Article pubs.acs.org/ac
L‑α-Phosphatidylglycerol
Chlorohydrins as Potential Biomarkers for
Chlorine Gas Exposure Petrus Hemström, Andreas Larsson, Linda Elfsmark, and Crister Åstot* The Swedish Defense Research Agency, FOI CBRN Defense and Security, 90182 Umeå, Sweden S Supporting Information *
ABSTRACT: Chlorine is a widely available toxic chemical that has been repeatedly used in armed conflict globally. The Organization for the Prohibition of Chemical Weapons (OPCW) have on numerous occasions found “compelling confirmation” that chlorine gas has been used against civilians in northern Syria. However, currently, there are no analytical methods available to unambiguously prove chlorine gas exposure. In this study, we describe the screening for chlorinated biomolecules by the use of mass isotope ratio filters followed by the identification of two biomarkers present in bronchoalveolar lavage fluid (BALF) from chlorine gas exposed mice. The relevance of these markers for human exposure was verified by their presence in in vitro chlorinated human BALF. The biomarkers were detectable for 72 h after exposure and were absent in nonexposed control animals. Furthermore, the biomarkers were not detected in humans diagnosed with chronic respiratory diseases. The potential chlorine specific markers were all chlorohydrins of unsaturated pulmonary surfactant phospholipids; phosphatidylglycerols, and phosphatidylcholines. Mass spectrometry fragmentation characteristics were favorable for the phosphatidylglycerol chlorohydrins, and they were therefore proposed as the best biomarker candidates.
C
enzyme myeloperoxidase. Myeloperoxidase is mainly present in neutrophil granulocytes, a type of white blood cell that is activated under inflammatory conditions.9 Chlorotyrosine and dichlorotyrosine have been proposed as biomarkers,8,10 but these compounds are also produced during the inflammatory response and may therefore be present in individuals not exposed to chlorine.11,12 In a study published during the peerreview process of this manuscript, 3-chloro tyrosine was shown to be elevated in mice for less than 24 h after chlorine gas (400 ppm) exposure.13 By the use of targeted analysis, elevated levels of chlorinated fatty acids in lung and plasma were shown and suggested as chlorine exposure markers.13 Biomarkers for myeloperoxidase derived HOCl have been reviewed by Winterbourn and Kettle.14 The biomarkers listed were chlorinated tyrosines, lipid chlorohydrins, anti-HOP reactivity, and glutathione sulfonamide. Since exposure to chlorine gas has been shown to induce an inflammatory response,15−17 the presence of inflammation cannot be used to eliminate false positives. For a chlorine gas biomarker to be considered unambiguous, it must not be produced in any significant quantities by endogenous reactions or responses. HOCl has been shown to react with alkene bonds,18,19 but the reaction is not able to compete with other more favorable
hlorine gas is an important industrial chemical used in large-scale chemical synthesis of, e.g., vinyl chloride (PVC), pesticides, aniline, dyes, and bleaches. Another vital use is as a germicide in drinking water and swimming pools. Chlorine gas was introduced as a chemical weapon by the German army in April 1915 near Ypres. More recently, the use of chlorine barrel bombs and bombings of chlorine transport and storage facilities have been reported in the Syrian/Iraq conflict.1 Compared to dedicated chemical warfare agents (CWA), the toxicity of chlorine gas is low and relatively high exposure levels are required for a fatal effect. Inhalation of the gas causes an acute lung injury characterized by pulmonary edema, pneumonitis, and hyperreactive airways that in some individuals develops into a persistent reactive airway dysfunction syndrome (RADS).2−6 Exposure to 50 ppm for more than 10 min is considered life-threatening or fatal in humans according to AEGL for airborne chemicals (a threshold level representing the general public including susceptible subpopulations such as elderly, children, and people with chronic diseases, e.g., asthma).7 Although there have been recent advances,8 there is still no known unambiguous biomarker for detecting chlorine gas exposure. Thus, there is a lack of tools for verifying the use of chlorine gas as a chemical weapon according to the criteria set by the The Organization for the Prohibition of Chemical Weapons (OPCW) for classical CWA, such as sarin and VX. Finding such biomarkers for chlorine gas is complicated by the in vivo production of hypochlorous acid (HOCl) by the © 2016 American Chemical Society
