ORIGINAL PAPER

Development and practical application of a library of CIDaccurate mass spectra of more than 2,500 toxic compoundsfor systematic toxicological analysis by LC–QTOF-MSwith 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)accurate mass spectra has been developed for efficient useof liquid chromatography in combination with hybridquadrupole time-of-flight mass spectrometry (LC–QTOF-MS) as a tool in systematic toxicological analysis. Themass spectra (Δm<3 ppm) of more than 2,500 illegal andtherapeutic drugs, pesticides, alkaloids, other toxic chem-icals and metabolites were measured, by use of an Agilent6530 instrument, by flow-injection of 1 ng of the puresubstances in aqueous ammonium formate–formic acid–methanol, with positive and negative electrospray-ionization (ESI), selection of the protonated or deproto-nated molecules [M+H]+ or [M−H]− by the quadrupole, andcollision induced dissociation (CID) with nitrogen ascollision gas at CID energies of 10, 20, and 40 eV. Thefragment mass spectra were controlled for structuralplausibility, corrected by recalculation to the theoreticalfragment masses and added to a database of accurate mass

data and molecular formulas of more than 7,500 toxicolog-ically relevant substances to form the “database and libraryof toxic compounds”. For practical evaluation, blood andurine samples were spiked with a mixture of 33 drugs atseven concentrations between 0.5 and 500 ng mL−1, pre-pared by dichloromethane extraction or protein precipita-tion, and analyzed by LC–QTOF-MS in data-dependentacquisition mode. Unambiguous identification by librarysearch was possible for typical basic drugs down to 0.5–2 ng mL−1 and for benzodiazepines down to 2–20 ng mL−1.The efficiency of the method was also demonstrated by re-analysis of venous blood samples from 50 death cases andcomparison with previous results. In conclusion, LC–QTOF-MS in data-dependent acquisition mode combinedwith an accurate mass database and CID spectra libraryseemed to be one of the most efficient tools for systematictoxicological analysis.

Keywords Accurate mass spectra library . Collision-induceddissociation . Liquid chromatography. Time of flight massspectrometry . Peak identification . Systematic toxicologicalanalysis

Introduction

Systematic toxicological analysis is the general search fortoxic compounds in a biological sample, for instance inhuman blood, urine, organ tissues, or hair, without anyinformation about presence and kind of poisons. It is one ofthe most difficult tasks of analytical chemistry because ofthe huge number of possible poisons and poison metabo-lites which may occur in low and very low concentrationsin the complicated matrix. It includes toxic gases, volatile

Published in the special issue Forensic Toxicology with Guest EditorsFrank T. Peters, Hans H. Maurer, and Frank Musshoff.

S. Broecker : S. Herre : F. Pragst (*)Institute of Legal Medicine, University Hospital Charité,Turmstraße 21, Building N,10559, Berlin, Germanye-mail: fritz.pragst@charite.de

B. WüstAgilent Technologies,Hewlett-Packard-Straße 8,76337, Waldbronn, Germany

J. ZweigenbaumAgilent Technologies, Inc.,2850 Centerville Road, BL3-2 3L11,Wilmington, DE 19808-1610, USA

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

substances, metal ions and, as the largest group, organiccompounds with low volatility such as illegal and thera-peutic drugs, pesticides, chemical reagents, and alkaloids.Up-to-date methods for systematic toxicological analysis oforganic compounds consist of suitable sample preparationwhich is able to extract as many poisons as possible fromthe matrix, and a combination of chromatography andmolecular spectrometry in order to separate the extractedmixture and to characterize the components. Widely usedmethod combinations are capillary gas chromatography–mass spectrometry (GC–MS) [1] and high-performanceliquid chromatography with photodiode array detection(HPLC–DAD) [2]. There is not yet any possibility ofdetermining the exact structure from these spectra. Therefore,substance identification is always based on comparison of theunknown spectrum with those in a library of spectra oftoxicologically relevant compounds.

In the last decade, several approaches have been made touse liquid chromatography in combination with massspectrometry (LC–MS or LC–MS–MS) with electrosprayionization (ESI) or atmospheric pressure chemical ioniza-tion (APCI) for systematic toxicological analysis [3–36].The application of single-stage quadrupole or ion-trap massspectrometers for this purpose is limited because ofdisturbances by high matrix burden and co-eluting peaks.By use of triple-quadrupole mass spectrometers or, morefavorably, hybrid triple-quadrupole linear ion-trap massspectrometers these problems have been solved but thesearch can only be performed as multi-targeted screening[7–18]. This means that substances can be detected only ifthey are a priory included in the method. The number ofsubstances in such a procedure is limited by the minimumdwell time for each multi reaction monitoring (MRM)transition included in one measurement cycle. Nevertheless,a powerful screening procedure with 700 substances in onechromatographic run has been developed by Dresen et al.with a hybrid triple-quadrupole linear ion-trap massspectrometer by limiting the detection time of every analyteto a chromatographic time window of 2 min, information-dependent acquisition (IDA) using the sensitive enhancedproduct ion scan of the instrument, and uniting the fragmentions from three collision energies in the trap to obtain onemixed mass spectrum (collision energy spread) [17].

Generally, LC screenings with MS–MS identificationconsist of a survey scan to detect the analytes and adependent scan for measurement of the corresponding MS–MS spectra which are submitted to library search foridentification. Survey scan and dependent scan can beaccomplished within the same analytical run by automaticselection of precursor ions and measurement of the MS–MSspectra immediately after their detection in the survey scan.In “data-dependent acquisition” this is determined only bythe actual MS data whereas “information-dependent acqui-

sition” is restricted to a list of preselected precursorsincluded in the method.

Mass spectra libraries for LC–MS with fragmentation byin-source collision and LC–MS–MS with fragmentation inthe collision cell between both MS units have beendescribed in several papers and contain between 301 and1,253 substances [13–16].

The availability of time-of-flight mass spectrometerswith much improved mass resolution and mass accuracy asdetectors in liquid chromatography (LC–TOF-MS) provid-ed new possibilities in the use of LC–MS for toxicologicalscreening in blood, urine, hair, meconium, and vitreoushumor [19–36]. The working principle of these instrumentsenables comprehensive recording of all data. Therefore,there is no a priori limitation or prediction of the substancesincluded in the search procedure. The increased massresolution also provides high selectivity for overlappingpeaks and high matrix burden. The most importantadvantage is that the molecular formula of an analyte isdirectly available from the accurate molecular mass and theisotope peak pattern. For substance identification by use ofthe molecular formula, theoretical databases of toxicolog-ically relevant compounds with up to 50,500 substancesincluding metabolites [29] or in-house databases with 100to 869 substances have been created [19–22, 24–28, 30,33–35].

Different search procedures are used in TOF methodology.The group of Ojanperä and Pelander used reversed targetsearch, which means the TOF file is searched for target massesincluded in the library [18–27]. Polettini et al. used forwardbasepeak search, which means the base mass peak of theunknown chromatographic peak in the analysis file wassearched after proton subtraction in their large database of50,500 compounds of toxicological interest and many phaseI and phase II metabolites [31, 32].

