A high-risk gut microbiota configuration associates with fatal hyperinflammatory immune and metabolic responses to SARS-CoV-2

Abstract

Protection against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and associated clinical sequelae requires well-coordinated metabolic and immune responses that limit viral spread and promote recovery of damaged systems. However, the role of the gut microbiota in regulating these responses has not been thoroughly investigated. In order to identify mechanisms underpinning microbiota interactions with host immune and metabolic systems that influence coronavirus disease 2019 (COVID-19) outcomes, we performed a multi-omics analysis on hospitalized COVID-19 patients and compared those with the most severe outcome (i.e. death, n = 41) to those with severe non-fatal disease (n = 89), or mild/moderate disease (n = 42), that recovered. A distinct subset of 8 cytokines (e.g. TSLP) and 140 metabolites (e.g. quinolinate) in sera identified those with a fatal outcome to infection. In addition, elevated levels of multiple pathobionts and lower levels of protective or anti-inflammatory microbes were observed in the fecal microbiome of those with the poorest clinical outcomes. Weighted gene correlation network analysis (WGCNA) identified modules that associated severity-associated cytokines with tryptophan metabolism, coagulation-linked fibrinopeptides, and bile acids with multiple pathobionts, such as Enterococcus. In contrast, less severe clinical outcomes are associated with clusters of anti-inflammatory microbes such as Bifidobacterium or Ruminococcus, short chain fatty acids (SCFAs) and IL-17A. Our study uncovered distinct mechanistic modules that link host and microbiome processes with fatal outcomes to SARS-CoV-2 infection. These features may be useful to identify at risk individuals, but also highlight a role for the microbiome in modifying hyperinflammatory responses to SARS-CoV-2 and other infectious agents.

Keywords: COVID-19; Microbiota; cytokines; inflammation; metabolites; tryptophan.

Figure 1.

Circulating immune mediators in COVID-19 patients. (a) PCA plot illustrating the differences in serum cytokine and inflammatory mediator levels in COVID-19 patients with different levels of severity. (b) Heatmap illustrates the serum immune mediators that are significantly increased (red), significantly decreased (blue), or remain unchanged (green). (c) Levels of the cytokines that are significantly different in patients with severe COVID-19 that survive (labeled “Severe”), compared to those with severe COVID-19 that have a fatal outcome (labeled “Fatal”). Results are expressed as mean and standard deviation. Differences between groups are calculated using the Kruskal–Wallis test and Dunn’s multiple comparison test (*p < .05, **p < .01, ***p < .001, ****p < .0001).

Figure 2.

Serum metabolites in COVID-19 patients. (a) PCA plot for the four conditions: control, mild/moderate, severe, fatal; (b) Barplot representing super pathways of the significant metabolites (LIMMA, FDR <0.05) between each comparison of conditions; (c) importance plot and confusion matrix from the random forest classifier between the four conditions.

Figure 3.

Serum metabolites in COVID-19 patients. (a) Heatmap representing metabolites from pathways of interest, listed at the bottom of the figure, divided according to group. Log fold change (LFC) for significant pairwise comparisons (LIMMA, FDR <0.05) are included. Sulfonated bile acids and metabolites of microbial origin are indicated. (b) Weighted co-expression network labeled for metabolites from pathways of interest. (c) Pathway enrichment analysis using Metabolon terms for communities 1, 3 and 5 (significant terms are displayed, gseapy, FDR <0.2). (d) Subset of metabolites of targeted pathways from co-expression network analysis. (e) Weighted co-expression network labeled for those metabolites that were significantly different between severe COVID-19 patients that survived versus those that died (LIMMA, FDR <0.05).

Figure 4.

Gut microbiome composition in COVID-19 patients. (a) Principal coordinate analysis of the genus-level microbiome composition of the three outcome groups of patients obtained using the Canberra distance measure. (b) Variation of the silhouette-Scores obtained, across for cluster sizes (k), for 50 iterations of k-means clustering of the first three dominant principal coordinates of the genus-level microbiome profiles. The principal coordinates of these two microbiome groups are demarcated in (c). The two microbiome groups exhibited distinct patterns of association with three COVID-19 disease severity outcome groups (d). (e) Volcano plot illustrates genera showing either significant (FDR ≤0.15, shown in blue) or nominally significant (P ≤ .05, shown in cyan) associations with PCo1. The x-axis shows the estimate of the linear-regression models (direction indicating the pattern of association) and y-axis shows the -logarithm of the p-value to the base 10. The genera associating with the high-risk MicrobiomeGroup1 are on the negative axis and those associating with low-risk MicrobiomeGroup2 are on the positive axis. Only those genera showing associations with P ≤ .05 are shown.

Figure 5.

Modules that positively correlate with severe and fatal COVID-19. Feature-to-feature positive association networks obtained using the ccrepe approach (Spearman correlations, 1000 iterations) for modules (or Module groups) that show (a) significantly positive (‘turquoise’) and (b) significantly negative (‘red’, ‘blue’, ‘yellow’, and ‘black’) associations with severe and fatal COVID-19. In (b) given the presence of features from four different modules, the location of the features belonging to the different modules are indicated in the smaller network representation in the lower left-hand corner. Microbiome, cytokine and metabolite features that are associated with severity and death are highlighted in different colors.  .