Changes in the upper airway microbiota in pediatric obstructive sleep apnea

Background: A few clinical studies have demonstrated that obstructive sleep apnea (OSA) is associated with dysbiosis of the oral and nasal microbiota. However, how upper airway microbial diversity, composition, and structure are altered in pediatric OSA has not been systemically explored. Methods: In total, 30 polysomnography (PSG)-conrmed OSA patients with adenoid hypertrophy, and 30 controls who did not snore or adenoid hypertrophy, were enrolled. Swabs from four surface oral tissue sites (tongue base, soft palate, both palatine tonsils, and the adenoids) and one nasal swab from both anterior nares were collected. The 16S ribosomal RNA (rRNA) V3–V4 region was sequenced to identify the microbial communities. Results: Alpha and beta diversity were not signicantly different between the pediatric OSA patients and controls at the ve upper airway sites. However, the microbial proles were signicantly different among the ve upper airway sites. The abundances of Haemophilus, Fusobacterium, and Porphyromonas were higher in the adenoids and tonsils of pediatric patients with OSA. Functional analysis revealed that the differential pathway between the pediatric OSA patients and controls involved amino acid metabolism and signal transduction. Conclusions: In this study, the upper airway microbiome of pediatric OSA patients only exhibited minor differences in composition compared with the controls. However, the microbiota data could be useful as a reference for studies on the upper airway microbiome or other relevant clinical phenomena. The data are presented as means and standard deviation; skewed data are presented as the median (IQR), and categorical data as the number (percentage). Differences in the baseline characteristics among the two groups were examined using Mann-Whitney U or Chi-square test as appropriate. This study revealed minor differences in the upper airway microbiota between pediatric OSA patients and controls. Large-scale metagenomic studies are warranted for in-depth examination of the upper airway microbiota in pediatric OSA patients. between pediatric OSA and controls in different sample sites. d The Chao1 index between pediatric OSA and controls in different sample sites. e Nonmetric multidimensional scaling analysis showed that the microbial beta diversity was not signicantly different. ANOSIM test was performed for comparing different groups. f Principal coordinates analysis plot based on the unweighted UniFrac distance depicting differences in the bacterial community between pediatric OSA and controls in different sample sites.


Background
Obstructive sleep apnea (OSA) is a highly prevalent sleep disorder (affecting about 1-4% of children worldwide that can cause cardiovascular and metabolic problems in later life (1). Pathogenic adenoid hypertrophy and enlarged tonsils are the causes of airway obstruction, which can cause OSA in children.
Hyperplasia of the tonsils and adenoids is correlated with the infectious diseases that can accompany OSA, such as acute otitis media (AOM) and secretory otitis media (SOM).
Many clinical studies have suggested that the adenoids and tonsils are pathogenic reservoirs, where 16S rRNA gene amplicon sequencing has revealed that the bacterial communities are complex, diverse, and highly variable. For example, Streptococcus pneumoniae, Haemophilus in uenzae, and Moraxella catarrhalis have been detected in the adenoids, but not the palatine tonsils, of AOM and SOM patients (2). The abundances of Haemophilus, Streptococcus, Neisseria, Capnocytophaga, Kingella, Moraxella, and Lachnospiraceae are higher in cases of tonsillar hypertrophy, while the abundances of Parvimonas, Bacteroidales [G-2], Aggregatibacter, and Atopobium are higher in patients with chronic tonsillitis (3). S. pneumoniae signi cantly affects the composition and diversity of the adenoid microbiota, and Acidobacteria is related to chronic tonsillitis in non-healthy-weight patients with OSA (4). However, the ndings are inconsistent. The upper airway cavity (i.e., oral and nasal cavity) is not a uniform ecosystem, instead comprising several fundamentally different niches with a high degree of microbial diversity. Thus, microbial communities in the mouth and nasal cavity may be differentially affected by OSA via intermittent hypoxia or other mechanisms. Considering the limited data on the diversity and abundance of the upper airway microbiota in pediatric OSA patients, we characterized the microbiota at four oral sites by nasal swab sampling, and compared them between pediatric OSA patients and controls.