However, the molecular formula of an eluted unknownsubstance in a chromatogram is only a first step of theidentification, because of the huge number of possibleisomers, as can easily be shown by use of chemicalsoftware. For instance, according to Molgen (molecularstructure generation [37]) for the nominal molecular massM=149, 27 different molecular formulas are theoreticallypossible if only the elements C, H, N, and O are included.TOF-MS with mass accuracy <3 ppm can clearly distin-guish between these 27 possibilities. One is C9H11NO withthe accurate molecular mass 149.084060. Based on therules of chemical bonds between these atoms, the softwaretheoretically calculates 25,895,621 structural isomers(stereoisomers not included). From these, 724 substancesare recorded in the Beilstein database and only 45 areincluded in the NIST register. The software “Chemspider”[38] shows structural formulas of 829 compounds with themolecular formula C9H11NO which are all reasonable,

102 S. Broecker et al.

among them, for instance, cathinone, N,N-dimethylbenza-mide, p-dimethylaminobenzaldehyde, 3,4-dimethylbenza-mide, p-aminopropiophenone and also heterocyclicsubstances such as 6-hydroxy-1,2,3,4-tetrahydroquinolineor N-methyl-6-hydroxy-2,3-dihydroindole. It is very im-portant to be aware of this almost unlimited structuraldiversity of organic chemistry when commencing system-atic toxicological analysis.

Therefore, much more evidence is required in order todistinguish between isomers. For this purpose, retentiontimes under defined chromatographic conditions have beenmeasured for 100 to 400 substances in some in-houselibraries [20–25, 27, 30, 33, 35]. Identification of metabo-lites in a metabolomic approach is another possibility forconfirming or rejecting a proposal from the theoreticaldatabase [20, 32].

Restriction to only toxicologically relevant substances intheoretical databases, retention times under standardizedconditions, and search for metabolites are very helpful.However, much more structure-specific information shouldbe obtained from collision-induced dissociation (CID)fragment spectra. For this reason, in-source collision-induced dissociation spectra have been measured in singlestage LC–TOF-MS procedures by some authors [33, 35];monitoring of drug class-specific CID mass fragments wasperformed for detection of nontarget analytes [35], andspecial software for fragmentation prediction was used inorder to differentiate between structural isomers in a targetdrug database by LC–QTOF-MS [26].

Hybrid quadrupole time-of-flight mass spectrometry(LC–QTOF-MS) has the advantage that, in contrast within-source CID, the fragment spectra are not disturbed bymatrix and co-eluting substances. A library of CID spectraof 319 substances measured with an LC–QTOF-MSinstrument at ten collision energies was described by Pavlicet al. [28]. Furthermore, for application in systematictoxicological analysis, the LC–QTOF-MS instrument canbe operated in a data-dependent acquisition mode(auto-MS–MS mode) in order to combine the advantageof TOF-MS for comprehensive data collection with themeasurement of accurate CID fragment spectra from allessential components of the sample after isolation of thecorresponding parent ions by the quadrupole (QTOF-MS).This technique was first applied to toxicological analysis byDecaestecker et al. for toxicological analysis [29, 30]although the advantages of accurate mass were not yetreally utilized.

In our study the accurate mass CID spectra of more than2,500 toxicologically relevant substances were measuredwith a hybrid quadrupole time-of-flight mass spectrometer(QTOF-MS) at three different collision energies andincluded in a TOF-MS database with molecular formulasof more than 7,500 substances from which the accurate

masses and isotope pattern are calculated. The practicalapplication of this “Library and Database of Toxic Com-pounds” was examined with spiked and real blood andurine samples using the auto-MS–MS mode.

Experimental

Chemicals and reagents

The solvents and chemicals used for the mobile phase werepurchased as follows: methanol (LC–MS grade), acetonitrile(LC–MS grade), and ammonium acetate (HPLC grade) fromFisher scientific (Schwerte, Germany), ammonium formate(LC–MS grade) from Agilent Technologies, water (HPLCgrade) and formic acid (99+% for analysis) from AcrosOrganics (Geel, Belgium). All other solvents and reagentsused for sample preparation were obtained from Merck(Darmstadt, Germany) in analytical grade purity.

Reference substances

The reference substances (more than 2,500) included in thelibrary of spectra were generously donated by numerouspharmaceutical manufacturers or purchased from chemicalcompanies such as LGC Promochem, Sigma, Radian,Riedel–de Haën, Dr Ehrenstorfer, or Lipomed. The UVspectra of the same substances were previously measuredby HPLC–DAD for the database “UV spectra of ToxicCompounds” [39] and a complete substance list is giventhere. The identity and sufficient purity of the substanceswas always monitored by mass spectrometry. In addition,52 deuterated standards of illegal drugs, legal opiates, andbenzodiazepines were included; these were purchased fromLGC Promochem (Wesel, Germany).

Instruments and software

All measurements were performed with a 6530 accurate-mass Q-TOF LC–MS instrument (Agilent Technologies,Santa Clara, USA). The Agilent 1200 SL series HPLCconsisted of a degasser, a thermostated HiP-ALS autosam-pler, a Bin Pump SL binary pump, and a TCC SL columnoven. The QTOF-MS instrument was operated with anelectrospray ion source ESI+Agilent Jet Stream Technologyin positive and negative ionization mode, a quadrupole forisolation of precursor ions with a mass window of 1.3 or 4m/z in MS–MS mode, a linear hexapole collision cell withnitrogen as collision gas and collision energy 0–40 eV, aTOF-MS with mass accuracy <3 ppm, mass resolution of5,000 to 10,000 (100 to 922m/z), a measuring frequency of10,000 transients s−1 and a detection frequency of 2 GHz(200,000 points/transient).

Development and practical application of a library of CID accurat 103

The data presented here used MassHunter AcquisitionB.02.01 with Service Pack 3 for the Agilent TOF andQTOF and MassHunter Qualitative Analysis B.03.01 withService Pack 3. In addition, the Personal CompoundDatabase and Library Software B.03.01 was used tointerrogate the database and library directly.

Measurement of library spectra

For measurement of the CID mass spectra for the library,1 mg mL−1 stock solutions of all compounds in methanolwere prepared and diluted to 1 μg mL−1. If solubility wasinsufficient, addition of formic acid, use of water oracetonitrile, or a larger volume of solvent was tried. Thespectra were measured by flow injection of 1 ng of eachsubstance (1 μL of 1 μg mL−1 solution) with 5 mmol L−1

ammonium formate and 0.01% formic acid in water–0.01%formic acid in methanol (50:50) as mobile phase.

The protonated or deprotonated molecules [M+H]+ or[M−H]− were selected by the quadrupole with a massresolution of 1.3m/z . Three MS–MS spectra weregenerated in product-ion-scan mode at CID energies of10, 20, and 40 eV. The QTOF conditions applied were: gastemperature 250 °C, gas flow 6 Lmin−1, nebulizer pressure35 psi, sheath gas temperature 300 °C, sheath gas flow10 Lmin−1, VCap voltage 3500 V, nozzle voltage 100 V,fragmentor voltage 150 V, mass range (MS and MS–MS)50–1700m/z depending on substance, scan rate 8 Hz in MSand MS–MS experiments, reference ions for mass calibra-tion: purine 121.050873 [M+H]+ and 119.036319 [M−H]−,HP-921=hexakis(1H,1H,3H-tetrafluoropropoxy)phosphazine922.009798 [M+H]+ and 966.000725 [M+HCO2]

−.

