Nasal microbiome and phageome profiles are associated with prospective respiratory viral infection risk in school-aged children

Abstract

Background.

Respiratory viral infections are common and can trigger asthma exacerbations in children. The roles of the nasal microbiome and phageome (viruses that infect microbes) are not well understood.

Objective.

To characterize the epidemiology of respiratory viral infections and the interplay between the nasal microbiome, phageome, and viral infections in school-aged children with asthma.

Methods.

We performed metagenomic sequencing and RT-qPCR detection of respiratory viruses on 375 nasal samples from 227 school-age children with asthma collected routinely three times over one year. Surveys on parent-reported cold and asthma symptoms were administered routinely every two months. We evaluated multi-kingdom changes to the nasal microbiome during infection. A sPLS-DA model identified microbial signatures associated with prospective viral infection risk.

Results.

Respiratory viruses were identified in 124 (33%) samples, with rhinovirus most prevalent. Cold and asthma symptoms within the prior 14 days had a sensitivity of 79% and 59%, respectively, for RT-qPCR-confirmed infection. Respiratory viral infection increased asthma symptoms and was accompanied by loss of nasal bacterial diversity and a reproducible bloom of pathobionts with no change in the mycobiome or phageome. A baseline bacteriome-dominated profile was protective (adjusted OR 0.41 [95% CI, 0.25 – 0.67]; P < 0.001), whereas phageome profiles increased risk (adjusted OR 3.74 [1.85 – 7.55]; P < 0.001) of viral infection. Specific phages inversely correlated with Staphylococcus epidermidis abundance, the most protective commensal against infection risk.

Conclusion.

The nasal microbiome and phageome exert opposing influences on respiratory viral infection risk, highlighting their potential roles in modulating susceptibility to viral infections.

Capsule summary

Distinct profiles of the nasal microbiome and phageome influence susceptibility to respiratory viral infections in school-aged children with asthma, highlighting potential targets for prevention of virus-induced symptoms including asthma exacerbations.

INTRODUCTION

Acute viral respiratory illnesses in school-aged children are common and carry large societal costs. Viral upper respiratory infections are a major cause of school absenteeism in school-aged children¹. For non-influenza related viral infections, the burden remains high, translating to 189 million missed school days annually, 126 million caregiver workdays lost due to childcare needs, and an estimated $14.5 billion/year societal cost². Viral illnesses also carry medical risk in this population, with viruses being identified in 63% of children presenting to the emergency department and in 55% of children admitted for community acquired pneumonia^3,4. Risk is especially high in school-aged children with asthma, with viral illnesses representing the most common cause of exacerbations⁵. Despite their prevalence and risk, interventions to reduce infections in school-aged children—including hand hygiene, disinfectant wipes, and masking—have had mixed results^6–8.

There is a clear interplay between the nasal microbiome, respiratory infections and allergic disease^9–11. But, to date much of the research in the respiratory microbiome has focused either on severity of viral infections or risk of respiratory infections in infants¹², whereas less is known regarding the impact of the nasal microbiome on prospective risk of viral infection in school age children. Viral infections themselves have been associated with alterations in the nasal microbiome in children, notably increases in the abundance of Streptococcus pneumoniae, Haemophilus influenzae, and Moraxella catarrhalis^13–15. Additionally, the nasal microbiome appears to impact illness course and severity in young children with viral infections^16–19. Some data suggests that components of the nasal microbiome may influence risk of viral infections in children^20–22. However, to date the majority of respiratory microbiome studies have utilized amplicon sequencing of the 16S rRNA gene which is unable to differentiate between bacteria at a species level (e.g. Staphylococcus epidermidis versus Staphylococcus aureus) and is solely focused on the bacteriome.

There is growing appreciation that viruses impact the incidence and severity of asthma. While much of the published literature has focused on eukaryotic viruses such as rhinovirus, influenza, or respiratory syncytial virus that infect humans, much less in known about the role of phages, which consist largely of prokaryotic viruses. The human microbiome is the community of microbes that live on and inside humans, and existing research has largely focused on bacteria and to a lesser extent archaea and fungi. The phageome is the matching community of viruses that specifically infect the human microbiome. These phages constantly attack, infect, kill, and sometimes re-program host-associated microbes such as bacteria, and act as a major force shaping which bacterial species thrive in places like the gut, skin, and airways^23,24. Because phages can also carry genes such as antibiotic resistance or virulence factors, shifts in the phageome may either promote or restrain harmful bacteria, indirectly influencing infection risk, inflammation, and treatment responses^25,26. In infants, phage signatures in stool, beyond bacterial signatures, have been shown to be independent predictors of developing asthma^27,28. In adults, the presence of certain eukaryotic viruses such as cytomegalovirus and Epstein-Barr virus in the airway has been associated with increased asthma severity, while the presence of Streptococcus phages appears protective²⁹. In asthma, the microbiome contains important contributions from the bacteriome, archaeome, mycobiome, as well as both the eukaryotic and prokaryotic virome, yet the virome and in particular the phageome has been relatively understudied, with few studies in school-aged children.

