Development and Practical Application of a Library of CID Accurate Mass

Anal Bioanal Chem (2011) 400:101–117 DOI 10.1007/s00216-010-4450-9

ORIGINAL PAPER

Development and practical application of a library of CID accurate mass spectra of more than 2,500 toxic compounds for systematic toxicological analysis by LC–QTOF-MS with data-dependent acquisition

Sebastian Broecker & Sieglinde Herre & Bernhard Wüst & Jerry Zweigenbaum & Fritz Pragst

Received: 30 September 2010 /Revised: 15 November 2010 /Accepted: 16 November 2010 /Published online: 3 December 2010 # Springer-Verlag 2010

Abstract A library of collision-induced dissociation (CID) data and molecular formulas of more than 7,500 toxicolog- accurate mass spectra has been developed for efficient use ically relevant substances to form the “database and library of liquid chromatography in combination with hybrid of toxic compounds”. For practical evaluation, blood and quadrupole time-of-flight mass spectrometry (LC–QTOF- urine samples were spiked with a mixture of 33 drugs at MS) as a tool in systematic toxicological analysis. The seven concentrations between 0.5 and 500 ng mL−1, pre- mass spectra (Δm<3 ppm) of more than 2,500 illegal and pared by dichloromethane extraction or protein precipita- therapeutic drugs, pesticides, alkaloids, other toxic chem- tion, and analyzed by LC–QTOF-MS in data-dependent icals and metabolites were measured, by use of an Agilent acquisition mode. Unambiguous identification by library 6530 instrument, by flow-injection of 1 ng of the pure search was possible for typical basic drugs down to 0.5– substances in aqueous ammonium formate–formic acid– 2ngmL−1 and for benzodiazepines down to 2–20 ng mL−1. methanol, with positive and negative electrospray- The efficiency of the method was also demonstrated by re- ionization (ESI), selection of the protonated or deproto- analysis of venous blood samples from 50 death cases and nated molecules [M+H]+ or [M−H]− by the quadrupole, and comparison with previous results. In conclusion, LC– collision induced dissociation (CID) with nitrogen as QTOF-MS in data-dependent acquisition mode combined collision gas at CID energies of 10, 20, and 40 eV. The with an accurate mass database and CID spectra library fragment mass spectra were controlled for structural seemed to be one of the most efficient tools for systematic plausibility, corrected by recalculation to the theoretical toxicological analysis. fragment masses and added to a database of accurate mass Keywords Accurate mass spectra library. Collision-induced dissociation . Liquid chromatography. Time of flight mass Published in the special issue Forensic Toxicology with Guest Editors spectrometry. Peak identification . Systematic toxicological Frank T. Peters, Hans H. Maurer, and Frank Musshoff. analysis S. Broecker : S. Herre : F. Pragst (*) Institute of Legal Medicine, University Hospital Charité, Turmstraße 21, Building N, 10559, Berlin, Germany Introduction e-mail: [email protected] Systematic toxicological analysis is the general search for B. Wüst toxic compounds in a biological sample, for instance in Agilent Technologies, Hewlett-Packard-Straße 8, human blood, urine, organ tissues, or hair, without any 76337, Waldbronn, Germany information about presence and kind of poisons. It is one of the most difficult tasks of analytical chemistry because of J. Zweigenbaum the huge number of possible poisons and poison metabo- Agilent Technologies, Inc., 2850 Centerville Road, BL3-2 3L11, lites which may occur in low and very low concentrations Wilmington, DE 19808-1610, USA in the complicated matrix. It includes toxic gases, volatile 102 S. Broecker et al. substances, metal ions and, as the largest group, organic sition” is restricted to a list of preselected precursors compounds with low volatility such as illegal and thera- included in the method. peutic drugs, pesticides, chemical reagents, and alkaloids. Mass spectra libraries for LC–MS with fragmentation by Up-to-date methods for systematic toxicological analysis of in-source collision and LC–MS–MS with fragmentation in organic compounds consist of suitable sample preparation the collision cell between both MS units have been which is able to extract as many poisons as possible from described in several papers and contain between 301 and the matrix, and a combination of chromatography and 1,253 substances [13–16]. molecular spectrometry in order to separate the extracted The availability of time-of-flight mass spectrometers mixture and to characterize the components. Widely used with much improved mass resolution and mass accuracy as method combinations are capillary gas chromatography– detectors in liquid chromatography (LC–TOF-MS) provid- mass spectrometry (GC–MS) [1] and high-performance ed new possibilities in the use of LC–MS for toxicological liquid chromatography with photodiode array detection screening in blood, urine, hair, meconium, and vitreous (HPLC–DAD) [2]. There is not yet any possibility of humor [19–36]. The working principle of these instruments determining the exact structure from these spectra. Therefore, enables comprehensive recording of all data. Therefore, substance identification is always based on comparison of the there is no a priori limitation or prediction of the substances unknown spectrum with those in a library of spectra of included in the search procedure. The increased mass toxicologically relevant compounds. resolution also provides high selectivity for overlapping In the last decade, several approaches have been made to peaks and high matrix burden. The most important use liquid chromatography in combination with mass advantage is that the molecular formula of an analyte is spectrometry (LC–MS or LC–MS–MS) with electrospray directly available from the accurate molecular mass and the ionization (ESI) or atmospheric pressure chemical ioniza- isotope peak pattern. For substance identification by use of tion (APCI) for systematic toxicological analysis [3–36]. the molecular formula, theoretical databases of toxicolog- The application of single-stage quadrupole or ion-trap mass ically relevant compounds with up to 50,500 substances spectrometers for this purpose is limited because of including metabolites [29] or in-house databases with 100 disturbances by high matrix burden and co-eluting peaks. to 869 substances have been created [19–22, 24–28, 30, By use of triple-quadrupole mass spectrometers or, more 33–35]. favorably, hybrid triple-quadrupole linear ion-trap mass Different search procedures are used in TOF methodology. spectrometers these problems have been solved but the The group of Ojanperä and Pelander used reversed target search can only be performed as multi-targeted screening search, which means the TOF file is searched for target masses [7–18]. This means that substances can be detected only if included in the library [18–27]. Polettini et al. used forward they are a priory included in the method. The number of basepeak search, which means the base mass peak of the substances in such a procedure is limited by the minimum unknown chromatographic peak in the analysis file was dwell time for each multi reaction monitoring (MRM) searched after proton subtraction in their large database of transition included in one measurement cycle. Nevertheless, 50,500 compounds of toxicological interest and many phase a powerful screening procedure with 700 substances in one I and phase II metabolites [31, 32]. chromatographic run has been developed by Dresen et al. However, the molecular formula of an eluted unknown with a hybrid triple-quadrupole linear ion-trap mass substance in a chromatogram is only a first step of the spectrometer by limiting the detection time of every analyte identification, because of the huge number of possible to a chromatographic time window of 2 min, information- isomers, as can easily be shown by use of chemical dependent acquisition (IDA) using the sensitive enhanced software. For instance, according to Molgen (molecular product ion scan of the instrument, and uniting the fragment structure generation [37]) for the nominal molecular mass ions from three collision energies in the trap to obtain one M=149, 27 different molecular formulas are theoretically mixed mass spectrum (collision energy spread) [17]. possible if only the elements C, H, N, and O are included. Generally, LC screenings with MS–MS identification TOF-MS with mass accuracy <3 ppm can clearly distin- consist of a survey scan to detect the analytes and a guish between these 27 possibilities. One is C9H11NO with dependent scan for measurement of the corresponding MS– the accurate molecular mass 149.084060. Based on the MS spectra which are submitted to library search for rules of chemical bonds between these atoms, the software identification. Survey scan and dependent scan can be theoretically calculates 25,895,621 structural isomers accomplished within the same analytical run by automatic (stereoisomers not included). From these, 724 substances selection of precursor ions and measurement of the MS–MS are recorded in the Beilstein database and only 45 are spectra immediately after their detection in the survey scan. included in the NIST register. The software “Chemspider” In “data-dependent acquisition” this is determined only by [38] shows structural formulas of 829 compounds with the the actual MS data whereas “information-dependent acqui- molecular formula C9H11NO which are all reasonable, Development and practical application of a library of CID accurat 103 among them, for instance, cathinone, N,N-dimethylbenza- masses and isotope pattern are calculated. The practical mide, p-dimethylaminobenzaldehyde, 3,4-dimethylbenza- application of this “Library and Database of Toxic Com- mide, p-aminopropiophenone and also heterocyclic pounds” was examined with spiked and real blood and substances such as 6-hydroxy-1,2,3,4-tetrahydroquinoline urine samples using the auto-MS–MS mode. or N-methyl-6-hydroxy-2,3-dihydroindole. It is very important to be aware of this almost unlimited structural diversity of organic chemistry when commencing system- Experimental atic toxicological analysis. Therefore, much more evidence is required in order to Chemicals and reagents distinguish between isomers. For this purpose, retention times under defined chromatographic conditions have been The solvents and chemicals used for the mobile phase were measured for 100 to 400 substances in some in-house purchased as follows: methanol (LC–MS grade), acetonitrile libraries [20–25, 27, 30, 33, 35]. Identification of metabo- (LC–MS grade), and ammonium acetate (HPLC grade) from lites in a metabolomic approach is another possibility for Fisher scientific (Schwerte, Germany), ammonium formate confirming or rejecting a proposal from the theoretical (LC–MS grade) from Agilent Technologies, water (HPLC database [20, 32]. grade) and formic acid (99+% for analysis) from Acros Restriction to only toxicologically relevant substances in Organics (Geel, Belgium). All other solvents and reagents theoretical databases, retention times under standardized used for sample preparation were obtained from Merck conditions, and search for metabolites are very helpful. (Darmstadt, Germany) in analytical grade purity. However, much more structure-specific information should be obtained from collision-induced dissociation (CID) Reference substances fragment spectra. For this reason, in-source collision- induced dissociation spectra have been measured in single The reference substances (more than 2,500) included in the stage LC–TOF-MS procedures by some authors [33, 35]; library of spectra were generously donated by numerous monitoring of drug class-specific CID mass fragments was pharmaceutical manufacturers or purchased from chemical performed for detection of nontarget analytes [35], and companies such as LGC Promochem, Sigma, Radian, special software for fragmentation prediction was used in Riedel–de Haën, Dr Ehrenstorfer, or Lipomed. The UV order to differentiate between structural isomers in a target spectra of the same substances were previously measured drug database by LC–QTOF-MS [26]. by HPLC–DAD for the database “UV spectra of Toxic Hybrid quadrupole time-of-flight mass spectrometry Compounds” [39] and a complete substance list is given (LC–QTOF-MS) has the advantage that, in contrast with there. The identity and sufficient purity of the substances in-source CID, the fragment spectra are not disturbed by was always monitored by mass spectrometry. In addition, matrix and co-eluting substances. A library of CID spectra 52 deuterated standards of illegal drugs, legal opiates, and of 319 substances measured with an LC–QTOF-MS benzodiazepines were included; these were purchased from instrument at ten collision energies was described by Pavlic LGC Promochem (Wesel, Germany). et al. [28]. Furthermore, for application in systematic toxicological analysis, the LC–QTOF-MS instrument can Instruments and software be operated in a data-dependent acquisition mode (auto-MS–MS mode) in order to combine the advantage All measurements were performed with a 6530 accurate- of TOF-MS for comprehensive data collection with the mass Q-TOF LC–MS instrument (Agilent Technologies, measurement of accurate CID fragment spectra from all Santa Clara, USA). The Agilent 1200 SL series HPLC essential components of the sample after isolation of the consisted of a degasser, a thermostated HiP-ALS autosam- corresponding parent ions by the quadrupole (QTOF-MS). pler, a Bin Pump SL binary pump, and a TCC SL column This technique was first applied to toxicological analysis by oven. The QTOF-MS instrument was operated with an Decaestecker et al. for toxicological analysis [29, 30] electrospray ion source ESI+Agilent Jet Stream Technology although the advantages of accurate mass were not yet in positive and negative ionization mode, a quadrupole for really utilized. isolation of precursor ions with a mass window of 1.3 or 4 In our study the accurate mass CID spectra of more than m/z in MS–MS mode, a linear hexapole collision cell with 2,500 toxicologically relevant substances were measured nitrogen as collision gas and collision energy 0–40 eV, a with a hybrid quadrupole time-of-flight mass spectrometer TOF-MS with mass accuracy <3 ppm, mass resolution of (QTOF-MS) at three different collision energies and 5,000 to 10,000 (100 to 922m/z), a measuring frequency of included in a TOF-MS database with molecular formulas 10,000 transients s−1 and a detection frequency of 2 GHz of more than 7,500 substances from which the accurate (200,000 points/transient). 104 S. Broecker et al.

