cfDNA correlates with endothelial damage after cardiac surgery with prolonged cardiopulmonary bypass and amplifies NETosis in an intracellular TLR9-independent manner

Cardiopulmonary bypass (CPB) provokes inflammation culminating in organ dysfunction and increased mortality. Recently, neutrophil extracellular traps (NETs) have been found to be involved in a variety of cardiovascular diseases promoting tissue and organ injury. Here, we aimed to elaborate the proinflammatory potential of circulating cell-free (cf)DNA in patients undergoing cardiac surgery with CPB. Plasma was collected pre- and postoperatively as well as at d1, d3, d5 and d8 after surgery. At d1, we found circulating cfDNA levels to be significantly increased in patients with prolonged CPB duration (>100 min) when compared to those with shorter CPB times (CPB < 100 min). Increased CPB duration yielded in higher levels of circulating mitochondrial (mt)DNA, soluble thrombomodulin (sCD141) and ICAM-1, reflecting endothelial damage. Positive correlation between cfDNA and sCD141 was demonstrated at all time points. Plasma and cfDNA from patients with CPB > 100 min induced NETs release by neutrophils from healthy donors which was not suppressed by inhibitors of intracellular toll-like receptor (TLR)9. DNA binding to neutrophils’ surface (s)TLR9 has been evidenced. Altogether, we demonstrate that elevated plasma cfDNA might be useful to assess CPB-mediated detrimental effects, including endothelial damage, in cardiac surgical patients with prolonged CPB duration. cfDNA-triggered NETosis is independent of classical TLR9 signaling.


Results
Patient demographics and clinical scores. Patients' baseline demographics, surgery information as well as physiologic parameters are summarized in Table 1. Among all patients twenty-two underwent cardiac surgery with CPB < 100 min and twenty-six patients underwent cardiac surgery with CPB > 100 min. Mean age at the time of investigation did not differ between the two groups. Both groups had comparable cardiovascular comorbidities such as hypertension, COPD and diabetes mellitus. Most of the patients included in the CPB < 100 min group underwent aortic valve replacement, whereas aortic valve replacement combined with coronary artery bypass grafting represented the most performed surgical procedure on patients with CPB > 100 min. Significant increase in aortic cross-clamping time, days of hospitalization and EuroSCOREs was found in patients with CPB > 100 min. Among the study population one patient died postoperatively at day 7. Table 2 summarizes the biochemical and clinical parameters of patients after on-pump surgery. Characteristics of patients undergoing off-pump artery bypass grafting (OPCAB) without CPB can be found as Supplementary  Table S1.
Elevated levels of circulating cfDNA and altered neutrophil TLR9 expression in patients with long-term CPB. In order to study the impact of CPB duration on the onset of inflammation, we first quantified levels of cfDNA in patients' plasma at the time of admission and postoperatively at defined time points. Plasma cfDNA levels strongly increased after surgery in patients undergoing cardiac surgery with CPB < 100 min (short-CPB) and CPB > 100 min (long-CPB), suggesting activation of proinflammatory pathways by the extracorporeal circuit (Fig. 1a). In patients with CPB > 100 min, cfDNA levels remained significantly elevated at least until day 5 and significant intergroup differences were already detected at day 1 after surgery. No significant increase of cfDNA was observed in patients undergoing OPCAB surgery ( Supplementary Fig. S1), indicating that cfDNA increase doesn't represent a common inflammatory response induced by the surgical trauma. Importantly, increase in circulating cfDNA was not related to a reduced plasma DNase I activity as no alternations in DNase I activity were detected over time (Fig. 1b). Thus, elevated cfDNA levels are attributed to an increased number of NETing neutrophils or damaged cells, respectively.