Better understanding of the relationship between all microbial kingdoms that comprise the nasal microbiome in school-aged children has been limited by the known high human DNA content of respiratory samples, severely limiting effective sequencing depth with metagenomics approaches leading to sequencing failures due to few or no microbial reads³⁰. We recently demonstrated that effective human DNA depletion can be performed without significant bias to nasal microbial communities, allowing for comprehensive characterization of bacterial, archaeal, fungal, and phage communities³¹. Here we describe the epidemiology of respiratory viral infections in school-aged children, the impact of viral infections on multi-kingdom nasal microbial communities, and the influence of the nasal microbiome and phageome on risk of prospective viral infection.

METHODS

Study Population and Study Procedures

This was an ancillary study based on the School Inner-City Asthma Intervention Study (SICAS-2: ClinicalTrials.gov; NCT02291302), a cluster-randomized, placebo-controlled trial of the effects of classroom-level high efficiency particulate air (HEPA) purifiers and school-level integrated pest management on asthma morbidity. Details of the study design and primary trial outcome have previously been published³². This study was approved by the Boston Children’s Hospital Institutional Review Board.

Briefly, between 2015 – 2020, school-aged children with active physician-diagnosed asthma and without major comorbid chronic disease were enrolled in the summer and followed for one academic school year. At the baseline screening and enrollment visit, demographics, symptoms, medication use, allergy skin testing to 14 common aeroallergens (ALK-Abelló) using the MultiTest II device (Lincoln Diagnostics), and spirometry was performed. Symptom surveys assessed (1) the presence of symptoms consistent with an upper respiratory tract infection (URI; “cold” symptoms) and (2) the number of days with asthma symptoms within the preceding two weeks. Surveys were administered at baseline and then by phone at pre-specified, routine intervals (every two months) throughout the school year. Nasal samples were collected longitudinally at three pre-specified time points per participant: once in the research clinic at the baseline summer screening visit, and twice at school—once in the early fall shortly after the start of the school year and once in late spring (Supplemental Figure 1). Neither the symptom surveys nor nasal sampling was triggered by acute symptoms; both followed a fixed schedule defined a priori in the study protocol. Samples included both an anterior nasal swab and a nasal blow sample³³. All samples were placed immediately on ice and frozen at −80°C within 4 hours of sample collection.

RT-qPCR for detection of respiratory viral infection in nasal blow samples

Each nasal blow sample underwent viral nucleic acid extraction and multiplexed RT-qPCR, with a Ct value of <35 considered positive. We assessed for a total of 19 common respiratory viruses, including adenovirus, coronavirus (HKU1, NL63, OC43), enterovirus (including D68), respiratory syncytial virus (RSV) A and B, influenza A, influenza A (H1N1), influenza A (H3N2), influenza B, parainfluenza 1, 2, 3, and 4, human metapneumovirus (hMPV), parechovirus, and rhinovirus. SARS-CoV-2 was not detected in nasal samples (all nasal samples collected prior to 2020).

Metagenomics sequencing of nasal swabs for microbiome characterization

Host DNA depletion was first performed as previously described³¹ prior to library preparation and sequencing on the Illumina NovaSeq platform targeting 10 Gb/sample. Profiling of metagenomes was processed with bioBakery 3 combined with bowtie2 with the hg38 reference database for mapping and removal of human reads^34,35. Taxonomic profiling was performed for non-viral communities using MetaPhlAn 3.0 while viral (phage) communities were profiled with Marker-MAGu v1.1³⁶. Community profiles, including abundance, taxonomic classification, and sample-related metadata were merged using phyloseq R package v1.48.0³⁷. A prevalence filter of less than 5 samples was applied to the raw data, and species identified as potential contaminants³⁸ were removed prior to downstream statistical analyses³⁹.

Statistical Analyses

Participant characteristics were summarized, overall and stratified by viral infection status. The performance of asthma symptom or reported colds as screening tools for viral detection was measured by estimating the sensitivity and specificity. To assess the impact of viral infection on asthma symptoms, we first ran a negative binomial mixed-effects model using glmmTMB R package v1.1.12, with the number of asthma symptoms within 14 days as the outcome⁴⁰. We then ran a model with any asthma symptom as a binary outcome using lme4 R package v1.1–36⁴¹.

Stacked bar plots were constructed to provide a preliminary understanding of the microbiome composition stratified by kingdom (Bacteria, Phages, and Fungi). Alpha and beta diversity indices were computed using the phyloseq R package and the vegan R package v2.7–1^37,42. Permutational multivariate analysis of variance (PERMANOVA) was performed with 10,000 permutations.

