Multiplex Quantification of Protein Toxins in Human ... - ACS Publications

Jul 13, 2015 - Food Matrices Using Immunoextraction and High-Resolution. Targeted Mass Spectrometry. Mathieu Dupré,. †. Benoit Gilquin,. §,∥,⊥. Franço...
0 downloads 0 Views 916KB Size
Download PDF
Article pubs.acs.org/ac

Multiplex Quantification of Protein Toxins in Human Biofluids and Food Matrices Using Immunoextraction and High-Resolution Targeted Mass Spectrometry Mathieu Dupré,† Benoit Gilquin,§,∥,⊥ François Fenaille,† Cécile Feraudet-Tarisse,‡ Julie Dano,‡ Myriam Ferro,§,∥,⊥ Stéphanie Simon,‡ Christophe Junot,† Virginie Brun,§,∥,⊥ and François Becher*,† †

CEA, DSV, iBiTec-S, Laboratoire d’études du métabolisme des médicaments, 91191 Gif-sur-Yvette, France CEA, DSV, iBiTec-S, Laboratoire d’études et de recherches en immunoanalyse, 91191 Gif-sur-Yvette, France § Université Grenoble Alpes, iRTSV-BGE, F-38000 Grenoble, France ∥ CEA, iRTSV-BGE, F-38000 Grenoble, France ⊥ INSERM, BGE, F-38000 Grenoble, France ‡

S Supporting Information *

ABSTRACT: The development of rapid methods for unambiguous identification and precise quantification of protein toxins in various matrices is essential for public health surveillance. Nowadays, analytical strategies classically rely on sensitive immunological assays, but mass spectrometry constitutes an attractive complementary approach thanks to direct measurement and protein characterization ability. We developed here an innovative multiplex immuno-LC-MS/MS method for the simultaneous and specific quantification of the three potential biological warfare agents, ricin, staphylococcal enterotoxin B, and epsilon toxin, in complex human biofluids and food matrices. At least 7 peptides were targeted for each toxin (43 peptides in total) with a quadrupole-Orbitrap highresolution instrument for exquisite detection specificity. Quantification was performed using stable isotope-labeled toxin standards spiked early in the sample. Lower limits of quantification were determined at or close to 1 ng·mL−1. The whole process was successfully applied to the quantitative analysis of toxins in complex samples such as milk, human urine, and plasma. Finally, we report new data on toxin stability with no evidence of toxin degradation in milk in a 48 h time frame, allowing relevant quantitative toxin analysis for samples collected in this time range.

S

protein biosynthesis through release of a specific adenine base from the essential 28S rRNA. Two isoforms of ricin have been reported, ricin D and ricin E, with identical RTA and a 15% difference in amino acid composition for RTB.6,7 The less toxic homologous protein R. communis agglutinin (RCA), also produced in the castor bean, has 93% and 84% sequence homology with the A- and B-chains of ricin D, respectively.7 Especially because of its widespread availability, ricin has been involved several times in human poisonings since the beginning of the 20th century.8 Staphylococcal enterotoxin B (SEB) is a pyrogenic exotoxin of ∼28 kDa produced by certain strains of Staphylococcus aureus.9 SEB is member of a family of several functionally related enterotoxins (SEs) that share sequence homology (SEB is 66% homologous with SEC 1, 2, and 3 subtypes).10 SEB has been

everal protein toxins naturally produced by microorganisms or plants are potential bioterrorism agents. According to their availability, ease of spread, toxicity, and ability to cause disease, ricin, staphylococcal enterotoxin B (SEB), epsilon toxin of Clostridium perf ringens (ETX), and botulinum neurotoxins (BoNTs) are considered as high-priority agents by the Centers for Disease Control and Prevention (CDC).1,2 In a situation of intentional release of any of these toxins, rapid discrimination and determination of the exposure doses are essential.3 A great advantage is therefore offered by methods identifying and quantifying several toxins simultaneously. Excluding BoNTs, which require a particular assay format of exquisite sensitivity due to their extreme toxicity and very low lethal dose,4 we focused here on quantification of ricin, SEB, and ETX. Ricin is a glycosylated protein of ∼62 kDa produced in the seeds of the castor bean Ricinus communis, composed of two disulfide-linked polypeptide chains: (i) a 30 kDa A-chain (RTA) and (ii) a 32 kDa B-chain (RTB).5 RTB is able to bind to terminal galactose residues on cell surfaces, allowing ricin internalization; then, RTA acts as a ribosome-inactivating protein by inhibiting © XXXX American Chemical Society

Received: May 21, 2015 Accepted: July 9, 2015

A

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

resolution tandem mass spectrometry on a Q-Orbitrap instrument. In contrast to most targeted MS strategies, a large number of selected peptides (>7 per toxin) were monitored for additional specificity. To reach high sensitivity in complex matrices, an advanced multiplexed immuno-affinity enrichment method with rapid on-bead digestion was implemented. Isotope-labeled toxin standards were produced and spiked early in the protocol to help compensate for any protein loss. We demonstrated the successful application of the assay to human biofluids and food matrices and reported new relevant data on toxin stability over time in assay conditions over a 48 h period.

