SARS-CoV-2 infection is associated with intestinal permeability, systemic inflammation, and microbial dysbiosis in hospitalized patients

Abstract

Coronavirus disease 2019 (COVID-19) and its associated severity have been linked to uncontrolled inflammation and may be associated with changes in the microbiome of mucosal sites including the gastrointestinal tract and oral cavity. These sites play an important role in host-microbe homeostasis, and disruption of epithelial barrier integrity during COVID-19 may potentially lead to exacerbated inflammation and immune dysfunction. Outcomes in COVID-19 are highly disparate, ranging from asymptomatic to fatal, and the impact of microbial dysbiosis on disease severity is unclear. Here, we obtained plasma, rectal swabs, oropharyngeal swabs, and nasal swabs from 86 patients hospitalized with COVID-19 and 12 healthy volunteers. We performed 16S rRNA sequencing to characterize the microbial communities in the mucosal swabs and measured concentrations of circulating cytokines, markers of gut barrier integrity, and fatty acids in the plasma samples. We compared these plasma concentrations and microbiomes between healthy volunteers and COVID-19 patients, some of whom had unfortunately died by the end of the study enrollment, and performed a correlation analysis between plasma variables and bacterial abundances. Rectal swabs of COVID-19 patients had reduced abundances of several commensal bacteria including Faecalibacterium prausnitzii and an increased abundance of the opportunistic pathogens Eggerthella lenta and Hungatella hathewayi. Furthermore, the oral pathogen Scardovia wiggsiae was more abundant in the oropharyngeal swabs of COVID-19 patients who died. The abundance of both H. hathewayi and S. wiggsiae correlated with circulating inflammatory markers including IL-6, highlighting the possible role of the microbiome in COVID-19 severity and providing potential therapeutic targets for managing COVID-19.IMPORTANCEOutcomes in coronavirus disease 2019 (COVID-19) are highly disparate and are associated with uncontrolled inflammation; however, the individual factors that lead to this uncontrolled inflammation are not fully understood. Here, we report that severe COVID-19 is associated with systemic inflammation, microbial translocation, and microbial dysbiosis. The rectal and oropharyngeal microbiomes of COVID-19 patients were characterized by a decreased abundance of commensal bacteria and an increased abundance of opportunistic pathogens, which positively correlated with markers of inflammation and microbial translocation. These microbial perturbations may, therefore, contribute to disease severity in COVID-19 and highlight the potential for microbiome-based interventions in improving COVID-19 outcomes.

Keywords: COVID-19; gut microbiome; host–pathogen interactions; human microbiome.

Fig 1

Plasma concentrations of cytokines in COVID-19 patients and healthy controls. The plasma concentrations of IL-6 (A), IL-2 (B), IL-10 (C), IL-17A (D), IL-23 (E), IL-8 (F), IL-1β (G), IL-12p70 (H), IL-18 (I), IFN-γ (J), and TNF-α (K) from hospitalized COVID-19 patients and controls. Tukey’s box shows the IQR; middle line, median; vertical lines, adjacent values (1st −1.5 IQR; 3rd quartile +1.5 IQR). Kruskal–Wallis tests were performed for all analytes but only followed by pairwise Dunn tests if the false discovery rate (FDR)-adjusted P-value was <0.05; thus, all plots showing pairwise comparisons had significant Kruskal–Wallis tests. Pairwise Dunn tests within each cytokine were also FDR-adjusted using the Benjamini–Hochberg method (47, 48). P-values <0.05 were considered significant.

