This study was approved by the Ethics Committee of Longgang ENT Hospital, Institute of ENT, and the Shenzhen Key Laboratory of ENT, and all parents of the children agreed and signed consent for this study.

A total of 64 children were included in this study. Overnight polysomnography was performed in all children. All testing was performed in the Sleep Physiology Laboratory in Longgang ENT Hospital, Institute of ENT, and the Shenzhen Key Laboratory of ENT with continuous observation of the child by a polysomnography technician skilled in pediatric polysomnography. A normal polysomnogram was defined by an apnea-hypopnea index (AHI) <1. Mild OSA was defined by an AHI of 1–5. Moderate OSA was defined by an AHI of 6–10. Severe OSAS was defined by an AHI >10 ²⁰. The inclusion criteria for OSA patients were as follows: 1) 2–13 years old and 2) AHI > 1; the inclusion criteria for control subjects were as follows: 1) 2–13 years old and 2) AHI < 1. The exclusion criteria were as follows: 1) do not meet the inclusion criteria and 2) the presence of a known genetic syndrome or any other chronic disease. The clinical characteristics of the all the subjects are shown in Table 1. After overnight polysomnography, 5 ml of blood was collected in the morning by using a serum separator tube. The serum samples were allowed to clot at room temperature for 2 h before centrifugation at 1,000 g for 15 min. The collected serum samples were stored at –80 °C before downstream assays.

The present study randomly selected 12 serum samples for TMT Quantitative Proteomics analysis. Among these 12 samples, three serum samples were from the non-OSA patients, three serum samples were from mild OSA patients, three serum samples were from moderated OSA patients, and three serum samples were from severe OSA patients. The clinical characteristics of the included subjects are shown in Supplemental Table S1.

The abundant serum proteins were depleted using High Select Top 14 spin columns (Thermo #A36370) following the manufacturer-provided protocol. Briefly, 10 μl of serum was applied to each column and incubated for 10 min with end-over-end rotation. Depleted samples were collected by centrifugation at 2000 g for 2 min. Ice-cold 100% trichloroacetic acid was added to 20% final concentration. The proteins were allowed to precipitate on ice for 60 min and then pelleted for 10 min at 20,000 g at 4°C. The pellets were washed twice with ice-cold acetone and allowed to dry at room temperature. Dry protein pellets were resuspended in 8 M urea and 50 mM ammonium bicarbonate (ambic). The proteins were quantified with a BCA Protein Assay Kit (Bio-Rad, United States). This was followed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and Coomassie Bright Blue staining to compare and analyze the protein expression consistency among samples. The poteins were reduced by the addition of dithiothreitol (DTT) to 100 mM and incubating in a boiled water bath for 30 min. The samples were then mixed with 200 µL of UA buffer (8 M urea with 150 mM Tris-HCl, pH 8.0), followed by centrifugation at 12,000 g for 15 min. After discarding the supernatant, the samples were incubated with 100 µL of indole-3-acetic acid (IAA) (50 mM IAA in UA) with agitation for 1 min, followed by incubation at room temperature in the dark for 30 min. After that, the samples were centrifuged at 12,000 g for 10 min. After discarding the supernatant, the samples were mixed with 100 µL of UA buffer, followed by centrifugation at 12,000 g for 10 min. After that, the samples were mixed with 100 µl of tetraethylammonium bromide buffer, followed by 14,000 g centrifugation for 10 min. After that, the samples were mixed with 40 µl of trypsin buffer with 1 min agitation, followed by 18 h incubation at 37 C. After centrifugation at 12,000 g for 10 min, the digested samples were acidified by the addition of a proper amount of 0.1% trifluoroacetic acid and were desalted c-18 Cartridge. The eluted peptides were dried in a centrifugal evaporator. After drying, the peptide samples were suspended on ice in 200 µL of the lysis buffer (4% SDS, 150 mM Tris-HCl, and 100 mM DTT, pH 7.8). The tissue was disrupted with agitation by a homogenizer and then immediately boiled for 5 min. The samples were further ultrasonicated and boiled again for another 5 min. The undissolved cellular debris was removed by centrifugation at 16,000 rpm for 15 min.