The data presented here used MassHunter Acquisition known from case histories and/or toxicological investiga- B.02.01 with Service Pack 3 for the Agilent TOF and tion with HPLC–DAD, GC–MS or immunoassay. QTOF and MassHunter Qualitative Analysis B.03.01 with Service Pack 3. In addition, the Personal Compound Sample preparation Database and Library Software B.03.01 was used to interrogate the database and library directly. Liquid–liquid extraction with dichloromethane

Measurement of library spectra The procedure described previously for HPLC–DAD [2] was only slightly changed. To 500 μL whole blood, serum, For measurement of the CID mass spectra for the library, or plasma in a 1.5-mL Eppendorf vial 100 μL 0.1 mol L−1 1mgmL−1 stock solutions of all compounds in methanol HCl (acidic extract) or 100 μL solution of Tris substance −1 were prepared and diluted to 1 μgmL . If solubility was (24.3 g in 1 L H2O, pH 9.0, basic extract) and 400 μL insufficient, addition of formic acid, use of water or CH2Cl2 were added. The mixture was vortex mixed for acetonitrile, or a larger volume of solvent was tried. The 1 min and centrifuged for 5 min at 13,200 rpm. The CH2Cl2 spectra were measured by flow injection of 1 ng of each layer (200 μL) was aspirated with a 200 μL Hamilton-type substance (1 μLof1μgmL−1 solution) with 5 mmol L−1 syringe and evaporated to dryness in a nitrogen stream at ammonium formate and 0.01% formic acid in water–0.01% 40 °C. The residue was dissolved in 100 μL ACN–0.1% formic acid in methanol (50:50) as mobile phase. HCOOH in water (35:65v/v) and 5 μL was injected for The protonated or deprotonated molecules [M+H]+ or LC–QTOF-MS measurement. [M−H]− were selected by the quadrupole with a mass resolution of 1.3m/z .ThreeMS–MS spectra were Protein precipitation with acetonitrile generated in product-ion-scan mode at CID energies of 10, 20, and 40 eV. The QTOF conditions applied were: gas Blood, serum, or plasma (100 μL) was placed in a 1.5-mL temperature 250 °C, gas flow 6 Lmin−1, nebulizer pressure Eppendorf vial and 400 μL acetonitrile was added. The 35 psi, sheath gas temperature 300 °C, sheath gas flow mixture was vortex mixed for 1 min and centrifuged for 10 Lmin−1, VCap voltage 3500 V, nozzle voltage 100 V, 5 min at 13,200 rpm. Then, 400 μL supernatant was fragmentor voltage 150 V, mass range (MS and MS–MS) separated and evaporated to dryness in a nitrogen stream at 50–1700m/z depending on substance, scan rate 8 Hz in MS 40 °C. The residue was reconstituted in 80 μL ACN–0.1 % and MS–MS experiments, reference ions for mass calibra- HCOOH in water (35:65v/v) and 5 μL was injected for tion: purine 121.050873 [M+H]+ and 119.036319 [M−H]−, LC–QTOF-MS measurement. HP-921=hexakis(1H,1H,3H-tetrafluoropropoxy)phosphazine + − 922.009798 [M+H] and 966.000725 [M+HCO2] . Preparation of urine samples

