Lung microbiota composition, respiratory mechanics, and outcomes in COVID-19-related ARDS

Abstract

Few data are available on the lung microbiota composition of patients with coronavirus disease 2019-related acute respiratory distress syndrome (C-ARDS) receiving invasive mechanical ventilation (IMV). Moreover, it has never been investigated whether there is a potential correlation between lung microbiota communities and respiratory mechanics. We performed a prospective observational study in two intensive care units of a university hospital in Italy. Lung microbiota was investigated by bacterial 16S rRNA gene sequencing, performed on bronchoalveolar lavage fluid samples withdrawn after intubation. The lung bacterial communities were analyzed after stratification by respiratory system compliance/predicted body weight (Crs) and ventilatory ratio (VR). Weaning from IMV and hospital survival were assessed as secondary outcomes. In 70 C-ARDS patients requiring IMV from 1 April through 31 December 2020, the lung microbiota composition (phylum taxonomic level, permutational multivariate analysis of variance test) significantly differed between who had low Crs vs those with high Crs (P = 0.010), as well as in patients with low VR vs high VR (P = 0.012). As difference-driving taxa, Proteobacteria (P = 0.017) were more dominant and Firmicutes (P = 0.040) were less dominant in low- vs high-Crs patients. Similarly, Proteobacteria were more dominant in low- vs high-VR patients (P = 0.013). After multivariable regression analysis, we further observed lung microbiota diversity as a negative predictor of weaning from IMV and hospital survival (hazard ratio = 3.31; 95% confidence interval, 1.52-7.20, P = 0.048). C-ARDS patients with low Crs/low VR had a Proteobacteria-dominated lung microbiota. Whether patients with a more diverse lung bacterial community may have more chances to be weaned from IMV and discharged alive from the hospital warrants further large-scale investigations.

Importance: Lung microbiota characteristics were demonstrated to predict ventilator-free days and weaning from mechanical ventilation in patients with acute respiratory distress syndrome (ARDS). In this study, we observed that in severe coronavirus disease 2019 patients with ARDS who require invasive mechanical ventilation, lung microbiota characteristics were associated with respiratory mechanics. Specifically, the lung microbiota of patients with low respiratory system compliance and low ventilatory ratio was characterized by Proteobacteria dominance. Moreover, after multivariable regression analysis, we also found an association between patients' microbiota diversity and a higher possibility of being weaned from mechanical ventilation and discharged alive from the hospital. For these reasons, lung microbiota characterization may help to stratify patient characteristics and orient the delivery of target interventions. (This study has been registered at ClinicalTrials.gov on 17 February 2020 under identifier NCT04271345.).

Registered at ClinicalTrials.gov, 17 February 2020 (NCT0427135).

Keywords: ARDS; COVID-19; microbiota; mortality; respiratory mechanics.

Fig 1

Lung microbiota composition in “low-Crs” and “high-Crs” patients. (A) Low-Crs patients (shown as red-colored circles and ellipse) had a bacterial community composition that differed from that of high-Crs patients (shown as blue-colored circles and ellipse), as assessed by permutational multivariate analysis of variance and visualized by principal coordinate analysis. (B) Biplot analysis showed the family-level bacterial taxa underlying the differences. (C) Rank abundance analysis of dominant family taxa revealed that the lung microbiota of low-Crs patients was enriched of Moraxellaceae and Lactobacillaceae, whereas that of (D) high-Crs patients was enriched with Paenibacillaceae, as assessed by the Kruskall–Wallis test.

Fig 2

Lung microbiota composition in “low-VR” and “high-VR” patients. (A) Low-VR patients (shown as red-colored circles and ellipse) had a bacterial community composition that differed from that of high-VR patients (shown as blue-colored circles and ellipse), as assessed by permutational multivariate analysis of variance and visualized by principal coordinate analysis. (B) Biplot analysis showed the family-level bacterial taxa underlying the differences. (C) Rank abundance analysis of dominant family taxa revealed that Paenibacillaceae were less dominant in low-VR patients than in (D) high-VR patients, as assessed by the Kruskall–Wallis test.

Fig 3

Graphic representation of respiratory mechanics, aerated tissue, and lung microbiota composition in two patients. Computed tomography scan images of lungs from two representative patients with (A) high Crs/PBW (0.61 mL/cmH₂O/kg) and VR (3.64) values or (B) low Crs/PBW (0.43 mL/cmH₂O/kg) and VR (1.65) values depicted the distribution of aerated lung tissue, which was detailed in the respective wave frequency graphs. For each patient, the bubble chart shows dominant families in the lung bacterial community

Fig 4

Correlation between serum procalcitonin and lung microbiota composition. Relationship assessment between lung microbiota-composing phyla and C-ARDS laboratory variables by the use of Spearman’s rank test showed the absence of correlation of Actinobacteria and Bacteroidetes with the (A) D-dimer, (B) LDH, or (C) PCT serum level. Firmicutes and Proteobacteria were, respectively, found to correlate negatively or positively only with the (C) PCT serum level.

Fig 5

Cumulative incidence of the primary outcome according to the Shannon diversity index. Kaplan–Meier curve analysis showed that patients with a high Shannon diversity index value (solid blue line) differed significantly from patients (dotted blue line) with a low Shannon diversity value regarding the median time from BAL sampling to being weaned from mechanical ventilation and alive (8 and 25 days, respectively, log-rank P = 0.0016).