Received: May 16, 2016 Accepted: September 27, 2016 Published: September 27, 2016 9972
DOI: 10.1021/acs.analchem.6b01896 Anal. Chem. 2016, 88, 9972−9979
Article
Analytical Chemistry targets of HOCl.14,20 The most favored lipid targets of HOCl are plasmalogens, 21 a class of vinyl ether containing phospholipids. The addition of HOCl to a plasmalogen generates an α-chlorinated fatty aldehyde and a lysophospholipid. The α-chlorofatty aldehyde is subsequently converted to an α-chlorofatty acid.22 The presence of α-chloro hexadecanoic acid in bronchoalveolar lavage fluid (BALF) from mice infected with Sendai virus has been shown by the Ford group23 by GC/ MS analysis of pentafluorobenzyl esters. However, concentrations were very low (15−50 pmol/L), and the virus induced inflammation only increased levels by a factor of 3−5. Numerous metabolomics, in particular lipidomics, studies in patients with asthma and other lung diseases have been published in recent years. Biomarkers in inflammatory lung disease were recently reviewed by Wheelock and colleagues24 but with no mention of chlorinated biomarkers. A review on chlorinated lipids in pathology concluded that, despite considerable interest in chlorinated lipids, evidence for their existence in vivo is mostly circumstantial.25 In this study, we applied an untargeted approach to detect chlorinated biomolecules present in BALF from mice exposed to inhaled chlorine. Our aim was to find biomarkers unequivocally proving chlorine gas exposure.
were approved by the regional ethics committee on animal experiments (Umeå, Sweden) according to national laws. Mice were exposed to 200 or 400 ppm chlorine gas during 15 min using a nose-only exposure system, as previously described.16,26 Healthy animals exposed to ambient air were used as controls. The animals were euthanized and tracheostomized before BALF was collected by injecting and withdrawing 4 × 1 mL ice cold HBSS; the first two aliquots were pooled and used as the sample (2 mL). The BALF was centrifuged at 1500 rpm for 10 min, and supernatants were frozen at −80 °C until analysis. BALF was collected at 2 h, 6 h, 12 h, 24 h, 48 h, 72 h, day 7, and day 14 postexposure to 200 ppm chlorine. White blood cells in the samples were counted in Bürker chambers before being cytocentrifuged onto slides. The cell preparations were stained with May-Grünwald Giemsa and differentiated using a light microscope. Bronchoalveolar Lavage Samples from Human Subjects. Human BALF samples were collected from healthy volunteers and subjects diagnosed with allergic asthma or COPD (both current and ex-smokers).27,28 Bronchoscopy was performed according to a previously described method.29 In brief, the subjects received topical anesthesia with lidocaine administered to the pharynx, epipharynx, and within the bronchial tree. BALF was collected (3 × 60 mL saline) and filtered through a nylon filter (100 μm pores) before being centrifuged at 400g for 15 min. Supernatants were frozen at −80 °C until analysis.30 Sample Preparation. Sample preparation was carried out by modifying a method devised at the metabolomics core facility at UC-Davies.31 The method used Matyash32 MTBE lipid extraction. In short, samples were thawed, and 200 μL was placed in 7 mL glass screwcap centrifuge tubes to which 600 μL of methanol and 2 mL of MTBE were added. The samples were vortexed for 10 s, allowed to rest for 10 min, and then