frequently involved in food poisoning episodes and studied because of its ability to cause temporary disabling disease at low doses even by inhalation exposure.11 Epsilon toxin (ETX) is produced by types B and D of the spore-forming C. perf ringens bacterium and belongs to the family of aerolysin pore-forming proteins.12 First secreted as a ∼33 kDa single-chain protein, Epsilon prototoxin is activated in ETX into isoforms of several sizes by protease cleavage. ETX is the third most potent clostridial toxin, but its precise mode of action has not been elucidated yet. ETX is released after bacterial contamination caused, for instance, by food poisoning or wound infections and is responsible for diverse intestinal diseases in humans and animals, i.e., gas gangrene, necrotic enteritis, lamb dysentery (type B), or enterotoxemia (type D).13 Up to now, a few multiplexed immunoassays have been developed for these biodefense toxins.14−17 Although easy to use, rapid, and cost-effective, immunoassays are commonly based on a final colorimetric or fluorescence detection unable to reveal potential antibody cross-reactivity in complex matrices.18,19 Therefore, complementary methods for confident multiplex toxin quantification would be beneficial. Mass spectrometry (MS) has generated much interest over the last two decades as a complementary approach to the analysis of biological warfare agents (BWAs),20−22 including ricin,23−28 SEB,29,30 and ETX toxins.31 It provides direct measurement and protein sequence characterization for unambiguous identification. The two most common ionization methods, electrospray ionization (ESI) and matrix-assisted laser desorption/ionization (MALDI), have been applied to BWA detection, by analysis of intact protein32,33 or tryptic peptides.25,31,34,35 MALDI-based assays have already been proposed for simultaneous MS detection of biodefense toxins,36,37 although targeted proteomics, based on liquid chromatography coupled to ESI and selected reaction monitoring (SRM) experiments, would be more efficient for quantitative analysis.38,39 Recently, alternative acquisition modes were introduced for multiplex targeted experiments, in relation to the significant progress achieved in high-resolution accurate-mass acquisition (HR/AM) instruments, with higher sensitivity and scan speed. Typically, the hybrid quadrupole-Orbitrap mass spectrometer operating in the parallel reaction monitoring (PRM) mode demonstrated high specificity in protein quantification experiments by simultaneous HR/AM detection of all potential fragment ions from preselected precursor ions.40,41 Efficient sample preparation prior to mass spectrometry quantification is nevertheless essential to counteract ion suppression/matrix effects in the diversity of complex environmental or biological samples that may be investigated in biodefense applications. In this regard, antibody-based affinity enrichment methods prior to proteomics experiments result in highly sensitive assays by combining both purification and concentration abilities.25,42 Although this procedure offers efficient enrichment of targeted proteins, variable recoveries due to analyte loss are observed depending on the investigated matrices. The PSAQ (Protein Standard for Absolute Quantification) strategy has recently been proposed to overcome the impact of this variability. For each protein target, an isotopically labeled protein analog (PSAQ) is added to the complex matrix. The protein analyte and the PSAQ are subject to the same sample handling, resulting in robust quantification.43,44 In this context, we report here a novel rapid multiplex LC-ESIMS/MS (PRM) method for the specific quantification of ricin, ETX, and SEB toxins in complex matrices, by targeted high-

■

EXPERIMENTAL SECTION Chemicals and Reagents. Phosphate-buffered saline (PBS) was from Lonza (DPBS without Ca2+ and Mg2+, Verviers, Belgium). N-2-Hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES), sodium phosphate monobasic (NaH2PO4) and dibasic (Na2HPO4), sodium acetate, bovine serum albumin (BSA), dithiothreitol (DTT), iodoacetamide (IAA), analytical grade formic acid (FA), trifluoroacetic acid (TFA), HPLC-grade acetonitrile (ACN), and SEB were from Sigma-Aldrich (Sigma Chemical Co., St Louis, MO, USA). Ricin was obtained by purification of R. communis seeds (Centre d’Etudes du Bouchet, Vert Le Petit, France).25 The prototoxin form of epsilon toxin (ETX) was purified from an overnight culture of C. perf ringens type D strain NCTC2062, and ETX was obtained upon trypsin activation as previously described.31 Ultrapure water was from a Milli-Q plus purifier (Millipore, Bedford, MA). Sequencinggrade modified trypsin was from Promega (Fitchburg, WI, USA). Mouse monoclonal antibodies Aric11, SEB27, and PεTX7 directed against RTA of ricin, enterotoxin staphylococcal B, and formalin-inactivated prototoxin, respectively, were produced by the Laboratoire d’Etudes et de Recherches en Immunoanalyse as previously described.31,45 Safety Considerations. Ricin, epsilon, and SEB toxins are highly toxic proteins. Even though low toxin amounts were needed for this work, strict adherence to safety rules for the handling of toxic substances is essential. All contact with the substances should be avoided. Isotopically Labeled Peptides and Protein Standards. Protein standards PSAQ were produced by Promise Advanced Proteomics (Grenoble, France) using cell-free extracts (RTS 500 ProteoMaster E. coli HY Kit, 5 Prime, Hamburg, Germany) and highly purified as previously described.43 Isotope labeling was performed using [13C6] L-lysine and [13C6] L-arginine (Eurisotop, Saint-Aubin, France) for SEB toxin which leads to a constant mass increment (+6.0201 Da) of fully tryptic SEB peptides. [13C6;15N2] L-lysine and [13C6;15N4] L-arginine were used to label RTA and ETX toxins, leading to a constant mass increment (+8.0142 and +10.0083 Da) of fully tryptic RTA and ETX peptides. PSAQ standards were checked for purity on SDSPAGE using Coomassie staining (>95% purity). N-terminal hexahistidine purification tags were not removed as they are not expected to modify PSAQ biochemical properties significantly. Isotope-labeled proteins were quantified by amino acid analysis. Isotope incorporation was verified by LC-MS and LC-SRM analysis and was found to be greater than 99%. Labeled peptides, used for initial MS experiments, were synthesized by Thermo Fisher Scientific (Bremen, Germany) (HEI[13C6;15N]PVLP[13C5;15N]NR, LTTGADV[13C5;15N]R, VGLP[13C5;15N]INQR for ricin RTA, ASDPSL[13C6;15N]K[13C6;15N2] for ricin RTB, and LV[13C5;15N]GQVSGSEWGEIPSYLAFP[ 1 3 C 5 ; 1 5 N]R, EI[ 1 3 C 6 ; 1 5 N]THNVPSB

