Metabolic changes during carbon monoxide poisoning: An experimental study

Abstract Carbon monoxide (CO) is the leading cause of death by poisoning worldwide. The aim was to explore the effects of mild and severe poisoning on blood gas parameters and metabolites. Eleven pigs were exposed to CO intoxication and had blood collected before and during poisoning. Mild CO poisoning (carboxyhaemoglobin, COHb 35.2 ± 7.9%) was achieved at 32 ± 13 minutes, and severe poisoning (69.3 ± 10.2% COHb) at 64 ± 23 minutes from baseline (2.9 ± 0.5% COHb). Blood gas parameters and metabolites were measured on a blood gas analyser and nuclear magnetic resonance spectrometer, respectively. Unsupervised principal component, analysis of variance and Pearson's correlation tests were applied. A P‐value ≤ .05 was considered statistically significant. Mild poisoning resulted in a 28.4% drop in oxyhaemoglobin (OHb) and 12‐fold increase in COHb, while severe poisoning in a 65% drop in OHb and 24‐fold increase in COHb. Among others, metabolites implicated in regulation of metabolic acidosis (lactate, P < .0001), energy balance (pyruvate, P < .0001; 3‐hydroxybutyrc acid, P = .01), respiration (citrate, P = .007; succinate, P = .0003; fumarate, P < .0001), lipid metabolism (glycerol, P = .002; choline, P = .0002) and antioxidant‐oxidant balance (glutathione, P = .03; hypoxanthine, P < .0001) were altered, especially during severe poisoning. Our study adds new insights into the deranged metabolism of CO poisoning and leads the way for further investigation.

oedema, unconsciousness and coma. 5 If patients survive, neurological sequelae can appear. 1 When poisoning is suspected, measurement of blood carboxyhaemoglobin (COHb) is performed. 6,7 Normal blood COHb levels range between 1% and 3%; however, up to 10% has been detected in active smokers. 8 Mild signs and symptoms of toxicity can be present at COHb levels ranging from 3% to 24%, 5,6 while loss of consciousness normally occurs at levels above 24%, and exposure to higher levels has been shown to be fatal. 1,6 The treatment of CO poisoning consists of O 2 therapies. While oxygen can be administered at normal pressure, normobaric oxygen therapy or as hyperbaric oxygen therapy, current evidence shows that treatment does not necessarily influence patient outcome. 9,10 Also, although COHb is the gold diagnostic standard, it is a poor predictor of the severity of acute toxicity. For example, studies have shown that the relationship between COHb levels and the severity of clinical symptoms is not well correlated because of the time between exposure cessation and the effects of supplemental oxygen treatments prior to COHb measurements. 6,7 In order to improve evaluation of the severity of CO intoxication, focus is needed on understanding the molecular mechanisms involved in its progression. 9,10 Under physiological conditions, endogenously produced CO is involved in cellular regulation of numerous physiological systems, including brain and muscle oxygen storage and utilization, relaxation of vascular and extra-vascular smooth muscle, modulation of synaptic neurotransmission, and anti-inflammatory, -apoptotic, anti-proliferative and anti-thrombotic processes. During exposure to abnormal exogenous CO levels, however, these physiological mechanisms are disrupted. Inhaled CO diffuses rapidly across the alveolar-capillary membrane and becomes excessively absorbed into blood and subsequently distributed throughout the body. The distribution of CO in the body is reflected by the binding of CO to haeme proteins (eg blood haemoglobin and tissue myoglobin).
Approximately 80%-90% of the absorbed CO binds to haemoglobin (Hb) with a 200-fold greater affinity than that of oxygen (O 2 ), forming COHb, which in turn, reduces blood oxygen-carrying capacity and induces impaired perfusion and tissue hypoxia. 1 Although all tissues are vulnerable to CO-induced hypoxic injury, the brain, heart and lungs are particularly vulnerable, because of their high O 2 demands. Some of the remaining CO binds to tissue myoglobin and neuroglobin, affecting brain and muscle oxygen storage and utilization. Although the molecular mechanisms of poisoning are not fully elucidated, evidence indicates that these events lead to derangements in ion channels, inactivation of mitochondrial enzymes, immune system activation, DNA damage and cell death. 1,2,5,9,[11][12][13][14][15][16] Until now, very few studies have investigated the effect of CO poisoning on the metabolome. 17,18 In a pilot metabolomics study, Ju et al 17 investigated blood metabolites from deceased patients because of poisoning and found altered fatty acid metabolism. While this study added new insights to the deranged mechanisms of CO poisoning, it was performed post-mortem. Hence, acute metabolite alterations reflecting ongoing toxicity may not have been reflected, because of the time elapsed from exposure cessation to sample collection. We have previously applied metabolomics to identify altered mechanisms in ischaemiareperfusion injury, 19-22 hyperoxia 23 and the progression to acute lung injury. 20,24,25 Therefore, in this current experimental study, we have been suggested that metabolomics can be used to detect metabolite changes related to the progression of acute CO poisoning. The aim was to explore the effects of mild and severe CO poisoning on arterial blood gas parameters and to associate potential metabolite alterations to increasing COHb levels. Because of the anatomical and immunological similarities between the porcine and human models, 26 pigs were used in this study and arterial blood samples were collected before and during exposure to exogenous CO. At the end of the study, the pigs were sacrificed by an overdose of Pentobarbital.