Processing of library spectra Urine samples were centrifuged for 5 min at 13,200 rpm. The supernatant (100 μL) in a 2-mL autosampler vial was The measured fragment spectra were controlled for structural diluted with 400 μL of 10 mmol L−1 ammonium acetate in plausibility and corrected by recalculation to the theoretical water (pH 6.8) and 5 μL was injected for LC–QTOF-MS masses by special processing before arrangement in the measurement. spectra library and database, which contains at its present state more than 7,500 toxic compounds, which means Sample measurement approximately 5,000 entries with only theoretical accurate mass data including isotope pattern and further 2,500 entries Chromatographic separation was performed at 50 °C with a with additional three accurate mass CID fragment spectra at Poroshell 120 EC-C18, 2.1×100 mm, 2.7 μm, column. For −1 10, 20, and 40 eV collision energy. gradient elution the mobile phases 10 mmol L NH4Ac in H2O (A) and methanol (B) were used with the time Blood and urine samples program: 0 min 10% B, linear to 50% B at 8 min, linear to 100% B at 20 min, constant 100% B to 23.9 min, back to Drug-free blood samples for validation of the method were 10% B at 24 min and equilibration for 3 min. The flow rate collected from volunteers among the laboratory staff. was 0.4 mL min−1. Furthermore, the method was applied to 50 blood samples The QTOF-MS instrument was operated under the from autopsy cases which were investigated in the Institute conditions: ion source ESI+Agilent Jet Stream Technology of Legal Medicine of the University Hospital Charité Berlin in positive ionization mode, quadrupole was used as an ion and for which exposure to therapeutic or illegal drugs was guide in MS mode and for selection of precursor ions with Development and practical application of a library of CID accurat 105

Δm/z=4 in MS–MS mode, collision cell without CID in pole with a mass window of Δm/z=1.3 and submitted to MS mode and with CID of precursor ions in MS–MS mode CID with nitrogen as collision gas in the collision cell at at mass dependent ramped CID energy (offset 4 eV, slope collision energies of 0, 10, 20, and 40 eV. In the range of 6 eV/100m/z), TOF-MS with a mass range of 100–1000m/z the substance peak between 20 and 30 spectra at each in MS mode and 50–600m/z in MS–MS mode. The scan collision energy were measured, with an accumulation rate rate was 4 Hz in MS and MS–MS experiments. The source of 1600 transients per spectrum, and were merged. The conditions were: gas temperature 320 °C, gas flow 8 L spectra before and after the peak were subtracted for min−1, nebulizer pressure 35 psi, sheath gas temperature background correction. For storage in the library, only the 380 °C, sheath gas flow 11 Lmin−1, VCap voltage 3000 V spectra acquired at collision energies of 10, 20, and 40 eV and nozzle voltage 0 V. The remaining instrument con- were used and only masses with an intensity above 100 ditions were the same as described above for measurement counts were imported into the library. The spectra were of library spectra. normalized to the largest peak. There was no other For systematic toxicological analysis the auto-MS–MS limitation of the number of fragment ions. mode (data-dependent acquisition) was used with a cycle The ions found in each spectrum were then identified by time of 1.1 s, 0.25 s measurement in MS mode, selection their chemical formula, and the exact mass of the ions was of three precursors which were fragmented in the calculated and used to replace the measured mass while following three MS–MS experiments, and active exclu- maintaining the relative abundance of each ion. This sion after one spectrum for 0.1 min. For this the three most provides an accurate mass library for routine matching of abundant masses were selected as precursors including fragment ions in real samples with the expected ions for the only singly-charged ions with a threshold abundance of proposed compound identification. 1000 counts. As an example the three spectra obtained for the anticoagulant warfarin are shown in Fig. 1. Whereas the [M+H]+ (m/z=309.11214) is one of the highest peaks at Results and discussion 10 eV it decreases strongly at 20 eV and is not seen at 40 eV, in favor of an increasing number and intensity of Collision-induced accurate mass spectra library fragment ions. The two most abundant fragments at m/z= 251.0702 and 163.0389 correspond to the fragmentation at Because the library of spectra was designed for systematic the α and β positions of the side chain, respectively (loss of toxicological analysis, it was important to include as many C3H6O and C10H10O). Depending on molecular mass and toxicologically relevant substances as possible. Selection structure, the degree of fragmentation at the three CID was based on the substance pool in the laboratory which the energies was different from compound to compound. For authors used for the database “UV Spectra of Toxic instance amphetamine and piracetam were already strongly Compounds” [2] but included also compounds without fragmented at 10 eV and totally disintegrated at 40 eV UV absorbance. In the first step of library development only whereas for morphine and strychnine the protonated positive and negative electrospray ionization was applied. molecule peak still dominated at 10 and 20 eV and For this reason compounds with insufficient ionization optimum fragmentation occurred only at 40 eV. Neverthe- under these conditions are still missing from the library and less, it was seen from the spectra that sufficient fragmen- will be added in a next step by use of atmospheric pressure tation was achieved for the large majority of the compounds chemical ionization (APCI). Table 1 shows the composition in order to provide structure-specific information. Some of of the library in its current state, on the basis of different the substances, for example digoxin and digitoxin provided groups of toxic compounds. Among them are 139 metab- only a very small ESI yield of [M+H]+, because they were olites (5.5%) which were also available as pure reference mainly ionized as sodium and potassium clusters [M+Na]+ substances. and [M+K]+.Thismustbekeptinmindinsample Measurement of the CID spectra was performed by flow screening for these highly toxic drugs. injection of 1 ng of each substance in 1 μL solvent in The spectra of more than 2,500 substances were added to mobile phase consisting of a 1:1 mixture of 5 mmol L−1 the “Personal Forensics/Toxicology Database” described ammonium formate+0.01% formic acid in water and 0.01% previously [40], which contains theoretically calculated formic acid in methanol, at a flow rate of 0.1 mL min−1. accurate mass data and molecular formulas of more than Chromatographic separation was not necessary because the 7,500 toxicologically relevant substances, to form the pure substances were measured and because possible “Personal Compound Database and Library of Toxic impurities were excluded by the quadrupole. The mono- Compounds”. The arrangement of the library and database isotopic ions [M+H]+ or [M−H]− as theoretically calculated is not finished but is steadily being extended by addition of from the structural formula were separated by the quadru- further compounds. 106 S. Broecker et al.