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry QDILVPANTTVEVIAYL[13C6;15N]K for ETX) and by Bachem (Burgdorf, Switzerland) (NTDTVTATTTHTVGTSIQATAK[13C6;15N2] and ALLTNDTQQEQK[13C6;15N2] for ETX). Sample Preparation. Antitoxin monoclonal antibodies (35 μg) were separately covalently linked to magnetic beads (75 μL, 30 mg·mL−1) (Dynabeads M-280 tosyl-activated, Life Technologies, Carlsbad, CA), preconditioned as mentioned in the protocol provided by the supplier, to obtain IgG-coupled bead solutions (100 μL). Samples (1 mL) in Eppendorf Protein LoBind Microcentrifuge tubes were spiked with 7.5 μL of PSAQ standard mixture containing 12.5 ng of RTA, 10 ng of SEB, and 15 ng of ETX proteins and then with IgG-coated bead solutions (20 μL). Samples were incubated for 60 min at room temperature with gentle shaking and then washed three times with PBS to eliminate weak nonspecific binding. After removing the washing solution, complex protein-IgG-beads were resuspended in 20 μL of 400 mM ammonium bicarbonate buffer pH 7.5 containing 0.05% RapiGest SF Surfactant (Waters Corp., Milford, MA) containing 185.1 pg of NTDTVTATTTHTVGTSIQATAK[13C6;15N2] and 62.2 pg of ASDPSL[13C6;15N]K[13C6;15N2]. Each sample was denatured and reduced by the addition of 5 μL of 6 mM DTT (in 50 mM NH4HCO3 buffer) at 95 °C for 45 min and alkylated by the addition of 5 μL of 18 mM IAA (in 50 mM NH4HCO3 buffer) in the dark at room temperature for 45 min. A total of 2 μL of 500 μg·mL−1 aqueous trypsin solution was added, and the enzymatic digestion was performed at 50 °C for 120 min. Five μL of 1 M HCl was then added to degrade RapiGest SF prior to further analysis. After 15 min of centrifugation at 10 000g, 15 μL samples were injected in the LC-MS system. Liquid Chromatography−Mass Spectrometry. LC-MS was performed on an Ultimate 3000 chromatography system coupled to a Q-Exactive Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Peptide separation was achieved in 25 min on a Zorbax C18-300 Å column (2.1 mm × 150 mm, 5 μm) (Agilent Technologies, Waldbronn, Germany). The mobile phases were (A) 0.1% formic acid in water and (B) 0.1% formic acid in ACN. After an isocratic step of 1 min at 5% phase B, a nonlinear three-step gradient from 5% to 55% B was run over the next 20 min with a mobile phase flow rate of 200 μL/min. Data were acquired in the positive ion mode. The detailed chromatographic conditions and MS parameters are described in the Supporting Information. For targeted experiments, the Q-Orbitrap was operated in the time-scheduled sequential PRM method with 13 separated windows. Several acquisition parameters were optimized. Quadrupole isolation was performed from 1 to 2 Th, meaning that isotopologue pairs of light and heavy peptides were isolated separately. Individual injection times were set from 125 to 500 ms depending on the number of coeluted peptides in a scheduled retention time window. The AGC target was constantly set at 1 × 106. Maximum C-trap accumulation time and Orbitrap resolution (from 17 500 to 140 000) were adjusted to reach chromatographic peaks with a minimum of 10 acquisition points. PRM data were processed with Pinpoint 3.0 (Thermo Fisher Scientific) and Xcalibur 2.2 (Quan Browser) software according to an optimized processing method. Further details are provided as Supporting Information.

objectives were: (i) to design multiplex immuno-affinity enrichment for efficient concentration of ricin, ETX, and SEB and to optimize LC-MS/MS peptide detection on a quadrupoleOrbitrap mass spectrometer using the PRM mode and (ii) to monitor the full protocol with labeled toxins for confident identification and quantification (Figure 1). In addition, time-toresults, a major criterion in biodefense, and assay specificity (through the monitoring of several signature peptides per toxin) were optimized.

Figure 1. Sample preparation protocol based on multiplexed immunoenrichment and early spiking of isotope-labeled toxins for quantification of ricin, SEB, and ETX.

Toxin Characterization and Identification of Unique Peptides. In-depth characterization of purified toxins (2 μg· mL−1) using bottom up LC-ESI-HRMS experiments with datadependent MS/MS acquisitions was first performed (detailed parameters are described in the Supporting Information). The Uniprot sequence entries P02879, P01552, and Q57398, respectively, referring to the D form of ricin, the enterotoxin type B, and the epsilon toxin, were considered. Consistency of toxin sequences and potential amino acid modifications were assessed first. Digestion of toxins without preliminary denaturation was achieved and highlighted the presence of ricin disulfide peptides including the disulfide-linked interchain part (Figure S1, Supporting Information). Ricin glycosylation at 2 potential sites was confirmed through detection and fragmentation of 4 glycopeptides (Figure S-2, Supporting Information). In this experiment, the SEB disulfide peptide was not observed in our conditions, probably because of the length of its amino acid sequence (Table S-1, Supporting Information). In reducing conditions, improved sequence coverages were obtained: 79% for ricin (29 out of 35 expected peptides with a mass above 400 Da and/or at least 4 residues considering no missed cleavage, while excluding the signal peptide), 86% for SEB (17 out of 21 peptides), and 94% for ETX (21 out of 23 peptides) (Table S-1 and Figure S-3, Supporting Information). As a consequence, reducing conditions were selected for peptide identification. The uniqueness of all identified peptides was checked by BLAST searches against the whole UniprotKB/Swiss-Prot database (http://blast.ncbi.nlm.nih.gov), in addition to sequence align-

■

RESULTS AND DISCUSSION Toxin identification and determination of contamination level in a single analysis are essential for fast confirmation of bioterrorism events and public health surveillance. In this regard, our C

