Ethylene, an early marker of systemic inflammation in humans

Ethylene is a major plant hormone mediating developmental processes and stress responses to stimuli such as infection. We show here that ethylene is also produced during systemic inflammation in humans and is released in exhaled breath. Traces of ethylene were detected by laser spectroscopy both in vitro in isolated blood leukocytes exposed to bacterial lipopolysaccharide (LPS) as well as in vivo following LPS administration in healthy volunteers. Exposure to LPS triggers formation of ethylene as a product of lipid peroxidation induced by the respiratory burst. In humans, ethylene was detected prior to the increase of blood levels of inflammatory cytokines and stress-related hormones. Our results highlight that ethylene release is an early and integral component of in vivo lipid peroxidation with important clinical implications as a breath biomarker of bacterial infection.

hypothesized that ethylene is formed as part of endogenous lipid peroxidation caused by the respiratory burst and can be detected as a gaseous signature of systemic inflammation in exhaled breath.
To test this hypothesis, we firstly utilized an isolated lipid model to consolidate basic chemical mechanisms of ethylene release by lipid peroxidation. We also investigated ethylene release and lipid peroxidation in isolated leukocytes activated by bacterial lipopolysaccharide (LPS) as a cellular model of infection. Finally, we evaluated ethylene release in vivo using the experimental human endotoxemia (LPS administration in healthy volunteers), a well-characterized model of systemic inflammation. We compared exhaled ethylene with established biomarkers of endogenous lipid peroxidation and aimed to clarify the temporal and mechanistic relationship between ethylene release and other biomarkers of the inflammatory stress response.

Results
Peroxidation of unsaturated lipids leads to ethylene formation. Initially, we investigated whether the in vitro peroxidation of mammalian lipids directly leads to formation of ethylene. We resuspended L-αphosphatidylethanolamine (PE) extracted from pig brain in water containing the surfactant Triton X-100 and treated this suspension with Fe(II)SO 4 , a catalyst of Fenton oxidation 22 . Brain PE consists of approximately 30% polyunsaturated and 25% monounsaturated fatty acids (Avanti polar lipids). An incubation of 30 minutes resulted in significant ethylene release (p = 0.02) (Fig. 1a).
Lipid peroxidation, ethylene release and inflammatory cytokine production in immune cells. In order to expand to biological systems relevant to infection and inflammation, we investigated lipid peroxidation, ethylene release and inflammatory cytokine production in human cellular endotoxemia models. Monocyte-derived dendritic cells (moDCs) acquired from healthy donor blood were utilized to study basic mechanisms of lipid peroxidation and cytokine release in response to either known inducers of lipid peroxidation or LPS. MoDCs contain the NADPH oxidase NOX2, which is activated via LPS, leading to a rapid and sustained production of ROS [23][24][25] . Two independent assays were used to monitor lipid peroxidation. First, we used Bodipy 581/591-C11, which is a fluorescent probe mimicking unsaturated fatty acids 26 . We also investigated formation of the highly specific lipid peroxidation product 8-isoprostane with an enzyme immunoassay (EIA) 27 . In addition, we measured secretion of IL-6, a prototype cytokine associated with cellular activation. As positive controls for lipid peroxidation, we treated moDCs with Fe(II)SO 4 .
Exposure of moDCs to Fe(II)SO 4 led to a significant increase in Bodipy fluorescent intensity at 510 nm (p < 0.001) (Fig. 2a). It also resulted in formation of 8-isoprostane (p < 0.001), confirming significant oxidative Lipid peroxidation in moDCs measured by EIA assay for 8-isoprostane. Cells were treated with either LPS or 100 µM Fe(II)SO 4 in serum-free RPMI-1640 (n = 3). (c) Total ethylene released during 1 hour stimulation of leukocytes with either 1 µg/ml LPS or 100 µM Fe(II)SO 4 , as detected by laser-based photoacoustic spectroscopy (n = 11). Data points show mean with SEM (panels a and b) or geometric mean with 95% CI (panel c), asterisks indicate significance compared to untreated samples. stress and lipid peroxidation in these cells (Fig. 2b). Treating the moDCs for 1 hour with LPS also led to a blue-shift of the Bodipy 581/591-C11 probe (p = 0.004) (Fig. 2a). This response was attenuated by the lipophilic antioxidant vitamin E (α-tocopherol) (p < 0.001, compared to LPS) and the iron chelator deferoxamine 28 (DFO) (p = 0.005, compared to LPS), suggesting that iron content plays a role in LPS-induced lipid peroxidation (see Supplementary Fig. S1). LPS-activation of the moDCs also induced 8-isoprostane release (p = 0.004) (Fig. 2b), further demonstrating the role of lipid peroxidation in the biological effect of LPS.