Processing of library spectra

The measured fragment spectra were controlled for structuralplausibility and corrected by recalculation to the theoreticalmasses by special processing before arrangement in thespectra library and database, which contains at its presentstate more than 7,500 toxic compounds, which meansapproximately 5,000 entries with only theoretical accuratemass data including isotope pattern and further 2,500 entrieswith additional three accurate mass CID fragment spectra at10, 20, and 40 eV collision energy.

Blood and urine samples

Drug-free blood samples for validation of the method werecollected from volunteers among the laboratory staff.Furthermore, the method was applied to 50 blood samplesfrom autopsy cases which were investigated in the Instituteof Legal Medicine of the University Hospital Charité Berlinand for which exposure to therapeutic or illegal drugs was

known from case histories and/or toxicological investiga-tion with HPLC–DAD, GC–MS or immunoassay.

Sample preparation

Liquid–liquid extraction with dichloromethane

The procedure described previously for HPLC–DAD [2]was only slightly changed. To 500 μL whole blood, serum,or plasma in a 1.5-mL Eppendorf vial 100 μL 0.1 mol L−1

HCl (acidic extract) or 100 μL solution of Tris substance(24.3 g in 1 L H2O, pH 9.0, basic extract) and 400 μLCH2Cl2 were added. The mixture was vortex mixed for1 min and centrifuged for 5 min at 13,200 rpm. The CH2Cl2layer (200 μL) was aspirated with a 200 μL Hamilton-typesyringe and evaporated to dryness in a nitrogen stream at40 °C. The residue was dissolved in 100 μL ACN–0.1%HCOOH in water (35:65v/v) and 5 μL was injected forLC–QTOF-MS measurement.

Protein precipitation with acetonitrile

Blood, serum, or plasma (100 μL) was placed in a 1.5-mLEppendorf vial and 400 μL acetonitrile was added. Themixture was vortex mixed for 1 min and centrifuged for5 min at 13,200 rpm. Then, 400 μL supernatant wasseparated and evaporated to dryness in a nitrogen stream at40 °C. The residue was reconstituted in 80 μL ACN–0.1 %HCOOH in water (35:65v/v) and 5 μL was injected forLC–QTOF-MS measurement.

Preparation of urine samples

Urine samples were centrifuged for 5 min at 13,200 rpm.The supernatant (100 μL) in a 2-mL autosampler vial wasdiluted with 400 μL of 10 mmol L−1 ammonium acetate inwater (pH 6.8) and 5 μL was injected for LC–QTOF-MSmeasurement.

Sample measurement

Chromatographic separation was performed at 50 °C with aPoroshell 120 EC-C18, 2.1×100 mm, 2.7 μm, column. Forgradient elution the mobile phases 10 mmol L−1 NH4Ac inH2O (A) and methanol (B) were used with the timeprogram: 0 min 10% B, linear to 50% B at 8 min, linearto 100% B at 20 min, constant 100% B to 23.9 min, back to10% B at 24 min and equilibration for 3 min. The flow ratewas 0.4 mL min−1.

The QTOF-MS instrument was operated under theconditions: ion source ESI+Agilent Jet Stream Technologyin positive ionization mode, quadrupole was used as an ionguide in MS mode and for selection of precursor ions with

104 S. Broecker et al.

Δm/z=4 in MS–MS mode, collision cell without CID inMS mode and with CID of precursor ions in MS–MS modeat mass dependent ramped CID energy (offset 4 eV, slope6 eV/100m/z), TOF-MS with a mass range of 100–1000m/zin MS mode and 50–600m/z in MS–MS mode. The scanrate was 4 Hz in MS and MS–MS experiments. The sourceconditions were: gas temperature 320 °C, gas flow 8 Lmin−1, nebulizer pressure 35 psi, sheath gas temperature380 °C, sheath gas flow 11 Lmin−1, VCap voltage 3000 Vand nozzle voltage 0 V. The remaining instrument con-ditions were the same as described above for measurementof library spectra.

For systematic toxicological analysis the auto-MS–MSmode (data-dependent acquisition) was used with a cycletime of 1.1 s, 0.25 s measurement in MS mode, selectionof three precursors which were fragmented in thefollowing three MS–MS experiments, and active exclu-sion after one spectrum for 0.1 min. For this the three mostabundant masses were selected as precursors includingonly singly-charged ions with a threshold abundance of1000 counts.

Results and discussion

Collision-induced accurate mass spectra library

Because the library of spectra was designed for systematictoxicological analysis, it was important to include as manytoxicologically relevant substances as possible. Selectionwas based on the substance pool in the laboratory which theauthors used for the database “UV Spectra of ToxicCompounds” [2] but included also compounds withoutUVabsorbance. In the first step of library development onlypositive and negative electrospray ionization was applied.For this reason compounds with insufficient ionizationunder these conditions are still missing from the library andwill be added in a next step by use of atmospheric pressurechemical ionization (APCI). Table 1 shows the compositionof the library in its current state, on the basis of differentgroups of toxic compounds. Among them are 139 metab-olites (5.5%) which were also available as pure referencesubstances.

Measurement of the CID spectra was performed by flowinjection of 1 ng of each substance in 1 μL solvent inmobile phase consisting of a 1:1 mixture of 5 mmol L−1

ammonium formate+0.01% formic acid in water and 0.01%formic acid in methanol, at a flow rate of 0.1 mL min−1.Chromatographic separation was not necessary because thepure substances were measured and because possibleimpurities were excluded by the quadrupole. The mono-isotopic ions [M+H]+ or [M−H]− as theoretically calculatedfrom the structural formula were separated by the quadru-

pole with a mass window of Δm/z=1.3 and submitted toCID with nitrogen as collision gas in the collision cell atcollision energies of 0, 10, 20, and 40 eV. In the range ofthe substance peak between 20 and 30 spectra at eachcollision energy were measured, with an accumulation rateof 1600 transients per spectrum, and were merged. Thespectra before and after the peak were subtracted forbackground correction. For storage in the library, only thespectra acquired at collision energies of 10, 20, and 40 eVwere used and only masses with an intensity above 100counts were imported into the library. The spectra werenormalized to the largest peak. There was no otherlimitation of the number of fragment ions.

The ions found in each spectrum were then identified bytheir chemical formula, and the exact mass of the ions wascalculated and used to replace the measured mass whilemaintaining the relative abundance of each ion. Thisprovides an accurate mass library for routine matching offragment ions in real samples with the expected ions for theproposed compound identification.

As an example the three spectra obtained for theanticoagulant warfarin are shown in Fig. 1. Whereas the[M+H]+ (m/z=309.11214) is one of the highest peaks at10 eV it decreases strongly at 20 eV and is not seen at40 eV, in favor of an increasing number and intensity offragment ions. The two most abundant fragments at m/z=251.0702 and 163.0389 correspond to the fragmentation atthe α and β positions of the side chain, respectively (loss ofC3H6O and C10H10O). Depending on molecular mass andstructure, the degree of fragmentation at the three CIDenergies was different from compound to compound. Forinstance amphetamine and piracetam were already stronglyfragmented at 10 eV and totally disintegrated at 40 eVwhereas for morphine and strychnine the protonatedmolecule peak still dominated at 10 and 20 eV andoptimum fragmentation occurred only at 40 eV. Neverthe-less, it was seen from the spectra that sufficient fragmen-tation was achieved for the large majority of the compoundsin order to provide structure-specific information. Some ofthe substances, for example digoxin and digitoxin providedonly a very small ESI yield of [M+H]+, because they weremainly ionized as sodium and potassium clusters [M+Na]+

and [M+K]+. This must be kept in mind in samplescreening for these highly toxic drugs.