| Experimental animals and instrumentation
A total of eleven Danish Landrace female pigs (mean ± standard deviation, SD; 48.4 ± 2.46 kg) were included in this study. The animals were initially anaesthetized using Zoletil (Tiletamine/Zolazepam) and once intravenous access was established, anaesthesia was continued with an intravenous infusion of Fentanyl and Propofol.
Intubation was performed by insertion of a 6.5 mm tube. The ventilator (Dameca DREAM) was set to a tidal volume equivalent of 8 mL/kg bodyweight, positive end-expiratory pressure of 5 cm H 2 O and fraction of inspired oxygen (FiO 2 ) to the lowest value needed to ensure an arterial partial pressure of oxygen (PaO 2 ) in the normal range of 9.6-13.7 kPa. Ventilation was set at a rate ensuring normal end-tidal CO 2 (ETCO 2 ) values between 4.7 and 6.0 kPa. Published guidelines for fluid therapy were used. 26,27 The pigs' vital signs were monitored during the experiment; urine production and temperature with a thermo-catheter inserted into the bladder, continuous

| Experimental protocol
Blood samples for metabolite measurements were drawn at baseline and during exposure to CO, but precisely 10 minutes after inhalation of CO. The delay was to allow dispersion of CO throughout the body until COHb values indicated mild poisoning, defined as ~30% COHb, and severe poisoning, defined as >60% COHb and the point of cardiac failure. Cardiac failure was defined as a 50% decrease in cardiac output compared to baseline. This was measured using  Note: Significance was assessed by means of ANOVA ( a ) with Tukey's post hoc test for multiple testing ( b ). A two-tailed P-value ≤ .05 was considered statistically significant.

| Data analysis
Spectral processing, metabolite identification and quantification were performed as previously described. 20  To determine differences between metabolite levels from samples collected at baseline and during mild and severe CO poisoning, standard biostatistical tests were applied. [19][20][21][22][23][24][25] We first employed a Shapiro-Wilk test to verify whether the data followed a normal distribution. Differences in group means were assessed by using analysis of variance (ANOVA) with Tukey's post hoc for multiple testing on logarithmically transformed data. Pearson correlation was applied to test the association of COHb levels and duration of exposure to CO with blood gas parameters and metabolites. A two-tailed P-value ≤ .05 was considered statistically significant.
Scatter and box-plots were used to visualize molecular changes from before and during poisoning. Relative concentrations are presented using means and standard deviations (SD). The quality of NMR data was assessed by correlating levels of fresh arterial blood glucose and lactate recorded on the ABL800 to serum glucose and lactate measured by NMR.

| Determination of CO poisoning
Carboxyhaemoglobin levels were measured in arterial blood collected from eleven Danish Landrace pigs at different time points before and during exposure to CO. To facilitate proper comparison, only blood gas data obtained from the samples used for metabolomics analyses are presented ( Table 1). The measured COHb was 2.9 ± 0.5% at baseline, 35.2 ± 7.9% at mild poisoning and 69.3 ± 10.2% at severe poisoning. Mild CO poisoning was achieved after 32 ± 13 minutes, while severe poisoning occurred at 64 ± 23 minutes from baseline.
Increased COHb and decreased oxyhaemoglobin (OHb) levels were noticed during mild and severe poisoning ( Figure 2,  increased COHb levels, low blood oxygenation and metabolic acidosis confirmed that poisoning was successful.