vortexed again. Phase separation was induced by adding 400 μL of MilliQ water, followed by vortexing and centrifuging the samples for 5 min at 2500 rpm (Heraeus Labofuge 300). The supernatant was collected, and 1 mL of fresh supernatant (prepared either prior to or in parallel with the samples by mixing and centrifuging the solvents in the same proportions as above but with Milli-Q water instead of the sample) was added. The samples were extracted again, and the supernatants were pooled, transferred to a 300 μL autosampler vial (after preliminary evaporation), and evaporated to dryness at 40 °C under a gentle nitrogen stream. The samples were dissolved in 10 or 50 μL of ethanol and diluted with Milli-Q to 100 μL (50/ 50 ethanol/water resulted in better sample stability but no loss in performance). For the initial screening, three samples from mice exposed to 400 ppm chlorine gas and two controls were used. For confirmation of the potential biomarkers, samples from three mice exposed to 200 ppm chlorine gas and one control were analyzed. Finally, a time series with one control and eight mice exposed to 200 ppm chlorine were analyzed. The time between chlorine exposure and sacrifice varied between 2 h and 14 days. Since HOCl is produced by granulocytes, such as neutrophils, eosinophils, and macrophages, during a normal inflammatory response, we also included BALF samples from mice with ongoing allergic or melphalan-induced neutrophilic airway inflammation as an additional control (see the Supporting Information for details). All sample handling was carried out using glass pipettes. Analysis. LC-HRMS was performed on a Bruker Impact HD Qq-TOF instrument equipped with a Captive nano ESI
■
EXPERIMENTAL SECTION Materials. Hank’s buffered salt solution (HBSS) purchased from Sigma was supplemented with 4.2 mM sodium hydrogen carbonate and 3 mM EDTA, and the pH was adjusted to 7.4. Chlorine gas (Cl2) at 1 mol % in nitrogen was obtained from Air Liquide (Germany). Acetonitrile and isopropanol were both ChemSolute UHPLC-MS grade from Th.Geyer (Renningen, Germany). Laboratory water was provided from a Milli-Q Advantage A10 system from Merck Millipore (Billerica, MA), who also supplied Merck Emsure ammonium acetate and LiChrosolv gradient grade HPLC methanol (Darmstadt, Germany). Sodium 1-palmitoyl-2-oleoyl-phosphatidylglycerol (Egg PG, Avanti 841138P) was from Avanti Polar Lipids Inc. (Alabaster, AL), and L-α-phosphatidylcholine (Egg PC, Sigma P-3556) was from Sigma. Meta-chloroperoxybenzoic acid (mCPBA), dichloromethane (ACS reagent), hydrochloric acid (ACS reagent, 37%), diethyl ether (puriss. p. a., ACS reagent), and methyl tert-butyl ether (MTBE) Chromasolv for HPLC were purchased from Sigma-Aldrich (St-Louis, MO). In Vitro Chlorination of Human BALF. 1% chlorine in nitrogen gas was passed through 1.5 mL of BALF from healthy volunteers at 100 mL/min, and aliquots were extracted for analysis after 5 and 10 min. As reference chemicals, 1 mg/mL solutions of L-α-phosphatidyl-DL-glycerol (Avanti, mix of PG isomers) and L-α-phosphatidylcholine in water were also treated with chlorine in the same manner. Chemical Synthesis of Reference Chemicals. Reference compounds of chloro-stearic and chloro-palmitic acids and the chlorohydrin of palmitoyl-oleyl phosphatidylglycerol were produced by chemical synthesis. The procedure and the chemical characterization of the products were supplied in the Supporting Information. Bronchoalveolar Lavage Sampling from Chlorine Exposed Animals. Animal experiments were performed using 10−11 week old female Balb/C mice obtained from Harlan Laboratories (Horst, Netherlands). The animals were kept under standard laboratory conditions (12 h daylight cycle, 22 °C, 30% relative humidity) and permitted access to food and water ad libitum. The experimental protocols and animal care 9973
DOI: 10.1021/acs.analchem.6b01896 Anal. Chem. 2016, 88, 9972−9979
Article
Analytical Chemistry