The spectra of more than 2,500 substances were added tothe “Personal Forensics/Toxicology Database” describedpreviously [40], which contains theoretically calculatedaccurate mass data and molecular formulas of more than7,500 toxicologically relevant substances, to form the“Personal Compound Database and Library of ToxicCompounds”. The arrangement of the library and databaseis not finished but is steadily being extended by addition offurther compounds.

Development and practical application of a library of CID accurat 105

No, or unsuitable, CID spectra were obtained fromapproximately 500 other substances under the experimentalconditions described, because of insufficient ESI yield forneutral compounds (e.g. halogenated pesticides, steroids,phenols), formation of sodium adducts (e.g. cardiac glyco-sides), insufficient stability in the ion source of thecompounds (organic phosphates such as E605 or organicnitrates such as erythritol tetranitrate), and formation of multicharged ions (e.g. antibiotics). Furthermore, dicationic speciessuch as pancuronium, vecuronium, or suxamethonium muststill be added to the library.

The CID accurate mass library was applied to chromato-graphic files acquired by use of different Agilent LC–QTOF-MS instruments of the series 6520, 6530, and 6540and proved to be fully suitable for peak identificationirrespective of the instrument.

Workflow of substance identification

Sample measurement in data-dependent acquisition mode(auto-MS–MS mode)

In principle, there are two workflows for use of LC–QTOF-MS in combination with the database and library of toxiccompounds for systematic toxicological analysis [41]. The“Targeted MS–MS” workflow measures the sample in afirst chromatographic run in single MS mode and performsa forward database search for possible compounds of

concern according to the algorithms described below inthe next two sections. The resulting positive list is thenvisually examined for quality of match and those com-pounds that appear as possibly present are confirmed in asecond chromatographic run under targeted MS–MS con-ditions in which the resulting MS–MS spectra are used tosearch the library for identification. This is a combinedforward and backward search with corresponding separatescores. Although this workflow has the advantage of highersensitivity it is rather complicated because of the two runsand the large number of false positives resulting from thefirst run which had to be introduced into the targeted run.

The auto-MS–MS workflow (data-dependent acquisition)is muchmore comfortable but it is less sensitive. The principleis shown in Fig. 2, with a real sample as an example. It canbe seen from the total ion chromatogram (Fig. 2a) and morein detail from the small part with higher time resolution(Fig. 2b) that the instrument is operated with steadyalternation of MS and MS–MS mode with a cycle time of1.1 s. In MS mode, for 0.25 s the full mass spectrum at thecorresponding retention time is recorded, three precursorions are selected and for each of these the mass-dependentcollision energy is chosen according to Eq. 1.

Collision energy¼ 4þ 0:06m=zð Þ eV ð1Þ

The quadrupole then selects these three masses insuccession and the CID accurate mass spectra are measured

Table 1 Composition of the CID accurate mass spectra library oftoxic compounds (August 2010). The arrangement in groups isslightly arbitrary since many substances have more than one effect

or medical use. Alkaloids and plant poisons such as aconitine, nicotineor strychnine were included into the groups according to their 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 addictionpotential, hypnotics

228 Cocaine, fentanyl, mescaline, pentobarbital, THC

2 Psychopharmaceuticals, benzodiazepines,CNS-active substances

289 Amitriptyline, carbamazepine, diazepam, fluoxetrine, perazine

3 Non-opioid analgesics, antirheumatics,antitussives, and similar

210 Ambroxol, etoricoxib, metamizole, paracetamol,

4 Further CNS-active substances,antihistaminics, antiallergics

154 Astemizole, benserazide, pyridostigmine, talastine

5 Cardiovascular agents 234 Amiodarone, digoxin, flecainide, metoprolol, propafenone

6 Diuretics, antidiabetics, anticoagulants,various drugs

174 Carbutamide, furosemide, glipizide, phenprocoumon

7 Steroids, hormones, vitamins, endogenoussubstances

211 Ascorbic acid, ethinylestradiol, flucinonide, stanazolol

8 Antibiotics, antimalarials, cytostatics,virustatics, and similar

256 Ampicillin, atazanavir, lincomycin, ofloxacin, sulfamerazin

9 Fungicides, disinfectants, adjuvants,diagnostics

178 Fluconazole, hexachlorophene, ioglicic acid, terbinafine,

10 Insecticides, acaricides, nematicides,and similar

219 Bromurone, fluazuron, parathion, propoxur, tetramethrin

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

106 S. Broecker et al.

during the residual time of the measurement cycle. Thesethree masses are excluded from MS–MS measurement for0.1 min in order to enable the acquisition of other co-eluting substances, and the instrument goes on to the nextcycle. The corresponding four mass spectra from the MSand MS–MS modes are shown in Fig. 2c to f.

Cycle time, time forMSmeasurement within the cycle, andnumber of precursor ions can be adjusted within specificlimits. Equation 1 takes into account that the collision energynecessary for sufficient fragmentation increases with increas-ing molecular mass. However, because the fragmentationalso depends strongly on the specific structure, this methodof automatic choice of the collision energy is not alwaysoptimum and will be improved in the future.

Furthermore, an abundance threshold (e.g. 1000 counts) isset for [M+H]+ or [M−H]− in order to include only essentialcomponents in measurement of CID spectra. Nevertheless,despite high concentration, the first CID spectrum of an ionmay be recorded at relatively low abundance in the leading

edge of a peak and may have only low quality. Therefore, theexclusion time chosen was shorter than the chromatographicpeak width in order enable a second acquisition of the sameion closer to the peak maximum.

Altogether, the auto-MS–MS file of the sample containsalso the CID fragment spectra in addition to the accuratemolecular masses of all essential constituents.

Peak identification

For post-run analysis of an LC–QTOF-MS file measured inauto-MS–MS mode, new efficient tools of the MassHuntersoftware are available. First, the background is removed bysubtraction of constant or slowly changing signals. Then, atool “Find Compounds” is applied which extracts theelution profile of each mass and groups masses with thesame elution profile to so-called “Compounds”. Such a“Compound” consists, for instance, of the protonatedmolecule [M+H]+, cluster ions with sodium, potassium or

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

80

100121.02841

163.03897

77.03858251.07027223.0753693.0334965.03858 205.0647951.02293

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

80

100 251.07027163.03897

147.08044

121.02841 223.07536 291.10157

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

80

100 163.03897

309.11214251.07027

147.08044291.10157

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

121.02841

163.03897

77.03858251.07027223.0753693.0334965.03858 205.0647951.02293

Library spectrum

m/z 40

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

121.02841

163.03897

77.03858251.07027223.0753693.0334965.03858 205.0647951.02293

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

80

100 251.07027163.03897

147.08044

121.02841 223.07536 291.10157

Library spectrum

m/z 40 60 80 100 120 140 160 180

60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

80

100 251.07027163.03897

147.08044

121.02841 223.07536 291.10157

Library spectrum

m/z 40 60 80 100 120 140 160 180 200 220 240 260 280 300 3200

20

40

60

80

100 163.03897

309.11214251.07027

147.08044291.10157

O O

OH O

CH3

10 eV

20 eV

40 eV80

100

80

100

Fig. 1 Accurate mass spectra of warfarin at CID energies of 10, 20, and 40 eV as stored in the library