Study design and population
We performed a cross-sectional study on patients with polysomnography (PSG)-con rmed pediatric OSA and controls who did not snore and were recruited from Shanghai Jiaotong University A liated Sixth People's Hospital and Shanghai Children's Hospital. The research protocols were independently approved by the ethics committees of the two aforementioned tertiary hospitals (approval nos. 2018-073 and 2019 R061-F01, respectively). Informed consent was obtained from participants aged ≥ 7 years, and from the legal guardians of those aged < 7 years. All pediatric OSA patients, but no controls, had adenoid hypertrophy. Weight and height were carefully measured using an electronic balance and plastic tape measure, while the participants were wearing light clothes and no shoes. Body mass index (BMI) was calculated as weight/height 2 (kg/m 2 ). The inclusion criteria were as follows: aged 4-12 years; obstructive apnea hypopnea index (OAHI) > 1event/h; no obvious dietary preferences; agreed to participate in the study; and willingness of parents to complete the questionnaire. The controls were children with congenital diseases who underwent general anesthesia in the hospital (e.g., for ear reconstruction, accessory ear, a cervical mass, or preauricular stula without infection). The exclusion criteria included a special diet (gluten-free, casein diet, or speci c carbohydrate diet); systemic disease (pulmonary, hepatic, renal, cardiovascular, gastrointestinal, or neurological disease), oral disease (dental caries or periodontal disease), or treatment for adenoid hypertrophy (tonsillectomy, adenoidectomy, corticosteroids, leukotriene antagonists); genetic/craniofacial syndromes, parents with obvious snoring or diagnosed with pediatric OSA; use of any medications, antibiotics or drugs to regulate the intestinal ora (prebiotics, synbiotics, or probiotics) during the previous month, or active infections (bacteria, fungi, or viruses); and pets in the home (a known source of bacteria). Ultimately, 30 pediatric OSA patients and 30 normal controls were enrolled.
De nitions of pediatric OSA All patients with pediatric OSA were monitored by overnight standard PSG (Alice 5; Respironics, Murrysville, PA, USA) at our sleep center. In detail, during sleep from 10 pm to 6 am, electroencephalogram, electrooculogram, genioglossus electromyogram, thoracic/abdominal movement, leg movement, and percutaneous oxygen saturation (at the ngertip) were recorded by sensors. A welltrained technician manually scored the polygraphic data (OAHI, rapid eye movement, non-rapid eye movement, oxygen desaturation index, average and minimum SpO 2 , and microarousal index) according to the 2012 guidelines of the American Academy of Sleep Medicine (5). OSA severity was categorized as mild (OAHI 1-5 events/h), moderate (OAHI 5-10 events/h), or severe (OAHI ≥10 events/h)(6).

Biospecimen collection and DNA extraction
We collected four swabs from the surfaces of oral tissue sites [tongue base, soft palate, both palatine tonsils, and the adenoids (nasopharynx site)], and one nasal swab from each anterior nare (Figure 1a).

Bioinformatics and statistical analyses
The raw 16S rRNA gene amplicon sequences were processed and analyzed using QIIME (8). First, the sequences were demultiplexed based on the barcodes assigned to each sample. Then, the demultiplexed pair-end sequences from each sample were quality-controlled (stitched, ltered, trimmed, and de-noised, with ambiguous/chimeric sequences being removed) using DADA2 in QIIME2 and clustered to generate an amplicon sequence variant table (9). Alpha diversity analysis of all samples was carried out using the Chao1 and Shannon diversity indices. Beta diversity was investigated through nonmetric multidimensional scaling (NMDS) analysis according to Bray-Curtis distance matrices. The bacterial composition (unweighted UniFrac distance) in the microbiome community was examined using the Mantel test. Differences in the relative abundance of taxa were identi ed based on the linear discriminant analysis effect size (LEfSe) (10). Regarding potential functional implications, Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was used to predict microbial metabolic pathways (11).
It was estimated that 20 subjects per group would be necessary to detect differences in unweighted pairwise distances with 90% power (12). Therefore, our study (30 subjects/group) had su cient power to detect differences in taxa between pediatric OSA patients and control subjects.
Statistical analyses were performed using the SPSS (version 22.0; IBM Corp., Armonk, NY, USA) and R studio (R Foundation for Statistical Computing, Vienna, Austria) software packages. Categorical variables are presented as numbers and percentages, and continuous variables as medians with interquartile range. The pediatric OSA and controls groups were compared using the Mann-Whitney U test or chisquare test, as appropriate. A two-sided p-value < 0.05 was considered signi cant.