fluid. In the search for novel biomarkers of chlorine gas exposure, BALF from chlorine exposed mice was analyzed by mass spectrometry, followed by software assisted screening for chlorinated biomolecules. Screening for Chlorinated Biomolecules Based on Mass Isotope Ratio. Initial screening for chlorinated biomolecules was performed on BALF samples from mice exposed to a high concentration of chlorine gas (400 ppm). High resolution MS data were acquired in both positive and negative ion modes, and the data sets were filtered on the basis of the unique 37Cl/35Cl isotope ratio. Separate settings were used to detect biomolecules with one or two chlorines (Figure 1a,b). Inhalation exposure of mice to 400 ppm chlorine for 15
spray ionization source. The mass spectrometer was operated in negative mode with 1600 V capillary voltage (1400 V in ESI+), 0 V end plate offset, 3.0 L/min dry gas, 150 °C dry temperature, and 0.2 bar nano Booster. Spectra were collected from m/z 110 to 1200. The acquisition speed was 1.96 spectra/ s in both the MS and MS/MS modes. Use of the nano Booster was necessary to obtain a stable spray in negative mode using the Captive spray source. Mass calibration was performed externally vs sodium formate clusters; external calibration cannot be made with the Captive nano spray source but requires a standard flow ESI source. Therefore, we used palmitic acid as a lock mass since it could be detected in all BALF samples and standards. Dipalmitoylphosphatidylcholine was used as a lock mass in positive mode. For the MS/MS experiments, the same chromatographic and source parameters were used but all product ions of the precursors were collected in parallel (parallel reaction monitoring, PRM).33,34 The isolation width was set at 4, and a collision energy of −40 eV was used throughout. The chromatographic system was a Dionex Ultimate 3000 RSLC − nano system with an Acclaim PepMap 100 C-18 column (150 × 0.075 mm, 2 μm particles), both from Thermo Scientific. The flow rate was 300 nL/min using gradient elution. The eluents used were A: 80% Milli-Q water/20% acetonitrile with 10 mM ammonium acetate; B: isopropanol/acetonitrile/1 M ammonium acetate, 90/10/1. The gradient was as follows: 10% B (0−3 min), linear increase 10−95% B (3−15 min), 95% B (15−25 min), linear decrease back to 10% B (25−26 min). The viscosity settings for the NC-pump were A: 107.9%; B: 150.4%. The samples (5 μL) were first preconcentrated on a trap column (PepMap 20x 0.1 mm Acclaim 100 C-18) using a separate loading pump at 7 μL/min. The eluent was 10 mM ammonium acetate (the autosampler was operated in μL pickup mode to minimize sample consumption; the transport solution was also 10 mM ammonium acetate). It should be noted that the first 3 min of the gradient program was run during sample loading and the trap went online at t = 3.1 min. The system dead time was approximately 7.5 min, resulting in a t0 of 10.5 min. Data Evaluation. Chromatographic data were evaluated using a chlorine isotope filter provided in the Bruker Compass data analysis software. The isotope filter settings were as follows: m/z difference 1.997 ± 0.002 and isotope intensity ratio 0.32 ± 10% (0.64 ± 20% for two chlorines). The resulting isotope cluster filtered chromatograms were scrutinized manually after smoothing in order to determine whether the peaks seen in the chromatogram could be attributed to any chlorine containing compounds. The manual inspection was time-consuming and dependent on search methodology. Bruker MetaboliteDetect software was also used.
Figure 1. (a) Negative ion isotope cluster filtered chromatogram for compounds containing one chlorine atom from mice exposed to 400 ppm chlorine. Isotope filter settings: m/z difference, 1.997 ± 0.002; isotope intensity ratio, 0.32 ± 10%. (b) Negative ion isotope cluster filtered chromatogram filtered for compounds containing two chlorine atoms. Isotope filter settings: m/z difference, 1.997 ± 0.002; isotope intensity ratio, 0.64 ± 20%. (c) Extracted ion chromatograms of the potential biomarkers found in the two chromatograms above.