| The impact of CO poisoning on blood metabolites
Increasing CO levels not only affected lactate and blood gas parameters, but also the composition of circulating metabolites. As depicted in Figure 3A, several metabolite intensities changed upon poisoning.
Lactate, alanine, succinate, pyruvate and glycerol signals increased, while acetate decreased, with lactate undergoing the most significant alterations.
Because lactate and glucose were measured by both arterial blood gas analyser (ABL) and NMR techniques, metabolomics data quality was assessed by correlating fresh blood lactate and glucose measurements recorded on the ABG to serum lactate and glucose measured by NMR spectroscopy. High degrees of correlation were found (coefficient of determination R-square, R 2 = .98; Figure 3B), indicating high data quality.
Unsupervised pattern recognition analysis discerned between samples collected before and during exposure to CO, as well as between the degrees of poisoning ( Figure 3C). While mild poisoning seemed to affect blood metabolites to a lesser extent, as reflected by the close proximity of baseline and mild samples, abnormally high COHb levels induced significant alterations, indicated by samples clustering separately along the first principal component (PC1).
Lactate, pyruvate, acetoin and fumarate were found to strongly correlate with COHb (Pearson correlation coefficient r ≥ .73, P < .0001, Figure 3E, Table 3), indicating a possible link between the degree of CO intoxication and derangements in the mechanisms involving these metabolites. One possible interpretation of altered reactions is provided in Figure 4. It is expected that some of the various modes of action differ quantitatively across species, and that some differences in doseresponse relationships between humans and animal models can be encountered because of differences in CO uptake/elimination and binding kinetics with haeme proteins. 29   In contrast, metabolites downstream of glycolysis including glyceraldehyde, pyruvate, alanine and lactate were affected. Increased lactate levels with concomitant decreases in blood O 2 and increases in COHb levels correspond with metabolic acidosis, a wellknown mechanism occurring during hypoxia caused by poisoning. 31 Lactate has previously been found to correlate to the degree of CO poisoning 1 and has been used as a marker of severe poisoning. 30 In this study, we also observed a high correlation between lactate and blood COHb. Hence, our results indicate successful experimental CO intoxication, giving confidence to our findings. In addition to could also be tested. The treatment of both mild and severe CO poisoning has not evolved greatly during the last decades and still consists of oxygen therapy, which in severe cases can be administered in a hyperbaric pressure 35 chamber. There is ongoing debate about whether oxygen therapy reduces the risk of developing neurological sequelae or mortality [8][9][10][11] or whether it might even be harmful, because of barotrauma, pulmonary oedema, seizures and free radical generation. 1,10 Therefore, there is still a need to understand the molecular basis of current treatments, which may combat the toxic effects of CO binding to haemoglobin, but fail to counteract the direct cellular actions of CO. In future studies, it would be interesting to explore the metabolic changes during a slightly less severe poisoning for a longer duration of time, including changes related to ongoing treatments until restoration of metabolite balance occurs.

| CON CLUS ION
Our experimental study adds new insights to the deranged metabolite mechanism of CO poisoning, including hypoxia, ROS formation and peroxidation. This study leads the way for further studies investigating this area. Ideally, studies looking into the effects of hyperbaric oxygen therapy versus normobaric oxygen should be considered. With further research, it may be possible to pinpoint markers that can be used to guide the intensity of oxygen therapy and thereby diminish some of the harmful side effects of oxygen therapy. F I G U R E 4 Simplified pathways showing metabolic changes as a consequence of increasing blood CO levels. Samples are visualized by scatter and box-plots and colour coded according to baseline (red), mild CO (green) and severe poisoning (blue). The arrows represent the direction of conversion of metabolites, while the lines represent metabolite proximity (solid lines: direct conversion; dashed line: at least one metabolite is missing because of not being measured by NMR). AMP & IMP, adenosine monophosphate and inosine monophosphate; 3-HBA, 3-hydroxybutyric acid; Ac-CoA, Acetyl-CoA enzyme; GPC, glycerophosphocholine