Development and practical application of a library of CID accurat 107

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Acquisition time (min)1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Abundancex 107

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Acquisition time (min)1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Abundancex 107

0

1

2

3

4 495.1797

237.1021

473.1972259.0839

275.0590

(m/z)200 240 280 320 360 400 440 480 520

Counts x106

0

0.4

0.8

1.2

1.6110.0963

138.0910

164.1175

349.2032

244.1424 289.2126

(m/z)40 120 200 280 360

Counts x103

0

0.4

0.8

1.2

1.6110.0963

138.0910

164.1175

349.2032

244.1424 289.2126

(m/z)40 120 200 280 360

Counts x103

0

0.4

0.8

1.2

1.6

2

2.4

2.8

3.2

194.0957

237.1020

(m/z)40 60 80 100 120 140 160 180 200 220 240 260

Counts x103

0

0.4

0.8

1.2

1.6

2

2.4

2.8259.0837

(m/z)50 100 150 200 250 300 350 400 450 500 550

Counts x105

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Acquisition time (min)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

Abundance x107

MSMS

MS

MS/MS(1)MS/MS(2)

MS/MS(3)

0

0.2

0.4

0.6

0.8

Acquisition time (min)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

Abundance x107

MSMS

MS

MS/MS(1)MS/MS(2)

MS/MS(3)

(a)

(b)

(c) (d)

(f)(e)

[M+H]+1 (d)

[M+H]+2 (e)

[M+H]+3 (f)[M+H]+1

25.0 eV

[M+H]+2

18.2 eV[M+H]+3

33.7 eV

108 S. Broecker et al.

ammonium, dimers of this ion species, the correspondingisotope peaks, and the CID spectra of [M+H]+ in theretention time range of the peak. The chromatogramobtained after this treatment is shown in Fig. 3 for a bloodsample after protein precipitation with acetonitrile. In thiscase, 6,256 such “Compounds” were found in the chro-matogram. In Fig. 3b, c, and d the mass spectrum of a peakat 1.850 min (identified as paracetamol), and thecorresponding CID fragment spectra are shown.

Search in database and library

The software tool “Identify Compounds” performs adatabase and library search for all “Compounds” by

application of a system of scores and exclusion criteria.For the MS data, identification is based on comparison ofthe exact value of the compound’s neutral mono-isotopicmass with the measured mass calculated from ionic m/zvalues of all detected specified adducts within, typically, a 3to 5 ppm mass error. In addition, identification relies onboth spacing and relative abundance of the isotopesdetected. Agreement with the database entries is assessedby use of a weighted score calculated from the mass match,the abundance match, and the spacing match. As a resultno, one, or more hits can be proposed. More than one hitoccurs for isomers or isobars with very close accuratemolecular masses present in the database.

As an additional tool, the “Molecular Formula Generator”can be applied. This calculates the molecular formulafrom the accurate mass and the isotope peak pattern of apeak not in the database. This is useful for intense peakswithout a database result or in order to cross-control adatabase results for the possibility of alternative molec-ular formulas with a better match to the experimentaldata. The “Molecular Formula Generator” also includesaccurate mass fragments in the calculation if an MS–MSspectrum was measured for the unknown peak, which isusually the case for sufficiently intense peaks in an auto-

Fig. 2 Poisoning case 994/09, with protein precipitation of the venousblood sample. Analysis of a sample by LC–QTOF-MS in auto-MS–MSmode. (a) Total ion chromatogram. (b) Part of the total ion chromatogrambetween 9.81 and 9.94 min shown with increased time resolution andindication of theMS and the threeMS–MS acquisitions. (c) Mass spectrumin MS mode at 9.845 min with indication of the three ions [M+H]1

+,[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 fragmentspectrum of [M+H]1

+. (e) CID fragment spectrum of [M+H]2+. (f) CID

fragment spectrum of [M+H]3+. The collision energies given in (d), (e),

and (f) were automatically calculated by use of Eq. 1

R

1.850 minParacetamol

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

1.2

1.0

0.8

0.6

0.4

0.2

0

Counts x 107

(a)

Countsx10

0

1

2

3

4

5

6

152.0691+

174.0517303.1305

325.1158190.0257

(m/z)140 180 220 260 300 340

5

0

1

2

3

4

5

6

152.0691

174.0517303.1305

325.1158190.0257

(m/z)140 180 220 260

[M+Na]+

[M+H]+

[M+K]+[2M+H]+

[2M+Na]+

0

1

2

3

4

5

6 152.0691(M+H)+

153.0728(M+H)+ 154.0719

(M+H)+

(m/z)150 151 152 153 154 155 156 157

Counts x105

0

1

2

3

4

5

6 152.0691

153.0728154.0719

(m/z)150 151 152 153 154 155 156 157

Counts x105

[M+H]+

[M+H]+[M+H]+

(c)(d)

0.2

0.4

0.6

0.8

1

(m/z)30 50 70 90 110 130 150 170

0

110.0605

152.0711

93.034665.0391

82.0653 134.0606

(m/z)30 50 70 90 110 130 150 170

(d)Countsx10

(b)

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

Fig. 3 Poisoning case 994/09, with protein precipitation of the venousblood sample. (a) Chromatogram measured in auto-MS–MS modeafter background removal and application of the tool “Find Com-pounds”. As a result, in this case 6256 “Compounds” were found. (b)

Single-stage mass spectrum of a peak at 1.850 min (paracetamol). This“Compound” consists of [M+H]+, [M+Na]+, and [M+K]+. (c) Zoomof (b) with the isotope peaks of [M+H]+. (d) The CID mass spectrumrecorded at 1.850 min

Development and practical application of a library of CID accurat 109

MS–MS file. This essentially increases the accuracy ofthe resulting molecular formula.

If the “Compound” also contains an MS–MS spectrum, asearch in the accurate mass spectra library can automati-cally be performed. The number of matching peaks, thenumber of non-matching peaks, agreement of accuratefragment masses, and the fragment abundance ratio areused as criteria for identification of the spectra. Also in thiscase, a weighted score from these criteria is used.Irrespective of the collision energy applied to the unknowncompound, all library spectra with 10, 20, and 40 eVcollision energy are included in the search procedure.Although the collision energy applied to the unknownmolecule after calculation by use of Eq. 1 is usuallydifferent from 10, 20, or 40 eV, the library spectrum withthe next higher or next lower energy was almost always hit1 of the library search. Moreover, hit 2 was usually theother neighboring spectrum of the same substance. This isbecause the match of the accurate fragment masses is astronger indication of identity than the abundance ratio.

By use of this search procedure 29 peaks in thechromatogram of Fig. 3 were identified by use of thedatabase and CID spectra and for further 266 peaks(score >75%, accuracy <5 ppm) only one or more proposalsfrom the database without a fitting spectrum from the librarywere obtained. Figure 4 shows the result of the searchprocedure for the peak at 1.850 min in the chromatogram ofFig. 3 which was identified as paracetamol.

