Neutrophil extracellular traps (NETs) are increased in the alveolar spaces of patients with ventilator-associated pneumonia

BALF specimens were collected from mechanically ventilated patients undergoing bronchoscopy for suspected VAP (n = 100) in intensive care units at Harborview Medical Center (Seattle, WA). VAP was defined according to the IDSA guidelines by a quantitative bacterial culture of > 10,000 CFU/ml in the BALF [21]. ARDS was defined according to the Berlin criteria as respiratory failure with acute onset in the setting of a known clinical insult, bilateral opacities on chest imaging, and hypoxemia with PaO₂/F_iO₂ ≤ 300 not fully explained by cardiac failure on the day of bronchoscopy [22]. The onset of ARDS was defined as the day of the first qualifying P/F ratio combined with an appropriate chest x-ray. Subjects were included if they were mechanically ventilated in the medical, surgical, or neurologic intensive care units and bronchoscopy was performed if the patient had new radiologic abnormalities, leukocytosis, purulent secretions, or fever. Subjects were excluded if they were less than 18 years of age, pregnant, prisoners, had a known diagnosis of HIV, previous bone marrow or solid organ transplant, or were admitted for severe burns or with metastatic cancer. The characteristics of this study population have been previously described [23]. This part of the study was approved by the University of Washington’s Human Subjects Committee and samples were obtained under a waiver of consent.

BALF was also obtained from healthy subjects undergoing research bronchoscopy as previously described [24]. This part of the study was also approved by the University of Washington’s Human Subjects Committee and samples were obtained after written consent.

The BALF specimen was filtered through a 70-μm cell strainer and then centrifuged at 300 × g for 5 min for critically ill subjects and for 10 mins at 4 °C for healthy subjects. The supernatant was collected as BALF and frozen at –80 °C. All samples were thawed only once to perform the following studies.

NETs were analyzed as described previously [14, 25]. Briefly, anti-human MPO (clone 4A4, AbD Serotec) and anti-DNA-HRP (Cell Death Detection ELISA Kit, Roche) were used as capture and detection antibodies. The NET-related markers peroxidase activity and cf-DNA were measured by colorimetric assays [14, 25]. Interleukin (IL)-8 concentrations were measured by immunoassay (Mesoscale Discovery, Rockville, MD). S100A8/A9 (calprotectin) levels were analyzed by enzyme-linked immunosorbent assay (ELISA; R&D Systems). Samples that fell below the lower limit of detection were assigned that value.

Bronchoalveolar lavage mitochondrial (mt) and genomic (g) DNA content was analyzed by measuring the copy number of two mitochondrial genes (MT-RNR2, MT-TL1) and two genomic genes (B2M, RNA18SN5). Briefly, 8 μl of BALF was mixed with 10 μl of powersybr Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) and 2 μl of forward (12.5 μm) and reverse (12.5 μm) primers for MT-RNR2, MT-TL1, B2M, or RNA18SN5 at a final volume of 20 μl (Additional file 1: Table S1). As a standard curve, synthesized oligonucleotides of the target sequence were used (Additional file 2: Table S2). Each quantitative real-time polymerase chain reaction (PCR) was run on a steponeplus™ Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Samples were denatured at 95 °C for 15 min prior to 40 cycles of amplification (95 °C for 15 s, 60 °C for 60 s). Sequence copy numbers were calculated by comparing C_T values of samples and a standard curve.

Statistics were performed on untransformed NET concentrations using nonparametric tests. For comparisons across multiple groups we performed the Kruskal-Wallis test and post-test Dunn’s to adjust for multiple comparisons. Correlations were performed using the Spearman’s rank-order correlation. For regression analyses, NET concentrations were log₁₀-transformed. We adjusted for age and gender given literature that supports both age and sex in affecting innate immune response, particularly in the lung [1, 2]. We further adjusted for ARDS as it could affect both NET formation and risk for VAP but would not necessarily be on the causal pathway between the two. Analyses were performed using GraphPad Prism (La Jolla, CA) and Stata 14 (College Station, Texas).

All mechanically ventilated subjects underwent bronchoscopy for clinical suspicion of VAP (n = 100). On average, subjects were 52 (±18) years of age and predominantly male (n = 79, 80%). Subjects were relatively evenly distributed amongst those without ARDS or VAP (n = 25), those with ARDS but without VAP (n = 19), those without ARDS but with VAP (n = 30), and those with both ARDS and VAP (n = 26). The average length of mechanical ventilation prior to bronchoscopy did not differ greatly by clinical condition and ranged from 4.3 to 7.8 days (Table 1). Of the subjects meeting ARDS criteria, the majority did so prior to the day of bronchoscopy with the exception of three subjects who met ARDS criteria on the day of bronchoscopy. The subjects overall had a high severity of illness measured by Acute Physiology and Chronic Health Evaluation (APACHE) III score (mean score across groups 75.8–87.7) and high mortality (rate across groups 21–36%). ARDS subjects were predominantly from surgical rather than medical ICUs, and there was a high proportion of trauma-related ARDS (Table 1). The four healthy subjects ranged in age from 18 to 22 and 75% were female.

