This study was approved by the Institutional Review Board of the Chung-Ang University Hospital, and written informed consent was obtained from the caregivers of all participants. The study was cross-sectional, comprising of 24 pediatric patients who underwent tonsillectomy with adenoidectomy. Patients with snoring or symptoms of sleep apnea were enrolled in this study, and those with recurrent tonsillitis or antibiotic medication history within the previous 4 weeks were excluded. Further, patients with underlying conditions, such as immunodeficiency, malignancy, or allergic asthma, were excluded from the study. Medical records, including results of the multiple allergen simultaneous test (MAST; LG Life Sciences, Seoul, Korea) [11,12], total serum immunoglobulin E (IgE) levels (group 1, negative; group 2, low titer (<2 positivity; ≤0.70 IU/mL); and group 3, high titer (≥2 positivity; 0.70 IU/mL) [13], and white blood cell (WBC) counts, and the OSA-18 questionnaire were reviewed.

The OSA-18 questionnaire assessed the quality of life, and disordered breathing and discomfort in pediatric sleep; it comprised of the following five domains: sleep disturbances, physical suffering, emotional distress, daytime problems, and parent or caretaker concern [14]. Blood tests and OSA-18 questionnaire were conducted a day before the surgery. Tonsillectomy with adenoidectomy was performed under general anesthesia by an otorhinolaryngologist, and LFs from the tonsils and adenoids (collected separately from each tissue), as well as the tonsil and adenoid tissues, including the surface mucosa and submucosal lymphoid tissues, were obtained as described previously [10]. Approximately 5 mL of regular saline was instilled onto the tonsil (under direct visualization) and adenoid (using a 30° endoscope) surfaces, and approximately 4 mL retrieved after 3 seconds. Next, the tonsil and adenoid tissues, including the surface mucosa and submucosal lymphoid tissues, were sampled using cutting forceps before surgery. The tissue samples were placed in a sterile container, stored on ice, and transferred to the laboratory within 1 hour. All samples were stored at –20°C until further analysis.

Total genomic DNA was extracted as described previously [15]. Microbial genomic DNA was extracted from the tonsil and adenoid biopsy specimens using the repeated-bead-beating method [16]. The barcoded forward primer 27F (5ʹ-XXXXXXXXXXGAGTTTGATCMTGGCTCAG-3ʹ, where “X” represents the barcode sequence) and reverse primer 338R (5ʹ-TGCTGCCTCCCGTAGGAGT-3ʹ), both ligated to an FLX Titanium adapter, were used for amplifying the V1–V2 region of the 16S rRNA gene. Diluted DNA template (8 μL, 1:7), primer mix (2 µL, 0.2 µM), and 2× Ex Taq polymerase premix (TaKaRa Bio, Shiga, Japan) were mixed for the polymerase chain reaction (PCR; total reaction volume, 50 μL). PCR conditions were set as previously described [13]. QIAquick PCR purification kit (Qiagen, Hilden, Germany) was used to purify amplicons, prepared as eight replicates. Quant-iT PicoGreen double-stranded DNA kit (Invitrogen, Carlsbad, CA, USA) was used to determine the concentration of DNA. Equimolar concentrations of the prepared amplicons for 48 tonsil and adenoid samples were subjected to pyrosequencing on the 454 GS-FLX Titanium sequencing platform (Roche Diagnostics, Manheim, Germany).

Analysis of tissue microbiota was performed as described previously [13]. Briefly, the reads were analyzed on the Quantitative Insights Into Microbial Ecology (QIIME) platform [16], and operational taxonomic units (OTUs) (≥97% identity) were selected using the Greengenes database [17]. After excluding unassigned or non-bacterial kingdom-assigned sequences and singletons, 6,015,107 reads were obtained (125,314±28,769 reads per sample). Alpha-diversity was measured using Shannon diversity index, Chao1, and observed OTUs. Beta-diversity was evaluated using principal component analysis (PCoA) based on weighted and unweighted UniFrac and Bray-Curtis distances.

The amounts of extracellular heat shock protein (Hsp)27, Hsp70, and interleukin-8 (IL-8) in the LFs were determined using a commercial enzyme-linked immunosorbent assay kit (R&D Systems, Minneapolis, MN, USA), according to the manufacturer’s instructions.

Mann-Whitney U-test and Spearman’s rank correlation were performed using GraphPad Prism software ver. 5.0 (GraphPad, San Diego, CA, USA). Linear discriminant analysis (LDA) coupled with an effect size measurement (LEfSe) was employed to distinguish the taxonomic characteristics between groups [18]. Alpha-diversity (within-sample diversity; Shannon diversity index, Chao1, and observed OTUs) was calculated for each sample in the QIIME pipeline. Beta-diversity (between-sample diversity) was calculated using the phylogenetic metrics (UniFrac) for microbial community and non-phylogenetic metrics (BrayCurtis) for microbial gene community in the QIIME pipeline. The calculated distance matrices were used to perform PCoA. The analysis of similarities (ANOSIM) method was used to predict the differences in microbial communities between groups, whereas permutational multivariate analysis of variance (PERMANOVA) was used to evaluate the dissimilarities based on site (adenoid and tonsil), sex, MAST, and total serum IgE levels.