The time for post-run analysis of one chromatogram,including automatic molecular feature extraction (MFE),database and library identification, manual exclusion of falsepositives, and generation of report, is altogether 10–30 min,depending on the complexity of the sample. The report containsa list of identified substances and for each of them a print-out ofthe chromatographic peak with MS and an MS–MS spectrumand scores of identification and visual comparison betweensample and database and library spectra.

Identification of metabolites

The chromatograms obtained from urine samples but alsofrom blood samples of poisoning cases are frequentlydominated by metabolite peaks. Unfortunately, metabo-lites were rarely available as reference substances for thelibrary. In order to fill this gap, a metabolite tool“Identify Metabolites” was developed in order to searchfor possible metabolites if the unknown peak couldbelong to a parent substance or vice versa to search for apossible parent substance if the peak originates from ametabolite. This tool calculates automatically the molec-ular formulas of relevant metabolites (e.g. hydroxylation= + O, demethylation = − CH2) as a basis for calculationof the accurate molecular mass and isotope peak patternand subsequent extraction of the corresponding “Com-pounds” from the analysis file by the MassHuntersoftware. The tool searches for products of all essentialphase I and phase II metabolism reactions and combina-tions between them. In looking for a parent substance, theopposite reactions are calculated. An example is shown inFig. 5 for an intoxication with tramadol, which wasidentified in the chromatogram by library search. Alto-gether 19 metabolites were found to result from demethy-lat ion, hydroxylat ion, N-oxide formation, andglucuronidation. Unambiguous structural identificationwas possible only for both demethyltramadols, becauseO-demethyltramadol is in the library. It is not possible todistinguish between the two or more isomers which wereformed or are possible for each metabolism route withoutadditional investigations. It must be remarked that themetabolite tool can only find but not really identifymetabolites, because CID spectra or retention times ofreference substances are not available. Nevertheless, thisexample shows that the metabolite tool is very useful insupport of identification of the drug and to identify intensepeaks in the chromatogram.

110.0609

152.0714

93.033865.0401 134.060382.0668 125.004377.0393

110.0609152.0714

93.033865.0401 134.060382.0668 125.004377.0393

110.0600

93.033565.0386152.070682.0651 134.0600

(m/z)

64 68 72 76 80 84 88 92 96 100 104 108 112 116 120 124 128 132 136 140 144 148 152

152.0706110.0600

93.033565.0386 134.0600

50

0

100

50

0

100

50

0

100

Abundance , %

Sample Spectrum

Hit 1, Match score 88 %

Hit 2, Match score 81 %

+ESI Product Ion (1.850 min) CID 13.1 V

Paracetamol) C8H9NO2 CID 10.0 V

Paracetamol) C8H9NO2 CID 20.0 V

110.0609

152.0714

93.033865.0401 134.060382.0668 125.004377.0393

110.0609152.0714

93.033865.0401 134.060382.0668 125.004377.0393

110.0600

93.033565.0386152.070682.0651 134.0600

152.0706110.0600

93.033565.0386 134.0600

50

0

100

50

0

100

50

0

100

Abundance , %

Sample Spectrum

Hit 1, Match score 88 %

Hit 2, Match score 81 %

+ESI Product Ion (1.850 min) CID 13.1 V

Paracetamol) C8H9NO2 CID 10.0 V

Paracetamol) C8H9NO2 CID 20.0 V

Fig. 4 Poisoning case 994/09, with protein precipitation of the venous blood sample. Result of the database and library search for the peak at1.850 min in Fig. 3

110 S. Broecker et al.

Accurate CID mass spectra of metabolites from poison-ing cases are collected in a separate library and will beadded to the database and library after accuracy control andcorrection as described above.

Analysis of spiked blood and urine samples

The application of LC–QTOF-MS in combination with thedatabase and library of toxic compounds was examinedwith regard to its advantages and restrictions for qualitativesubstance identification in blood and urine samples. Forthis purpose drug-free samples were spiked with 33 drugsat seven concentrations between 0.5 and 500 ng mL−1

(Table 2), analyzed according to the procedures describedin the “Experimental” section, and submitted to a search inthe database and library. All test substances were present inthe library and had the structural prerequisites for efficientionization by ESI.

No signals were obtained from analysis of the reagentblank of the sample preparations. Different blood andurine samples from five volunteers who did not take anydrugs with the exception of drinking coffee or tea wereanalyzed. Besides caffeine and its metabolites, piperine(ingredient of pepper), benzododecinium (disinfectant,stabilizer), diethyl phthalate (plasticizer), carnitine (aminoacid), adenosine, and thiamine (vitamin B1) were identified.

The chromatogram obtained from a blood samplespiked at a concentration of 2 ng mL−1 after extractionwith CH2Cl2 is shown in Fig. 6. Table 2 summarizes theresults from all spiked samples. There were two differentmethods of identification. At higher concentrations the

substance is detected in the MS file by the database andalso identified from the fragment spectrum measured inthe auto-MS–MS mode. This method is indicated by “I” inTable 2 and was possible at 2 ng mL−1 for 26 of the 33compounds extracted from blood samples (c.f. Fig. 6). Atstill lower concentrations often no MS–MS spectrum isrecorded for the compound, because more matrix massesare preferentially selected for MS–MS measurement bythe software or because the MS signal of the compound isbelow the corresponding lower limit (1000 counts).Nevertheless, the substance is still detected during theMS phase of the measurement cycle and is proposed as aresult of the accurate mass database search. This isindicated by “D” in Table 2, and confirmation of thesubstance was achieved by repetition of the analysis intargeted MS–MS mode and library search of the obtainedfragment spectrum. Because of the 5 to 10 times highersensitivity in the targeted MS–MS analysis, the spectrawere always suitable for identification when the substancewas detectable in the MS data of the auto-MS–MS mode.This way of identification was possible for six compoundsin Fig. 6. At still lower concentrations the substance wasalso not proposed as a hit from the database search althoughthe peak could still be seen in the extracted chromatogram of[M+H]+. This is the case for Δ9-tetrahydrocannabinol whichis therefore not detected in Fig. 6.

From the auto-MS–MS results of Table 2 follows thattypical basic drugs such as amitriptyline, cocaine, orstrychnine were identified (I) in the CH2Cl2 extracts fromblood down to 0.5 ng mL−1. The six benzodiazepines wereidentified with lower sensitivity between 0.5 ng mL−1

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

Counts x 107

0.0

0.2

0.4

0.6

0.8

1.0 Tramadol

12

3 4

5+67

8

9

101211

1314

15,16

17

18

19

N

OH

O

CH3

CH3

CH3

OH

O

+ESI EIC(20.0000-30.0000) Scan Frag=150.0V 944_Poroshell_ACN_5µl_2.d ***ZERO ABUNDANCE***

Fig. 5 Poisoning case 994/09, with protein precipitation of the venousblood sample. Application of the tool “Find Metabolites” to the peak at7.089 min, which was identified as tramadol (T) by library search.Nineteen metabolites were suggested, which result from demethylation,didemethylation, tridemethylation, hydroxylation, or formation of the N-oxide, combinations of these, and combination with glucuronidation.Peak identification by accurate mass: 17=O-demethyl-T (identified also

by library); 19=N-demethyl-T, 11=N,O-didemethyl-T, distinguishedfrom 16 by retention time; 16=N,N-didemethyl-T; 15=tridemethyl-T,13,18=hydroxy-T and T-N-oxide; 1,3,10,14=isomeric hydroxydemethyl-T; 2,12=isomeric hydroxydidemethyl-T; 4=glucuronide of demethyl-T;5,7=glucuronides of di-demethyl-T; 6=glucuronide of hydroxydemethyl-T; 8=glucuronide of hydroxydidemethyl-T; 9=glucuronide of hydroxy-T