Differential abundance analysis was conducted using the lme4 R package⁴¹. To assess the effect of viral infection on nasal microbial communities within each kingdom, linear mixed-effects models were used with centered log ratio transformed relative abundance as the outcome and viral infection as the binary predictor. Volcano plots were generated for the results of differential abundance analyses.

To identify which group of microbial/phage features best predicts future viral infection risk, the sparse Partial Least Squares Discriminant Analysis (sPLS-DA) method was implemented for feature selection using the mixOmics R package v6.28.0. Microbes or phages were tuned from 10 to 30 for 5-fold 50 repeated cross-validation⁴³. The optimal number of features was selected based on area under the ROC curve (AUC).

To determine the impact of microbiome component and phage component on future risk of viral infection, we ran mixed-effects models using the sPLS-DA components as the predictors, and viral infection ascertained by RT-qPCR as the binary outcome. To evaluate the individual effect of each selected microbe or phage on future viral infection, univariate and multivariate generalized linear mixed-effects models were performed for the binary viral infection with presence/absent of the signature as a predictor. Spearman’s correlation between each selected microbe and phage was depicted using the ComplexHeatmap R package v2.21.2⁴⁴.

To address missing data in covariates, a K-nearest neighbors (KNN) algorithm was applied to impute missing values using the VIM R package v6.2.2 with k = 5⁴⁵. This approach included baseline demographics, socioeconomic status, home zip code, and school.

All multivariate analyses were conducted with covariate adjustment including age, male, race and ethnicity, annual household income over $25,000, outcome measurement during the school year, and for symptom outcomes included allergic sensitization. Models examining the effect of microbiome and phage signatures on future viral infection additionally adjusted for HEPA intervention status, as this cohort was derived from a clinical trial (ClinicalTrials.gov NCT02291302).

Two-sided p-values less then 0.05 were considered significant. Multiple testing correction was performed using R package qvalue v2.36.0, and a q-value less than 0.1 was considered significant. All analyses were conducted using R version 4.4.1⁴⁶. Additional details are provided in the Online Supplement. Sequencing data is available through the NCBI Sequence Read Archive under Bioproject PRJNA1255658.

RESULTS

A total of 227 school-aged children with asthma (mean age 8.63 ± 1.98 years; 52% male; 17% non-Hispanic White, Table 1) contributed 375 nasal-blow specimens (220 summer, 58 fall, 97 spring, see Supplemental Figure 1) over the study period (Figure 1 for Study Flow Diagram). Respiratory viruses were detected in 124 samples (33%) from 104 participants (46%). Viral infections were more common in younger students (8.17 ± 1.93 vs 8.81 ± 1.98 years, P = 0.026) and those with more poorly controlled asthma (mean asthma symptom days 3.04 ± 3.97 vs. 1.72 ± 3.33 days, P = 0.010; ACT score > 19 indicating well controlled asthma, 84.9% vs. 68.6%, P = 0.009, comparing those without vs. with viral infections).

Table 1.

Characteristics of study participants overall and stratified by viral infection status ascertained by RT-qPCR testing of nasal samples.

	Overall	Virus not detected	Virus detected	P-value
N	227	152	70
Age, years (mean (SD))	8.63 (1.98)	8.81 (1.98)	8.17 (1.93)	0.026
Male (%)	118 (52.0)	75 (49.3)	40 (57.1)	0.349
Non-Hispanic White (%) 1	38 (16.7)	30 (19.7)	7 (10.0)	0.086
Race (%) 2				0.363
Asian	6 (2.6)	5 (3.3)	1 (1.4)
Black	55 (24.2)	33 (21.7)	22 (31.4)
American Indian or Alaska Native	1 (0.4)	1 (0.7)	0 (0.0)
White	96 (42.3)	66 (43.4)	28 (40.0)
Other	17 (7.5)	9 (5.9)	7 (10.0)
Hispanic (%) 3	126 (55.5)	82 (53.9)	42 (60.0)	0.390
Income over $25,000 (%) 4	112 (49.3)	75 (49.3)	33 (47.1)	0.719
Grade (%) 5				0.479
K	18 (7.9)	11 (7.2)	7 (10.0)
1	30 (13.2)	15 (9.9)	15 (21.4)
2	46 (20.3)	32 (21.1)	13 (18.6)
3	42 (18.5)	27 (17.8)	14 (20.0)
4	35 (15.4)	25 (16.4)	8 (11.4)
5	33 (14.5)	24 (15.8)	9 (12.9)
6	10 (4.4)	8 (5.3)	2 (2.9)
7	6 (2.6)	5 (3.3)	1 (1.4)
8	6 (2.6)	4 (2.6)	1 (1.4)
Allergic sensitization (%) 6	135 (59.5)	92 (60.5)	41 (58.6)	0.900
Asthma symptom days in the past two weeks (mean (SD))	2.10 (3.56)	1.72 (3.33)	3.04 (3.97)	0.010
ACT score (mean (SD)) 7	22.39 (2.90)	22.76 (2.73)	21.64 (3.20)	0.009
ACT score > 19 (%) 7	181 (79.7)	129 (84.9)	48 (68.6)	0.015
Steroid use within 12 months (%)	3 (1.3)	2 (1.3)	1 (1.4)	1.000
Antibiotic use within 12 months (%)	8 (3.5)	5 (3.3)	2 (2.9)	1.000
Daily controller (%)	126 (55.5)	84 (55.3)	39 (55.7)	1.000
FeNO (mean (SD))	20.63 (19.33)	21.45 (19.00)	19.69 (20.71)	0.569
FEV 1 /FVC ratio (mean (SD))	84.97 (7.94)	85.28 (7.85)	84.48 (8.28)	0.501
% predicted FEV1 (mean (SD))	99.04 (17.50)	98.23 (17.97)	101.15 (16.75)	0.264
% predicted FVC (mean (SD))	98.77 (15.45)	97.57 (15.23)	101.42 (15.94)	0.095