Table 1 Composition of the CID accurate mass spectra library of or medical use. Alkaloids and plant poisons such as aconitine, nicotine toxic compounds (August 2010). The arrangement in groups is or strychnine were included into the groups according to their effect. slightly arbitrary since many substances have more than one effect Metabolites are in the same group as their parent drugs

Group no. Effect or use Number of compounds Examples

1 Illegal drugs, substances with addiction 228 Cocaine, fentanyl, mescaline, pentobarbital, THC potential, hypnotics 2 Psychopharmaceuticals, benzodiazepines, 289 Amitriptyline, carbamazepine, diazepam, fluoxetrine, perazine CNS-active substances 3 Non-opioid analgesics, antirheumatics, 210 Ambroxol, etoricoxib, metamizole, paracetamol, antitussives, and similar 4 Further CNS-active substances, 154 Astemizole, benserazide, pyridostigmine, talastine antihistaminics, antiallergics 5 Cardiovascular agents 234 Amiodarone, digoxin, flecainide, metoprolol, propafenone 6 Diuretics, antidiabetics, anticoagulants, 174 Carbutamide, furosemide, glipizide, phenprocoumon various drugs 7 Steroids, hormones, vitamins, endogenous 211 Ascorbic acid, ethinylestradiol, flucinonide, stanazolol substances 8 Antibiotics, antimalarials, cytostatics, 256 Ampicillin, atazanavir, lincomycin, ofloxacin, sulfamerazin virustatics, and similar 9 Fungicides, disinfectants, adjuvants, 178 Fluconazole, hexachlorophene, ioglicic acid, terbinafine, diagnostics 10 Insecticides, acaricides, nematicides, 219 Bromurone, fluazuron, parathion, propoxur, tetramethrin and similar 11 Herbicides 305 Atrazine, dalapon, glyphosate, diuron, imazapyr, paraquat 12 Carcinogens, chemical reagents, fragrances 32 Aniline, acrylamide, diethyl phthalate, pyridine 13 Deuterated standards 52 Cocaethylene-D3, diazepam-D5, fentanyl-D5, haloperidol-D4

No, or unsuitable, CID spectra were obtained from concern according to the algorithms described below in approximately 500 other substances under the experimental the next two sections. The resulting positive list is then conditions described, because of insufficient ESI yield for visually examined for quality of match and those com- neutral compounds (e.g. halogenated pesticides, steroids, pounds that appear as possibly present are confirmed in a phenols), formation of sodium adducts (e.g. cardiac glyco- second chromatographic run under targeted MS–MS con- sides), insufficient stability in the ion source of the ditions in which the resulting MS–MS spectra are used to compounds (organic phosphates such as E605 or organic search the library for identification. This is a combined nitrates such as erythritol tetranitrate), and formation of multi forward and backward search with corresponding separate charged ions (e.g. antibiotics). Furthermore, dicationic species scores. Although this workflow has the advantage of higher such as pancuronium, vecuronium, or suxamethonium must sensitivity it is rather complicated because of the two runs still be added to the library. and the large number of false positives resulting from the The CID accurate mass library was applied to chromato- first run which had to be introduced into the targeted run. graphic files acquired by use of different Agilent LC– The auto-MS–MS workflow (data-dependent acquisition) QTOF-MS instruments of the series 6520, 6530, and 6540 is much more comfortable but it is less sensitive. The principle and proved to be fully suitable for peak identification is shown in Fig. 2, with a real sample as an example. It can irrespective of the instrument. be seen from the total ion chromatogram (Fig. 2a) and more in detail from the small part with higher time resolution Workflow of substance identification (Fig. 2b) that the instrument is operated with steady alternation of MS and MS–MS mode with a cycle time of Sample measurement in data-dependent acquisition mode 1.1 s. In MS mode, for 0.25 s the full mass spectrum at the (auto-MS–MS mode) corresponding retention time is recorded, three precursor ions are selected and for each of these the mass-dependent In principle, there are two workflows for use of LC–QTOF- collision energy is chosen according to Eq. 1. MS in combination with the database and library of toxic compounds for systematic toxicological analysis [41]. The Collision energy ¼ðÞ4 þ 0:06m=z eV ð1Þ “Targeted MS–MS” workflow measures the sample in a first chromatographic run in single MS mode and performs The quadrupole then selects these three masses in a forward database search for possible compounds of succession and the CID accurate mass spectra are measured Development and practical application of a library of CID accurat 107

Library spectrum 100 163.03897 O O 80 10 eV CH3 309.11214 60 251.07027 OH O 40 20 147.08044 291.10157 0 m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 Library spectrum

100 163.03897 251.07027

80 20 eV 60

40 20 147.08044 121.02841 223.07536 291.10157 0 m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 Library spectrum 121.02841 100

80 40 eV 163.03897

60 40 77.03858 20 223.07536 251.07027 65.03858 93.03349 51.02293 205.06479 0 m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320

Fig. 1 Accurate mass spectra of warfarin at CID energies of 10, 20, and 40 eV as stored in the library during the residual time of the measurement cycle. These edge of a peak and may have only low quality. Therefore, the three masses are excluded from MS–MS measurement for exclusion time chosen was shorter than the chromatographic 0.1 min in order to enable the acquisition of other co- peak width in order enable a second acquisition of the same eluting substances, and the instrument goes on to the next ion closer to the peak maximum. cycle. The corresponding four mass spectra from the MS Altogether, the auto-MS–MS file of the sample contains and MS–MS modes are shown in Fig. 2c to f. also the CID fragment spectra in addition to the accurate Cycle time, time for MS measurement within the cycle, and molecular masses of all essential constituents. number of precursor ions can be adjusted within specific limits. Equation 1 takes into account that the collision energy Peak identification necessary for sufficient fragmentation increases with increasing molecular mass. However, because the fragmentation For post-run analysis of an LC–QTOF-MS file measured in also depends strongly on the specific structure, this method auto-MS–MS mode, new efficient tools of the MassHunter of automatic choice of the collision energy is not always software are available. First, the background is removed by optimum and will be improved in the future. subtraction of constant or slowly changing signals. Then, a Furthermore, an abundance threshold (e.g. 1000 counts) is tool “Find Compounds” is applied which extracts the set for [M+H]+ or [M−H]− in order to include only essential elution profile of each mass and groups masses with the components in measurement of CID spectra. Nevertheless, same elution profile to so-called “Compounds”. Such a despite high concentration, the first CID spectrum of an ion “Compound” consists, for instance, of the protonated may be recorded at relatively low abundance in the leading molecule [M+H]+, cluster ions with sodium, potassium or 108 S. Broecker et al.

1.6 Abundance 1.4 x 107 (a) 1.2 1 0.8 0.6 0.4 0.2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Acquisition time (min)

Abundance x107 1.6 MS MS 1.4 MS (b) 1.2 1 0.8 0.6 0.4 MS/MS(2) MS/MS(1) MS/MS(3) 0.2 0 9.81 9.82 9.83 9.84 9.85 9.86 9.87 9.88 9.89 9.9 9.91 9.92 9.93 9.94 Acquisition time (min)

3 + Counts x10 6 [M+H] (f) Counts x10 3 1.6 110.0963 [M+H]+ 4 (c) 495.1797 1 (d) 25.0 eV + [M+H] 2 (e) 1.2 3 138.0910 237.1021 164.1175 2 0.8 349.2032 473.1972 259.0839 1 0.4 + 244.1424 289.2126 275.0590 [M+H] 1 (d) 0 0 200 240 280 320 360 400 440 480 520 40 120 200 280 360 (m/z) (m/z)

3.2 Counts x103 Counts x105 2.8 259.0837 2.8 (e) 194.0957 (f) [M+H]+ 2.4 [M+H]+ 2.4 2 3 18.2 eV 2 33.7 eV 2 1.6 1.6 1.2 1.2 0.8 0.8 237.1020 0.4 0.4 0 0 40 60 80 100 120 140 160 180 200 220 240 260 50 100 150 200 250 300 350 400 450 500 550 (m/z) (m/z) Development and practical application of a library of CID accurat 109