TMT reagents were used for labeling of peptides according to the manufacturer’s instructions (Thermo Fisher Scientific). Each aliquot (50 µg of peptide equivalent) was reacted with one tube of the TMT reagent. The sample labeling was as follows: Group A1, 128C; Group A2, 129N; Group A3, 129C; Group B1, 130C; Group B2, 130C; Group B3, 121N; Group C1, 131C; Group C2, 127N; Group C3, 132C; Group D1, 133N; Group D2, 133C; and Group D3, 134N. Equal amounts of TMT labeling peptides were mixed in each group, and then, HPRP (Pierce™ High-pH Reversed-Phase Peptide Fractionation Kit, Thermo Fisher, Waltham, MA, United States) was used to fractionate peptides after drying. The samples were eventually collected into 10 components. Each component of the peptides was stored at −80 C for liquid chromatography–mass spectrometry (LC-MS) analysis.

The redissolved peptide solution was taken for LC-MS/MS analysis, and each fractional component of the sample was injected once for a total of 10 times for mass spectrometry analysis. The high-performance liquid chromatography liquid-phase system Easy-nLC 1,200 was used for separation. Buffer solutions: A, 0.1% formic acid solution, and B, 80% acetonitrile solution. The chromatographic column was balanced with 100% A solution. The sample was loaded onto the chromatographic column, that is, trap column (20 mm × 100 μm, 5 μm-C18, Dr. Maisch GmbH). Then, the C-18 column (75 μm × 150 mm, 5 μm-c-18, Dr. Maisch GmbH) was used for separation, and the velocity was at 300 nL/min. The relative liquid gradient was as follows: linear-gradient of liquid B, 2–5% for 0–2 min; linear gradient of liquid B, 5–28% for 2–71 min; linear gradient of liquid B, 28–40% for 71–79 min; linear gradient of liquid B, 40–100% for 79–81 min; and liquid B maintained at 90% for 81–90 min. After chromatographic separation, the peptides were analyzed using a Q-Exactive HF-X mass spectrometer (Thermo Scientific, Waltham, MA, United States) with the following parameters: analysis duration, 90 min; detection method, positive ion; and scanning range of parent ions, 300–1800 m/z. The mass charge ratios of polypeptides and polypeptide fragments were collected as follows: 20 fragment profiles (MS2 scan, HCD) were collected after each full scan. The resolution of the first-level mass spectrometry was 60,000 at m/z 200, the AGC target was 3e6, and the first-level maximum IT was 50 ms. The resolution of the second-level mass spectrometry was 45,000 at m/z 200, the AGC target was 1e5, and the second-level maximum IT was 50 ms. The MS2 activation type was HCD; the isolation window was 1.6 m/z, and the normalized collision energy was 30.

The raw files produced from LC-MS/MS were imported into proteome discoverer software (version 2.4). The protein database (192,367 proteins) was sourced from uniport-uniprot-homo-192367-20200629.fasta. The search parameters were set as follows: Type, Reporter ion MS2; isobaric labels, TMTpro 16plex; Enzyme, Trypsin; Reporter mass tolerance, 0.005 Da; Max Missed cleavages, 2; Peptide Tolerance, 10 ppm; MS/MS Tolerance, 0.02 Da; Fix modifications, Carbamidomethyl (C); Variable modifications, Oxidation (M), Acetyl (Protein N-term), Deamidation (N,Q), TMTpro (K), TMTpro (N-term); Database, uniport-uniprot-homo-192367-20200629.fasta; Database pattern: Target-Reverse; and Percolator (FDR) ≤ 0.01; Razor and unique peptides were used for protein quantification.

Perseus software and R statistical computing software were used to analyze the bioinformatics data. Differentially expressed proteins were screened with the cutoff of a ratio fold-change of >1.20 and p-values < 0.05. In our study, differentially expressed proteins were compared among different groups. Combining the comparative analysis of variance (ANOVA), t-test, and FDR (Benjamini–Hochberg), all qualitative and quantitative protein analysis results were obtained. Hierarchical clustering was adopted to categorize expression data together according to the protein level. Huge amounts of data are produced by mass spectrometry technology in proteomics, which represents all the biological processes of the organism. The aim of bioinformatics analysis was to find the source and mechanism for biological changes. UniProt Knowledgebase (UniProtKB)/Swiss-Prot, Gene Ontology (GO) enrichment, Kyoto Encyclopedia of Genes and Genomes (KEGG), and protein interaction network analysis were adopted.