We first comprehensively characterized NET concentrations (MPO-DNA complexes) and other markers of NETosis (peroxidase and cf-DNA) in BALF from critically ill subjects. NETs were detectable in BALF from all subjects with a median (interquartile range (IQR)) of 223 (40.6–766) U/mL. Of the patients included in the study, 69 subjects were not on antibiotics for 48 h prior to bronchoscopy. Thirty-one subjects were on antibiotics, but these antibiotics had not changed 48 h prior to the bronchoscopy. We compared the level of NETs by MPO-DNA complex in subjects on antibiotics with those not on antibiotics and there was no significant difference (p = 0.13). In healthy subjects, NETs were undetectable in all four BALF samples. Further markers of NETosis were not obtained in the healthy subjects. Peroxidase activity (6.40 (1.34–15.5) mU/mL) and cf-DNA (87.74 (20.0–253) ng/mL) varied widely across critically ill subjects. We have previously shown that both nuclear and mitochondrial DNA are present in NETs, with mitochondrial DNA being more proinflammatory [14]. The range of nuclear DNA content by 18S RNA and β-2 microglobulin ranged from 7881 (441–49,257) copies/μL and 7026 (297–39,765) copies/μL, respectively. Mitochondrial DNA content was also highly variable, ranging from 65.3 (25.3–253) copies/μL for 16S RNA and 1229 (333–3863) copies/μL for MT-TL1. There was a strong correlation between NET concentration and peroxidase activity and cf-DNA (p < 0.0001; Table 2). However, correlations between content of 18S, β-2 microglobulin, mitochondrial 16S, and MT-TL1 DNA content and NETs were either only weakly associated or independent. These data show that, across a spectrum of mechanically ventilated patients with and without pneumonia or ARDS, NETs and other markers of NETosis were present and highly variable.

We next wanted to identify whether there were correlations between NETs and markers of NETosis and VAP. We compared NET concentrations between subjects with or without VAP further stratified by ARDS status. We found that MPO-DNA complex, peroxidase, and cf-DNA concentrations were significantly different between these groups by Kruskal-Wallis (all p < 0.0001; Fig. 1). After adjusting for multiple comparisons, MPO-DNA complexes in subjects with ARDS and VAP were higher than those with ARDS alone (p < 0.05; Fig. 1a). We further tested for associations between increased NETs and risk of VAP by logistic regression. For each log₁₀ increase in MPO-DNA complex concentration, there was an increased risk of VAP (odds ratio (OR) 3.03, 95% confidence interval (CI) 1.69–5.43; p < 0.0001). Multiple logistic regression to adjust for age, gender, and ARDS status showed that increased log₁₀ (MPO-DNA) was again associated with VAP (OR 3.75, 95% CI 1.91–7.35; p < 0.0001). When we restricted the analysis to patients without VAP, MPO-DNA complexes were not associated with ARDS (OR 2.09, p = 0.11). Peroxidase activity and cf-DNA were also associated with increased odds of VAP (Table 3).

Neither nuclear DNA content by 18S or β-2 microglobulin DNA nor mitochondrial DNA content by 16S and MT-TL1 DNA were statistically different amongst these populations (Additional file 3: Figure S1). Additionally, we did not identify any associations between markers of NETosis and ARDS severity or mortality. These results show that NET concentrations differ amongst subjects with ARDS and/or VAP and that the presence of VAP is strongly associated with NET concentrations.

We subsequently determined whether MPO-DNA and markers of NETosis were associated with other measures of neutrophil activation or local inflammation in the alveolar space. Calprotectin, also known as S100A8/A9, is a cytosolic protein measure of neutrophil activation [26]. The chemokine IL-8 is a known marker of inflammation in ARDS and VAP and a strong neutrophil chemoattractant [27]. We found that both MPO-DNA complex (r = 0.88, p < 0.0001), peroxidase (r = 0.81, p < 0.0001), and cf-DNA (r = 0.86, p < 0.0001) were positively correlated with calprotectin (Fig. 2a–c). There was a similarly strong correlation to levels of IL-8 (Fig. 2d–f). This indicates a strong link between NETosis and local inflammation and neutrophil degranulation.

Quantitative bacterial cultures for each subject provided us with the ability to identify associations between the burden of bacteria present in the alveolar space and concentrations of NETs. With each higher quartile of bacterial content, there was an increase in log-transformed concentration of MPO-DNA complexes (β = 0.30, p < 0.0001), peroxidase (β = 0.24, p < 0.0001), and cf-DNA concentrations (β = 0.19, p < 0.0001; Fig. 3). Because bacterial content is linked to the presence VAP, we wanted to test whether bacterial burden within clinical groups with VAP was still associated with increased NETosis. In subjects with both ARDS and VAP or VAP alone, we dichotomized patients into high and low colony count groups. We found that amongst subjects with both VAP and ARDS, MPO-DNA concentration and peroxidase activity were significantly higher in subjects with higher bacterial colony counts (p = 0.03 and p = 0.02, respectively; Additional file 4: Figure S2A–C). Similarly, MPO-DNA concentration and cf-DNA were higher with higher bacterial colony counts in subjects with VAP alone (Additional file 4: Figure S2D–F). When we compared concentrations of markers of NETosis and bacterial quality by gram stain in all critically ill subjects, there was no difference in MPO-DNA concentration, peroxidase activity, or cf-DNA concentration (Additional file 5: Figure S3A–C). Taken together, increasing bacterial burden is associated with increased NETosis.