Recent work has demonstrated that mitochondrial DNA (mtDNA) might represent an ingredient of NETs structures among others 17 and its proinflammatory property in DAMP-associated inflammation is well documented 20 . We therefore questioned if mtDNA might also be detected in circulating cfDNA in patients' plasma. Real time PCR analyses of DNA isolated from plasma samples revealed that mtDNA levels were significantly increased in patients with long-CPB vs. those with short-CPB at admission, indicating that patients included in the long-CPB group display a more severe disease and are more prone for systemic inflammation (Fig. 1c). Circulating mtDNA levels markedly increased postoperatively in patients with long-term CPB, especially at day 3 and day 8 after surgery. At these times, significant intergroup differences were observed. mtDNA is known to activate neutrophils through TLR9/p38 MAPK 22 and it has already been reported that it might upregulate TLR9 expression 23 . Here, patients undergoing cardiac surgery with CPB > 100 min displayed significantly reduced neutrophil TLR9 baseline levels at admission vs. patients with short-CPB, although TLR9 levels were found to become significantly upregulated immediately after surgery as well as at day 1 (Fig. 2). However, neutrophil TLR9 levels remained clearly reduced at day 3 and day 5 after surgery when compared to patients with shorter CPB duration, beside higher levels of circulating mtDNA. Compared to healthy controls (median 21.7, IQR 6.7-67.22), all patients had significantly elevated neutrophil TLR9 levels at admission time (**p < 0.01 vs. short-CPB (median 84.64; IQR 73.4-95.33); *p < 0.05 vs. long-CPB (median 62.63; IQR 31.9-88.1)).
Positive correlation between cfDNA and endothelial damage in patients with long CPB duration. Taken into account that cfDNA might include large amounts of NETs which represent potential inducers of endothelial damage, we further assessed plasma concentrations of soluble thrombomodulin (sCD141) and ICAM-1 which represent common molecular markers of endothelial cell injury (Fig. 3a,b). Whereas sCD141 levels in patients with short-CPB didn't change over time, sCD141 levels clearly increased after surgery in patients with long CPB duration. Significant intergroup differences were observed immediately after surgery as well as at day 3. Additionally, markedly elevated levels of circulating ICAM-1 were quantified in plasma samples of patients with CPB > 100 min with significant intergroup difference found at day 1 after surgery. In contrast, proinflammatory IL-6 ( Fig. 3c) and CRP (Fig. 3d) levels, although tending to be higher in the long-CPB group, did not differ between patients of the two groups. Similarly, no changes in leukocyte counts (Fig. 3e) were found. In support of our hypothesis, cfDNA levels in patients with long CPB duration were significantly positively correlated with sCD141 at all times after surgery despite the limited number of patients included (Fig. 4a). Further on, neutrophil TLR9 expression showed significant negative correlation with sCD141 at day 1 and day 8 after surgery (Fig. 4b), arguing for a protective role of TLR9 on endothelial dysfunction. Importantly, sCD141 and cfDNA did not correlate in patients undergoing surgery with CPB < 100 min (data not shown).
Circulating cfDNA contributes to patients' plasma-induced NETosis. Inflammatory cytokines are known to play a critical role for NETosis 24 and circulating mtDNA was found to amplify inflammation 19 by inducing NETs release via TLR9 activation amongst others 25 . Neutrophils isolated from healthy volunteers, which were already reported to be unstimulated 26 , were activated in vitro with pooled plasma from volunteers as well as from cardiac surgery patients (five patients of each group), and cfDNA/NETs were quantified in culture supernatants (Fig. 5a). cfDNA/NETs were significantly increased and highest in supernatants of cells stimulated with plasma from patients with long CPB duration when compared to control groups. To elicit the role of mtDNA  Fig. 5b, only stimulation of neutrophils with DNA isolated from plasma of patients with long CPB duration significantly increased cfDNA/NETs levels in culture supernatants. This finding implicates a role of mtDNA for NETosis induction and is consistent with our data showing increase in NETs formation by plasma of patients with long-term CPB. Degradation of plasma cfDNA by DNase I markedly impaired the onset of plasma-induced NETosis (Fig. 5c). Nevertheless, cytokines largely contribute to NETs formation, as plasma DNA degradation did not completely inhibit NETosis.