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

25 NCE). Optimal precursor ions (peptide charge states) and collision energies were defined on the basis of the intensity of fragment ions, processed as indicated above. Summed XIC signals were compared for final selection of 7, 7, and 8 peptides for ricin (3 for RTA, 4 for RTB), SEB, and ETX analysis (Tables S-1 and S-2, Supporting Information), respectively. As the monitoring of a large number of peptides further increases confidence in toxin identification, 2 cysteine-containing peptides (peptide 502-512 from RTB, peptide 134-140 from ETX) and one methionine-containing peptide (peptide 101-119 from ETX) were included in the selection. Additionally, 2 common peptides with RCA (peptides 75-83 and 249-269 from RTA) were monitored as additional signatures of a potential contamination by a crude preparation of ricin castor seeds.47 This high number of monitored signature peptides ensures maximal sequence coverage for ricin, SEB, and ETX toxins at 24%, 35%, and 51%, respectively, for highly specific detection, reducing the risk of false-positive identifications. Peptides were quite uniformly distributed, which could also be useful for further toxin sequence integrity evaluation. Design of the Multiplexed Targeted LC-MS/MS Method. The aim of the project was to develop a unique method for the simultaneous detection of ricin, SEB, and ETX. Thus, the selected peptides and their isotope-labeled counterparts were included in one common inclusion list. The LC gradient (Figure S-6A, Supporting Information) was carefully optimized to minimize possible coelutions. However, in order to maintain an acceptable separation time (25 min), it was not possible to prevent coelutions among the 43 monitored peptides, i.e., the 24 toxin peptides and 19 from labeled standards. In this situation, it was necessary to establish precisely suitable fill times based on the number of eluted peptides to define cycle times compatible with collection of at least 10 data points per peptide.40 The sequential acquisition method, where each target parent ion is fragmented and analyzed in the Orbitrap individually and sequentially, was used. In contrast to simultaneous (several target ions in one single isolation window) or multiplexed (target ions isolated separately but accumulated together) acquisition strategies, the sequential method potentially leads to less mass interference, thanks to narrower quadrupole isolation windows, and to differentiation of common b fragment ions from labeled and unlabeled peptides (i.e., b ions are unlabeled since isotopic labeling is on the C-terminal arginine or lysine) (Figure S-6B, Supporting Information). Depending on peptide retention times, 13 targeted MS2 windows were optimized with lengthy fill times between 100 and 500 ms for high sensitivity.48 Orbitrap resolution was adjusted from 17 500 to 140 000 to have equal scanning time and fill time. Detailed parameters are presented as Supporting Information. Multiplexed Sample Preparation. Application to all complex matrices potentially found in bioterrorism investigations requires an efficient sample pretreatment prior to targeted proteomics. An immuno-enrichment step combined with rapid on-bead digestion was optimized for optimal detection of the 3 toxins together, as summarized in Figure 1 and Figure S-7, Supporting Information. Simultaneous Toxin Immuno-Enrichment Procedure. Immunoaffinity extraction has been successfully used as a sample purifier for the analysis of BWAs from complex samples.25,31,42,49,50 Immunoextraction combines many advantages, including specific purification and concentration of proteins from highly complex samples and usability thanks to the properties of the magnetic beads. Several capture conditions were evaluated

ment with homologous proteins (RCA120 and SEs, especially SEC) to evaluate sequence similarity (Table S-1, Supporting Information). Additionally, previous experimental data on RCA120 and SEs differentiation were taken into account.35,46 For instance, the peptide VGLPINQR, which has already been used for ricin quantification independently of RCA120, was observed. Regarding SEB, most identified peptides allow differentiation from other SEs. Multiplex Targeted LC-MS/MS Detection (PRM Mode). Targeted proteomics traditionally involves a triple quadrupole instrument operated in the SRM mode, according to high sensitivity and high multiplexing abilities. Thus, ricin detection was initially performed on a low-resolution triple quadrupole (Quantum Ultra from Thermo Fisher Scientific), monitoring the three most abundant fragment ions of selected peptides. However, the approach suffers from background interference in complex mixtures because of the low resolution of the quadrupolar mass analyzer, as illustrated for the peptide HEIPVLPNR (Figure S-4, Supporting Information). This interference was related to the release of peptides with similar m/z from the high concentration of on-bead coated homemade antibodies and the short LC separation time used in this study (detailed in the Sample Preparation section). Targeted MS/MS experiments were then performed on a quadrupole-Orbitrap mass spectrometer using the targeted PRM mode.40,41 In this configuration, all fragment ions are simultaneously detected with high resolution and high mass accuracy (HR/AM), demonstrating highly specific PRM transitions and drastic noise reduction in the detected signal, especially in complex matrices.41 In our conditions, no interference was observed for peptide HEIPVLPNR with PRM, illustrating the higher detection specificity (Figure S-4, Supporting Information). Also, it should be noted that peptide selection and the final PRM assay can be performed on the same instrument, which simplifies method development. The PRM mode was therefore selected, and method optimization for sensitive detection of toxin signature peptides is discussed in the following sections. Data Processing Refinements. PRM targeted MS2 experiments on labeled and unlabeled toxin solutions at 2 μg·mL−1 were first performed to assign y and b fragment ions for each signature peptide. Isotope peaks were considered since high intensity was observed for heavy isotopes (M + 1, M + 2, M + 3), particularly for high-mass fragment ions. All fragment ion signals were then individually checked at a 5 ppm mass tolerance using PinPoint Software to remove those with irregular profiles in terms of ion abundance or chromatographic peak shape compared to signal from purified toxins. All PRM noninterfered signals from selected fragments were summed to provide one extracted ion chromatogram (XIC) for each targeted peptide. This allowed up to a 10-fold final increase in the XIC area compared with extraction of monoisotopic mass only, with no additional interference or noise in the control sample, as illustrated in Figure S-5, Supporting Information, for an RTA ricin peptide. The same procedure was applied to PRM transitions of all selected peptides for further establishment of calibration curves in the complex matrices investigated. Selection of High-Responding Peptides. For optimal assay sensitivity, selection was based not only on peptide precursor ion signals but also on fragmentation yield. Thus, LC-ESI-PRM analysis of the most intense charge states (among 2+, 3+, and 4+) from all unique peptides identified (Table S-1, Supporting Information) in the 2 μg·mL−1 samples of labeled and unlabeled toxins was performed at different collision energies (from 10 to D

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Table 1. Detection and Quantification Limits for the Best Responding Peptides of Each Toxin in PBS/BSA Buffer, Milk, and Serum Samplesa PBS/BSA

serum

toxin

peptides

LOD

LOQ

LOD

LOQ

LOD

LOQ

ETX toxin

SQSFTC[Carboxyamidomethyl]K ALLTNDTQQEQK ASYDNVDTLIEK VELDGEPSMNYLEDVYVGK LVGQVSGSEWGEIPSYLAFPR EITHNVPSQDILVPANTTVEVIAYLK LTTGADVR HEIPVLPNR VGLPINQR LSTAIQESNQGAFASPIQLQR VLYDDNHVSAINVK IEVYLTTK VTAQELDYLTR LYEFNNSPYETGYIK

1 0.25 0.25 0.1 0.1 0.25 0.25 0.25 0.25 0.1 0.1 0.1 0.1 0.25

1 1 N/A 0.5 1 0.5 0.5 0.5 0.5 1 0.5 0.5 0.5 1

0.5 0.5 1 0.25 0.5 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.5

0.5 1 1 0.5 0.5 0.5 1 1 0.5 1 0.5 1 1 2.5

1 2.5 2.5 0.5 1 0.5 0.25 0.5 0.5 0.25 2.5 2.5 5 10

1 N/A N/A 1 1 1 0.5 2.5 2.5 1 N/A N/A N/A N/A

ricin toxin

SEB toxin

a

milk

Expressed in ng·mL−1. N/A = not applicable (detection not sensitive enough).