The LPS treatment of moDCs resulted in significant concentration-dependent increases in IL-6 secretion with a maximum effect observed at 1 µg/ml LPS (p = 0.03) (see Supplementary Fig. S2a), while exposure to iron did not induce IL-6 production (p = 0.55) (see Supplementary Fig. S2b). Furthermore, deferoxamine or α-tocopherol did not affect IL-6 secretion either (p = 0.31; p = 0.43, respectively). Cytotoxicity data from the MTT assay indicated that, although LPS induces lipid peroxidation, this does not affect cell viability (p = 0.57), whereas Fe(II)SO 4 does (71% viability of control, p = 0.024). Addition of α-tocopherol or deferoxamine did not affect viability (p = 0.50; p = 0.75, respectively). These results suggest that lipid peroxidation is a physiological response to LPS exposure. Our observations confirm that LPS induces both lipid peroxidation and effective cellular signaling to activate cytokine release, however lipid peroxidation alone is insufficient to activate cytokine production.
To explore the direct link between LPS, lipid peroxidation and ethylene release in human leukocytes, we stimulated peripheral blood leukocytes (PBLs) with LPS or Fe(II)SO 4 for 1 hour and measured ethylene release by laser-based photoacoustic spectroscopy 29 (Fig. 2c). Compared to untreated cells, both LPS and Fe(II)SO 4 caused a significant increase in ethylene release (p = 0.04; p = 0.001, respectively). These data are supportive of the paradigm that human leukocytes produce ethylene upon LPS stimulation as a product of lipid peroxidation.

LPS causes ethylene formation in healthy volunteers.
In order to determine whether systemic inflammation leads to lipid peroxidation and ethylene release in vivo, we utilized the experimental human endotoxemia model. This is a well characterized, controlled and reproducible human model of systemic inflammation elicited by intravenous LPS administration 30 . In all 8 subjects that underwent endotoxemia, breath ethylene levels, detected by laser-based photoacoustic spectroscopy 29 increased rapidly after LPS administration and reached maximum levels after 60 minutes (Fig. 3a). In similar fashion, plasma levels of 8-isoprostane increased after LPS administration (Fig. 3b). Peak levels of 8-isoprostane were detected about 30 minutes after peak ethylene levels in breath. There was also a correlation between peak plasma 8-isoprostane and ethylene levels (r = 0.76, p = 0.03). To investigate the placebo effect, 8-isoprostane in plasma samples from healthy volunteers injected with phosphate buffered saline (PBS) in a previous study with identical setting as reported here, was also measured. In contrast to LPS injection, saline treatment did not result in increase of 8-isoprostane (Fig. 3b).
Interestingly, breath ethylene showed a biphasic response with levels decreasing below baseline 8 hours post-LPS, although this did not reach statistical significance. This might be caused by an increased antioxidant response, for example upregulation of glutathione levels or higher expression of ROS sequestering proteins. Another potential explanation could be auto-oxidation of NOX2. As ROS are known to react with proteins, often inactivating them in the process 25, 31 , this could self-limit the oxidative stress induced by LPS signaling.
In order to explore ethylene release in the context of broader stress responses, we also determined plasma levels of the stress hormones epinephrine, norepinephrine and cortisol (Fig. 4d). There were no changes in norepinephrine levels over time, but epinephrine and cortisol levels increased significantly compared to their baseline, reaching peak levels at 90 and 180 minutes after LPS administration, respectively. In addition, ethylene levels at various time-points (45, 90 and 120 minutes after LPS administration) correlated with peak levels of epinephrine (r = 0.89, p = 0.007; r = 0.85, p = 0.007; r = 0.74, p = 0.04), but not with norepinephrine or cortisol.