Development and practical application of a library of CID accurat 111

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 I I I I I I

B-Prec. Db I I I I I I

U-Dil. – – D I I I I

Amitriptyline B-Extr. I I I I I I I

B-Prec. D I I I I I I

U-Dil. – – I I 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. – – I I I I I

Clonazepam B-Extr. – D D I I I I

B-Prec. – – D I I I I

U-Dil. – – – – D I I

Clozapine B-Extr. I I I I I I I

B-Prec. I I I I I I I

U-Dil. – D I 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. – – I I I I I

Codeine B-Extr. I I I I I I I

B-Prec. D I I I I I I

U-Dil. – I I I I I I

Diazepam B-Extr. I I I I I I I

B-Prec. D I I I I I I

U-Dil. – – – I I I I

Flunitrazepam B-Extr. D D I I I I I

B-Prec. – D D I I I I

U-Dil. – – – I I I I

Hydrocodone 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

Ketamine 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

Lorazepam B-Extr. D D D I I I I

B-Prec. – D D D I I I

U-Dil. – – – D D I I

Methadone 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

Methamphetamine B-Extr. I I I I I I I

B-Prec. – – D I I I I

U-Dil. – – D I I I I

3,4-Methylendioxyamphetamine B-Extr. – I I I I I I

B-Prec. – D I 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. – – D I 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

112 S. Broecker et al.

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. – D I I I I I

Nitrazepam B-Extr. – D I I I I I

B-Prec. – – D I I I I

U-Dil. – D D D D I I

Oxazepam B-Extr. D D I I I I I

B-Prec. – D D I I I I

U-Dil. – – D D D I I

Oxycodone 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

Pethidine 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

Phencyclidine B-Extr. I I I I I I I

B-Prec. D I I I I I I

U-Dil. – – I I I I I

Phentermine B-Extr. – D I I I I I

B-Prec. – – – D D I I

U-Dil. – – – D I I I

Proadifen B-Extr. I I I I I I I

B-Prec. D I I I I I I

U-Dil. – I I I I I I

Strychnine B-Extr. I I I I I I I

B-Prec. I I I I I I I

U-Dil. – – D I I I I

Temazepam B-Extr. D I I I I I I

B-Prec. D D I I I I I

U-Dil. – – D D I I I

Δ9-Tetrahydrocannabinol B-Extr. – – – D D I I

B-Prec. – – – – I I I

U-Dil. – – – – – I I

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. – D I 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. – I I I 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 waterbWorkflow of identification: D=database proposal from MS data, no MS–mass spectrum in auto-MS–MS file, confirmed by targeted MS–MS and library insecond run; I=Identification by MS–mass spectrum from auto-MS–MS data

Development and practical application of a library of CID accurat 113

(diazepam) and 20 ng mL−1 (clonazepam, lorazepam) andwere found by the database at 0.5 or 2 ng mL−1. It wasfound that for oxazepam and lorazepam the main reasonfor the lower efficiency of CID spectra generation is thepreferred formation of sodium clusters instead of [M+H]+.

These lower limits were one or two concentration levelshigher for protein precipitation and two or three levelshigher for diluted urine. Database detection was generallyone or two levels more sensitive than identification bylibrary spectrum. Δ9-Tetrahydrocannabinol was onlydetected in the CH2Cl2 extract at 20 ng mL−1 and delivereda CID spectrum only at 200 ng mL−1.

It is obvious that lower limits of identification as shownin Table 2 depend on sample preparation (extraction yieldand purity of the extract) and on the amount of sampleinjected. In these investigations very simple preparationmethods were applied and only drugs from 12.5 μL (CH2Cl2extract) or 5 μL blood (protein precipitation), or from 1 μLurine were measured. This different portion of the sample isthe main reason for the differences between the three samplepreparations, besides the higher purity of the CH2Cl2 extract.

Application to real samples

Venous blood samples from 50 death cases with knownillegal drug abuse or therapeutic drug intake which hadpreviously been investigated by HPLC–DAD, GC–MS, andimmunoassay were re-analyzed with LC–QTOF-MS incombination with the accurate mass database and CIDspectra library according to the methods described above.One case is described in detail.

Case 994/09

The 54 year old woman died at home after a strongcoughing fit. Resuscitation attempts were without success.She had suffered from unilateral paralysis after a strokeand several vertebral disc prolapses and had been a chainsmoker. She had been taking clarithromycin A for fourdays because of the beginning of pneumonia. Thefollowing drugs had been prescribed or were found inthe apartment: acetylsalicylic acid, baclofen, beclometa-son dipropionate, bisacodyl, carbamazepine, clemastine,codeine, estriol, fenoterol, formoterol, ipratropium, lactu-lose, levomepromazine, loperamide, metoprolol, mirtaza-pin, propiverine, ranitidin, tilidine, and tizanidine.Toxicological analysis was requested for exclusion ofpoisoning.

By HPLC–DAD and GC–MS investigation of venousblood the following result was obtained: codeine(1.1 μg mL−1), tramadol (8.4 μg mL−1), carbamazepine(9.6 μg mL−1), carbamazepine-10,11-epoxide (0.8 μg mL−1),ranitidine (0.7 μg mL−1), and levomepromazine(0.3 μg mL−1). It was concluded that the severe overdoseof codeine and tramadol had at least contributed to the causeof death.

The chromatogram obtained by LC–QTOF-MS from thevenous blood sample after protein precipitation, afterapplication of the tools “Find Compounds” and “IdentifyCompounds” is shown in Fig. 7. The chromatograms andmass spectra shown in Figs. 2, 3, 4 and 5 were alsoobtained from the same analysis file. The previouslydetected compounds were identified by this method with

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

Counts x 105

0.0

0.2

0.6

0.4

0.8

1.0

1.2

1

23 4

5

6 7

8

9

10

11

12 13

14

15 16

1718

19

20

21 22

23

24

25

26

27

28

29

30

31

32

x 3

x 3

x 3

Fig. 6 Chromatogram obtained from a serum sample spiked with2 ng mL−1: 1=3,4-methylenedioxyamphetamine; 2=methamphet-amine; 3=methylenedioxymethamphetamine; 4=3,4-methylenedioxy-ethamphetamine (MDEA); 5=phentermine; 6=oxycodone; 7=codeine;8=metoprolol; 9=tramadol; 10=hydrocodone; 11=strychnine; 12=co-caine; 13=meperidine (pethidine); 14=phencyclidine (PCP); 15=cit-alopram; 16=carbamazepine; 17=nitrazepam; 18=clonazepam;19=ketamine; 20=flunitrazepam; 21=oxazepam; 22=lorazepam;

23=alprazolam; 24=zolpidem; 25=temazepam; 26=methadone; 27=di-azepam; 28=verapamil; 29=trazodone; 30=amitriptyline; 31=cloza-pine; 32=proadifen; 33=Δ9-tetrahydrocannabinol (not detected).Substances 5, 16, 17, 20, 21, and 22 shown also in threefoldmagnification in the upper part of the figure were only detected byaccurate mass and isotope pattern. All other substances were identifiedby library search of the CID fragment spectrum

114 S. Broecker et al.

the exception of ranitidine which could not be confirmed.In addition, baclofen, clarithromycin, clemastine, clomipr-amine, fentanyl, loperamide, midazolam, methiomeprazine,morphine, nortilidine, paracetamol, paroxetine, propiverine,tilidine, and tizanidine were also seen. Furthermore, in theCH2Cl2 extract lidocaine, nicotine, and ofloxacine wereidentified. These results were mainly in agreement with thedrugs known from the case history. Smoking habits of thediseased were confirmed by detection of nicotine and

cotinine. This was the death case in this study with thehighest number of identified substances.

