Sputum production and salivary microbiome in COVID-19 patients reveals oral-lung axis

• Published in: PloS one Vol. 19; no. 7; p. e0300408
• Main Authors: Lu, Korina Yun-Fan, Alqaderi, Hend, Bin Hasan, Saadoun, Alhazmi, Hesham, Alghounaim, Mohammad, Devarajan, Sriraman, Freire, Marcelo, Altabtbaei, Khaled
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 25-07-2024; Public Library of Science (PLoS)
• Subjects: Actinomyces; Actinomyces - isolation & purification; Adult; Aged; Analysis; Bacteria; Biology and Life Sciences; Biomarkers; Biomarkers - blood; Blood; COVID-19; COVID-19 - microbiology; Cytokines; Cytokines - blood; Cytokines - metabolism; Demographic variables; Development and progression; Female; Gene sequencing; Health aspects; Hospitalization; Hospitals; Humans; Immune response; Infection; Infections; Learning algorithms; Leptotrichia; Lung - microbiology; Lung - virology; Lung diseases; Lungs; Machine Learning; Male; Mann-Whitney U test; Medicine and Health Sciences; Microbial activity; Microbiomes; Microbiota; Microorganisms; Middle Aged; Mouth - microbiology; Mouth - virology; Observational learning; Pandemics; Patients; Performance measurement; Performance prediction; Prevotella; Prevotella - isolation & purification; Regression analysis; Respiratory diseases; Ribosomal RNA; RNA, Ribosomal, 16S - genetics; rRNA 16S; Saliva; Saliva - microbiology; SARS-CoV-2 - isolation & purification; Severe acute respiratory syndrome coronavirus 2; Sputum; Sputum - microbiology; Statistical analysis; Streptococcus - isolation & purification; Variables; Viral diseases; Viral infections; γ-Interferon; Kuwait; United States > US
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0300408

SARS-CoV-2, a severe respiratory disease primarily targeting the lungs, was the leading cause of death worldwide during the pandemic. Understanding the interplay between the oral microbiome and inflammatory cytokines during acute infection is crucial for elucidating host immune responses. This study aimed to explore the relationship between the oral microbiome and cytokines in COVID-19 patients, particularly those with and without sputum production. Saliva and blood samples from 50 COVID-19 patients were subjected to 16S ribosomal RNA gene sequencing for oral microbiome analysis, and 65 saliva and serum cytokines were assessed using Luminex multiplex analysis. The Mann-Whitney test was used to compare cytokine levels between individuals with and without sputum production. Logistic regression machine learning models were employed to evaluate the predictive capability of oral microbiome, salivary, and blood biomarkers for sputum production. Significant differences were observed in the membership (Jaccard dissimilarity: p = 0.016) and abundance (PhILR dissimilarity: p = 0.048; metagenomeSeq) of salivary microbial communities between patients with and without sputum production. Seven bacterial genera, including Prevotella, Streptococcus, Actinomyces, Atopobium, Filifactor, Leptotrichia, and Selenomonas, were more prevalent in patients with sputum production (p<0.05, Fisher's exact test). Nine genera, including Prevotella, Megasphaera, Stomatobaculum, Selenomonas, Leptotrichia, Veillonella, Actinomyces, Atopobium, and Corynebacteria, were significantly more abundant in the sputum-producing group, while Lachnoanaerobaculum was more prevalent in the non-sputum-producing group (p<0.05, ANCOM-BC). Positive correlations were found between salivary IFN-gamma and Eotaxin2/CCL24 with sputum production, while negative correlations were noted with serum MCP3/CCL7, MIG/CXCL9, IL1 beta, and SCF (p<0.05, Mann-Whitney test). The machine learning model using only oral bacteria input outperformed the model that included all data: blood and saliva biomarkers, as well as clinical and demographic variables, in predicting sputum production in COVID-19 subjects. The performance metrics were as follows, comparing the model with only bacteria input versus the model with all input variables: precision (95% vs. 75%), recall (100% vs. 50%), F1-score (98% vs. 60%), and accuracy (82% vs. 66%).