tetano
Editor, Senior Moderator
J Transl Med
. 2026 Aug 7;24(1):1027.
doi: 10.1186/s12967-026-08697-8.
Longitudinal profiling of upper respiratory tract microbiota and metabolome in hospitalized COVID-19 convalescents: a 3-year prospective cohort study
Zhijin Zhang[SUP] #[/SUP][SUP] 1 2 [/SUP], Leyi Gao[SUP] #[/SUP][SUP] 1 2 [/SUP], Lujia Guan[SUP] 1 2 [/SUP], Zhenbei Qian[SUP] 1 2 [/SUP], Jieqiong Li[SUP] 3 4 5 [/SUP], Zhaohui Tong[SUP] 6 7 8 9 10 [/SUP]
Affiliations
Background: Long COVID is characterized by persistent, far-reaching effects in convalescent individuals, with pulmonary diffusion impairment emerging as a clinically impactful sequela affecting more than one-third of this population. The salivary microbiome and metabolome, reflecting the oral-lung axis, offer a window into the mechanisms underlying this condition. However, systematic longitudinal evidence on their long-term dynamics after infection and their predictive value for persistent pulmonary diffusion impairment remains scarce.
Methods: In this prospective cohort, we profiled the salivary bacterial microbiome (16S rRNA sequencing) and metabolome (untargeted LC-MS/MS) in 424 COVID-19 convalescents at 2 (T1) and 3 (T2) years post-discharge, alongside 106 demographically matched healthy controls. To explore whether 2-year salivary multiomics signatures were associated with 3-year pulmonary diffusion status, microbial and metabolic features were ranked using random forest mean decrease in accuracy and used to train 10 machine-learning classifiers after stratified training/internal validation splitting. Because this modeling strategy was exploratory, we further performed a repeated stability-selection analysis across 100 stratified resampling iterations to identify reproducibly selected salivary multiomics features.
Results: COVID-19 convalescents exhibited sustained, interrelated salivary microbiome dysbiosis and metabolic reprogramming at 3 years post-infection. The microbial perturbations were characterized by reduced alpha diversity, a shift in phylogenetic dominance from Bacteroidota to Actinobacteriota, and a marked expansion of Proteobacteria at the 2-year follow-up. The microbial co-occurrence networks also became sparser, suggesting diminished stability. Metabolomic profiling revealed upregulation of the TCA cycle, purine/pyrimidine metabolism, arginine biosynthesis, and other pathways at the 2-year follow-up, with a discernible trend toward recovery by year 3. Notably, 36.6% of patients presented with persistent pulmonary diffusion dysfunction at the 3-year follow-up. Leveraging 2-year salivary multiomics signatures, we developed an exploratory proof-of-concept model for predicting 3-year pulmonary diffusion dysfunction. The CatBoost classifier achieved the best overall performance, achieving an area under the curve of 0.808 in the internal validation set; key predictive features included genera Catonella and Actinomyces, and metabolites adenosine 3',5'-diphosphate, triiodothyronine sulfate and betaine. In an exploratory stability-selected analysis, a conservatively tuned CatBoost model based on repeatedly selected features achieved an internal validation AUC of 0.798.
Conclusions: This research provides the first longitudinal characterization of the salivary bacterial microbiome and metabolome in COVID-19 convalescents up to 3 years post infection. Furthermore, we developed a novel predictive model for post-SARS-CoV-2 pulmonary diffusion impairment based on salivary multiomics features, which may represent a promising screening tool for identifying high-risk individuals.
Keywords: Convalescence; Long COVID; Metabolomics; Microbiomics; Saliva.
. 2026 Aug 7;24(1):1027.
doi: 10.1186/s12967-026-08697-8.
Longitudinal profiling of upper respiratory tract microbiota and metabolome in hospitalized COVID-19 convalescents: a 3-year prospective cohort study
Zhijin Zhang[SUP] #[/SUP][SUP] 1 2 [/SUP], Leyi Gao[SUP] #[/SUP][SUP] 1 2 [/SUP], Lujia Guan[SUP] 1 2 [/SUP], Zhenbei Qian[SUP] 1 2 [/SUP], Jieqiong Li[SUP] 3 4 5 [/SUP], Zhaohui Tong[SUP] 6 7 8 9 10 [/SUP]
Affiliations
- PMID: 42581368
- DOI: 10.1186/s12967-026-08697-8
Background: Long COVID is characterized by persistent, far-reaching effects in convalescent individuals, with pulmonary diffusion impairment emerging as a clinically impactful sequela affecting more than one-third of this population. The salivary microbiome and metabolome, reflecting the oral-lung axis, offer a window into the mechanisms underlying this condition. However, systematic longitudinal evidence on their long-term dynamics after infection and their predictive value for persistent pulmonary diffusion impairment remains scarce.
Methods: In this prospective cohort, we profiled the salivary bacterial microbiome (16S rRNA sequencing) and metabolome (untargeted LC-MS/MS) in 424 COVID-19 convalescents at 2 (T1) and 3 (T2) years post-discharge, alongside 106 demographically matched healthy controls. To explore whether 2-year salivary multiomics signatures were associated with 3-year pulmonary diffusion status, microbial and metabolic features were ranked using random forest mean decrease in accuracy and used to train 10 machine-learning classifiers after stratified training/internal validation splitting. Because this modeling strategy was exploratory, we further performed a repeated stability-selection analysis across 100 stratified resampling iterations to identify reproducibly selected salivary multiomics features.
Results: COVID-19 convalescents exhibited sustained, interrelated salivary microbiome dysbiosis and metabolic reprogramming at 3 years post-infection. The microbial perturbations were characterized by reduced alpha diversity, a shift in phylogenetic dominance from Bacteroidota to Actinobacteriota, and a marked expansion of Proteobacteria at the 2-year follow-up. The microbial co-occurrence networks also became sparser, suggesting diminished stability. Metabolomic profiling revealed upregulation of the TCA cycle, purine/pyrimidine metabolism, arginine biosynthesis, and other pathways at the 2-year follow-up, with a discernible trend toward recovery by year 3. Notably, 36.6% of patients presented with persistent pulmonary diffusion dysfunction at the 3-year follow-up. Leveraging 2-year salivary multiomics signatures, we developed an exploratory proof-of-concept model for predicting 3-year pulmonary diffusion dysfunction. The CatBoost classifier achieved the best overall performance, achieving an area under the curve of 0.808 in the internal validation set; key predictive features included genera Catonella and Actinomyces, and metabolites adenosine 3',5'-diphosphate, triiodothyronine sulfate and betaine. In an exploratory stability-selected analysis, a conservatively tuned CatBoost model based on repeatedly selected features achieved an internal validation AUC of 0.798.
Conclusions: This research provides the first longitudinal characterization of the salivary bacterial microbiome and metabolome in COVID-19 convalescents up to 3 years post infection. Furthermore, we developed a novel predictive model for post-SARS-CoV-2 pulmonary diffusion impairment based on salivary multiomics features, which may represent a promising screening tool for identifying high-risk individuals.
Keywords: Convalescence; Long COVID; Metabolomics; Microbiomics; Saliva.