Values presented as mean (standard deviation) for continuous variables and number (%) for categorical variables. Two-tailed p-values were estimated using two-sample t-test for continuous variables and Pearson chi-square test for categorical variables. Viral infection status was unavailable for 5 participants though they contributed nasal samples for microbiome analysis.

Abbreviations: ACT = Asthma Control Test. FeNO = fractional exhaled nitric oxide. FEV₁ = forced expiratory volume in the first second, a measure of lung function. FVC = forced vital capacity, a measure of lung function.

¹There were 4 (1.8%) missing values in non-Hispanic White.

²There were 52 (22.9%) unknown or refused to say in race.

³There were 4 (1.8%) missing values in Latino.

⁴There was 74 (32.6%) missing values in household income, of which 56 (24.7%) caretakers responded that they did not know their household income, and 17 (7.5%) declined to answer. Income threshold of $25,000 chosen as it is the state poverty level.

⁵There was 1 (0.4%) missing value in grade.

⁶There were 5 (2.2%) missing values. Allergic sensitization to any aeroallergen determined through allergy skin testing to 14 common aeroallergens (ALK-Abelló) using the MultiTest II device (Lincoln Diagnostics).

⁷Higher ACT scores (> 19) indicate good control. There were 8 (3.5%) missing values in ACT scores.

Figure 1. Flow diagram of study population.

SICAS-2 cohort and nasal sampling flow. Boxes show numbers of participants and nasal samples at each stage. Per protocol, nasal samples were collected routinely three times a year (once during the baseline screening visit in the summer, and twice during the school year in the fall and spring/summer). Nasal samples collected during the school year were collected during the school day when the teacher briefly excused the child from class; if the child was absent from school that day, or study procedures could not be completed, then there was missingness in nasal biospecimen collection. Final analysis sets: (i) cross-sectional infection status and respiratory symptoms (n = 218 children), (ii) cross-sectional infection status and nasal microbiome (n = 224 children), and (iii) nasal microbiome and prospective nasal blow samples to verify infection by RT-qPCR (n = 139 children). Abbreviations: RT-qPCR = reverse transcription quantitative polymerase chain reaction. SICAS-2 = School Inner City Asthma Intervention Study cohort.

Ten distinct respiratory viruses were identified (Table 2). Rhinovirus predominated with 108 (28.8%) nasal samples testing positive. Sixteen samples had more than one virus detected. The median [interquartile range] number of days between nasal biospecimen collection and matched symptom survey was 0 [0 – 0] days in our study; respiratory symptom surveys focused on symptoms in the preceding two weeks. Parent/caregiver report of a “cold” identified children with a positive viral test with high sensitivity (79%) but low specificity (19%). In contrast, reported asthma symptoms had modest ability to discriminate viral positivity (sensitivity 59%, specificity 60%). For parent-reported “cold,” the positive predictive value was 32% and the negative predictive value was 66%. For asthma symptoms, the positive predictive value was 42% and the negative predictive value was 75%.

Table 2. Prevalence of respiratory viruses detected in nasal samples.

We tested 375 nasal samples for common respiratory viruses using reverse transcription quantitative polymerase chain reaction (RT-qPCR). Viruses are listed from most to least frequently detected. For each virus, we also report the proportion of virus-positive samples for which a parent/caregiver reported (1) upper respiratory infection symptoms or (2) asthma symptoms occurring within 14 days before or after the nasal sample was collected.