Discussion
Building on the well-recognized role of ethylene as a stress hormone in plants [1][2][3][4] , and previous findings linking ethylene to mammalian oxidative stress 9, 12-16 , we explored ethylene as a novel gaseous signature of the inflammatory response associated with bacterial infection. We have found that treating isolated leukocytes with LPS resulted in significant ethylene release caused by the respiratory burst. Furthermore, we show that humans produce ethylene as part of a systemic inflammatory response to bacterial endotoxin, and that this event precedes the classical inflammatory cytokine response. Collectively, these results signify ethylene as a novel biomarker for infection and oxidative stress with potentially important clinical implications.
We have conducted experiments with isolated mammalian lipids to decipher potential chemical mechanisms of ethylene formation during lipid peroxidation. Our data are relevant to both the general concept of oxidative modification of lipid membranes and to the exact process of ethylene formation during lipid peroxidation. Firstly, lipids within biological membranes are relatively vulnerable to oxidative attack compared to other cellular environments due to the high partition coefficient of oxygen in lipid membranes (≈5-fold compared to aqueous environment 36 ) and due to the lack of lipophilic antioxidants (except for α/γ-tocopherol) protecting these compartments. Secondly, we propose that, during oxidative stress, ethylene production proceeds by decomposition of fatty acids by ROS. This occurs spontaneously and is not catalyzed by enzymes. In this point, this mechanism would resemble the spontaneous isomerization of 7-dehydrocholesterol to previtamin D3 in the skin induced by UV-light 37,38 .
In order to test this novel paradigm to infection, we used leukocyte activation by LPS as a model of the inflammatory and oxidative response to infection. Leukocytes are principle mediators of host defense during all phases of infection and basic mediators of the innate immune response. They mount a strong oxidative response against invading bacteria, which is crucial for their elimination, and also as a pathological mechanism of tissue injury during uncontrolled oxidative stress 39 .
The overall aim of these cellular model experiments was to show ethylene release in an in vitro inflammation model and to utilize this model to forecast in vivo human phenomena. We have done this by characterizing lipid peroxidation and inflammatory cytokine release. Our experiments confirmed both intracellular lipid peroxidation to well-known stimuli and a clinically relevant oxidative stress biomarker response in the release of 8-isoprostane. Relevant to the inflammatory paradigm, activation of leukocytes by LPS also resulted in oxidative stress, 8-isoprostane release and ethylene production.
The human relevance of these discoveries was tested in a well characterized, controlled and reproducible model of human systemic inflammation. Remarkably, gaseous ethylene and 8-isoprostane in plasma were detected in such subjects for the first time and with important characteristics. Firstly, ethylene production started to increase rapidly following endotoxin administration to humans. It paralleled 8-isoprostane in plasma, thus providing an additional link between our in vitro and human data. Ethylene release appeared as part of the broader inflammatory and stress response, as we found it to correlate significantly with epinephrine and IL-10 release. The correlation with the archetypal anti-inflammatory cytokine IL-10 could indicate a potential link between endogenous lipid peroxidation and a central host defense response in triggering anti-inflammatory mechanisms. Ethylene production occurred earlier than the pro-and anti-inflammatory cytokine responses. This suggests the potential for ethylene as an early biomarker of oxidative stress in infection. For instance, the possibility of real-time breath analysis could be directly relevant to patients in intensive care units at risk of ventilator associated pneumonia 40 .
Our observation that ethylene is released prior to inflammatory cytokines also raises the exciting possibility that lipid peroxidation or ethylene itself could be involved in mounting the cytokine response. Signaling by lipid peroxidation products has been investigated before 41 and there are some indications that ethylene itself could have a signaling function. For instance, several cell lines (of both mouse and human origin) responded with increased Ca 2+ levels and altered gene expression after incubation with an ethylene-generating system 42 . In addition, ethylene has profound neurological effects and was widely used as an anesthetic in the 1930s, although this required very high concentrations 43 . Further studies are required to elucidate if ethylene is only a byproduct of lipid peroxidation or whether it has a signaling function.