Patients' plasma-induced, ROS-independent NETosis does not depend on intracellular TLR9 signaling. As demonstrated above, plasma cfDNA triggers in vitro NETosis in human neutrophils. Therefore, we further examined whether TLR9 is involved in the mechanism of plasma-induced NETs formation. Two common inhibitors of TLR9 signaling, ODN A151 and chloroquine were used for cell culture experiments. Surprisingly, blockage of intracellular TLR9 signaling didn't suppress plasma-induced NETosis, but rather partially increased NETs release by cultured neutrophils (Fig. 6a,b). Plasma-driven NETosis also did not depend on ROS (Fig. 6a), as in contrast to PMA-triggered NETosis (Fig. 6c), no diminution of NETs formation in the presence of the NADPH inhibitor DPI could be observed. Of note, chloroquine also did not affect PMA-induced NETosis (Fig. 6c).
Recently, a functional form of the TLR9 receptor on neutrophils' surface has been reported 27 . We further investigated a possible implication of neutrophil surface TLR9 (sTLR9) for patients' plasma-triggered NETosis. Neutrophil sTLR9 expression was evidenced by immunofluorescence ( Fig. 7a) and flow cytometry (Fig. 7b). In general, most TLR9 was expressed intracellularly. Treatment of neutrophils with the TLR9 agonist CpG ODN 2006-FITC and subsequent staining of the cell membrane with fluorescent wheat germ agglutinin revealed increase of double positive cells, suggesting DNA binding to a surface receptor. However, CpG ODN 2006-FITC was also internalizied by the cells after 3 hours of incubation and chloroquine did not block uptake of oligonucleotides (Fig. 7c). Notably, similar to our findings demonstrating increased plasma-triggered NETosis in the presence of chloroquine, chloroquine significantly increased the amount of NETs in culture supernatants after cell stimulation with plasma-derived cfDNA (Fig. 7d). However, chloroquine alone did not influence NETosis (not shown). As illustrated in Fig. 7e, CpG ODN 2006 showed colocalization with neutrophil sTLR9, whereby only cells displaying high levels of sTLR9 strongly bound to CpG ODN. Hence, CpG ODN 2006, which is similar to bacterial DNA, provoked strong increase in NETs release which was further reinforced by chloroquine.

Discussion
Despite major improvements in surgical techniques, the use of CPB is still a key requisite helping to bypass circulation out of the heart and lung during the operation. Morbidity associated with CPB can be attributed to the generalized inflammatory response induced by blood-artificial surfaces interactions during extracorporeal circulation and the ischemia/reperfusion implications, including exacerbated inflammatory response resembling SIRS. It is widely recognized that the CPB-induced inflammation is largely mediated by activated neutrophils which shed NETs that have marked effects on inflammation. In this regard, recent evidence has emerged suggesting that NETs might have a role in infectious as well as noninfectious diseases, including, but not limited to, sepsis, SIRS 24,28,29 . In the current study, we demonstrate that circulating cfDNA might represent a feasible early biomarker to assess the onset of CPB-induced inflammation and endothelial damage in patients undergoing open   heart surgery with CPB. In addition, we provide evidence for plasma cfDNA-triggered NETosis by a mechanism which is different from the classical TLR9 signaling.
Here, we found that cfDNA levels were significantly increased and highest in patients undergoing cardiac surgery with long-term CPB when compared to those with shorter extracorporeal circulation time. We further speculated that increase in cfDNA might indicate massive neutrophil activation due to prolonged bypass times resulting in excessive NETosis. This assumption was supported by increased creatinine values and increased risk for adverse outcomes found in patients with long CPB duration. However, levels of proinflammatory IL-6, which is also a potent inducer of NETosis 30 , did not differ significantly between groups. On the other hand, proinflammatory biomarkers such as IL-6 and CRP were recently reported to be inappropriate to assess systemic inflammation after cardiac surgery 31 . Importantly, patients with CPB > 100 min displayed significant increase in circulating mtDNA being in line with data reported by other groups 32,33 . As mtDNA did not directly correlate with cfDNA plasma levels, they might not only represent NETs components 17,34 but rather become continuously released from necrotic tissue. Indeed, NETs were found to damage the host tissue, contribute to the development of autoimmunity and lead to other dysfunctional outcomes, including metastasis, thrombosis and inappropriate coagulation 29 . The detrimental effects of long-term CPB and NETs formation are supported by our observation showing a positive correlation between cfDNA and sCD141 in patients' plasma starting immediately after surgery and persisting until at least day 8. However, due to technical and logistic issues, a very small number of patients were included in the mentioned anlayses representing a major limitation of this study. Hence, although our data strongly suggest a link between cfDNA and endothelial damage, further analyses including a larger number of patients or an established animal model are necessary to confirm our findings.
