key: cord-0738133-o1n5ylkx
authors: Hurst, Jillian H; McCumber, Alexander W; Aquino, Jhoanna N; Rodriguez, Javier; Heston, Sarah M; Lugo, Debra J; Rotta, Alexandre T; Turner, Nicholas A; Pfeiffer, Trevor S; Gurley, Thaddeus C; Moody, M Anthony; Denny, Thomas N; Rawls, John F; Clark, James S; Woods, Christopher W; Kelly, Matthew S
title: Age-related changes in the nasopharyngeal microbiome are associated with SARS-CoV-2 infection and symptoms among children, adolescents, and young adults
date: 2022-03-05
journal: Clin Infect Dis
DOI: 10.1093/cid/ciac184
sha: 00c2da5ddbb8a6f349b7582e991f9fcec7ac8544
doc_id: 738133
cord_uid: o1n5ylkx

BACKGROUND: Children are less susceptible to SARS-CoV-2 infection and typically have milder illness courses than adults, but the factors underlying these age-associated differences are not well understood. The upper respiratory microbiome undergoes substantial shifts during childhood and is increasingly recognized to influence host defense against respiratory pathogens. Thus, we sought to identify upper respiratory microbiome features associated with SARS-CoV-2 infection susceptibility and illness severity. METHODS: We collected clinical data and nasopharyngeal swabs from 285 children, adolescents, and young adults (<21 years of age) with documented SARS-CoV-2 exposure. We used 16S ribosomal RNA gene sequencing to characterize the nasopharyngeal microbiome and evaluated for age-adjusted associations between microbiome characteristics and SARS-CoV-2 infection status and respiratory symptoms. RESULTS: Nasopharyngeal microbiome composition varied with age (PERMANOVA, p<0.001, R (2)=0.06) and between SARS-CoV-2-infected individuals with and without respiratory symptoms (PERMANOVA, p=0.002, R (2)=0.009). SARS-CoV-2-infected participants with Corynebacterium/Dolosigranulum-dominant microbiome profiles were less likely to have respiratory symptoms than infected participants with other nasopharyngeal microbiome profiles (odds ratio: 0.38, 95% confidence interval: 0.18–0.81). Using generalized joint attributed modeling, we identified nine bacterial taxa associated with SARS-CoV-2 infection and six taxa that were differentially abundant among SARS-CoV-2-infected participants with respiratory symptoms; the magnitude of these associations was strongly influenced by age. CONCLUSIONS: We identified interactive relationships between age and specific nasopharyngeal microbiome features that are associated with SARS-CoV-2 infection susceptibility and symptoms in children, adolescents, and young adults. Our data suggest that the upper respiratory microbiome may be a mechanism by which age influences SARS-CoV-2 susceptibility and illness severity.

In contrast to most other respiratory viruses [1] , children appear to be less susceptible to infection with severe acute respiratory virus 2 (SARS-CoV-2), and typically have milder illness courses than adults. In a recent meta-analysis of 32 studies that included 41,640 children and adolescents and 268,945 adults, SARS-CoV-2 infection susceptibility was estimated to be 46% lower among children and adolescents relative to adults [2] . Further, a higher incidence of SARS-CoV-2 infection has been observed with increasing age, even among infants, children, and adolescents [3] . We previously demonstrated that up to one-third of SARS-CoV-2-infected children and adolescents are asymptomatic [4] , and the vast majority of children who develop symptoms report mild respiratory symptoms [4, 5] . Additionally, COVID-19 hospitalization rates and mortality among children are substantially lower than among adults of all ages [6] . These data suggest that changes in host biological or immunological factors that occur with age modify susceptibility to and severity of SARS-CoV-2 infection.

Given that the upper respiratory microbiome undergoes substantial shifts in early childhood [7, 8] , and is increasingly recognized to play a key role in the pathogenesis of respiratory virus infections [9, 10], we hypothesized that age-associated changes in the upper respiratory microbiome might contribute to differences in SARS-CoV-2 susceptibility and illness severity among children and adults.