In conclusion, we now show that humans release ethylene during systemic inflammation. The in vitro experiments suggested that blood leukocytes could be one of the cell types contributing to this. We could connect the chemical and biological factors causing the ethylene release in response to an inflammatory stimulus, showing that oxidation of unsaturated fatty acids plays a role in this reaction. Together, our results identify ethylene as a promising and novel biomarker for early onset of infection in the clinic.

Methods
Experiments with isolated lipid preparations. L-α-phosphatidylethanolamine from porcine brain (brain PE; 840022 P, Avanti), cis -4,7,10,13,16,19-docosahexaeinoic acid (DHA; Sigma, St. Louis, USA), cis-9-Octadecenoic (oleic) acid (Sigma, St. Louis, USA) and palmitic acid (Sigma, St. Louis, USA) were suspended in milliQ H 2 O containing 0.1% Triton X-100, which was deoxygenated by bubbling N 2 through it for 15 minutes. Isolated lipid samples were placed into plastic closed-cap T25 flasks (3055, Corning, Corning, USA) that were previously equilibrated by exposure to filtered air for >48 hours at 37 °C. Brain PE, oleic acid and palmitic acid were suspended at 1.4 g/l, while DHA was suspended at 0.7 g/l. To initiate lipid peroxidation, Fe(II)SO 4 (Sigma, St. Louis, USA) was added to a final concentration of 500 µM in 3 ml total volume. After Fe(II)SO 4 addition, the flasks were closed to allow gas accumulation in the headspace of the samples for 30 minutes before connecting to the ethylene detector.
For ethylene measurements, PBLs in serum-free RPMI-1640 (Thermo Fisher Scientific, Waltham, USA) were placed into plastic closed-cap T25 flasks (3055, Corning, Corning, USA) that were previously equilibrated by exposure to filtered air for >48 hours at 37 °C. Fe(II)SO 4 (Sigma, St. Louis, USA) was added to a final concentration of 500 µM in 4 ml total volume. LPS was added to a final concentration of 1 µg/ml in 4 ml total volume. After stimulation, the flasks were closed to allow gas accumulation in the headspace of the samples for 60 minutes before connecting to the ethylene detector.
Experimental human endotoxemia. The study procedures are described in detail elsewhere 45 . Briefly, subjects were admitted to the research unit of the intensive care department where a cannula was placed in an antecubital vein to permit infusion of 0.9% NaCl solution; the subjects received 1.5 L 0.9% NaCl during one hour starting one hour before endotoxin infusion (pre-hydration) followed by 150 ml/h until 6 hours after endotoxin infusion and 75 ml/h until the end of the experiment. The radial artery was cannulated using a 20-gauge arterial catheter (Angiocath, Becton Dickinson, Sandy, Utah) and connected to an arterial pressure monitoring set (Edwards Lifesciences LLC, Irvine, CA, USA) to allow the continuous monitoring of blood pressure and blood sampling. Heart rate was recorded using a 3-lead electrocardiogram. Purified LPS (US Standard Reference Endotoxin Escherichia Coli O:113) obtained from the Pharmaceutical Development Section of the National Institutes of Health (Bethesda, MD), supplied as a lyophilized powder, was reconstituted in 5 ml saline 0.9% for injection and vortex-mixed for at least 20 minutes after reconstitution. The LPS solution was administered as an intravenous bolus injection at a dose of 2 ng/kg body weight in one minute at t = 0 hours. Subjects were continuously monitored up till 8 hours after LPS administration.
Breath collection. Each subject provided two single mouth-exhaled breath samples collected separately in 1-liter Mylar bags (ABC ballonnen, Zeist, The Netherlands) with 40 seconds pause in between samples. Breath samples were collected using a hand held breath sampler that is custom-built according to the ATS Guidelines and validated previously 46,47 . The exhaled breath line consists of a bacterial filter (Air Safety Limited, Lancashire, UK), a discard bag (150 ml) for collection of the dead space, a pressure meter and ends with the sampling bag. The pressure meter provides a back pressure of 10 cm H 2 O that ensures soft palate closure and prevents nasal contamination and was calibrated for 50 ml/s exhalation flow. After a few tests on the exhalation procedure, the subjects were asked to exhale while having visual feedback from 3 LEDs (two green and one red) indicating suitable pressure/exhalation flow rate for the sampling (when both green LEDs were on). Breath samples were collected before LPS administration (approximately 5 minutes), every 15 minutes from 0 to 2 hours after LPS administration, and every hour from 4-8 hours after LPS administration. In addition, ambient air samples were collected every hour.