	Positive Test N (%)	URI Symptoms N (%)	Asthma Symptoms N (%)
Any virus 1	124 (33.1)	81 (65.3)	67 (54.0)
Rhinovirus 2, 3, 5, 6, 8, 9	108 (28.8)	73 (67.6)	58 (53.7)
Adenovirus 2, 7, 8	16 (4.3)	12 (75.0)	9 (56.2)
Enterovirus (pan-assay) 3	5 (1.3)	4 (80.0)	3 (60.0)
Metapneumovirus 9	3 (0.8)	1 (33.3)	3 (100.0)
Influenza A (pan-assay) 4,9	2 (0.5)	1 (50.0)	2 (100) .0
Parainfluenza 1 6	2 (0.5)	1 (50.0)	0 (0.0)
Bocavirus 7	1 (0.3)	1 (100.0)	0 (0.0)
Coronavirus OC43	1 (0.3)	0 (0.0)	1 (100.0)
Enterovirus (D68) 5	1 (0.3)	0 (0.0)	1 (100.0)
Influenza A (H1N1) 4	1 (0.3)	0 (0.0)	1 (100.0)
Influenza A (H3N2) 9	1 (0.3)	1 (100.0)	1 (100.0)
Influenza B (pan-assay) 8	1 (0.3)	0 (0.0)	1 (100.0)
Parainfluenza 3	1 (0.3)	0 (0.0)	1 (100.0)
RSV A 7	1 (0.3)	1 (100.0)	0 (0.0)

¹16 samples from 14 children had more than one virus detected.

²Rhinovirus and Adenovirus were found together in 5 samples.

³Rhinovirus and Enterovirus (pan-assay) were found together in 5 samples.

⁴Influenza A (pan-assay) and Influenza A (H1N1) were found together in 1 sample.

⁵Rhinovirus and Enterovirus (D68) were found together in 1 sample.

⁶Rhinovirus and Parainfluenza 1 were found together in 1 sample.

⁷Adenovirus, Bocavirus, and RSV A were found together in 1 sample.

⁸Rhinovirus, Adenovirus and Influenza B (pan-assay) were found together in 1 sample.

⁹Rhinovirus, Metapneumovirus, Influenza A (pan-assay) and Influenza A (H3N2) were found together in 1 sample.

Abbreviations: URI = upper respiratory infection.

Mixed effects model revealed that contemporaneous viral infection was associated with greater asthma burden. Viral positivity corresponded to a 117% increase in asthma symptom rate (IRR [95% CI] = 2.17 [1.31 – 3.58]; P = 0.003). Odds of any asthma symptoms were 2.89-fold higher among virus positive children (95% CI = 1.48 – 5.64; P = 0.002) after adjustment for covariates.

A total of 360 nasal samples (along with 15 negative reagent-only and 15 positive mock community samples) underwent metagenomics sequencing. The average sequencing depth was 93.46 million reads per sample. The most abundant species, stratified by kingdom, are depicted in Figure 2. Microbial species richness was highest in phages (881 unique species), followed by bacteria (572 unique species) and fungi (9 unique species). No archaea were detected. Negative reagent-only controls were used to assess for the possibility of contamination. Three out of the 15 negative control samples failed sequencing with no detected microbial reads. The most prevalent and abundant microbes detected in the negative controls included Cupriavidus spp., Cutibacterium acnes, and Corynebacterium pseudodiphtheriticum (Supplemental Figure 2). Statistical models³⁸ identified 15 bacterial species and 8 phage species as potential contaminants (Supplemental Table 1); these were removed from downstream analyses.

Figure 2. Species-level composition of the nasal microbiome in school age children, stratified by microbial kingdom.

Each stacked bar along the x-axis represents an individual participant’s nasal microbiome, and the y-axis shows the relative abundance of detected microbes. Within each kingdom, species are displayed in order of relative abundance (ranging from 0% to 100%) for that participant, with microbial taxa displayed in the legend in order of greatest to least abundance. No archaeal species were detected. Bacterial communities were highly diverse, although many participants showed dominance by a single species. The most abundant bacterial taxa were Dolosigranulum pigrum, Corynebacterium pseudodiphtheriticum, and Corynebacterium accolens. Fungal communities were less diverse, with Malassezia restricta and Malassezia globosa predominating. Phage communities were more variable; phage taxa are labeled by viral species genome bin (vSGB) identifier and annotated in parentheses with their predicted microbial host at the genus level to provide biological context.