In this study, we used 16S ribosomal RNA (rRNA) gene amplicon sequencing to characterize the nasopharyngeal microbiomes of 285 children, adolescents, and young adults with close contact with a SARS-CoV-2-infected individual and to identify microbiome features associated with SARS-CoV-2 infection and with the presence of respiratory symptoms among SARS-CoV-2-infected individuals.

A c c e p t e d M a n u s c r i p t

The Duke Biospecimens from RespirAtory Virus-Exposed Kids (BRAVE Kids) study is a prospective cohort study of children, adolescents, and young adults (<21 years of age) with confirmed SARS-CoV-2 infection or close contact with an individual with confirmed SARS-CoV-2 infection, as previously described [4] . Exposure, sociodemographic, and clinical data are collected at enrollment, and we record symptoms occurring up to 14 days prior to and 28 days after study enrollment.

Nasopharyngeal samples are collected with nylon flocked swabs (Copan Italia, Brescia, Italy) and placed into RNAProtect (Qiagen, Hilden, Germany) prior to storage at -80°C. Participants are classified as SARS-CoV-2-infected if the virus is detected in either a clinical or research PCR assay. For the analyses presented herein, we considered SARS-CoV-2-infected individuals to have respiratory symptoms if they reported cough, rhinorrhea, nasal congestion, shortness of breath, sore throat, or anosmia at any point between 14 days prior to enrollment through 28 days after enrollment.

The Duke Microbiome Core Facility extracted DNA from nasopharyngeal samples using Powersoil Pro extraction kits (Qiagen). DNA concentrations were determined using Qubit dsDNA high-sensitivity assay kits (Thermo Fisher Scientific). Bacterial community composition was characterized by PCR amplification of the V4 variable region of the 16S rRNA gene [11] . Equimolar 16S rRNA PCR products were quantified and pooled prior to sequencing. Sequencing was performed by the Duke Sequencing and Genomic Technologies Core Facility on an Illumina MiSeq instrument configured for 250 basepair paired-end sequencing. All samples were included in a single sample processing run with negative extraction and PCR controls. We analyzed raw sequences using DADA2 version 1.16 [12] A c c e p t e d M a n u s c r i p t and assigned taxonomy to amplicon sequence variants (ASVs) using the expanded Human Oral Microbiome Database version 15.1 [13] . We identified and removed presumed reagent contaminant ASVs (n=35; Supplemental Table 1 ) based on presence in negative control samples or negative correlation with DNA concentration using the frequency method (threshold=0.10) implemented in the decontam R package version 1.12 [14] . We excluded samples with less than 1,000 sequencing reads after quality filtering and contaminant removal. We obtained a median [interquartile range (IQR)] of 24,360 (18, 371 ) high-quality sequencing reads from the 285 samples included in these analyses. Sequencing reads were classified into 1,854 ASVs representing 202 bacterial genera from 8 phyla.

We calculated nasopharyngeal microbiome alpha diversity (Shannon diversity index and number of unique ASVs) using the phyloseq R package version 1.36 [15] . We fit linear regression models to evaluate associations between patient characteristics and microbiome alpha diversity measures. The number of unique ASVs was not normally distributed and was log-transformed for these analyses.

We used the microbiome R package version 1.8.0 [16] to generate centered log-ratio (CLR)transformed sample counts to evaluate between-sample compositional differences [17] . We used kmedoids clustering and the Calinski-Harabasz index to classify samples into distinct nasopharyngeal microbiome profiles. We evaluated associations between patient characteristics and nasopharyngeal microbiome composition with PERMANOVA using the adonis function within the vegan R package version 2.5.7 [18] . To evaluate associations between patient characteristics and the relative abundances of specific ASVs within the nasopharyngeal microbiome, we used generalized joint attribute modeling (GJAM) implemented in the gjam R package version 2.3.5 [19] . Analyses conducted in gjam were limited to ASVs present in at least 5% of samples. We adjusted for participant age (as a continuous variable) and assessed the significance of interaction terms in all A c c e p t e d M a n u s c r i p t analyses to evaluate for interactive relationships between age and the relative abundances of specific ASVs on SARS-CoV-2 infection and SARS-CoV-2-associated respiratory symptoms. Our findings in all analyses were not substantively changed when we additionally adjusted for sex and race (data not shown). All analyses were conducted in R version 4.1 [20] .