Fig 2

Plasma concentrations of gut barrier damage markers and circulating fatty acids in COVID-19 patients and healthy controls. Plasma concentrations for I-FABP (A), zonulin (B), LBP (C), and sCD14 (D) measured by enzyme-linked immunosorbent assay (ELISA) and circulating fatty acids propionic acid (E), decanoic acid (F), butyric acid (G), nonanoic acid (H), and isovaleric acid (I) measured by liquid chromotagraphy-tandem mass spectrometry ( LC-MS/MS). Comparisons between groups were performed by Kruskal–Wallis tests for all analytes, but only followed by Dunn post hoc tests if adjusted P-values were below 0.05; thus, all plots showing pairwise comparisons had significant Kruskal–Wallis tests. Pairwise Dunn tests within each variable were also FDR-adjusted using the Benjamini–Hochberg method (48). P-values <0.05 were considered significant.

Fig 3

Correlations between cytokines, markers of gut barrier damage, fatty acids, hospital lab results, and SARS-CoV-2 viral loads. Spearman correlation plot; asterisks denote FDR-adjusted P-values <0.05 (A). Representative correlations of interest for hospital labs are shown in (B), gut barrier damage in (C), short-chain fatty acids in (D), and SARS-CoV-2 viral loads in (E). Lines represent the simple linear regression, and shaded areas represent the 95% confidence intervals.

Fig 4

Identifying biomarkers most important in separating COVID-19 patients and healthy controls. sPLS-DA of plasma biomarkers, hospital labs, and swab SARS-CoV-2 viral loads from oropharyngeal and nasal swabs used to identify the variables most important in discriminating patients who died or survived or healthy control. The ordination of the samples is shown in component 1 and component 2 and colored by their group (A). Variables contributing to those components are shown below (B) and indicate each variable’s loading value and direction. Each variable is colored by the group that had the highest median value.

Fig 5

Rectal, oropharyngeal, and nasal microbial communities in COVID-19 patients and healthy controls. V3V4 16S rRNA gene sequencing was used to characterize the microbial communities in healthy controls and hospitalized COVID-19 patients who survived or died in rectal swabs (A), oropharyngeal swabs (B), and nasal swabs (C). Alpha diversity was measured by the Shannon diversity index, and significance was tested by the Mann–Whitney U test for two group comparisons or Kruskal–Wallis tests with Dunn post hoc comparisons for three groups. Beta diversity was determined by the Bray–Curtis distance at the amplicon sequence variant (ASV) level and tested by PERMANOVA; ellipses represent 95% confidence intervals. Beta-dispersion was tested using the permutest function in the vegan R package with the group variances determined by the Bray–Curtis distance matrix; box plots show the Bray–Curtis distances between each sample in the same group. Factors affecting the microbial composition were determined by PERMANOVA; P-values for each variable are shown next to the bar plot.

Fig 6

Differentially abundant taxa in the rectal, oropharyngeal, and nasal swabs of patients with severe COVID-19. Differentially abundant taxa between hospitalized COVID-19 patients and healthy controls were identified using MaAsLin2 for each sample type at the phylum, genus, and species level—requiring an FDR-adjusted P-value less than 0.05 to be considered significant. (A) Differentially abundant taxa were identified in rectal swabs; due to a limited number of collected rectal swabs, patients from the survived and deceased groups were combined for comparison to healthy controls. (B) Differentially abundant taxa identified in oropharyngeal and nasal swabs. Red cells indicate an enrichment of that taxa in the reference group (listed first on the x-axis) relative to the comparison group (listed second on the x-axis), and blue cells indicate a depletion of that taxa. Asterisks denote FDR-adjusted P-values as follows: ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001. (C) Representative box plots showing the relative abundances of several taxa of interest from the rectal and oropharyngeal swabs.

Fig 7

Bacteria enriched in severe COVID-19 are correlated with pro-inflammatory cytokines, markers of gut barrier damage, and SARS-CoV-2 viral loads. Heat maps depicting the correlation between the relative abundance of bacterial species enriched in rectal swabs (A), oropharyngeal swabs (B), and markers of inflammation, tight junction permeability, and oropharyngeal SARS-CoV-2 viral loads. Spearman’s rank correlation tests were used for statistical analysis; asterisks denote FDR-adjusted P-values as follows: ^*P < 0.05; ^**P < 0.01.