Furtheron, if mtDNA, which was also elevated in patients with CPB > 100 min, directly contributes to endothelial damage or is rather a consequence of it cannot be definitely assessed. Activation of the immune system by mtDNA is based on the recognition and binding of unmethylated CpG dinucleotides to endosomal TLR9. In recent years, neutrophils have been found to express sTLR9 which is potentially able to engage DAMPs, such as mtDNA released from necrotic cells during sterile inflammation. Till date, only a few studies exist describing the function of neutrophil sTLR9 27,35 and they largely propose an involvement in detrimental effects by provoking massive proinflammatory cytokine production. Thus, sTLR9-expressing neutrophils are key players in inducing rapid inflammation 36 , promoting sepsis or SIRS 37 . Of interest, we found that plasma-triggered NETosis does not involve intracellular TLR9 signaling. Treatment of cells with inhibitors of TLR9 signaling, e.g. chloroquine, strongly increased plasma-mediated NETosis, suggesting that blockage of intracellular TLR9 signaling might strengthen signaling mediated by sTLR9. Indeed, we demonstrate here for the first time that neutrophils expressing sTLR9 bind high amounts of extracellular DNA, yielding in NETosis, again pointing for a role of sTLR9 as a mediator of rapid immune responses. sTLR9 and NETosis, respectively, were not suppressed by the inhibitors used in this study. Thus, our results are in line with the findings reported by Lindau et al. demonstrating that the signaling pathway mediated by sTLR9 is different from the classical TLR9 pathway 27 . As NETs are known to stimulate TLR9 in dendritic cells 38 , we suggest that cfDNA, enclosing NETs, might amplify the inflammatory response by activating sTLR9 and NETosis in adjacent neutrophils, whereby a possible involvement of other DNA sensors is not negligible. Mechanistically, cfDNA-triggered NETosis is independent on ROS production. In this regard, ROS-independent NETosis has already been reported as a very rapid process stimulated by Leishmania amazonensis 39 amongst others. Additionally, patient plasma-triggered NETosis does not solely depend on cfDNA and other proinflammatory cytokines and DAMPs are also implicated in this process.
It has been found that mtDNA upregulates TLR9 expression in macrophages and also neutrophils 23,27 . Similarly, plasma from patients with acetaminophen-induced acute liver failure upregulated neutrophil TLR9 expression in vitro and this increase was abrogated by preincubation with DNase I 40 . Here, neither an in vivo correlation between TLR9 and mtDNA nor an in vitro receptor upregulation by patients' plasma or plasma cfDNA could be detected (unpublished results). Although patients with CPB > 100 min displayed rapid TLR9 upregulation after surgery, it has to be taken into account that we did not differentiate between endosomal and sTLR9 expression representing a further major limitation or our study. Thus, in vivo sTLR9 regulation, especially after prolonged bypass, cannot be surely excluded. On the other hand, negative correlation of total TLR9 and sCD141 might indicate TLR9-linked endothelial protection. On the basis of our data, we speculate that increased NETosis in patients with prolonged CPB duration is rather due to an upregulation of neutrophil sTLR9 whereby only a minor implication of endosomal TLR9 can be assumed.
In summary, cfDNA represents an early biomarker for CPB-induced inflammation and a potential mediator of endothelial damage after cardiac surgery with prolonged bypass duration. To our knowledge, our study demonstrates for the first time that cfDNA boosts inflammation, amplifying NETosis by a novel mechanism which is independent on endosomal TLR9 and ROS and at least partially depends on sTLR9. However, the therapeutic value of cfDNA-targeting molecules/enzymes and the functional role of neutrophil sTLR9 in inflammation should be validated in future clinical and experimental studies using animal CPB models.