Concurrent viral infection was accompanied by significant shifts in the nasal bacteriome, but not the mycobiome or phageome. Bacteria alpha-diversity analyses demonstrated lower species richness (−9.96 [−16.07 – −3.85], P = 0.001) and Shannon diversity (−0.15 [−0.29 – −0.01], P = 0.036) among virus-positive samples. Bacterial community structure, assessed by beta-diversity distances, differed by infection status using both Bray-Curtis (PERMANOVA R² 0.010, P < 0.001) and Horn–Morisita distances (PERMANOVA R² 0.012, P = 0.001). 16 bacterial species were differentially abundant with viral infection (Figure 3; Supplemental Table 2); 11 species from the Proteobacteria, Actinobacteria, and Firmicutes phyla increased in relative abundance including both commensals such as Dolosigranulum pigrum (CLR transformed relative abundance 2.16 [0.60 – 3.73], P = 0.007) and known pathobionts such as Haemophilus influenzae (1.07 [0.44 – 1.70], P = 0.001), Moraxella catarrhalis (1.13 [0.35 – 1.92], P = 0.005), and Streptococcus pneumoniae (0.86 [0.22 – 1.49], P = 0.008). Respiratory viral infection also increased the odds of detecting the presence of Haemophilus influenzae (OR [95% CI] = 3.64 [1.49 – 8.86]; P = 0.005; q = 0.014) and Moraxella catarrhalis (OR [95% CI] = 2.20 [1.06 – 4.59]; P = 0.035; q = 0.052) but not Streptococcus pneumoniae (OR [95% CI] = 0.98 [0.22 – 4.34], P = 0.976) in the nares. Bacterial species that decreased in relative abundance during viral infection include 5 from the Proteobacteria, Actinobacteria and Firmicutes phyla.

Figure 3. Change in the differential abundance of nasal bacterial species during respiratory viral infection.

In this volcano plot the x-axis represents the centered log-ratio (CLR) change of species relative abundance in nasal samples with respiratory viral infection. Positive values on the x-axis represent microbial species with increased relative abundance in the context of viral infection, whereas negative values represent decreased relative abundance. The y-axis represents the statistical significance of that change in relative abundance. Bacteria above the horizontal line represent those with a statistically significant change using a false discovery rate threshold of q <0.1. Bacterial species with increased abundance in the presence of a respiratory virus infection include known pathobionts in the Proteobacteria and Firmicutes phyla such as Haemophilus influenzae, Moraxella catarrhalis, and Streptococcus pneumoniae, as well as commensals such as Dolosigranulum pigrum. There were no differences in the abundance of fungi and phages with viral infection (all q-values > 0.1 after adjustment for multiple testing).

Baseline nasal microbiome and phage profiles were strongly predictive of prospective viral infection. Using an sPLS-DA model, we identified separate microbiome and phageome signatures for prospective risk of viral infection. Model performance was good, yielding a cross-validated AUC of 0.79 for the microbiome signature and 0.80 for the phage signature for predicting viral infection. In multivariable mixed-effects logistic regression, simultaneously adjusting for both signatures as well as covariates, the primary microbiome component was protective (adjusted OR [95% CI] = 0.41 [0.25 – 0.67]; P < 0.001), whereas the primary phage component conferred increased risk (adjusted OR [95% CI] = 3.74 [1.85 – 7.55]; P < 0.001) of infection (Table 3). Within the primary microbiome component, Staphylococcus epidermidis, Streptococcus cristatus, and Actinomyces hongkongensis contributed most to protection (Figure 4A). At the individual-species level after covariate adjustment, nasal presence of Staphylococcus epidermidis (OR [95% CI] = 0.20 [0.06 – 0.65]; P = 0.008) and Streptococcus cristatus (OR [95% CI] = 0.28 [0.09 – 0.89]; P = 0.032) conferred protection against viral infection (Supplemental Table 3). Phages with the predicted bacterial hosts Dolosigranulum, Haemophilus, Moraxella, and Staphylocccus contributed most to the signature conferring increased risk (Figure 4B). After adjusting for covariates, nasal presence of two individual phages predicted increased risk of viral infection, vSGB22521 (predicted host Dolosigranulum; OR [95% CI] = 3.48 [1.16 – 10.46]; P = 0.026) and vSGB5757 (predicted host Moraxella; OR [95% CI] = 6.60 [1.16 – 37.34]; P = 0.033). In exploratory analyses, we evaluated a correlation heatmap (Figure 5) to better understand how constituents of the phage and microbiome signatures interacted. vSGB22521 and vSGB22520, both associated with increased risk of viral infection, were correlated with decreased abundance of Staphylococcus epidermidis and Malassezia restricta while correlated with increases in Streptococcus pneumoniae and Haemophilus influenzae, which are both associated with increased viral infection risk. vSGB589, with Staphylococcus genus as the predicted host, was protective against viral infection, positively correlated with Staphylococcus epidermidis abundance, and negatively correlated with Streptococcus pneumoniae and Haemophilus influenzae abundance. There were correlations between phages and bacteria outside of the predicted host, suggesting either broader host range than predicted or ecological interactions between host bacteria after phage predation.

Table 3.

Univariate and multivariate effects of nasal microbiome and phageome signatures on prospective respiratory viral infection risk.