8-isoprostane assay. 8-isoprostane levels in supernatants from moDCs and in plasma samples from healthy volunteers undergoing experimental human endotoxemia were determined using an enzyme immune assay (EIA) according to the manufacturer's instructions (Cayman Chemical, Ann Arbor, USA).
SCIENTIfIC REPoRTs | 7: 6889 | DOI:10.1038/s41598-017-05930-9 Cytokine/chemokine/stress hormone assays. For the determination of circulating cytokine levels from healthy volunteers undergoing experimental human endotoxemia, EDTA anticoagulated blood was centrifuged immediately after collection (2000g, 10 minutes, 4 °C), after which plasma was stored at −80 °C until analysis. Concentrations of TNF-α, IL-6, IL-8, MCP-1, IL-1RA, and IL-10 were measured using a simultaneous Luminex Assay according to the manufacturer's instructions (Milliplex, Millipore, Billerica, MA, USA). Circulating levels of epinephrine, norepinephrine, and cortisol were determined as described previously 45 . Ethylene detection. Ethylene was detected by a CO 2 laser-based photoacoustic detector (ETD-300, Sensor Sense) as described previously 29 . Briefly, the ETD-300 uses a CO 2 laser emitting light in the 10 μm region where ethylene presents the strongest absorption features. The energy from the laser light causes local heating of the gas sample and by switching the light on and off with a certain frequency, rapid heating/cooling occurs, resulting in a periodical pressure change (a sound wave). The amplitude of this sound wave is proportional to the absorbed energy, i.e., the concentration of the ethylene in the absorption cell, and this can be detected with high-sensitivity microphones. The ETD-300 is capable of measuring ethylene on-line in the 300 pptv (pptv = parts-per-trillion volume, 1:10 12 ) range within a 5 seconds time scale. All gas samples were led through a KOH-CaCl 2 scrubber before entering the detector to remove water and carbon dioxide. The ethylene released in the headspace of the lipid samples during the 30 minute incubation was transported to the ethylene detector using clean air as carrier gas at a constant flow of 3 L/hr. Each sample was measured for 10 minutes in stop and flow mode of the instrument 48 . The PBL samples were measured in the same way, but were incubated for 60 minutes. For ethylene detection in breath of healthy volunteers, the content of the sampling bags was introduced into the ethylene detector with a small pump at a constant flow of 3 L/h. Each sample was measured for about 8 minutes (the steady state C 2 H 4 concentration was reached within 2 min) and an averaged value of the last 2 minutes was determined. The ethylene concentration in exhaled breath was measured in duplicate for each subject and the average value used for analysis. All sampling bags were analyzed within 24 hours. The stability of exhaled ethylene in the Mylar bags was tested prior the endotoxemia experiment from 0 to 24 hours. 15 mL of ethylene from a certified calibrated mixture of 1 ppmv ethylene in air (Linde Gas Benelux, Netherlands) was injected into a Mylar bag filled with 1 L exhaled breath. Each bag was analyzed directly after injection of ethylene, and randomly after 1, 2, 4, 8, 16 or 24 hours. The experiment was performed in triplicate and showed that ethylene concentration in the bags was preserved within a 2.6% error range.

Statistical analysis.
In vitro results are presented as mean ± SEM or geometric mean with 95% CI of at least three different donors/independent experiments. Repeated measures one-way ANOVA with Bonferroni post-hoc tests, paired 2-sided Student's t-tests or Wilcoxon matched-pairs signed rank tests were applied to assess significance. In vivo data are represented as median and interquartile range. Friedman tests with Dunn's post-hoc test were applied to assess significance for the in vivo data. Furthermore, Pearson correlation on log-transformed data was used to investigate the relationship between parameters. A value of p < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001).
Data availability. All data is available from the authors upon request.