min is highly toxic, and for biomarkers to be clinically relevant, they must be measurable after exposure to sublethal concentrations. To refine the biomarker selection, a second set of BALF samples from mice exposed to a lower chlorine concentration (200 ppm) was analyzed and compounds not detected in all exposed individuals or found in controls were not further considered. The negative ion HRMS data were initially given most attention owing to the expected presence of chlorofatty acids. In a chromatogram filtered for a single chlorine, three candidate peaks were identified at 21.1, 22.1, and 23.5 min (Figure 1a). The candidates showed both a typical chlorine isotope pattern and reasonable peak shape in ion extracted chromatograms (Figure 1c). The first two abundant peaks were preliminary identified as chloro-palmitic acid (m/z 289, C16H30O2Cl) and chloro-stearic acid (m/z 317, C18H34O2Cl), whereas a peak at
■
RESULTS AND DISCUSSION The exact mechanism of chlorine gas toxicity is unknown, but both HOCl and chlorine have been suggested as the reactive species. In the lung, inhaled chlorine reacts with the lung epithelium, which is covered by epithelial lining fluid, consisting of water, lipids, and proteins. At physiological pH, chlorine is converted to hydrogen chloride (HCl) and HOCl when dissolved in water.35 However, the lipids constituting the pulmonary surfactant are present as mono- and doublemembrane layers, forming a lipophilic phase where chlorine can dissolve and react with biomolecules in the epithelial lining 9974
DOI: 10.1021/acs.analchem.6b01896 Anal. Chem. 2016, 88, 9972−9979
Article
Analytical Chemistry
Figure 2. Structures of the potential biomarkers: (a) chlorohydrin of palmitoyl-oleyl phosphatidylglycerol (16:0−18:1 PG-HOCl), (b) chlorohydrin of palmitoyl-oleyl phosphatidylcholine (16:0−18:1 PC-HOCl), (c) dichlorohydrin of palmitoyl-linoleyl phosphatidylglycerol (16:0−18:2 PG(HOCl)2), (d) chlorohydrin of oleic acid (m/z fragment 333), (e) 9-chloro fatty acid, and (f) α-chloro fatty acid.
23.5 min with m/z of 799.5 remained unidentified. When the data set was filtered for double chlorination, a peak at 22.5 min with m/z of 849.5 was added to the list of potential biomarkers (Figure 1b). In an attempt to automate the search procedure, we used Bruker add-on software to compare two full scan LCHRMS chromatograms (i.e., exposed and nonexposed). Although this software reported 400 unique features (m/z and time readings in negative ESI) in BALF from a chlorine exposed mouse, only two additional potential biomarkers were actually found (m/z 315.2 and 771.5). These two compounds were detected in all exposed samples but with relatively low intensities and were not considered as promising biomarker candidates. Biomarker Identification. For the first two peaks (m/z = 289 and 317), chloro-stearic and chloro-palmitic acids were the only plausible candidates suggested by the software. Their identity was further supported by synthesis (see the Supporting Information for details). Since these chloro-fatty acids have previously been shown in BALF from mice not exposed to chlorine gas,23 we decided to focus on the two unknowns at m/ z = 799.5 and 849.5 judged as promising unambiguous markers for chlorine exposure. The unknown with m/z of 799 was identified as the chlorohydrin of a pulmonary surfactant phospholipid: palmitoyl-oleyl phosphatidylglycerol (16:0−18:1 PG-HOCl, Figure 2a) based on its MS/MS spectrum and elemental composition of C40H77O11ClP (Figure 3, Table 1). In the product ion spectrum for the compound with m/z = 799 (Figure 3), favorable loss of hydrogen chloride was detected as the base peak of m/z = 763. The chlorohydrin of oleic acid was detected as a fragment ion of m/z = 333 (inset) with 335/333 ratio close to 0.33, confirming the chlorine content. The low abundance of the oleic acid chlorohydrin fragment can be explained by the subsequent loss of hydrogen chloride, producing the m/z = 297 fragment ion. The m/z = 255 fragment ion corresponds to palmitate. The second unknown compound had an exact mass of 849.443, corresponding to an elemental composition of C40H76Cl2O12P. From the mass difference between the two unknown compounds, it seems likely that they were closely
Figure 3. MS/MS spectra of unknown biomarker with m/z 799 from BALF of a mouse exposed to 400 ppm chlorine gas (top) and in vitro chlorinated phospatidylglycerol standard (bottom). The insets show expanded plots for the chlorinated fatty acid fragment at m/z 333.