R Fig. 2 Poisoning case 994/09, with protein precipitation of the venous application of a system of scores and exclusion criteria. blood sample. Analysis of a sample by LC–QTOF-MS in auto-MS–MS mode. (a) Total ion chromatogram. (b) Part of the total ion chromatogram For the MS data, identification is based on comparison of between 9.81 and 9.94 min shown with increased time resolution and the exact value of the compound’s neutral mono-isotopic indication of the MS and the three MS–MS acquisitions. (c) Mass spectrum mass with the measured mass calculated from ionic m/z + in MS mode at 9.845 min with indication of the three ions [M+H]1 , + + values of all detected specified adducts within, typically, a 3 [M+H]2 and [M+H]3 at m/z=349.2032, 237.1021, and 495.1797 automatically selected for MS–MS measurement (d) CID fragment to 5 ppm mass error. In addition, identification relies on + + spectrum of [M+H]1 . (e) CID fragment spectrum of [M+H]2 . (f) CID both spacing and relative abundance of the isotopes + fragment spectrum of [M+H]3 . The collision energies given in (d), (e), detected. Agreement with the database entries is assessed and (f) were automatically calculated by use of Eq. 1 by use of a weighted score calculated from the mass match, the abundance match, and the spacing match. As a result ammonium, dimers of this ion species, the corresponding no, one, or more hits can be proposed. More than one hit isotope peaks, and the CID spectra of [M+H]+ in the occurs for isomers or isobars with very close accurate retention time range of the peak. The chromatogram molecular masses present in the database. obtained after this treatment is shown in Fig. 3 for a blood As an additional tool, the “Molecular Formula Generator” sample after protein precipitation with acetonitrile. In this can be applied. This calculates the molecular formula case, 6,256 such “Compounds” were found in the chro- from the accurate mass and the isotope peak pattern of a matogram. In Fig. 3b, c, and d the mass spectrum of a peak peak not in the database. This is useful for intense peaks at 1.850 min (identified as paracetamol), and the without a database result or in order to cross-control a corresponding CID fragment spectra are shown. database results for the possibility of alternative molecular formulas with a better match to the experimental Search in database and library data. The “Molecular Formula Generator” also includes accurate mass fragments in the calculation if an MS–MS The software tool “Identify Compounds” performs a spectrum was measured for the unknown peak, which is database and library search for all “Compounds” by usually the case for sufficiently intense peaks in an auto-

Cpd 6256: 19.820: +ESI ECC Scan Frag=150.0V 944_Poroshell_ACN_5µl_2.d

Counts x 107 1.2 (a)

1.0

0.8 1.850 min 0.6 Paracetamol

0.4

0.2

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Time, min

Counts x105 Counts x105 6 152.0691 6 (M+H)+[M+H]+ 110.0605 (d) (b) (c) 1 (d) 5 152.0691 5 [M+H]++ 0.8 4 4 152.0711 0.6 3 3 0.4 2 2 174.0517 [M+Na]+ 303.1305 190.0257 + 325.1158 153.0728153.0728 0.2 65.0391 93.0346 1 [2M+H] 1 + [M+K]+ [2M+Na]+ [M+H](M+H)+154.0719154.0719 [M+H](M+H)++ 82.0653 134.0606 0 0 0 140 180 220 260 300 340 150 151 152 153 154 155 156 157 30 50 70 90 110 130 150 170 (m/z) (m/z) (m/z)

Fig. 3 Poisoning case 994/09, with protein precipitation of the venous Single-stage mass spectrum of a peak at 1.850 min (paracetamol). This blood sample. (a) Chromatogram measured in auto-MS–MS mode “Compound” consists of [M+H]+, [M+Na]+, and [M+K]+. (c) Zoom after background removal and application of the tool “Find Com- of (b) with the isotope peaks of [M+H]+. (d) The CID mass spectrum pounds”. As a result, in this case 6256 “Compounds” were found. (b) recorded at 1.850 min 110 S. Broecker et al.

MS–MS file. This essentially increases the accuracy of Identification of metabolites the resulting molecular formula. If the “Compound” also contains an MS–MS spectrum, a The chromatograms obtained from urine samples but also search in the accurate mass spectra library can automati- from blood samples of poisoning cases are frequently cally be performed. The number of matching peaks, the dominated by metabolite peaks. Unfortunately, metabo- number of non-matching peaks, agreement of accurate lites were rarely available as reference substances for the fragment masses, and the fragment abundance ratio are library. In order to fill this gap, a metabolite tool used as criteria for identification of the spectra. Also in this “Identify Metabolites” was developed in order to search case, a weighted score from these criteria is used. for possible metabolites if the unknown peak could Irrespective of the collision energy applied to the unknown belong to a parent substance or vice versa to search for a compound, all library spectra with 10, 20, and 40 eV possible parent substance if the peak originates from a collision energy are included in the search procedure. metabolite. This tool calculates automatically the molec- Although the collision energy applied to the unknown ular formulas of relevant metabolites (e.g. hydroxylation molecule after calculation by use of Eq. 1 is usually =+O,demethylation=− CH2) as a basis for calculation different from 10, 20, or 40 eV, the library spectrum with of the accurate molecular mass and isotope peak pattern the next higher or next lower energy was almost always hit and subsequent extraction of the corresponding “Com- 1 of the library search. Moreover, hit 2 was usually the pounds” from the analysis file by the MassHunter other neighboring spectrum of the same substance. This is software. The tool searches for products of all essential because the match of the accurate fragment masses is a phase I and phase II metabolism reactions and combina- stronger indication of identity than the abundance ratio. tions between them. In looking for a parent substance, the By use of this search procedure 29 peaks in the opposite reactions are calculated. An example is shown in chromatogram of Fig. 3 were identified by use of the Fig. 5 foranintoxicationwithtramadol,whichwas database and CID spectra and for further 266 peaks identified in the chromatogram by library search. Alto- (score >75%, accuracy <5 ppm) only one or more proposals gether 19 metabolites were found to result from demethy- from the database without a fitting spectrum from the library lation, hydroxylation, N-oxide formation, and were obtained. Figure 4 shows the result of the search glucuronidation. Unambiguous structural identification procedure for the peak at 1.850 min in the chromatogram of was possible only for both demethyltramadols, because Fig. 3 which was identified as paracetamol. O-demethyltramadol is in the library. It is not possible to The time for post-run analysis of one chromatogram, distinguish between the two or more isomers which were including automatic molecular feature extraction (MFE), formed or are possible for each metabolism route without database and library identification, manual exclusion of false additional investigations. It must be remarked that the positives, and generation of report, is altogether 10–30 min, metabolite tool can only find but not really identify depending on the complexity of the sample. The report contains metabolites, because CID spectra or retention times of a list of identified substances and for each of them a print-out of reference substances are not available. Nevertheless, this the chromatographic peak with MS and an MS–MS spectrum example shows that the metabolite tool is very useful in and scores of identification and visual comparison between support of identification of the drug and to identify intense sample and database and library spectra. peaks in the chromatogram.

Abundance , % +ESI Product Ion (1.850 min) CID 13.1 V 100 110.0609 Sample Spectrum 50 152.0714 65.0401 77.0393 82.066893.0338 125.0043 134.0603 0 Paracetamol) C8H9NO2 100 110.0600110.0609 Hit 1, Match score 88 % 152.0706 CID 10.0 V 152.0714 50 65.0401 77.0393 82.0668 93.0338 125.0043 134.0603 65.0386 93.0335 134.0600 0 100 Paracetamol) C8H9NO2 Hit 2, Match score 81 % CID 20.0 V 110.0600 50 93.0335 65.0386 0 82.0651 134.0600 152.0706 64 68 72 76 80 84 88 92 96 100 104 108 112 116 120 124 128 132 136 140 144 148 152 (m/z)

Fig. 4 Poisoning case 994/09, with protein precipitation of the venous blood sample. Result of the database and library search for the peak at 1.850 min in Fig. 3 Development and practical application of a library of CID accurat 111

+ESI EIC(20.0000-30.0000) Scan Frag=150.0V 944_Poroshell_ACN_5µl_2.d ***ZERO ABUNDANCE*** Counts x 107 1.0 18 Tramadol O H3C N OH 0.8 17 H3C

0.6 OH 4 15,16 19 0.4 3 14 9 13 O 7 CH3 0.2 5+6 10 1 2 8 11 12 0.0 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8 5.2 5.6 6.0 6.4 6.8 7.2 7.6 8.0 8.4 8.8 Time, min