detected with higher intensity with 80% ACN (data not shown). This procedure was extended to other toxins, where 0.05% RapiGest SF was found to provide the best peptide signals. Protein digestion is typically done overnight, which is not compatible with the fast decision-making required with potentially contaminated samples. Therefore, shorter incubation times were evaluated using standard digestion conditions (37 °C, 0.2 μg of trypsin). Release of most peptides was observed in 2 h, with lowered signals (up to a factor of 2) for few peptides compared with overnight incubation (Figure S-10, Supporting Information). With the objective of maintaining optimal sensitivity with a 2 h digestion time, the amount of trypsin and the proteolysis temperature were investigated. Overall, minor differences were observed between the investigated conditions (Figure S-10, Supporting Information). The protocol at 50 °C using 0.5 μg of trypsin per sample was finally selected as it generally showed a slightly more efficient release of most peptides. Toxin Quantification. Proteins are usually quantified using isotope-labeled synthetic peptides (AQUA for Absolute QUAntification),53 and initial experiments in this study were performed accordingly. Although previously applied to quantification of BWAs including toxins, AQUA only leads to normalization of LC-MS detection.44,54 Bias in quantification could therefore be observed, mostly resulting from significant variations of proteolytic digestion efficiency or analyte purification recovery. The PSAQ strategy (Protein Standard Absolute Quantification) based on in vitro-synthesized isotopelabeled full-length protein standards was successfully introduced to circumvent these limitations.43 On the condition that PSAQ and targeted analytes share the same biochemical properties, these internal standards truly mimic analyte behavior in any sample preparation. This methodology, previously applied to quantification of proteins including some staphylococcal enterotoxins,46,51 was extended here to ricin and ETX. For safety reasons, ricin PSAQ was produced for the RTA only. Consequently, one RTB AQUA peptide and PSAQ toxins, when available, were used in the final LC-MS/MS method (Table S-1, Supporting Information). The labeled peptides resulting from trypsin digestion of PSAQ proteins were investigated. Similar profiles for PSAQ and purified toxins were obtained in terms of peptide ion relative intensity and

here, typically the chemistry of antibody coating on beads (covalent binding with tosyl-activated beads or protein G affinity), beads, antibody amounts, and nature of the buffer (PBS or HEPES). Also, two strategies for bead coating were considered for simultaneous toxin purification, (i) using separate coating protocols leading to individual IgG-coupled beads or (ii) using a shared coating protocol resulting in one type of magnetic bead presenting the three antibodies. Finally, the highest peptide signal was reached with the separate coating protocol, 0.01 M PBS buffer and 20 μL of beads/sample corresponding to 7 μg of antibody/sample (Figures S-8 and S-9, Supporting Information). Recovery of captured toxins from beads was then optimized. Acidic pH elution was first investigated as it was previously used successfully in the lab for BWAs including ricin.25 However, when applied to the simultaneous detection of the 3 toxins, it resulted in a significant and variable loss of signal for epsilon toxin as previously observed31 (Figure S-9, Supporting Information). As an alternative, direct enzymatic digestion performed on beads leads to the release of proteolytic peptides instead of the whole intact proteins. This approach was finally selected as it resulted in (i) homogeneous recoveries for all toxins with better signal for ETX and (ii) a faster result compared with acidic elution (Figure S-9, Supporting Information). In the final conditions, efficiency of the multiplex toxin immunocapture was evaluated as between 15% and 25% ± 9%, depending on the toxin, from 1 mL of PBSBSA buffer (50-fold sample volume concentration) (Table S-4, Supporting Information). These recoveries are consistent with previous work on single toxin analysis.31,51 Optimization of Protein Denaturation and Fast Proteolysis. Protein denaturation and enzymatic digestion are crucial steps, especially for toxins that are generally known to be resistant to proteases because of their compact structures. Proteomics conditions were carefully optimized. It has been demonstrated that the use of a surfactant such as RapiGest SF for protein denaturation has a critical influence on tryptic digestion efficiency, notably in the case of ETX31 and ricin.52 To confirm these results when using our on-bead protocol, ricin analysis with trypsin digestion assisted either by organic solvent (80% ACN) or by surfactant (0.25%, 0.1%, and 0.05% RapiGest) was evaluated. Improved detection of monitored peptides was generally observed with the RapiGest SF, even though two RTA peptides (LEQLAGNLR and YTFAFGGNYDR) were E

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry Table 2. Method Evaluation in Complex Samplesa toxins at 1 ng·mL−1 area ratio toxin peptides VGLPINQR [ricin]

VLYDDNHVSAINVK [SEB]

EITHNVPSQDILVPA NTTVEVIAYLK [ETX]

a

blanks

matrices

toxin area avg (n = 2)

ISTD area avg (n = 2)

avg

CV, %

ISTD area avg (n = 2)

PBS/BSA milk serum plasma urine tap water PBS/BSA milk serum plasma urine tap water PBS/BSA milk serum plasma urine tap water

1.0 × 105 1.0 × 105 4.1 × 104 4.4 × 104 1.1 × 105 1.8 × 105 2.1 × 105 2.0 × 105 nd 1.1 × 104 1.3 × 105 1.0 × 105 2.5 × 105 9.1 × 104 2.3 × 104 3.0 × 104 1.1 × 105 4.5 × 105

1.3 × 106 9.5 × 104 9.0 × 104 8.2 × 104 7.9 × 105 1.2 × 106 2.6 × 106 5.8 × 106 1.3 × 106 1.7 × 106 3.4 × 106 2.6 × 106 3.2 × 106 1.9 × 106 8.2 × 105 7.3 × 105 2.2 × 106 2.4 × 106

0.08 1.08 0.44 0.53 0.14 0.15 0.08 0.04 nd 0.01 0.04 0.04 0.08 0.05 0.03 0.04 0.05 0.19

5 7 32 18 4 15 24 6 nd 52 2 19 11 6 2 7 18 14

1.3 × 106 1.0 × 105 8.2 × 104 1.0 × 105 8.7 × 105 2.0 × 106 2.2 × 106 6.3 × 106 1.6 × 106 2.2 × 106 3.6 × 106 2.7 × 106 2.8 × 106 2.3 × 106 6.6 × 105 8.6 × 105 2.5 × 106 2.5 × 106

Signal detected for toxins spiked together at 1 ng·mL−1 in buffer, milk, serum, plasma, urine, and tap water.