Table 1. Exact Mass of Potential Biomarker Ions
9975
exact mass
molecular formula
error (ppm)
289.1949 317.2249 799.4875 849.4429
C16H30ClO2 C18H34ClO2 C40H77ClO11P C40H76Cl2O12P
−3.0 1.3 2.8 3.3 DOI: 10.1021/acs.analchem.6b01896 Anal. Chem. 2016, 88, 9972−9979
Article
Analytical Chemistry related, with HOCl added to one additional double bond. Thus, the m/z = 849 unknown was tentatively identified as the dichlorohydrin of palmitoyl-linoleyl phosphatidylglycerol (16:0−18:2 PG-(HOCl)2 Figure 2c). This identification is consistent with the MS/MS fragmentation (Figure 4).
Figure 5. Chromatogram showing identical retention times of mouse BALF (200 ppm chlorine) extracted phosphatidylglycerol chlorohydrins with m/z 799 and 849 (top) compared to those of in vitro chlorinated standards (bottom).
Supporting Information. In phosphatidylcholines, the positive charge is fixed to the choline. The lack of a mobile proton reduces the low energy fragmentation of the molecules. The only fragment ion detected was the phosphocholine moiety (m/z 184). The requirement for high energy collision activation reduced the sensitivity of targeted MS/MS analysis. Thus, no further studies were performed in positive ion mode. The pulmonary surfactant is essential for proper functioning of mammalian lungs by lowering the surface tension at the water air interface. The most abundant phospholipid species in the pulmonary surfactant is 16:0−16:0 phosphatidylcholine (PC), which constitutes roughly 40% of all pulmonary surfactant lipids. Another 40% is made up of unsaturated PC species. In rats, the most common of these unsaturated PCs is 16:0−16:1 PC, whereas in humans, 16:0−16:1 PC and 16:0− 18:1 PC are the most abundant unsaturated surfactant lipids. Phosphatidylglycerols (PG) generally make up about 20% of the surfactant lipids, but their composition varies widely between species; in rats, the main species are 16:0−16:0 PG, 16:0−18:1 PG, and 16:0−18:2 PG.36 Selectivity and Time Course of Identified Biomarkers. Our aim was to identify biomarkers that could be used to unequivocally prove human exposure to chlorine gas. Although discovered in a murine system, the identified biomarkers are expected to be present in human victims because the pulmonary surfactants of the human lung also include the precursor PGs. To test this hypothesis, BALF from healthy human volunteers was collected and exposed to chlorine gas in vitro. The presence of the PG chlorohydrins in exposed human BALF indicated that they are relevant as biomarkers for human chlorine exposure analysis. (Figure 6). Chlorine exposed mice developed acute airway inflammation within 24 h that mainly consisted of neutrophils (Figure 7). These granulocytes are well-known to produce endogenous
Figure 4. MS/MS spectra of unknown biomarker with m/z 849 from BALF of a mouse exposed to 400 ppm chlorine (top) and in vitro chlorinated phosphatidylglycerol standard (bottom).
The identities of the biomarkers as 16:0−18:1 PG-HOCl and 16:0−18:2 PG-(HOCl)2 were confirmed by analyses of reference standards (mix of PG chlorohydrins). Both 16:0− 18:1 PG-HOCl and 16:0−18:2 PG-(HOCl)2 showed identical MS/MS spectra and retention times as the unknown compounds 799 and 849 (Figures 3, 4, and 5). The standard also contained chlorohydrins of other phosphatidylglycerols, such as 18:0−18:1 PG-HOCl and 18:0−18:2 PG-(HOCl)2. A pure reference standard of 16:0−18:1 PG-HOCl was subsequently produced by chemical synthesis. This enabled standard addition quantification of the level in a pooled BALF sample from mice exposed to 200 ppm chlorine to 54 nM (Figure S3). The sensitivity of the biomarker analysis suffered from heavy ion suppression in the exposed BALF samples. It had a major impact on the limit of detection that was estimated to be
