Dysbiosis: An Indicator of COVID-19 Severity in Critically Ill Patients

Abstract

Here, we examined the dynamics of the gut and respiratory microbiomes in severe COVID-19 patients in need of mechanical ventilation in the intensive care unit (ICU). We recruited 85 critically ill patients (53 with COVID-19 and 32 without COVID-19) and 17 healthy controls (HCs) and monitored them for up to 4 weeks. We analyzed the bacterial and fungal taxonomic profiles and loads of 232 gut and respiratory samples and we measured the blood levels of Interleukin 6, IgG, and IgM in COVID-19 patients. Upon ICU admission, the bacterial composition and load in the gut and respiratory samples were altered in critically ill patients compared with HCs. During their ICU stay, the patients experienced increased bacterial and fungal loads, drastic decreased bacterial richness, and progressive changes in bacterial and fungal taxonomic profiles. In the gut samples, six bacterial taxa could discriminate ICU-COV(+) from ICU-COV(-) cases upon ICU admission and the bacterial taxa were associated according to age, PaO2/FiO2, and CRP levels. In the respiratory samples of the ICU-COV(+) patients, bacterial signatures including Pseudomonas and Streptococcus were found to be correlated with the length of ICU stay. Our findings demonstrated that the gut and respiratory microbiome dysbiosis and bacterial signatures associated with critical illness emerged as biomarkers of COVID-19 severity and could be a potential predictor of ICU length of stay. We propose using a high-throughput sequencing approach as an alternative to traditional isolation techniques to monitor ICU patient infection.

Keywords: bacterial and fungal microbiome; composition and load; gut and lung; mechanical ventilation; severe COVID-19 cases in ICU.

Figure 1

Diagram of participant groups and number of samples collected and processed. The study population consisted of three groups of subjects: 17 HCs, 32 patients admitted to the ICU before the SARS-CoV-2 pandemic (ICU-COV(−)), and 53 patients with COVID-19 admitted to the ICU (ICU-COV(+)). Fecal samples were collected from HCs, rectal swabs from all ICU patients, and tracheal aspirates and blood from ICU-COV(+) patients. HCs and ICU-COV(+) patients were monitored for up to 4 weeks and ICU-COV(−) patients were monitored for up to 2 weeks. The reason the number of samples decreased over time was that the patients were either discharged or deceased before their samples were collected, and the reason the number of samples analyzed is lower than those collected was that some samples were not amplifiable by PCR.

Figure 2

Gut bacterial taxa associated with ICU-COV(+) patients and their predictive value. Comparison of the gut microbiome sequence data between ICU-COV(+) and ICU-COV(−) patients who did not receive antibiotics or corticosteroids before ICU admission using MaAsLin2 identified six bacterial species (FDR < 0.22). ICU = intensive care unit; COV(−) = COVID-19-negative; COV(+) = COVID-19-positive; RS = rectal swab; TA = tracheal aspirate.

Figure 3

Dynamics of the gut and respiratory bacterial load and diversity. (a) Bacterial load in critically ill patients as assessed by qPCR of the V4 region of the 16S rRNA gene at ICU admission and over time spent in the ICU. * p < 0.05, ** p < 0.001, *** p < 0.0001, **** p < 0.0001, Mann–Whitney test unless three groups were tested, then Kruskal–Wallis test was applied (a). (b) Bacterial richness as assessed by the Chao1 index analysis of the 16S rRNA sequences upon ICU admission, and progressive decrease in this richness in both the gut and respiratory microbiome from admission day to 2 weeks for ICU-COV(−) patients, up to 4 weeks for ICU-COV(+) patients, one fecal sample at baseline and one 4 weeks later for HCs. a. Kruskal–Wallis test and ANOVA test for more than two groups and non-parametric or parametric data, respectively. (c) Beta-diversity analyses using unweighted UniFrac Principal Coordinate Analysis (PCoA). T0 = baseline; T1W = sample collected 1 week after baseline; T2W = sample collected after 2 weeks after baseline; T4W = samples collected after 4 weeks after baseline. ** p < 0.005, 5 PERMANOVA test (adonis2 function). ICU = intensive care unit; COV(−) = COVID-19-negative; COV(+) = COVID-19-positive; RS = rectal swab; TA = tracheal aspirate.

Figure 4

Bacterial taxonomic profiling over time in the ICU. (a) Significant changes in the gut microbiome of the ICU-COV(−) patients at baseline, and 1 and 2 weeks after ICU admission. (b) Significant changes in the gut microbiome of ICU-COV(+) patients at baseline, and 2 and 4 weeks after admission. (c) Significant changes in the tracheal aspirates of ICU-COV(+) patients at baseline, and 2 and 4 weeks after ICU admission. Intra-group variability across the three time points was evaluated using MaAsLin2. Only significant results have been plotted (FDR < 0.05). ICU = intensive care unit; COV(−) = COVID-19-negative; COV(+) = COVID-19-positive; RS = rectal swab; TA = tracheal aspirate. Species inside most affected over time are highlighted in red boxes.

Figure 5

Association between clinical and laboratory data and gut microbiome composition data at baseline using MaAsLin2. Only significant results are shown (FDR < 0.25 are specified for each association). Gut microbiome data are plotted in raw counts.

Figure 6

Association between clinical/laboratory data and microbiome composition data of the tracheal aspirate at baseline for ICU-COV(+) patients using MaAsLin2. Only the most significant results are shown (FDR < 0.05 are specified for each association). Respiratory microbiome data are plotted in raw counts.

Figure 7

Dysbiosis scores calculated for each sample and association with clinical data. (a) Dysbiosis scores in critically ill patients were calculated based on the weighted UniFrac distance metrics of the 16S rRNA sequence data of rectal swab samples, using fecal samples of HCs as reference data. * p < 0.05, ** p < 0.001, *** p < 0.0001, Mann–Whitney test. (b) Correlation between dysbiosis scores and clinical data such as the PaO2/FiO2 ratio, length of ward stay before ICU, and duration of pre-ICU clinical symptoms (Spearman correlation test). The correlations were performed using rectal swab samples obtained at baseline for both ICU groups.