% below 13% (Table S-3, Supporting Information). Mean absolute bias error was between 1.7% and 9.1%. Finally, for each toxin, similar slopes (variation below 20%) were observed for calibration curves obtained individually with the best responding peptides, illustrating efficient proteolysis protocol normalization by PSAQ and acceptable quantification ability whatever the peptide used (Figure S-12, Supporting Information). Sequence Coverage at Low Concentration. At a toxin concentration of 1 ng·mL−1 in buffer samples with 0.5% BSA, close to the limit of quantification, it was observed that 21 out of 24 peptides were still detected for the 3 toxins, resulting in attractive sequence coverages from 24% to 42% as illustrated in Figure S-13, Supporting Information. Thus, unambiguous identification is achieved with a first indication of toxin integrity. For instance, the 2 chains of ricin, which are both required for toxic action,5 are monitored through at least 3 peptides per chain. Moreover, the detection of 3 RTA peptides and 4 RTB peptides ensures the distinction with RCA. For SEB, all detected peptides ensured discrimination from other staphylococcal enterotoxins. Finally, for ETX, several peptides common to the different size isoforms were detected, which means the method is able to monitor the toxin whatever its activation state. Application to Complex Samples. Several beverages and biological samples were evaluated. Milk and tap water were selected for the food safety context. Urine, serum, and plasma were investigated as human biofluid matrices for diagnosis. For the 5 matrices, sensitivity was evaluated with toxins spiked at a concentration close to the LOQ determined in buffer, i.e., 1 ng· mL−1. The signals detected for the best responding peptide in all investigated matrices are reported in Table 2 and Figure S-14, Supporting Information. For the 3 toxins, signal areas tended to decrease in relation to matrix complexity, but remained well detectable (i.e., signal intensity greater than 1.0 × 104), with no signal detected in any of the blank (zero) samples. Sequence coverage, determined in the milk sample, was found similar to the PBS buffer samples (Figure S-13, Supporting Information). Only SEB detection was significantly impacted in human serum and

fragmentation pattern (Figures S-6 and S-11, Supporting Information). The possibility of spiking the PSAQ proteins at an early stage of the sample preparation (before immunocapture, Figure 1) and monitoring as much as possible the analytical protocol was then evaluated. PSAQ proteins were recovered well from Ab-coated beads, but at a slightly different yield to purified toxins (Table S-4, Supporting Information). For ricin, this was most probably related to the absence of the B-chain in the PSAQ protein, although the Ab was directed to the RTA and/or the presence of glycosylated residues. For SEB and ETX, these discrepancies may originate from slight folding differences between the PSAQ proteins (which are fused to an N-terminal His tag) and the targeted proteins. To compensate for the differences in recovery, quantification relies additionally on normalized peak area ratios from calibration curves generated with purified toxins. Analytical Performances. Analytical performances, including repeatability, quantification limits, linearity, and specificity, were evaluated. Detection and Quantification Limits and Reproducibility. The PRM method was validated for the multiplex quantification of ricin, SEB, and ETX with spiking experiments at concentrations ranging from 0.10 (detection limit, i.e., signal intensity greater than 1.0 × 104) to 50 ng·mL−1, in PBS buffer with 0.5% BSA first. For the best responding peptides (4−5 per toxin), linear calibration curves were obtained between calibration levels at 0.5 and 50 ng·mL−1 (Figure S-12A, Supporting Information). Quantification limits (LOQ), defined by accuracy of 80−120%, were therefore determined at 0.5 ng· mL−1 for the 3 toxins (Table 1). These values compare favorably with previous results from single toxin analysis using a similar sample preparation protocol with LC-SRM detection.31,35 At LOQ, the signal was still detected with at least 10 data points collected per peptide. Repeatability was successfully validated for the three toxins with control samples at 2 and 15 ng·mL−1 showing coefficients of variation (CV) lower than 9% and concentrations of calibration curve standards measured with CV F

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry plasma, with undetectable or lower signal compared with the milk sample (Table 2). Signals from the more concentrated PSAQ SEB (working concentration at 10 ng/mL) were also impacted, but still detectable. This sensitivity loss for SEB was reported in previous studies11,14 and is likely due to interactions between SEB and serum/plasma proteins. However, especially for ricin, area ratios between unlabeled and labeled peptides were quite different between buffer and complex matrix samples. In fact, the PSAQ protein does not fully compensate the difference in ricin recovery among distinct matrices, probably due to the absence of RTB or glycosylation on PSAQs as mentioned above. In-matrix calibration samples prepared with purified toxin are required to achieve accurate quantification. Standard curves generated in milk and human serum are depicted in Figures 2 and

Figure 3. Evaluation of SEB toxin stability in PBS/BSA, tap water, and milk. Normalized signal obtained. (A) For peptide VLYDDNHVSAINVK, after various toxin incubation times (3, 6, 12, 24, and 48 h) in the three matrices. (B) For five SEB peptides distributed along the SEB sequence, before and after a 48 h incubation in the same 3 mixtures.

proteases) and (ii) an absence of toxin loss in collection tubes in water samples, for at least 48 h.

■

CONCLUSION We have successfully developed a multiplex immunoaffinity extraction LC-MS/MS assay to identify and quantify simultaneously 3 category B toxins. To ascertain quickly the presence of toxin in collected samples, we developed a fast method able to reach quantification limits at or below 1 ng·mL−1 within 5 h. This mass spectrometry-based assay constitutes a complementary approach to immunological methods for toxin quantification. An exquisite specificity is provided by the combination of immunocapture and targeted quantification at high resolution in the PRM mode, with sensitivity similar to or better than that of previous methods using the SRM mode. Although 2 unique peptides are commonly required to provide a positive result, our method allows the detection of 4 to 7 peptides per toxins depending on toxin concentration and type of sample. Thus, even at low toxin concentrations, sequence coverages from 20% to 42% were obtained through at least the best responding peptides (4−5 per protein), ensuring unambiguous toxin identification and differentiation from proteins with similar sequences, while providing information about toxin integrity. Precise quantification was accomplished in complex matrices with isotope-labeled proteins and calibration curves prepared with purified toxins. While labeled tryptic peptides are added at late stages of the analytical workflow, labeled proteins can be spiked earlier in the test samples, thereby maximizing monitoring of the sample process. Although it is difficult to obtain a recombinant protein with conformation and post-translational modifications similar to those of the native toxins, as we observed in this work for the toxic and glycosylated ricin protein, such entities hitherto are the best source of internal standard. Further improvement of the assay could include the production and screening of several toxin antibodies and selecting those providing most similar immuno-extraction recovery between labeled recombinant proteins and native toxins. To increase assay sensitivity further, especially due to the need to confirm positive ELISA results at low concentrations, nanoliquid chromatography could be implemented instead of the standard LC system together with the use of a large initial sample volume.