Methods
Patients and sample collection. This prospective pilot study was approved by the Ethics Committee of the Medical Faculty of the University of Cologne (#15-393). Written informed consent was obtained from all participants at the time of admission. Forty-eight patients undergoing any kind of on-pump cardiac surgery between April 2016 and March 2017, aged 18 years or older were enrolled in this study. Exclusion criteria were withdrawal of patients' consent, infectious diseases and malignancy. In addition, patients with known preexisting acute or chronic immunological disorders or systemic immunosuppression were excluded from the study. The preoperative risk for operative mortality was evaluated by means of the additive and logistic EuroSCORE 41 . Additionally, the ICU scores SAPS II and TISS 42,43 were calculated for each patient. Included patients were stratified according to the time on CPB (cut-off: 100 min) 44 . In addition, fifteen patients undergoing off-pump coronary artery bypass (OPCAB) surgery were included (#13-230).
Fourteen healthy volunteer blood donors (4 male, 10 female) aged between 21 and 60 years served as controls. Exclusion criteria were age under 18 years, recent or ongoing infection, cardiovascular disorders, immunosuppressive therapy or pregnancy.
Blood samples were collected into tubes containing lithium heparin (Sarstedt, Cologne, Germany) at the day of admission, immediately after surgery as well as postoperatively at day 1, day 3, day 5 and day 8. In some cases, no blood samples were collected from patients or patients were released from the hospital before day 8 after surgery, respectively. Blood samples were immediately used for neutrophil isolation and subsequent analyses. Plasma samples were harvested by centrifugation (10 min at 3000 × g) and stored at −80 °C until further processing. Isolation of human neutrophils, culture conditions and cell activation. Human neutrophils were separated from whole blood by discontinuous density-gradient centrifugation on Percoll (Millipore) 45 . In brief, blood was carefully layered onto the percoll gradient and centrifuged at 2000 rpm for 30 min without break. After removal of PBMCs, erythrocytes were lysed using isotonic ammonium chloride solution (155 mM NH 4 Cl, 10 mM KHCO 3 , 0.1 mM EDTA) at 4 °C for 10 min. The purity and viability of neutrophils were >95% as examined by flow cytometry analysis and trypan blue exclusion, respectively.

Determination of TLR9 expression by flow cytometry.
To assess TLR9 protein expression in neutrophils, 1 × 10 6 neutrophils were fixed in 4% paraformaldehyde (PFA), permeabilized in methanol and stained with PE-labelled rat anti-human TLR9 antibody (clone eB72-1665, BD Pharmingen) in 0.1% Triton X-100 and 3% FCS Figure 6. Patient plasma-triggered NETs formation occurs independently of intracellular TLR9 signaling and ROS. (a) Neutrophils from healthy volunteers (1 × 10 6 ) were pre-cultured with different concentration of ODN A151 (n = 7), chloroquine (CQ) (n = 5) and DPI (n = 5) for 30 min before medium supplementation with 20% pooled plasma from three healthy controls, and four patients with short-CPB or long-CPB, respectively. After incubation for 3.5 hours at 37 °C, cfDNA/NETs were quantified in culture supernatants by Pico green DNA staining. *p < 0.05; **p < 0.01; ***p < 0.001. (b) Freshly isolated neutrophils from controls were seeded on poly-D-lysine coated coverslips and allowed to adhere for 30 min in the presence of the inhibitors CQ (10 µg/ ml) and DPI (10 µM). Then, cells were stimulated with control plasma (healthy volunteers), pooled plasma of patients with short and long CPB duration for 3.  in cell wash buffer (BD) for 30 min in the dark. After washing two times with 0.1% Triton X-100 and 3% FCS in cell wash buffer (BD), cells were analyzed by flow cytometry using a MACSQuant Analyzer (Miltenyi Biotec) or FACSCalibur (BD). For this, neutrophils were gated using forward and side scatter. For surface TLR9 staining, non-permeabilized cells were incubated with the same antibody in cell wash buffer with 3% FCS for 30 min on ice. Cells were washed twice and immediately analyzed by flow cytometry.
