs that mediate associations between Corynebacterium sp., S. epidermidis, and pet sensitization. Our study provides novel insight into airway host-microbe relationships in pet allergen sensitization.

Key Messages.

• Cat sensitization is associated with lower nasal microbial diversity and differential nasal bacterial composition.

• Nasal abundances of Corynebacterium sp. and Staphylococcus epidermidis are associated with absence of cat and/or dog sensitization

• These protective associations are causally mediated by reduced nasal expression of genes related to IgE-hypersensitivity and mast cell function.

Abbreviations:

DEG

differentially expressed gene

FDR

false discovery rate

ICS/LABA

inhaled corticosteroid/long-acting beta agonist

LDA

linear discriminant analysis

LEfSe

Linear discriminant effect size

PERMANOVA

permutational multivariate analysis of variance

sIgE

specific IgE

Treg

regulatory T cell