For the validation studies, serum samples from 64 subjects were used for the determination of protein levels. The protein levels of anthrax toxin receptor 1 (ANTXR1), collectin-10 (COLEC10), neural cell adhesion molecule 1 (NCAM1), vanin 1 (VNN1) and chondroitin sulfate proteoglycan 4 (CSPG4) in the serum samples were determined by the respective enzyme-linked immunosorbent assay (ELISA) kits from CUSABIO (Wuhan, China), and the protein level of mannosidase alpha class 1A member 1 (MAN1A1) in the serum samples was determined by Reddnot Biotech (Kelowna, Canada). In brief, 100 µl of serum samples was added into respective precoated ELISA plates. The plates were incubated for 2 h at 37°C. After removing the liquid from each well, the plates were incubated with 100 µl of biotin-antibody (1x) for 1 h at 37 °C. After washing three times, the plates were incubated with 100 µl of horseradish peroxidase-avidin (1x) 1 h at 37 °C. After washing five times, the plates were incubated with the 3,3′,5,5′-tetramethylbenzidine substrate for 30 min at 37 °C. Then, 50 μL of the stop solution was added to each well to stop the reaction. The optical density was detected at a 450 nm wavelength within 15 min. Then, the stop solution was added. A standard curve was constructed by plotting a graph of the absorbance of each reference standard against its corresponding levels and used to determine each of the protein levels.

The data for ELISA were presented as mean ± standard deviation. The data analysis was performed using GraphPad Prism Software (GraphPad Software 7.0, La Jolla, United States). Significant differences among different groups were analyzed using one-way ANOVA, followed by Bonferroni’s post-hoc tests. The correlation between AHI and protein levels was analyzed by Pearson correlation analysis. p < 0.05 was considered statistically significant.

In the study, we obtained 59,578 peptide spectrum matching (PSM) numbers and 8,544 unique peptides, which were mapped to 754 protein groups with 752 being quantified as proteins. The details of the identified proteins are shown in Supplemental Table S2.

The box plot of normalized density (box plots of log2 protein intensity average for each sample used for TMT-based proteomic analysis) is shown in Figure 1. Figure 2 presents the histogram of intensity for each sample used for TMT-based proteomic analysis. In addition, the Pearson correlation coefficients of the samples exhibited high repeatability via assessment of relative quantitative proteins (Supplemental Figure S1).

The hierarchical cluster analysis and volcano plots of differentially expressed proteins between different groups are shown in Figures 3, 4. As shown in Figure 3, a total of 23 proteins (six upregulated and 17 downregulated) were significantly changed in the mild OSA group (Figures 3A,B), a total of 31 proteins (13 upregulated and 18 downregulated) were significantly changed in the moderate OSA group (Figures 3A,C,D), and a total of 53 proteins (four upregulated and 49 downregulated) were significantly changed in the severe group when compared to the non-OSA group (Figures 3E,F).

In addition, a total of 21 proteins (12 upregulated and nine downregulated) were significantly changed in the moderate OSA group (Figures 4A,B), and a total of 17 proteins (four upregulated and 13 downregulated) were significantly changed in the severe group when compared to the mild OSA group (Figures 4C,D). A total of 34 proteins (10 upregulated and 24 downregulated) were significantly changed in the severe group when compared to the moderate group (Figures 4E,F).

The most enriched terms of GO enrichment analysis between different groups are shown in Figures 5, 6. For the differentially expressed proteins between mild OSA and non-OSA groups, proteins were mainly enriched in “myeloid leukocyte activation,” “leukocyte activation involved in the immune response,” “regulation of body fluid levels,” etc. under the category of biological process; were enriched in “secretory granule,” “vesicle,” “extracellular membrane-bounded organelle,” “intracellular vesicle,” etc. under the category of cellular component; and were enriched in “monosaccharide binding,” “cation binding,” “peptidase regulator activity,” etc. under the category of molecular function. For the differentially expressed proteins between moderate OSA and non-OSA groups, proteins were mainly enriched in “regulation of immune response,” “leukocyte mediated immunity,” “humoral immune response,” etc. under the category of biological process; were enriched in “secretory granule,” “vesicle,” “extracellular membrane-bounded organelle,” etc. under the category of cellular component; and were enriched in “thyroid hormone binding,” “complement binding,” “calcium-dependent protein binding,” etc. under the category of molecular function (Figure 5B). For the differentially expressed proteins between severe OSA and non-OSA groups, proteins were mainly enriched in “cell activation,” “humoral immune response,” “multicellular organism development,” etc. under the category of biological process; were enriched in “extracellular membrane-bounded organelle,” “extracellular exosome,” “vesicle,” etc. under the category of cellular component; and were enriched in “serine hydrolase activity,” “cell adhesion molecule binding,” “receptor binding,” etc. under the category of molecular function.