Two hundred eighty-five children, adolescents, and young adults were included in these analyses ( Table 1) . Participants were classified as SARS-CoV-2-exposed but uninfected (n=74, 26%); SARS-CoV- .051]. There were no significant differences in the prevalences of comorbidities or recent receipt of antibiotics or probiotics in these groups.

We first sought to describe changes in nasopharyngeal microbiome diversity that occur with age from infancy through early adulthood, and by SARS-CoV-2 infection and symptom status. Median 

Five bacterial genera accounted for more than 80% of the sequencing reads identified in nasopharyngeal samples: Corynebacterium (26%), Staphylococcus (21%), Moraxella (15%), Dolosigranulum (13%), and Streptococcus (5%). Nasopharyngeal microbiome composition varied with age (PERMANOVA, p<0.001, R 2 =0.06); specifically, increasing age was associated with decreases in the CLR-transformed abundances of the bacterial genera Moraxella (Spearman's rank correlation; ρ=-0.40, p<0.0001) and Dolosigranulum (ρ=-0.32, p<0.0001) and increases in the CLR-transformed abundances of Corynebacterium (ρ=0.24, p<0.0001) and Staphylococcus (ρ= 0.45, p<0.0001) ( Figure   2 ). Two bacterial genera -Lawsonella and Peptoniphilus -were highly prevalent in participants 12 years of age or older (78% and 69%, respectively), but were identified in only 21% and 19% of A c c e p t e d M a n u s c r i p t

To further characterize differences in nasopharyngeal microbiome composition by age and SARS-CoV-2 status, we used unsupervised clustering to classify nasopharyngeal samples into seven distinct microbiome profiles (Figure 3 Table 2; Kruskal-Wallis test, p<0.0001); however, there were no significant differences in other patient characteristics by microbiome profile. SARS-CoV-2 infection prevalence varied from 69% to 82% by microbiome profile, with the lowest prevalence observed among participants with a Moraxelladominant microbiome profile and the highest prevalence seen among participants with a Corynebacterium-dominant microbiome profile. However, in analyses adjusting for age, there were no significant associations between nasopharyngeal microbiome profile and SARS-CoV-2 infection.

Among SARS-CoV-2-infected participants, the prevalence of respiratory symptoms varied from 35% to 67% by nasopharyngeal microbiome profile, with the lowest prevalence seen among participants with a Corynebacterium/Dolosigranulum-dominant microbiome profile and the highest prevalence A c c e p t e d M a n u s c r i p t

We next used generalized joint attribute modeling (GJAM) to evaluate associations between specific

ASVs and SARS-CoV-2 infection and SARS-CoV-2-associated respiratory symptoms. GJAM allows for the concurrent evaluation of distinct types of data derived from observations of ecological systems,

where attributes of the system may be interdependent. Because we observed associations between age and nasopharyngeal microbiome composition, as well as associations between age and SARS-CoV-2 infection and respiratory symptoms, we used GJAM to separately evaluate associations between specific bacterial ASVs and SARS-CoV-2 infection and SARS-CoV-2-associated respiratory symptoms in the context of interactions of these variables with participant age. We identified nine

ASVs that were associated with SARS-CoV-2 infection ( Table 3) ; for eight of these ASVs, the magnitude of the association varied by participant age. For example, the relative abundance of ASV1163 (Corynebacterium propinquum) decreased with increasing participant age and was also higher among SARS-CoV-2-infected participants than uninfected participants independent of age.

However, the difference in the relative abundance of ASV1163 between SARS-CoV-2-infected and uninfected participants decreased with age, such that the negative association between the relative abundance of ASV1163 and SARS-CoV-2 infection was primarily observed among young children ( Figure 4A ). We next used GJAM to identify ASVs associated with respiratory symptoms among participants with confirmed SARS-CoV-2 infection. We identified six ASVs that were differentially abundant among SARS-CoV-2-infected participants with respiratory symptoms ( A c c e p t e d M a n u s c r i p t

We would like to thank the Duke University School of Medicine for use of the Microbiome Core Facility, which performed the DNA extractions and library preparations for this research, and the Duke Sequencing and Genomic Technologies Core Facility, which sequenced these libraries. We offer our sincere gratitude to the children and families who participated in this research.