Patient demographics
In total, 30 pediatric OSA patients and 30 control subjects were included in our study. No signi cant differences in age, gender, or BMI were observed between the two groups (all p > 0.05, Table 1). The detailed clinical and PSG data of the pediatric OSA patients are presented in Table 1. In total, 298 swab samples (60 from the tongue base, 60 from the soft palate, 60 from the palatine tonsils, 58 from the adenoids, and 60 from the anterior nares) were eligible for 16s RNA analysis. The data are presented as means and standard deviation; skewed data are presented as the median (IQR), and categorical data as the number (percentage). Differences in the baseline characteristics among the two groups were examined using Mann-Whitney U or Chi-square test as appropriate.
Differential upper airway microbiota diversity was observed between the pediatric OSA patients and controls for the adenoids; p = 0.07 for nares; p = 0.25 for the palate; p = 0.93 for the tongue; p = 0.11 for the tonsils) (Figure 1b).
The Chao1 and Shannon diversity indices were not signi cantly different between the pediatric OSA and control groups at the various upper airway sites (Chao1 index: p = 0.50 for the adenoids; p = 0.06 for nares; p = 0.27 for the palate; p = 0.99 for the tongue, p = 0.12 for the tonsils; Shannon diversity index: p = 0.68 for the adenoids; p = 0.21 for nares; p = 0.72 for the palate; p = 0.49 for the tongue, p = 0.34 for the tonsils) (Figure 1c and 1d). Overall, the microbiota alpha-diversity values were similar, indicating comparable alpha diversity of the microbiomes at these upper airway sites between the pediatric OSA and control groups.
Microbiota beta-diversity was compared between the pediatric OSA and control groups at the various upper airway sites by principal coordinate analysis (PCoA) and NMDS analysis. The latter analysis did not indicate signi cant dissimilarity (P Stress = 0.11) in the microbial communities of the different upper airway sites (Figure 1e). On the PCoA plot, the microbial communities of the different upper airway sites did not clearly differ between the groups (Figure 1f). However, the pathogenic bacteria from the ve upper airway sites were different between the groups (Figure 1f).

Bacterial composition identi ed from the different upper airway sites between pediatric OSA patients and controls
In Figure 2 a-d, the average composition of bacterial communities at the phylum, family, genus, and species levels were presented. At phylum level, Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, and Fusobacteria were signi cantly different in groups of different upper airway sites (Figure 2a). At family level, Streptococcaceae, Prevotellaceae, Pasteurellaceae, Neisseriaceae, and Moraxellaceae were signi cantly different in these groups (Figure 2b). At genus level, Streptococcus, Haemophilus, Neisseria, Prevotella_7, and Moraxella were signi cantly different in these groups ( Figure  2c). At species level, Streptococcus_sp. Dolosigranulum_sp. Haemophilus_sp. and Moraxella_sp. were signi cantly different in these groups (Figure 2d).

Differences in taxa between the pediatric OSA and control groups at the various upper airway sites
We utilized LEfSe analysis to compare the microbiota between the pediatric OSA and control groups at the various upper airway sites (Figure 3a-e). We identi ed important taxonomic differences between the pediatric OSA and control groups based on a logarithmic linear discriminant analysis score (log10) > 2.0. The LEfSe results suggested remarkable differences in upper airway microbiota between the pediatric OSA and control groups. We were particularly interested in differences in taxa at the genus level. In adenoid, the relative abundances of the genera Haemophilus, Fusobacterium, Porphyromonas, Prevotella, Treponema, Agathobacter, Parvimonas, Campylobacter, and Faecallbacterium were higher in the pediatric OSA than control group, whereas the relative abundances of genera Actinobacillus, Burkholderiales, Ruminococcaceae_UCG_005, Eikenella, and Romboutsia were higher in the controls than pediatric OSA (Figure 3a). In nares, only the relative abundances of the genera Haemophilus, Porphyromonas, and Capnocytophaga were higher in the pediatric OSA than control group (Figure 3b). In palate, the relative abundances of the genera Haemophilus, Actinobacillus, Porphyromonas, Fusobacterium, Prevotella, Streptobacillus, and Campylobacter were higher in the pediatric OSA than control group, whereas the relative abundances of genera Lachnoanaerobaculum, Abiotrophia, Trichococcus, Rothia, and Actinomyces were higher in the controls than pediatric OSA (Figure 3c). In tongue, the relative abundances of the genera Porphyromonas, Fusobacterium, Haemophilus, Capnocytophaga and Prevotella were higher in the pediatric OSA than control group, whereas the relative abundances of genera Abiotrophia, Trichococcus, Lautropia, Alloprevotella, and Streptococuus were higher in the controls than pediatric OSA (Figure 3d). In tonsils, the relative abundances of the genera Fusobacterium, Haemophilus, Porphyromonas, Moraxella, Neisseria, Aggregatibacter, Treponema, Prevotella, Parvimonas, Streptobacillus, Campylobacter, and Collinsella were higher in the pediatric OSA than control group, whereas only the relative abundance of genera Trichococcus was higher in the controls than pediatric OSA (Figure 3e).

