Circulating microbial cell-free DNA is associated with inflammatory host-responses in severe pneumonia

Abstract

Host inflammatory responses predict worse outcome in severe pneumonia, yet little is known about what drives dysregulated inflammation. We performed metagenomic sequencing of microbial cell-free DNA (mcfDNA) in 83 mechanically ventilated patients (26 culture-positive, 41 culture-negative pneumonia, 16 uninfected controls). Culture-positive patients had higher levels of mcfDNA than those with culture-negative pneumonia and uninfected controls (p<0.005). Plasma levels of inflammatory biomarkers (fractalkine, procalcitonin, pentraxin-3 and suppression of tumorigenicity-2) were independently associated with mcfDNA levels (adjusted p<0.05) among all patients with pneumonia. Such host-microbe interactions in the systemic circulation of patients with severe pneumonia warrant further large-scale clinical and mechanistic investigations.

Keywords: pneumonia.

Figure 1:. Plasma microbial cell-free DNA levels are elevated in culture-positive pneumonia compared with culture-negative pneumonia and uninfected controls.

A: Patients with culture-positive pneumonia had higher levels of plasma mcfDNA Molecules per Microliter (MPMs) compared to patients with culture-negative pneumonia, who in turn had also higher plasma mcfDNA levels compared to uninfected controls (pairwise comparisons post hoc adjusted by Benjamini-Hochberg method). B: Types of mcfDNA (bacterial, fungal or viral) detected in culture-positive, culture-negative pneumonia and in uninfected controls. The radius of pie charts scales quadratically proportional to the sum of mcfDNA MPMs detected within each patient subgroup. The proportion of viral mcfDNA was significantly higher in the culture-negative (17.7%) compared to the culture-positive pneumonia (1.7%) group (p<0.0001 for z test of comparison of proportions).

Figure 2:. Circulating mcfDNA is associated with host inflammatory responses in patients with pneumonia. A. Linear regression results for association between microbial cell-free DNA (mcfDNA) and host response biomarkers.

We built linear regression models of plasma biomarkers (outcomes) against plasma mcfDNA levels (predictor) in unadjusted as well as adjusted models for a priori selected potential confounders: i) a surrogate of the microbial inoculum (culture-positive vs. negative classification), ii) degree of lung injury as depicted radiographically by radiographic severity index (RSI) and by the epithelial injury biomarker receptor for advanced glycation end product (RAGE), and iii) host innate immunity status (age, chronic obstructive pulmonary disease and immunosuppression). Estimated regression coefficient, 95% confidence intervals and p-values for significance of mcfDNA MPMs for each regression model are reported. B. Patients with pneumonia assigned to the hyperinflammatory subphenotype had significantly higher mcfDNA compared to hypoinflammatory patients (median 7,731, interquartile range-IQR, MPMs, [3,100-79,849] vs. 546 [0-4,609] respectively, p<0.05). We assigned patients to the hyper- vs. hypo-inflammatory subphenotype based on a parsimonious predictive model utilizing levels of angiopoietin-2, procalcitonin, TNFR1 and bicarbonate (see supplemental methods). Ang-2, angiopoietin-2; CI, confidence interval; IL, interleukin; RAGE, receptor for advanced glycation end product; ST-2, suppression of tumorigenicity-2; TNFR-1, tumour necrosis factor receptor 1.