For the differentially expressed proteins between moderate and mild OSA groups, proteins were mainly enriched in “positive regulation of immune response,” “humoral immune response,” “regulation of immune response,” etc. under the category of biological process; were enriched in “blood microparticle,” “collagen-containing extracellular matrix,” “extracellular membrane-bounded organelle,” etc. under the category of cellular component; and were enriched in “peptidase regulator activity,” “calcium-dependent protein binding,” “enzyme inhibitor activity” etc. under the category of molecular function. For the differentially expressed proteins, between severe and mild OSA groups, proteins were mainly enriched in “leukocyte mediated immunity,” “cell activation,” and “myeloid leukocyte activation” under the category of biological process; were enriched in “extracellular membrane-bounded organelle,” “extracellular exosome,” “vesicle” etc. under the category of cellular component; and were mainly enriched in “glycolipid binding,” “serine hydrolase activity,” and “cation binding” under the category of molecular function. For the differentially expressed proteins between severe OSA and moderate OSA groups, proteins were mainly enriched in “leukocyte mediated immunity,’ “cell activation,” “cell migration,” etc. under the category of biological process; were mainly enriched in “secretory granule,” “extracellular membrane-bounded organelle,” “extracellular exosome,” etc. under the category of cellular component; and were mainly enriched in “monosaccharide binding,” “peptidase regulator activity”, “cell adhesion molecule binding,” etc. under the category of molecular function.

The differentially expressed proteins between severe OSA and non-OSA groups were mainly enriched in “other glycan degradation,” “various types of N-glycan biosynthesis,” “N-glycan biosynthesis,” and so on (Figure 7A). The differentially expressed proteins between moderate OSA and non-OSA groups were mainly enriched in “Hedgehog signaling pathway” and “Complement and coagulation cascades” (Figure 7B). The differentially expressed proteins between severe OSA and non-OSA groups were mainly enriched in “other glycan degradation,” “pantothenate and CoA biosynthesis,” “PI3K-Akt signaling pathway,” and so on (Figure 7C). The differentially expressed proteins between moderate and mild OSA groups were mainly enriched in “Hedgehog signaling pathway,” “Complement and coagulation cascades,” and “Thyroid hormone synthesis” (Figure 7D). The differentially expressed proteins between severe OSA and mild OSA groups were mainly enriched in “Pantothenate and CoA biosynthesis,” “B cell receptor signaling pathway,” “Osteoclast differentiation,” and so on (Figure 7E). The differentially expressed proteins between severe OSA and moderate OSA groups were mainly enriched in “Tyrosine metabolism” and “Other glycan degradation” (Figure 7F).

The differentially expressed proteins between mild OSA and non-OSA groups were associated with natural killer-cell mediated cytotoxicity, phagosomes, osteoclast differentiation, and so on (Figure 8A). The differentially expressed proteins between moderate OSA and non-OSA groups were associated with glycerophospholipid metabolism, the B cell receptor signaling pathway, phagosomes, and so on (Figure 8B). The differentially expressed proteins between severe OSA and non-OSA groups were associated with leukocyte transendothelial migration, focal adhesion, phagosomes, and so on (Figure 8C). The differentially expressed proteins between moderate OSA and mild OSA groups were associated with the Hedgehog signaling pathway, complement and coagulation cascades, and phagosomes (Figure 8D). The differentially expressed proteins between severe OSA and mild OSA groups were associated with osteoclast differentiation, the B cell receptor signaling pathway, metabolic pathways, and so on (Figure 8E). The differentially expressed proteins between severe OSA and moderate OSA groups were associated with the leukocyte transendothelial migration, tyrosine metabolism, extracellular matrix (ECM)–receptor interaction, and so on (Figure 8F).

The protein expression levels of ANTR1, COLEC10, CSPG4, MAN1A1, NCAM1, and VNN1 in the serum levels from non-OSA and OSA patients were further determined by ELISA. The protein levels of ANTXR1, COLEC10, NCAM1, and VNN1 were significantly decreased in the moderate and severe OSA patients when compared to the non-OSA patients (Figures 9A,B,E,F); the VNN1 protein level was also higher in the severe OSA patients than in non-OSA patients (Figure 9F). The protein levels of CSPG4 and MAN1A1 were significantly increased in the severe OSA patients when compared to the non-OSA patients (Figures 9C,D). The correlation analysis showed that the protein levels of ANTXR1, COLEC10, NCAM1, and VNN1 were inversely correlated with AHI in the recruited subjects (Figures 10A,B,E,F), while the protein level of MAN1A1 was positively correlated with AHI (Figure 10D), and no significant correlation was detected between CSPG4 protein and AHI (Figure 10C).