Figure 2. Calibration curves obtained for the 5 best responding peptides from ETX toxin spiked in milk at purified toxin concentrations ranging from 0.5 to 100 ng·mL−1. The signal from each peptide was normalized to the corresponding PSAQ peptide. The mean slope of 0.095 was calculated from the 5 calibration curves, with a variation of 15%.

S-12B,C, Supporting Information. In comparison with buffer, similar linearity ranges and LOQs were obtained for ricin, ETX toxins, and SEB (in milk only for the latter) (Table 1). Coefficients of variation of calibration curve standards were below 17%, and mean absolute biases were between 2.6% and 10.7% (Table S-3, Supporting Information). In both matrices, the assay was therefore reproducible. Toxin Stability in Food Matrices. The time frame from sample contamination to assay includes the time between contamination and sampling (most probably unknown), transport to the lab, and storage. To our knowledge, there is limited information on toxin stability in matrices, especially in food.55 As a consequence, there is considerable interest in identifying potential degradation over time. Toxins were spiked in buffer with BSA, tap water, and milk at 20 ng·mL−1, and tubes were left at room temperature on the laboratory bench for 3 to 48 h. Normalized signals (with labeled standards spiked on the day of analysis) were then compared with those obtained from samples spiked immediately, to evaluate if sensitivity and also toxin integrity were affected. No significant loss of signal was observed for the best SEB responding peptide with time (Figure 3A) or for any of the other high responding peptides (Figure 3B). Similar results were obtained for ricin and ETX toxins (Figure S-15, Supporting Information). This means that in our conditions there was: (i) no evidence of toxin degradation (such as by milk G

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

■

(25) Duriez, E.; Fenaille, F.; Tabet, J. C.; Lamourette, P.; Hilaire, D.; Becher, F.; Ezan, E. J.Proteome.Res. 2008, 7, 4154−4163. (26) Fredriksson, S.-Å.; Hulst, A. G.; Artursson, E.; de Jong, A. L.; Nilsson, C.; van Baar, B. L. M. Anal. Chem. 2005, 77, 1545−1555. (27) Kalb, S. R.; Barr, J. R. Anal. Chem. 2009, 81, 2037−2042. (28) Ö stin, A.; Bergström, T.; Fredriksson, S.-Å.; Nilsson, C. Anal. Chem. 2007, 79, 6271−6278. (29) Callahan, J. H.; Shefcheck, K. J.; Williams, T. L.; Musser, S. M. Anal. Chem. 2006, 78, 1789−1800. (30) Schlosser, G.; Kačer, P.; Kuzma, M.; Szilágyi, Z.; Sorrentino, A.; Manzo, C.; Pizzano, R.; Malorni, L.; Pocsfalvi, G. Appl. Environ. Microbiol. 2007, 73, 6945−6952. (31) Seyer, A.; Fenaille, F.; Feraudet-Tarisse, C.; Volland, H.; Popoff, M. R.; Tabet, J. C.; Junot, C.; Becher, F. Anal. Chem. 2012, 84, 5103− 5109. (32) Fagerquist, C. K.; Zaragoza, W. J.; Sultan, O.; Woo, N.; Quinones, B.; Cooley, M. B.; Mandrell, R. E. Appl. Environ. Microbiol. 2014, 80, 2928−2940. (33) Sospedra, I.; Soler, C.; Manes, J.; Soriano, J. M. J.Chromatogr.A 2012, 1238, 54−59. (34) Dupuis, A.; Hennekinne, J. A.; Garin, J.; Brun, V. Proteomics 2008, 8, 4633−4636. (35) McGrath, S. C.; Schieltz, D. M.; McWilliams, L. G.; Pirkle, J. L.; Barr, J. R. Anal. Chem. 2011, 83, 2897−2905. (36) Alam, S. I.; Kumar, B.; Kamboj, D. V. Anal. Chem. 2012, 84, 10500−10507. (37) Kull, S.; Pauly, D.; Stormann, B.; Kirchner, S.; Stammler, M.; Dorner, M. B.; Lasch, P.; Naumann, D.; Dorner, B. G. Anal. Chem. 2010, 82, 2916−2924. (38) Picard, G.; Lebert, D.; Louwagie, M.; Adrait, A.; Huillet, C.; Vandenesch, F.; Bruley, C.; Garin, J.; Jaquinod, M.; Brun, V. J. Mass Spectrom. 2012, 47, 1353−1363. (39) Picotti, P.; Aebersold, R. Nat. Methods 2012, 9, 555−566. (40) Gallien, S.; Duriez, E.; Crone, C.; Kellmann, M.; Moehring, T.; Domon, B. Mol. Cell. Proteomics 2012, 11, 1709−1723. (41) Peterson, A. C.; Russell, J. D.; Bailey, D. J.; Westphall, M. S.; Coon, J. J. Mol. Cell. Proteomics 2012, 11, 1475−1488. (42) Chenau, J.; Fenaille, F.; Ezan, E.; Morel, N.; Lamourette, P.; Goossens, P. L.; Becher, F. Anal. Chem. 2011, 83, 8675−8682. (43) Brun, V.; Dupuis, A.; Adrait, A.; Marcellin, M.; Thomas, D.; Court, M.; Vandenesch, F.; Garin, J. Mol. Cell. Proteomics 2007, 6, 2139− 2149. (44) Brun, V.; Masselon, C.; Garin, J.; Dupuis, A. J. Proteomics 2009, 72, 740−749. (45) Prigent, J.; Panigai, L.; Lamourette, P.; Sauvaire, D.; Devilliers, K.; Plaisance, M.; Volland, H.; Creminon, C.; Simon, S. PLoS One 2011, 6, e20166. (46) Hennekinne, J. A.; Brun, V.; De Buyser, M. L.; Dupuis, A.; Ostyn, A.; Dragacci, S. Appl. Environ. Microbiol. 2009, 75, 882−884. (47) Schieltz, D. M.; McWilliams, L. G.; Kuklenyik, Z.; Prezioso, S. M.; Carter, A. J.; Williamson, Y. M.; McGrath, S. C.; Morse, S. A.; Barr, J. R. Toxicon 2015, 95, 72−83. (48) Gallien, S.; Bourmaud, A.; Kim, S. Y.; Domon, B. J. Proteomics 2014, 100, 147−159. (49) Becher, F.; Duriez, E.; Volland, H.; Tabet, J. C.; Ezan, E. Anal. Chem. 2007, 79, 659−665. (50) Chenau, J.; Fenaille, F.; Simon, S.; Filali, S.; Volland, H.; Junot, C.; Carniel, E.; Becher, F. Anal. Chem. 2014, 86, 6144−6152. (51) Adrait, A.; Lebert, D.; Trauchessec, M.; Dupuis, A.; Louwagie, M.; Masselon, C.; Jaquinod, M.; Chevalier, B.; Vandenesch, F.; Garin, J.; Bruley, C.; Brun, V. J. Proteomics 2012, 75, 3041−3049. (52) Norrgran, J.; Williams, T. L.; Woolfitt, A. R.; Solano, M. I.; Pirkle, J. L.; Barr, J. R. Anal. Biochem. 2009, 393, 48−55. (53) Gerber, S. A.; Rush, J.; Stemman, O.; Kirschner, M. W.; Gygi, S. P. Proc. Natl. Acad. Sci. U. S. A. 2003, 100, 6940−6945. (54) Villanueva, J.; Carrascal, M.; Abian, J. J. Proteomics 2014, 96, 184− 199. (55) Jackson, L. S.; Zhang, Z.; Tolleson, W. H. J. Food Sci. 2010, 75, T65−T71.