The sequencing dataset supporting the conclusions of this study is available in the Sequence Read

Archive (PRJNA703574). The statistical files and script used for data analyses are also publicly available (https://github.com/alexmccumber/BRAVE_Kids). M a n u s c r i p t differences in the mean relative abundance of ASV1163 (Corynebacterium propinquum) among SARS-CoV-2-infected participants relative to uninfected participants in different age categories. The gold line was constructed using the GJAM estimates for the association of SARS-CoV-2 infection with the relative abundance of ASV1163 (intercept) and the association of the interaction term between SARS-CoV-2 infection and age with the relative abundance of ASV1163 (slope). Higher relative abundances of ASV1163 were observed in SARS-CoV-2-infected compared to uninfected participants across all ages but these differences were more pronounced in young children. B) Differences in mean relative abundance of ASV336 (Moraxella lincolnii) between SARS-CoV-2-infected participants with respiratory symptoms and SARS-CoV-2-infected participants without respiratory symptoms are depicted by age category. Dark (light) grey bars represent age categories in which ASV336 was more (less) abundant among SARS-CoV-2-infected participants with respiratory symptoms compared to SARS-CoV-2-infected participants without respiratory symptoms. The purple line was constructed using the GJAM estimates for the association of SARS-CoV-2-associated respiratory symptoms with the relative abundance of ASV336 (intercept) and the association of the interaction term between respiratory symptoms and age with the relative abundance of ASV336 (slope). The difference in the mean relative abundance of ASV336 between SARS-CoV-2-infected participants with and without respiratory symptoms differed by age, such that this ASV was less abundant in the context of SARS-CoV-2-associated respiratory symptoms among young children and more abundant in the context of SARS-CoV-2-associated respiratory symptoms in older age groups. Lines were fit using the regression coefficients generated using GJAM. Age is shown as a categorical variable only for graphical representation; all statistical analyses included age as a continuous variable.

M a n u s c r i p t Other comorbidities included hypertension (n=5), congenital heart disease (n=3), chronic neurological disorder (n=3), chronic kidney disease (n=2), and malignancy (n=1)

Innate Immunity to Respiratory Infection in Early Life. Front Immunol

Susceptibility to SARS-CoV-2 Infection Among Children and Adolescents Compared With Adults: A Systematic Review and Meta-analysis

COVID-19 Stats: COVID-19 Incidence, by Age Group -United States

SARS-CoV-2 Infections Among Children in the Biospecimens from Respiratory Virus-Exposed Kids (BRAVE Kids) Study

Severe Acute Respiratory Syndrome Coronavirus

Infection in Children and Adolescents: A Systematic Review

Hospitalization Rates and Characteristics of Children Aged <18 Years Hospitalized with Laboratory-Confirmed COVID-19 -COVID-NET, 14 States

Dynamics of the nasal microbiota in infancy: A prospective cohort study

Early respiratory microbiota composition determines bacterial succession patterns and respiratory health in children. American journal of respiratory and critical care medicine

Association between the respiratory microbiome and susceptibility to influenza virus infection

Nasopharyngeal Microbiota, Host Transcriptome and Disease Severity in Children with Respiratory Syncytial Virus Infection

Meeting report: the terabase metagenomics workshop and the vision of an Earth microbiome project

DADA2: high-resolution sample inference from Illumina amplicon data

Construction of habitat-specific training sets to achieve specieslevel assignment in 16S rRNA gene datasets. Microbiome

Simple statistical identification and removal of contaminant sequences in marker-gene and metagenomics data. Microbiome

phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data

It's all relative: analyzing microbiome data as compositions

The vegan package. Community ecology package

Generalized joint attribute modeling for biodiversity analysis: Median-zero, multivariate, multifarious data

A c c e p t e d M a n u s c r i p t M a n u s c r i p t A c c e p t e d M a n u s c r i p t