	Univariate1			Multivariate2
	OR	95%CI	p-value3	OR	95%CI	p-value3
Microbiome signature	0.35	[0.23, 0.55]	<0.001	0.41	[0.25, 0.67]	<0.001
Phage signature	3.94	[2.19, 7.10]	<0.001	3.74	[1.85, 7.55]	<0.001
Age, years	0.86	[0.71, 1.04]	0.112	1.14	[0.85, 1.53]	0.365
Male sex	2.33	[1.07, 5.10]	0.034	3.02	[0.92, 9.91]	0.069
Non-Hispanic White Race/Ethnicity	2.57	[0.93, 7.14]	0.070	3.46	[0.76, 15.72]	0.109
Annual Household Income over $25,000	1.37	[0.61, 3.07]	0.446	0.60	[0.16, 2.21]	0.440
HEPA intervention arm	0.87	[0.42, 1.79]	0.700	0.91	[0.30, 2.72]	0.862
Observation During School Year	0.97	[0.40, 2.35]	0.952	1.16	[0.28, 4.79]	0.834

¹Results from univariate generalized linear mixed-effects models assessing one variable at a time.

²Results from multivariate generalized linear mixed-effects models controlling for other covariates.

³Two-tailed p-value for odds ratio (OR).

Abbreviations: HEPA = High Efficiency Particulate Air

Figure 4. Microbiome and phageome signatures associated with prospective risk of respiratory viral infections.

Loading plots from the sPLS-DA applied to the microbiome component (A) and the phageome component (B) to discriminate risk of future viral infection status. Colors show the viral infection status with the highest mean abundance within each feature (microbial species). The directionality (positive vs. negative) on the x-axis reflects the direction of between-group separation along the latent component while the absolute magnitude indicates the contribution to increased risk (orange) or decreased risk (blue).

Figure 5. Correlation heatmap of species present in microbiome and phageome signatures for prospective risk of respiratory viral infection.

Heatmap shows pairwise Spearman correlations between species present in the phage-based signature (rows) and the bacterial microbiome signature (columns), with hierarchical clustering applied to both dimensions. Microbes are labeled as either associated with protective or increased risk for prospective respiratory viral infection. Each phage’s predicted bacterial host genus (shown in parentheses) included for biological context. Phages vSGB22521 and vSGB22520 were two influential contributors to the phage signature and were associated with increased risk of viral infection. These phages were correlated with decreased abundance of Staphylococcus epidermidis and Malassezia restricta (both associated with protection against viral infection) and increases in Streptococcus pneumoniae and Haemophilus influenzae (both associated with increased risk of viral infection). Phage vSGB589 was protective against viral infection, positively correlated with Staphylococcus epidermidis abundance, and negatively correlated with Haemophilus influenzae and Streptococcus pneumoniae abundance. These demonstrate the complex interactions between the nasal phageome, nasal microbiome, and respiratory viral infection risk.

DISCUSSION

In a prospective study of school-aged children with asthma we demonstrate that respiratory viral infection increased asthma symptoms and is accompanied by loss of nasal bacterial diversity and a reproducible bloom of pathobionts. A nasal microbiome signature dominated by commensal bacteria was shown to be protective against viral infection, whereas a nasal phageome signature independently conferred increased risk of viral infection.

In school-aged children with asthma, exposure to respiratory viruses is common. In our cohort, nearly a third of the nasal samples contained a respiratory viral pathogen with rhinovirus being the most common. Most existing studies on respiratory viral infections in school age children focus on those presenting for medical care and therefore with severe illness; our samples were collected routinely irrespective of reported symptoms. Other studies in infants with routine nasal sample collection have shown that rhinovirus is the most common infection⁴⁷. We have recently shown that on average of 3 (range 0–13) viruses are detected per classroom bioaerosol sample from this cohort, with rhinovirus being the most common and detected in 89.5% of classroom samples⁴⁸. While viral infections are a well-known cause of asthma exacerbations requiring care^5,49, our work expands on this, showing that molecularly confirmed viral infections correlate strongly with reported worsened asthma symptoms in the community setting.

We highlight the sensitivity of cold symptoms as a screening tool for viral infections in school age children with asthma. A population-based study that recruited children attending daycare along with their siblings and parents, students and teachers from a high school, and physicians with weekly collection of nasal swabs and multiplexed detection of respiratory viruses found that 17.5% of samples tested were positive for a virus. In that study between 69% to 74% of all laboratory confirmed infections were asymptomatic; children were less likely to have symptoms compared to adults⁵⁰. In our cohort more nasal samples were found to be positive for viral detection (33%) and cold symptoms were more common (79% symptomatic). These differences could be related to the school- age profile of the children in our cohort, the timing of samples and symptom screening questionnaires, or the fact that our cohort included only children with asthma. In a study where adults were experimentally infected with rhinovirus-16, adults with asthma were not more likely to be symptomatic with infection than healthy controls though baseline sputum eosinophils were correlated with increased symptoms⁵¹. This question has been less studied in children, and it remains unclear if children with asthma are more likely to experience symptoms during viral infection than those without.