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b01900.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: 33-1-69-08-13-15. Fax: 33-1-69-08-59-07. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Guillaume Picard and Mathilde Louwagie for technical and scientific assistance. This project was supported by the joint ministerial program of R&D against CBRNE risks.

■

REFERENCES

(1) http://www.bt.cdc.gov/ (accessed May 14, 2015). (2) Mantis, N. J. Adv. Drug Delivery Rev. 2005, 57, 1424−1439. (3) Balali-Mood, M.; Moshiri, M.; Etemad, L. Toxicon 2013, 69, 131− 142. (4) Boyer, A. E.; Moura, H.; Woolfitt, A. R.; Kalb, S. R.; McWilliams, L. G.; Pavlopoulos, A.; Schmidt, J. G.; Ashley, D. L.; Barr, J. R. Anal. Chem. 2005, 77, 3916−3924. (5) Olsnes, S. Toxicon 2004, 44, 361−370. (6) Woo, B. H.; Lee, J. T.; Lee, K. C. Protein Expression Purif. 1998, 13, 150−154. (7) Araki, T.; Funatsu, G. Biochim. Biophys. Acta, Protein Struct. Mol. Enzymol. 1987, 911, 191−200. (8) Worbs, S.; Kohler, K.; Pauly, D.; Avondet, M. A.; Schaer, M.; Dorner, M. B.; Dorner, B. G. Toxins 2011, 3, 1332−1372. (9) Pinchuk, I. V.; Beswick, E. J.; Reyes, V. E. Toxins 2010, 2, 2177− 2197. (10) Krakauer, T.; Stiles, B. G. Virulence. 2013, 4, 759−773. (11) LeClaire, R. D.; Hunt, R. E.; Bavari, S.; Estep, J. E.; Nelson, G. O.; Wilhelmsen, C. L. Toxicol. Pathol. 1996, 24, 619−626. (12) Popoff, M. R. FEBS J. 2011, 278, 4602−4615. (13) Stiles, B. G.; Barth, G.; Barth, H.; Popoff, M. R. Toxins 2013, 5, 2138−2160. (14) Jenko, K. L.; Zhang, Y.; Kostenko, Y.; Fan, Y.; Garcia-Rodriguez, C.; Lou, J.; Marks, J. D.; Varnum, S. M. Analyst 2014, 139, 5093−5102. (15) Pauly, D.; Kirchner, S.; Stoermann, B.; Schreiber, T.; Kaulfuss, S.; Schade, R.; Zbinden, R.; Avondet, M. A.; Dorner, M. B.; Dorner, B. G. Analyst 2009, 134, 2028−2039. (16) Shlyapnikov, Y. M.; Shlyapnikova, E. A.; Simonova, M. A.; Shepelyakovskaya, A. O.; Brovko, F. A.; Komaleva, R. L.; Grishin, E. V.; Morozov, V. N. Anal. Chem. 2012, 84, 5596−5603. (17) Weingart, O. G.; Gao, H.; Crevoisier, F.; Heitger, F.; Avondet, M. A.; Sigrist, H. Sensors 2012, 12, 2324−2339. (18) Chen, H. Y.; Tran, H.; Foo, L. Y.; Sew, T. W.; Loke, W. K. Anal. Bioanal. Chem. 2014, 406, 5157−5169. (19) Ler, S. G.; Lee, F. K.; Gopalakrishnakone, P. J.Chromatogr.A 2006, 1133, 1−12. (20) Boyer, A. E.; Gallegos-Candela, M.; Lins, R. C.; Kuklenyik, Z.; Woolfitt, A.; Moura, H.; Kalb, S.; Quinn, C. P.; Barr, J. R. Molecules 2011, 16, 2391−2413. (21) Demirev, P. A.; Fenselau, C. J. Mass Spectrom. 2008, 43, 1441− 1457. (22) Tevell Åberg, A.; Bjornstad, K.; Hedeland, M. Biosecur. Bioterrorism 2013, 11 (Suppl 1), S215−S226. (23) Bergström, T.; Fredriksson, S.-Å.; Nilsson, C.; Åstot, C. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2015, 974, 109−117. (24) Brinkworth, C. S. Anal. Chem. 2010, 82, 5246−5252. H

DOI: 10.1021/acs.analchem.5b01900 Anal. Chem. XXXX, XXX, XXX−XXX