Fig. 5 Poisoning case 994/09, with protein precipitation of the venous by library); 19=N-demethyl-T, 11=N,O-didemethyl-T, distinguished blood sample. Application of the tool “Find Metabolites” to the peak at from 16 by retention time; 16=N,N-didemethyl-T; 15=tridemethyl-T, 7.089 min, which was identified as tramadol (T) by library search. 13,18=hydroxy-T and T-N-oxide; 1,3,10,14=isomeric hydroxydemethyl- Nineteen metabolites were suggested, which result from demethylation, T; 2,12=isomeric hydroxydidemethyl-T; 4=glucuronide of demethyl-T; didemethylation, tridemethylation, hydroxylation, or formation of the N- 5,7=glucuronides of di-demethyl-T; 6=glucuronide of hydroxydemethyl- oxide, combinations of these, and combination with glucuronidation. T; 8=glucuronide of hydroxydidemethyl-T; 9=glucuronide of hydroxy-T Peak identification by accurate mass: 17=O-demethyl-T (identified also

Accurate CID mass spectra of metabolites from poison- substance is detected in the MS file by the database and ing cases are collected in a separate library and will be also identified from the fragment spectrum measured in added to the database and library after accuracy control and the auto-MS–MS mode. This method is indicated by “I” in correction as described above. Table 2 and was possible at 2 ng mL−1 for 26 of the 33 compounds extracted from blood samples (c.f. Fig. 6). At Analysis of spiked blood and urine samples still lower concentrations often no MS–MS spectrum is recorded for the compound, because more matrix masses The application of LC–QTOF-MS in combination with the are preferentially selected for MS–MS measurement by database and library of toxic compounds was examined thesoftwareorbecausetheMSsignalofthecompoundis with regard to its advantages and restrictions for qualitative below the corresponding lower limit (1000 counts). substance identification in blood and urine samples. For Nevertheless, the substance is still detected during the this purpose drug-free samples were spiked with 33 drugs MS phase of the measurement cycle and is proposed as a at seven concentrations between 0.5 and 500 ng mL−1 result of the accurate mass database search. This is (Table 2), analyzed according to the procedures described indicated by “D” in Table 2, and confirmation of the in the “Experimental” section, and submitted to a search in substance was achieved by repetition of the analysis in the database and library. All test substances were present in targeted MS–MS mode and library search of the obtained the library and had the structural prerequisites for efficient fragment spectrum. Because of the 5 to 10 times higher ionization by ESI. sensitivity in the targeted MS–MS analysis, the spectra No signals were obtained from analysis of the reagent were always suitable for identification when the substance blank of the sample preparations. Different blood and was detectable in the MS data of the auto-MS–MS mode. urine samples from five volunteers who did not take any This way of identification was possible for six compounds drugs with the exception of drinking coffee or tea were in Fig. 6. At still lower concentrations the substance was analyzed. Besides caffeine and its metabolites, piperine also not proposed as a hit from the database search although (ingredient of pepper), benzododecinium (disinfectant, the peak could still be seen in the extracted chromatogram of stabilizer), diethyl phthalate (plasticizer), carnitine (amino [M+H]+. This is the case for Δ9-tetrahydrocannabinol which acid), adenosine, and thiamine (vitamin B1) were identified. is therefore not detected in Fig. 6. The chromatogram obtained from a blood sample From the auto-MS–MS results of Table 2 follows that spiked at a concentration of 2 ng mL−1 after extraction typical basic drugs such as amitriptyline, cocaine, or with CH2Cl2 isshowninFig.6.Table2 summarizes the strychnine were identified (I) in the CH2Cl2 extracts from results from all spiked samples. There were two different blood down to 0.5 ng mL−1. The six benzodiazepines were methods of identification. At higher concentrations the identified with lower sensitivity between 0.5 ng mL−1 112 S. Broecker et al.

Table 2 Identification of drugs in spiked blood and urine samples by use of LC–QTOF-MS in auto-MS–MS mode

Substance Sample prep.a Concentration (ng mL−1)

0.5 2 5 20 50 200 500

Alprazolam B-Extr. Ib III I I I B-Prec. Db III I I I U-Dil. ––DI I I I Amitriptyline B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––II I I I Carbamazepine B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. D I I I I I I Citalopram B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––II I I I Clonazepam B-Extr. – DDI I I I B-Prec. ––DI I I I U-Dil. ––––DI I Clozapine B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. – DI I I I I Cocaine B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. ––II I I I Codeine B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. – III I I I Diazepam B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. –––III I Flunitrazepam B-Extr. D D I I I I I B-Prec. – DDI I I I U-Dil. –––III I Hydrocodone B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––DI I I I Ketamine B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––DI I I I Lorazepam B-Extr. D D D I I I I B-Prec. – DDD I I I U-Dil. –––DDI I Methadone B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. – III I I I Methamphetamine B-Extr. I I I I I I I B-Prec. ––DI I I I U-Dil. ––DI I I I 3,4-Methylendioxyamphetamine B-Extr. – III I I I B-Prec. – DI I I I I U-Dil. D D D D D I I 3,4-Methylenedioxyethamphetamine B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––DI I I I 3,4-Methylenedioxymethamphetamine B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. D D D I I I I Development and practical application of a library of CID accurat 113

Table 2 (continued)

Substance Sample prep.a Concentration (ngmL−1)

0.5 2 5 20 50 200 500

Metoprolol B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. – DI I I I I Nitrazepam B-Extr. – DI I I I I B-Prec. ––DI I I I U-Dil. – DDD D I I Oxazepam B-Extr. D D I I I I I B-Prec. – DDI I I I U-Dil. ––DD D I I Oxycodone B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––DI I I I Pethidine B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. – III I I I Phencyclidine B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. ––II I I I Phentermine B-Extr. – DI I I I I B-Prec. –––DDI I U-Dil. –––DI I I Proadifen B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. – III I I I Strychnine B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. ––DI I I I Temazepam B-Extr. D I I I I I I B-Prec. D D I I I I I U-Dil. ––DD I I I Δ9-Tetrahydrocannabinol B-Extr. –––DDI I B-Prec. ––––II I U-Dil. –––––II Tramadol B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. I I I I I I I Trazodone B-Extr. I I I I I I I B-Prec. D I I I I I I U-Dil. – DI I I I I Verapamil B-Extr. D I I I I I I B-Prec. I I I I I I I U-Dil. D D I I I I I Zolpidem B-Extr. I I I I I I I B-Prec. I I I I I I I U-Dil. – III I I I a Sample preparation: B-Extr. = liquid–liquid extraction of blood with dichloromethane; B-Prec. = protein precipitation of blood with acetonitrile; U-Dil. = 1:5 dilution of urine with water b Workflow of identification: D=database proposal from MS data, no MS–mass spectrum in auto-MS–MS file, confirmed by targeted MS–MS and library in second run; I=Identification by MS–mass spectrum from auto-MS–MS data 114 S. Broecker et al.