We observed decreased bacterial diversity during viral infection, consistent with prior reports^17,52–54, although we build on prior work by simultaneously interrogating the nasal mycobiome and phageome. Work in younger children has shown a direct relationship between nasal microbiome alpha diversity and time to next self-reported viral infection, implying increased diversity may be protective⁵⁵. Yet, given the dynamic nature of the nasal microbiome, it remains challenging to know if decreased diversity and richness contributed to risk of infection or are the result of it⁵⁶. More convincing is the expansion of pathobionts during viral infections, specifically S. pneumoniae, H. influenzae, and M. catarrhalis which are common causes of secondary bacterial pneumonias after viral infection^13–15. Haemophilus and Moraxella have also been shown to upregulate markers of T_H2 inflammation thus providing insight into their potential role in increased asthma symptom burden⁵². Mechanistically, viruses alter the respiratory epithelium making bacterial infection more likely⁵⁷. Influenza infection impairs respiratory ciliary function, leading to decreased mucociliary clearance and increased respiratory bacterial density⁵⁸. Rhinoviruses, the most prevalent viruses in our study, have been shown to alter the binding of S. aureus, S. pneumoniae, and H. influenzae to nasal epithelial cells⁵⁹, perhaps in part explaining the change in bacteriome profiles after viral infection.

Beyond the change in the nasal microbiome after infection, we demonstrated that a baseline microbiome signature prevalent with common commensals such as Staphylococcus epidermidis strongly protect against subsequent development of viral infection. These findings align with mechanistic studies showing that Staphylococcus epidermidis limits influenza via Embp-mediated binding and augments type-III interferon responses^60,61. While nasal commensal augmentation has been proposed for microbes such as Dolosigranulum pigrum, Staphylococcus epidermidis has to date received little attention.

In a novel component of this study, we showed that a baseline phageome signature conferred increased risk of subsequent viral infection, independent of the protective nasal microbiome signature. In our study, we see that presence of the phages with large contributions to the phageome signature (vSGB22521 and vSGB22520) were negatively correlated with the abundance of protective commensals such as Staphylococcus epidermidis and Malassezia restricta and increase in abundance of pathogens Streptococcus pneumoniae and Haemophilus influenzae. The role of phages in relation to viral infections has been understudied. However, work in infants suggests that, independent of the bacteriome, the phageome predicts future development of asthma^27,28. Phages may directly or indirectly alter nasal bacterial commensal populations, the virulence of these bacteria, or directly impact the human most immune system^62–64. The exact interplay between phage, bacterial host and human host is a promising area of future investigation.

Interpretation of our results should take into consideration several strengths and limitations. This represents a relatively large sample of a comparatively understudied demographic, school-aged children with asthma. Another strength is our use of metagenomic rather than amplicon sequencing on human DNA depleted samples, providing species level taxonomic resolution and interrogation of multiple microbial kingdoms, including the phageome. Viral infections were verified with molecular testing of nasal biospecimens. However, there are a few limitations. Nasal samples were collected only three times during the year due to feasibility, and the timing of the nasal samples did not overlap strongly with peak viral season occurring in the winter, likely impacting the diversity and prevalence of observed viruses. Our symptom surveys asked about respiratory symptoms within a two week period; some children with reported cold symptoms may have had virus present earlier in the course of illness but undetectable at the time of our nasal sample collection. Specific cycle thresholds from PCR testing of biospecimens does not distinguish between active infection, residual genetic material from a previous, cleared or asymptomatic infection, or infectiousness although we use a cycle threshold of 35 to denote infection, similar to many clinical microbiology testing guidelines^65–69. Our study included children with asthma in one region of the United States and did not include healthy children without asthma, which could limit generalizability.

Our study demonstrates that in school-aged children with asthma, cold symptoms may offer utility as a screening test for respiratory viral infections. Viral infections strongly increase asthma symptom burden, with major changes to the nasal bacteriome including pathobiont expansion. Nasal microbiome and phageome signatures are strongly associated with susceptibility to these infections, though with opposing effects. Further research is warranted to elucidate the interactions between respiratory viruses, microbiota, and phages, which may ultimately inform targeted microbiome- or phage-based interventions to improve respiratory health in children with asthma.

Key Messages:

• Alterations in the nasal bacteriome—but not the mycobiome or phageome—accompany viral infections

• The nasal microbiome and phageome have opposing effects on the risk of respiratory viral infections in asthmatic children and could represent novel targets for preventing virus-induced symptoms including asthma exacerbations.

Abbreviations

RT-qPCR

Reverse Transcription quantitative Polymerase Chain Reaction

AUC

Area Under the receiver operating characteristic Curve

OR

Odds Ratio

IRR

Incidence Rate Ratio