Discussion
This is the rst study to provide an overview of the differences in upper airway microbiota between pediatric OSA patients and controls. We observed minor microbial differences between pediatric OSA patients and controls at the various upper airway sites using microbiome 16S rRNA gene sequencing. We also investigated the bacterial functions. Although our sample size provided su cient power, we found no associations between the microbial features and OSA at any upper airway site. Only the microbial pro les differed signi cantly among the ve oral and nasal sites. These results suggest that the microbial communities in the upper airway might be resistant or resilient to disturbances caused by OSA.
Many studies have focused on the effects of environmental factors on the oral and nasal microbiota, such as cigarette smoking and alcohol consumption (7,13,14). Exposure to these factors could lead to a higher abundance of opportunistic pathogens in the upper respiratory tract (7,13,14). Sleep can also affect microbial abundance in the pediatric oral cavity (15). OSA is a common chronic sleep disorder characterized by intermittent hypoxia during sleep. The abundance of oral microbiota changes during the anoxia and reoxygenation cycle. Microbiome pro ling has played an integral role in understanding the development and exacerbation of OSA (16). In our previous study, we reported that the microbiota (Firmicutes, Proteobacteria, Bacteroidetes, Fusobacteria, and Actinobacteria) on the buccal mucosa in pediatric OSA patients was altered (17). However, in this study, no associations between microbial features and OSA were observed at any of the oral sites. This is similar to cigarette smoking, which only affects the oral microbiota in the buccal mucosa (7). One study showed that the salivary microbial composition was stable after antibiotic use at various follow-up timepoints (18). Thus, oral microbial organisms may be tolerant to many adverse conditions (19). This could partially explain why the microbial communities in the upper airway are resistant or resilient to disturbances such as intermittent hypoxia.
Data on the effect of OSA on nasal or nasopharynx microbiota are sparse, and the results are inconsistent. One study found that the nasal microbiome of adult patients with severe OSA was enriched with Streptococcus, Prevotella, and Veillonella (20). Another study reported no difference in the composition of the nasopharynx microbiome between mild-to-moderate adult OSA patients and those without OSA (21). No previous study has explored the composition of the nasal microbiome in pediatric OSA patients, while one study found no signi cant difference in nasopharynx or nasal cavity composition or diversity between patients with chronic rhinosinusitis and controls (22). The adenoids and tonsils are lymphoid tissues; hypertrophy thereof plays an important role in the onset and development of pediatric OSA, but few studies have explored the adenoid or tonsil microbiome in OSA to determine whether it participates in hypertrophy. Higher abundances of the genera Haemophilus, Fusobacterium, and Porphyromonas were found in the adenoids and tonsils of our pediatric OSA group. Haemophilus is the main pathogenic bio lm bacteria constituting the adenoid reservoir (23,24). Fusobacteria were detected at a tonsillar site in pediatric OSA patients (25) and associated with the AHI (20). The abundance of Porphyromonas is higher in patients with OSA; it profoundly affects the likelihood of developing OSA-related cardiovascular diseases (26). The roles of these genera in the development of the adenoids and tonsils need to be further explored.
Several limitations of this study should be acknowledged. First, we used 16S rRNA sequencing (rather than deep-shotgun sequencing) to detect microbial diversity, which is not capable of in-depth analysis of the function of the upper airway microbiome. Second, although the sample size was su cient to detect differences in microbes between the case and control groups, our cohort was relatively small. Thus, caution is required when interpreting the data. Third, the normal controls were classi ed as such based on a questionnaire rather than objective sleep parameters (i.e., standard PSG