1.2 Counts x 105 21 22 26

1.0 24 10 12 13 x 3 20 0.8 1718 31 8 5 x 3 0.6 29 x 3 27 11 15 16 25 0.4 3 19 30 4 6 9 14 2 7 23 32 0.2 28 1 0.0 3 4 567891011121314 15 16 17 18 Time, min

Fig. 6 Chromatogram obtained from a serum sample spiked with 23=alprazolam; 24=zolpidem; 25=temazepam; 26=methadone; 27=di- 2ngmL−1: 1=3,4-methylenedioxyamphetamine; 2=methamphet- azepam; 28=verapamil; 29=trazodone; 30=amitriptyline; 31=cloza- amine; 3=methylenedioxymethamphetamine; 4=3,4-methylenedioxy- pine; 32=proadifen; 33=Δ9-tetrahydrocannabinol (not detected). ethamphetamine (MDEA); 5=phentermine; 6=oxycodone; 7=codeine; Substances 5, 16, 17, 20, 21, and 22 shown also in threefold 8=metoprolol; 9=tramadol; 10=hydrocodone; 11=strychnine; 12=co- magnification in the upper part of the figure were only detected by caine; 13=meperidine (pethidine); 14=phencyclidine (PCP); 15=cit- accurate mass and isotope pattern. All other substances were identified alopram; 16=carbamazepine; 17=nitrazepam; 18=clonazepam; by library search of the CID fragment spectrum 19=ketamine; 20=flunitrazepam; 21=oxazepam; 22=lorazepam;

(diazepam) and 20 ng mL−1 (clonazepam, lorazepam) and Case 994/09 were found by the database at 0.5 or 2 ng mL−1.Itwas found that for oxazepam and lorazepam the main reason The 54 year old woman died at home after a strong for the lower efficiency of CID spectra generation is the coughing fit. Resuscitation attempts were without success. preferred formation of sodium clusters instead of [M+H]+. She had suffered from unilateral paralysis after a stroke These lower limits were one or two concentration levels and several vertebral disc prolapses and had been a chain higher for protein precipitation and two or three levels smoker. She had been taking clarithromycin A for four higher for diluted urine. Database detection was generally days because of the beginning of pneumonia. The one or two levels more sensitive than identification by following drugs had been prescribed or were found in library spectrum. Δ9-Tetrahydrocannabinol was only the apartment: acetylsalicylic acid, baclofen, beclometa- −1 detected in the CH2Cl2 extract at 20 ng mL and delivered son dipropionate, bisacodyl, carbamazepine, clemastine, a CID spectrum only at 200 ng mL−1. codeine, estriol, fenoterol, formoterol, ipratropium, lactu- It is obvious that lower limits of identification as shown lose, levomepromazine, loperamide, metoprolol, mirtaza- in Table 2 depend on sample preparation (extraction yield pin, propiverine, ranitidin, tilidine, and tizanidine. and purity of the extract) and on the amount of sample Toxicological analysis was requested for exclusion of injected. In these investigations very simple preparation poisoning. methods were applied and only drugs from 12.5 μL(CH2Cl2 By HPLC–DAD and GC–MS investigation of venous extract) or 5 μL blood (protein precipitation), or from 1 μL blood the following result was obtained: codeine urine were measured. This different portion of the sample is (1.1 μgmL−1), tramadol (8.4 μgmL−1), carbamazepine the main reason for the differences between the three sample (9.6 μgmL−1), carbamazepine-10,11-epoxide (0.8 μgmL−1), −1 preparations, besides the higher purity of the CH2Cl2 extract. ranitidine (0.7 μgmL ), and levomepromazine (0.3 μgmL−1). It was concluded that the severe overdose Application to real samples of codeine and tramadol had at least contributed to the cause of death. Venous blood samples from 50 death cases with known The chromatogram obtained by LC–QTOF-MS from the illegal drug abuse or therapeutic drug intake which had venous blood sample after protein precipitation, after previously been investigated by HPLC–DAD, GC–MS, and application of the tools “Find Compounds” and “Identify immunoassay were re-analyzed with LC–QTOF-MS in Compounds” is shown in Fig. 7. The chromatograms and combination with the accurate mass database and CID mass spectra shown in Figs. 2, 3, 4 and 5 were also spectra library according to the methods described above. obtained from the same analysis file. The previously One case is described in detail. detected compounds were identified by this method with Development and practical application of a library of CID accurat 115

+ESI EIC(20.0000-30.0000) Scan Frag=150.0V 944_Poroshell_ACN_5µl_2.d ***ZERO ABUNDANCE*** 2 2 1.5 Counts x 106 13 7 9 11 12 16 18

1.2 1 4 X 21 22 24 0.9

0.6 14 3 8 0.3 6 10 20 5 19 23 15 17 2 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Time, min

Fig. 7 Poisoning case 994/09, with protein precipitation of the venous zepine-9,10-epoxide; 11=nortilidine; 12=carbamazepine; blood sample. Substances identified by library search. 1=paracetamol; 13=hydrocortisone; 14=methiomeprazine; 15=paroxetine; 16=clo- 2=baclofen; 3=morphine; 4=cotinine; 5=tizanidine; 6=codeine glucu- mipramine; 17=tilidine; 18=midazolam; 19=loperamide; 20=fentanyl: ronide; 7=O-demethyltramadol; 8=codeine; 9=tramadol; 10=carbama- 21=clemastine; 22=clarithromycin; 23=clemastine; 24=propiverine

the exception of ranitidine which could not be confirmed. cotinine. This was the death case in this study with the In addition, baclofen, clarithromycin, clemastine, clomipr- highest number of identified substances. amine, fentanyl, loperamide, midazolam, methiomeprazine, In the same way all 50 cases were investigated. An morphine, nortilidine, paracetamol, paroxetine, propiverine, overview of the results is given in Table 3. Between 1 and tilidine, and tizanidine were also seen. Furthermore, in the 10 (mean 3.5, median 4) substances per sample, and a total

CH2Cl2 extract lidocaine, nicotine, and ofloxacine were of 94 different substances had previously been identified identified. These results were mainly in agreement with the and semi-quantitatively determined by general screening for drugs known from the case history. Smoking habits of the unknown substances using HPLC–DAD [2] and by GC– diseased were confirmed by detection of nicotine and MS analysis in selected-ion-monitoring mode for illegal

Table 3 Application of LC–QTOF-MS in auto-MS–MS mode to venous blood samples from 50 death cases and comparison with the results described in previous case reports

LC–QTOF-MS Case report

Different substances 125 94 Total positive results 311 178 Frequency of detection per substance: range 1–15 1–10 mean 2.4 1.9 median 1 1 Substances per case Total range 1–24 1–10 mean 5.8 3.5 median 6 4 Only found by LC–QTOF-MS range 0–14 – mean 3.2 – median 2 – Not confirmed by LC–QTOF-MS range 0–2 – mean 0.3 – median 0 – Substances not detected by LC–QTOF-MS Diphenylmethoxyacetic acid, furosemide, ibuprofen, pentobarbital, phenobarbital, salicylic acid, thiopental 116 S. Broecker et al. drugs and substances suspected from case histories. Both which suitable MS–MS spectra for library search were −1 the basic CH2Cl2 extract and the supernatant from protein obtained were 0.5 to 20 ng mL , depending on structure precipitation of the blood sample were analyzed by LC– and matrix. Detection by database search of the MS data QTOF-MS. The ionization mode was restricted to positive was found to be approximately one order of magnitude ESI. All substances included in the evaluation were more sensitive but in this case a second run in targeted MS– identified by library search of CID fragment spectra which MS mode is necessary to obtain CID fragment spectra for were exclusively measured in auto-MS–MS mode. Only unambiguous identification and for distinguishing between metabolites present in the library were included and no structural isomers. specific search for metabolites with the metabolite tool as Because of the comprehensive data acquisition and the described above was performed. higher sensitivity, clearly more substances were identified Under these conditions, a total of 125 different sub- in blood samples from death cases than in previous routine stances were identified by LC–QTOF-MS, 31 more than investigations by the authors using HPLC–DAD and GC– known from the toxicological reports. The number of MS. This makes LC–QTOF-MS to one of the most efficient positive results increased from 178 to 311, because many methods in systematic toxicological analysis. drugs were also detected more frequently. The number of Further work is in progress to include substances with low identified substances per case was between 0 and 14 (mean ESI yield by extension of the library to atmospheric pressure 3.2, median 4) higher than described in the toxicological chemical ionization (APCI), to optimize the choice of the CID reports. The reasons for this are the much higher sensitivity energy during auto-MS–MS measurement, and to perform in comparison with HPLC–DAD, less disturbing effect of automatic search for metabolites in the post run data- overlapping chromatographic peaks, and the inclusion of processing algorithm. Furthermore, a way of approximate substances without UV absorption. estimation of the concentrations from the peak areas and a On the other hand, some substances were also not detected system of internal standards will be developed to assess the by LC–QTOF-MS, for example furosemide, ibuprofen, toxicological relevance of the identified peaks from which barbiturates or salicylic acid. These substances are preferen- many may be in the therapeutic or sub-therapeutic range. tially ionized in negative-ESI mode which was not performed in this study. In other cases, the method was less sensitive than Acknowledgement The authors thank Agilent Technologies, Inc. the GC–MS analysis. For instance, Δ9-tetrahydrocannabinol (Santa Clara, California, USA) for generous technical support of these investigations and in particular Dr Peter Stone and Dr Frank (THC) could not be detected in blood samples previously Kuhlmann for helpful cooperation and discussions. They are also tested positive by GC–MS. For cannabinoids and for grateful to Mrs Jana Küchler, Mr Bruno Feyerabend, Mrs Karolin confirmation of low-dose benzodiazepines an additional Heinze, Mr Henrik Petszulat, Mrs Franziska Wiechert, and Mrs Sina targeted MS–MS run must be recommended. Wuttig for their thorough and efficient experimental assistance.