Quantification of cfDNA and NETs. Plasma levels of cfDNA were quantified by Quant-iT Pico Green dsDNA assay by following the manufacturer's instructions (Invitrogen GmbH, Darmstadt, Germany) 46 . For the quantification of NETs in cultured neutrophils, cells were suspended and centrifuged at 300 × g for 5 min at 4 °C. Supernatants containing NETs were collected and quantified by Pico Green staining. The fluorescence intensity reflects the amounts of DNA and was measured at excitation and emission wavelengths of 485 nm and 530 nm, respectively in a microplate reader (Victor3, PerkinElmer). A standard calibration curve by means of defined calf thymus DNA (Sigma Aldrich) amounts ranging from 0 to 2 µg/ml has been used in all analyses.
Measurement of plasma DNaseI activity. DNase I activity in patients' plasma was measured by the method previously described with some modifications 47  Heparinase I treatment of samples. Digestion of plasma samples using heparinase I (Sigma Aldrich) was performed as previously described 48 . In brief, heparinase I (50 U) was dissolved in 50 mM Tris (pH 7.5) and 1 mM CaCl 2 , and stored at −20 °C. 5 µl of freshly isolated DNA was added to 19 µl of heparinase I and samples were incubated for 2 hours at 25 °C. 5 µl of the treated DNA solution was used as template in subsequent real time PCR reactions.

Quantification of mitochondrial and nuclear DNA in plasma samples by Real time PCR.
Relative mitochondrial DNA (mtDNA) copy numbers were quantified by Real time PCR and normalized by simultaneous detection of nuclear DNA (nucDNA). The ABI StepOne Plus Real-Time PCR System (Applied Biosystems) was used to amplify GAPDH, the nuclear gene, and the mtDNA-encoded ATPase 8 gene by using previously published primer sequences 49 . PCR was performed in 12.5 µl of total reaction volume containing 5 µl treated DNA (10 ng-20 ng), 6.25 µl 2 × SYBR Green PCR master mix (Applied Biosystems) and 800 nM forward and reverse primers. The following thermal cycling conditions were used: denaturation step at 95 °C for 10 min, 40 cycles of 60 sec at 60 °C and 15 sec at 95 °C. All samples were analyzed in duplicate. The average threshold cycle number (Ct) values were determined in the same quantitative PCR run. The threshold was set in the same level for each run. The level of mtDNA was calculated relative to GAPDH as followed: relative copy number 2 ∆Ct , whereby ∆Ct = Ct nuc − Ct mit . Immunofluorescence staining and fluorescence microscopy. For NETs staining, freshly isolated neutrophils (1 × 10 6 ) were seeded on poly-D-lysine coated coverslips and incubated in the presence of 10 µM DPI or 10 µg/ml chloroquine for 30 min. Then, cells were stimulated with serum or phorbol 12-myristate 13-acetate (PMA, Sigma Aldrich) for 3.5 hours. After that, samples were fixed with 4% PFA, blocked and incubated with a polyclonal rabbit anti-myeloperoxidase antibody (ab9535, 1:100, Abcam) and a secondary goat anti-rabbit Alexa Fluor 488-conjugated antibody (1:1000; Cell Signaling Technology). To demonstrate binding of DNA to neutrophils' surface, PMN were incubated with FITC-conjugated CpG ODN 2006 (2 µM, Invivogen). Plasma membrane was stained with Rhodamine Wheat Germ Agglutinin (5 µg/ml, Vector Laboratories). For colocalization studies of DNA and surface TLR9, neutrophils were fixed and incubated with PE-conjugated anti-TLR9 antibody (clone eB72-1665, Invitrogen). For intracellular TLR9 staining, cells were permeabilized with PBS + 1% FCS + 0.3% TritonX-100 prior incubation with anti-TLR9 antibody. Isotype control was used as control. All cells were counterstained with DAPI (Cayman Chemical) and specimens were mounted in Dako fluorescent mounting medium (Dako). Cells were finally visualized by immunofluorescence microscopy (Eclipse Ti-U 100, Nikon) using the NIS Elements BR 3.10 software package. Neutrophil sTLR9 surface expression and colocalization with FITC-CpG-ODN 2006 were confirmed by confocal microscopy using the Zeiss LSM 710 Inverted Meta confocal microscope (Carl Zeiss).