In a recent study posted to the bioRxiv* preprint server, researchers evaluated the association between guild-level microbiome signature and the severity and prognosis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections by a genome-level metagenomic analysis.
Background
Angiotensin-converting enzyme 2 (ACE-2) is an essential receptor for SARS-CoV-2 host invasion and is widely expressed in the gut. Coronavirus disease 2019 (COVID-19)-associated inflammation may cause gut microbiome disruptions, which may further potentiate the hyperinflammatory state among SARS-CoV-2-infected individuals. However, the association between intestinal microbiome alterations and prognosis of SARS-CoV-2 infections is not well-characterized.
About the study
In the present study, researchers analyzed HQMAGs (high-quality metagenome-assembled genomes) sequenced from stool samples to assess the association between the guild-level microbiome signature and the severity and prognosis of SARS-CoV-2 infections.
Between May and September 2020, 330 fecal specimens were obtained from 300 hospitalized SARS-CoV-2-positive patients, with diagnoses confirmed by quantitative reverse-transcription-polymerase chain reaction (RT-qPCR) analysis. Metagenomic sequencing analysis was performed to evaluate the gut microbiome profiles, and non-repeating HQMAGs (n=2568) were reconstructed, of which 33 HQMAGs showed differential distributions based on COVID-19 severity.
Individuals were categorized as mild, moderate, and critical/severe patients by their SARS-CoV-2 infection severity. A random forest (RF) algorithm-based machine learning (ML) classifier was developed to investigate whether the participants could be classified into the disease severity groups based on the 33 HQMAGs. Further, co-abundance analysis was applied to the 33 HQMAGs to assess gut microbiome interactions and determine the structures of the guilds (coherent functional groups) formed by them.
A genome-centered analysis of the metagenomes of the guilds was performed to explore the genetic basis of the associations between the two guilds and COVID-19 severity. Furthermore, guild-level microbiome index (GMI) values were calculated based on the mean difference in abundance between the guilds.
To investigate whether the obtained guild-level signature at hospital admission can be associated with COVID-19 prognosis, the RF-based regression analysis was performed. The team investigated if the microbiome signature at hospital admission can predict severe COVID-19 outcomes and whether the signature could be applied to other COVID-19 cohorts.
The validity of the study findings was determined by applying them to another study comprising 24 and 14 mild/moderate and critical/severe COVID-19 patients, respectively. The team investigated whether patients from differing COVID-19 severity groups could be classified based on the 33 HQMAGs and whether the signature could distinguish between SARS-CoV-2-positive SARS-CoV-2-negative individuals based on data from 66, 69, and nine SARS-CoV-2-positive patients, SARS-CoV-2-negative persons, and non-COVID-19 community-acquired pneumonia (CAP) patients, respectively.
Results
RF-based modeling based on the 33 HQMAGs classified participants by COVID-19 severity and co-abundance network analysis demonstrated the formation of two guilds by the 33 HQMAGs. Guild 1 comprised more short-chain fatty acid biosynthesis genes, and fewer antibiotic resistance and virulence genes, than Guild 2. The 33 HQMAGs at hospital admission could predict COVID-19 prognosis within a week of hospitalization. Further, Guild 1 was found to dominate over Guild 2 at hospital admission, and the dominance was predictive of critical COVID-19 outcomes such as death.
RF models could discriminate SARS-CoV-2-positive patients from healthy individuals, SARS-CoV-2-negative patients, and CAP patients in three different datasets. Of 2,568 HQMAGs, 48 had ≥five percent variability depending on COVID-19 severity, of which 17 were significantly higher among mild COVID-19 patients and formed Guild 1, including Faecalibacterium prausnitzii, Romboutsia timonensis, Clostridium, Ruminococcus, Allisonella histaminiformans, Negativibacillus, Acutalibacteraceae, Lachnospiraceae, and Coprococcus.
Of 48 HQMAGs, 31 were more abundant among critical/severe COVID-19 patients, of which 16 demonstrated significant differences by severity and formed Guild 2, including Enterococcus, Lactobacillus, Akkermansia muciniphila, Acutalibacteraceae, Barnesiella intestinihominis, Anaerotignum, Dore, Clostridium_M bolteae, Lachnospiraceae, Intestinibacter bartlettii, Ruthenibacterium lactatiformans, and Phascolarctobacterium faeciu. The two guilds correlated negatively with each other, indicative of a potentially competitive relationship between the guilds.
Guild 2 had more virulence factor (VF) genes, whereas Guild 1 had more antibiotic resistance genes (ARGs). After one week of hospitalization, GMI values correlated positively with interleukins (IL)-5,12p70, blood lymphocyte percentage and absolute counts, total cholesterol, calcium, and albumin levels. GMI scores correlated negatively with neutrophil percentage, D-Dimer, fibrinogen(B), fibrin-degradation products, total and direct bilirubin, glucose, and lactate dehydrogenase (LDH) levels.
Further, GMI values at hospitalization of the three deceased individuals were significantly lesser than those of the four discharged critical COVID-19 patients. The findings indicated that gut microbiome signature in the early stage may reflect COVID-19 outcomes among hospitalized patients, and GMI can be used as a predictor of COVID-19 severity.
Conclusion
Overall, the study findings highlighted the association between genome-centered guild-level gut microbiome signatures and COVID-19 severity and prognosis, which may aid in the early identification of COVID-19 patients and estimate the risk of severity outcomes.
*Important notice
bioRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