Twenty-four patients were enrolled in this study (mean age, 6.29 years; standard deviation [SD], 6.29 years; age range, 4–12 years). Sixteen were boys and eight were girls. Nine (37.5%) had allergic rhinitis (AR). Among the OSA-18 questionnaire domains, sleep disturbance, physical suffering, and caregiver concerns were the highest-rated, followed by daytime problems and emotional distress. The mean±SD levels of IL-8, Hsp27, and Hsp70 detected in the adenoid LFs were 216.97±77.61, 328.39±222.47, and 4,023.94±2,708.82 pg/mL, respectively, whereas those detected in the tonsil LFs were 189.68±80.07, 438.77±204.17, and 4,756.17±2,747.32 pg/mL, respectively (Table 1).

Tissue samples from the tonsils revealed a predominance of Proteobacteria (mostly the genus Haemophilus), Firmicutes (dominated by the genus Streptococcus), and Bacteroidetes (dominated by the genus Prevotella). Tissue samples from the adenoids were dominated by Proteobacteria (mostly the genus Haemophilus), Firmicutes (dominated by the genus Streptococcus), and Fusobacteria (dominated by the genus Fusobacterium). At the genus level, Streptococcus was abundant, while Neisseria was present to a lesser extent in the adenoids than in the tonsils (Fig. 1A and B). LEfSe was performed to obtain the cladogram representation and to identify the predominant bacteria in tonsils and adenoids. We found Neisseria to be more abundant in the tonsils than in the adenoids (Fig. 1C).

Chao1, observed OTUs, and the Shannon diversity index were used to compare the alpha-diversity of the microbiome between the adenoids and tonsils. Neither bacterial community richness nor diversity was significantly different between the adenoid and tonsil microbiomes (P=0.219 for Chao1, P=0.323 for observed OTUs, and P=0.584 for Shannon diversity) (Fig. 2). We evaluated the correlations of the alpha-diversity of the entire adenotonsillar microbiome with clinical characteristics. Bacterial diversity, as represented by Shannon diversity, differed significantly between males and females; males had a more diverse adenotonsillar microbiome than females (P=0.045). The score of the emotional stress domain in the OSA-18 questionnaire (P=0.047, rho=–0.287 for Chao1; P=0.014, rho=–0.350 for observed OTUs) and the IL-8 concentration in the tonsil LFs (P=0.028, rho=–0.310 for Chao1; P=0.001, rho=–0.438 for observed OTUs) were negatively correlated with the alpha-diversity of the adenotonsillar microbiome. This suggested that higher levels of emotional stress in OSA or IL-8 concentrations in the tonsil LFs were associated with lower richness of the adenotonsillar bacterial community (Table 2). However, the alphadiversity of the adenotonsillar microbiome based on MAST positivity or total serum IgE levels revealed no significant difference in Chao1, observed OTUs, and Shannon diversity (Supplementary Fig. 1).

Since the UniFrac distance considers the phylogenetic relatedness of OTUs with the Bray-Curtis distance, which quantifies the compositional dissimilarity between two different sites based on the counts at each site, it is only a measure of abundance [19]. We generated PCoA plots for the Bray-Curtis, unweighted UniFrac, and weighted UniFrac distances of the phylogenetic tree. PCoA, which showed a dissimilarity between the adenoids and tonsils, was not significant (P=0.167 for Bray-Curtis, P=0.607 for unweighted UniFrac, and P=0.281 for weighted UniFrac) (Fig. 3). Next, we performed ANOSIM and PERMANOVA to identify the characteristics responsible for beta-diversity, and found that the total serum IgE levels were a significant factor (Supplementary Table 1). When the tonsils or adenoids were individually evaluated to explore possible correlations between beta-diversity and clinical characteristics, no significant factor was identified, with the exception of the total IgE level (data not shown). The relationship of clinical characteristics with the betadiversity of the entire adenotonsillar microbiome was further analyzed based on the average weighted UniFrac, unweighted UniFrac, and Bray-Curtis distances, which represented intraindividual dissimilarities in the adenotonsillar microbiota. The Hsp27 level in the tonsil LFs (P=0.038, rho=0.288 for Bray-Curtis) and WBC counts (P=0.001, rho=0.610 for unweighted UniFrac; P=0.037, rho=0.426 for Bray-Curtis) were responsible for beta-diversity (Table 3).