Conclusions References

It has been shown in this study that the advantages of TOF 1. Pfleger K, Maurer HH, Weber A (2007) Mass spectral and GC data mass spectrometry for systematic toxicological analysis on drugs, poisons, pesticides, pollutants and their metabolites, 3rd described previously [19–28, 31–35] can be exploited to a revised and enlarged edition. Wiley, Weinheim much greater extent by using a hybrid quadrupole TOF 2. Pragst F, Herzler M, Erxleben B-T (2004) Clin Chem Lab Med 42:1325–1340 instrument in combination with a large database and library 3. Maurer HH (2005) Clin Biochem 38:310–318 of toxic compounds which contains CID accurate mass 4. Maurer HH (2007) Anal Bioanal Chem 388:1315–1325 spectra in addition to theoretical accurate mass data of a 5. Wood M, Laloup M, Samyn N, Mar Ramirez FM, De Bruijn EA, – wide variety of drugs and poisons. Relatively simple Maes RAA, De Boeck G (2006) J Chromatogr A 1130:3 15 6. Peters FT (2010) Clin Biochem Aug 13. [Epub ahead of print] sample-preparation techniques from blood and urine and 7. Marquet P (2002) Ther Drug Monit 24:125–133 LC–QTOF-MS measurement in data-dependent MS–MS 8. Vernisse N, Marquet P, Duchoslav E, Dupuy JL, Lachatre G acquisition (auto-MS–MS mode) provide accurate MS and (2003) J Anal Toxicol 27:7–14 MS–MS data in a single chromatographic run for substance 9. Marquet P, Saint-Marcoux F, Gamble TN, Leblanc JC (2003) J Chromatogr B Anal Technol Biomed Life Sci 789:9–18 identification by use of efficient software tools. The 10. Jansen R, Lachatre G, Marquet P (2005) Clin Biochem 38:362–372 accurate-mass CID fragment spectra proved to be very 11. Sauvage FL, Saint-Marcoux F, Duretz B, Deporte D, Lachatre G, specific and enabled peak identification with high accuracy. Marquet P (2006) Clin Chem 52:1735–1742 The library was successfully applied to analysis files 12. Sauvage FL, Picard N, Saint-Marcoux F, Gaulier JM, Lachatre G, Marquet P (2009) J Sep Sci 32:3074–3083 measured by other QTOF-MS instrument series of the 13. Weinmann W, Wiedemann A, Eppinger B, Renz M, Svoboda M same manufacturer. The lower limits of identification at (1999) J Am Soc Mass Spectrom 10:1028–1037 Development and practical application of a library of CID accurat 117

14. Dresen S, Kempf J, Weinmann W (2006) Forensic Sci Int 161:86–91 31. Polettini A, Gottardo R, Pascali JP, Tagliaro F (2008) Anal Chem 15. Mueller CA, Weinmann W, Dresen S, Schreiber A, Gergov M 80:3050–3057 (2005) Rapid Commun Mass Spectrom 19:1332–1338 32. Liotta E, Gottardo R, Bertaso A, Polettini A (2010) J Mass 16. Dresen S, Gergov M, Politi L, Halter C, Weinmann W (2009) Spectrom 45:261–271 Anal Bioanal Chem 395:2521–2526 33. Lee HK, Ho CS, Iu YP, Lai PS, Shek CC, Lo YC, Klinke HB, 17. Dresen S, Ferreiros N, Gnann H, Zimmermann R, Weinmann W Wood M (2009) Anal Chim Acta 649:80–90 (2010) Anal Bioanal Chem 396:2425–2434 34. Stroh JG, Petucci CJ, Brecker SJ, Nogle LM (2008) J Sep Sci 18. Gergov M, Ojanpera I, Vuori E (2003) J Chromatogr B 795:41–53 31:3698–3703 19. Gergov M, Boucher B, Ojanperä I, Vuori E (2001) Rapid 35. Kaufmann A, Butcher P, Maden K, Widmer M (2007) Anal Chim Commun Mass Spectrom 15:521–526 Acta 586:13–21 20. Pelander A, Ojanperä I, Laks S, Rasanen I, Vuori E (2003) Anal 36. Nielsen MK, Johansen SS, Dalsgaard PW, Linnet K (2010) Chem 75:5710–5718 Forensic Sci Int 196:85–92 21. Ojanperä I, Pelander A, Laks S, Gergov M, Vuori E, Witt M 37. http://molgen.de (molecular structure generation), assessed 20 (2005) J Anal Toxicol 29:34–40 August 2010 22. Ojanperä S, Pelander A, Pelzing M, Krebs I, Vouri E, Ojanperä I 38. www.chemspider.com, assessed 20 August 2010 (2006) Rapid Commun Mass Spectrom 20:1161–1167 39. Pragst F, Herzler M, Herre S, Erxleben B-T, Rothe M (2001) 23. Kolmonen M, Leinonen A, Pelander A, Ojanperä I (2007) Anal UV-Spectra of toxic compounds. Database of photodiode array Chim Acta 585:94–102 UV spectra of illegal and therapeutic drugs, pesticides, ecotoxic 24. Pelander A, Ristimaa J, Rasanen I, Vuori E, Ojanperä I (2008) substances and other poisons. Verlag Dieter Helm, Heppen- Ther Drug Monit 30:717–724 heim; Suppl Vol (2008) Edition Toxicological Chemistry, 25. Pelander A, Ristimaa J, Ojanperä I (2010) J Anal Toxicol 34:312–318 Berlin 26. Tyrkkö E, Pelander A, Ojanperä I (2010) Drug Test Anal 2:259–270 40. Stone P, Zweigenbaum J (2009) An application kit for the 27. Ristimaa J, Gergov M, Pelander A, Halmesmäki E, Ojanperä I screening of samples for analytes of forensic and toxicological (2010) Anal Bioanal Chem 398:925–935 interest using TOF or Q-TOF LC/MS with a personal forensics/ 28. Pavlic M, Libiseller K, Oberacher H (2006) Anal Bioanal Chem toxicology database. Agilent application note 5990-4252EN, 386:69–82 Agilent Technologies, Inc 29. Decaestecker TN, Clauwaert KM, Van Bocxlaer JF, Lambert WE, 41. Broecker S, Herre S, Pragst F, Kuhlman F, Wüst B, Van den Eeckhout EG, Van Peteghem CH, De Leenheer AP Zweigenbaum J (2010) Toxicological screening with the (2000) Rapid Commun Mass Spectrom 14:1787–1792 agilent LC/MS-QTOF and the personal compound database 30. Decaestecker TN, Vande Casteele SR, Wallemacq PE, Van and the Broecker, Herre and Pragst accurate mass spectral Peteghem CH, Defore DL, Van Bocxlaer JF (2004) Anal Chem library. Agilent Application Note 5990-6419EN Agilent Tech- 76:6365–6373 nologies, Inc