In the same way all 50 cases were investigated. Anoverview of the results is given in Table 3. Between 1 and10 (mean 3.5, median 4) substances per sample, and a totalof 94 different substances had previously been identifiedand semi-quantitatively determined by general screening forunknown substances using HPLC–DAD [2] and by GC–MS analysis in selected-ion-monitoring mode for illegal

+ESI EIC(20.0000-30.0000) Scan Frag=150.0V 944_Poroshell_ACN_5µl_2.d ***ZERO ABUNDANCE***

2 2

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Time, min

0.0

0.3

0.9

0.6

1.2

1.5 Counts x 106

1

2

3

4

5

7

6 8

9 11

10

12

14

13

1715

16

X

18

19 20

21 22

23

24

Fig. 7 Poisoning case 994/09, with protein precipitation of the venousblood sample. Substances identified by library search. 1=paracetamol;2=baclofen; 3=morphine; 4=cotinine; 5=tizanidine; 6=codeine glucu-ronide; 7=O-demethyltramadol; 8=codeine; 9=tramadol; 10=carbama-

zepine-9,10-epoxide; 11=norti l idine; 12=carbamazepine;13=hydrocortisone; 14=methiomeprazine; 15=paroxetine; 16=clo-mipramine; 17=tilidine; 18=midazolam; 19=loperamide; 20=fentanyl:21=clemastine; 22=clarithromycin; 23=clemastine; 24=propiverine

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

Development and practical application of a library of CID accurat 115

drugs and substances suspected from case histories. Boththe basic CH2Cl2 extract and the supernatant from proteinprecipitation of the blood sample were analyzed by LC–QTOF-MS. The ionization mode was restricted to positiveESI. All substances included in the evaluation wereidentified by library search of CID fragment spectra whichwere exclusively measured in auto-MS–MS mode. Onlymetabolites present in the library were included and nospecific search for metabolites with the metabolite tool asdescribed above was performed.

Under these conditions, a total of 125 different sub-stances were identified by LC–QTOF-MS, 31 more thanknown from the toxicological reports. The number ofpositive results increased from 178 to 311, because manydrugs were also detected more frequently. The number ofidentified substances per case was between 0 and 14 (mean3.2, median 4) higher than described in the toxicologicalreports. The reasons for this are the much higher sensitivityin comparison with HPLC–DAD, less disturbing effect ofoverlapping chromatographic peaks, and the inclusion ofsubstances without UV absorption.

On the other hand, some substances were also not detectedby LC–QTOF-MS, for example furosemide, ibuprofen,barbiturates or salicylic acid. These substances are preferen-tially ionized in negative-ESI mode which was not performedin this study. In other cases, the method was less sensitive thanthe GC–MS analysis. For instance, Δ9-tetrahydrocannabinol(THC) could not be detected in blood samples previouslytested positive by GC–MS. For cannabinoids and forconfirmation of low-dose benzodiazepines an additionaltargeted MS–MS run must be recommended.

Conclusions

It has been shown in this study that the advantages of TOFmass spectrometry for systematic toxicological analysisdescribed previously [19–28, 31–35] can be exploited to amuch greater extent by using a hybrid quadrupole TOFinstrument in combination with a large database and libraryof toxic compounds which contains CID accurate massspectra in addition to theoretical accurate mass data of awide variety of drugs and poisons. Relatively simplesample-preparation techniques from blood and urine andLC–QTOF-MS measurement in data-dependent MS–MSacquisition (auto-MS–MS mode) provide accurate MS andMS–MS data in a single chromatographic run for substanceidentification by use of efficient software tools. Theaccurate-mass CID fragment spectra proved to be veryspecific and enabled peak identification with high accuracy.The library was successfully applied to analysis filesmeasured by other QTOF-MS instrument series of thesame manufacturer. The lower limits of identification at

which suitable MS–MS spectra for library search wereobtained were 0.5 to 20 ng mL−1, depending on structureand matrix. Detection by database search of the MS datawas found to be approximately one order of magnitudemore sensitive but in this case a second run in targeted MS–MS mode is necessary to obtain CID fragment spectra forunambiguous identification and for distinguishing betweenstructural isomers.

Because of the comprehensive data acquisition and thehigher sensitivity, clearly more substances were identifiedin blood samples from death cases than in previous routineinvestigations by the authors using HPLC–DAD and GC–MS. This makes LC–QTOF-MS to one of the most efficientmethods in systematic toxicological analysis.

Further work is in progress to include substances with lowESI yield by extension of the library to atmospheric pressurechemical ionization (APCI), to optimize the choice of the CIDenergy during auto-MS–MS measurement, and to performautomatic search for metabolites in the post run data-processing algorithm. Furthermore, a way of approximateestimation of the concentrations from the peak areas and asystem of internal standards will be developed to assess thetoxicological relevance of the identified peaks from whichmany may be in the therapeutic or sub-therapeutic range.

Acknowledgement The authors thank Agilent Technologies, Inc.(Santa Clara, California, USA) for generous technical support of theseinvestigations and in particular Dr Peter Stone and Dr FrankKuhlmann for helpful cooperation and discussions. They are alsograteful to Mrs Jana Küchler, Mr Bruno Feyerabend, Mrs KarolinHeinze, Mr Henrik Petszulat, Mrs Franziska Wiechert, and Mrs SinaWuttig for their thorough and efficient experimental assistance.

References

1. Pfleger K, Maurer HH, Weber A (2007) Mass spectral and GC dataon drugs, poisons, pesticides, pollutants and their metabolites, 3rdrevised and enlarged edition. Wiley, Weinheim

2. Pragst F, Herzler M, Erxleben B-T (2004) Clin Chem Lab Med42:1325–1340

3. Maurer HH (2005) Clin Biochem 38:310–3184. Maurer HH (2007) Anal Bioanal Chem 388:1315–13255. Wood M, Laloup M, Samyn N, Mar Ramirez FM, De Bruijn EA,

Maes RAA, De Boeck G (2006) J Chromatogr A 1130:3–156. Peters FT (2010) Clin Biochem Aug 13. [Epub ahead of print]7. Marquet P (2002) Ther Drug Monit 24:125–1338. Vernisse N, Marquet P, Duchoslav E, Dupuy JL, Lachatre G

(2003) J Anal Toxicol 27:7–149. Marquet P, Saint-Marcoux F, Gamble TN, Leblanc JC (2003) J

Chromatogr B Anal Technol Biomed Life Sci 789:9–1810. Jansen R, Lachatre G, Marquet P (2005) Clin Biochem 38:362–37211. Sauvage FL, Saint-Marcoux F, Duretz B, Deporte D, Lachatre G,

Marquet P (2006) Clin Chem 52:1735–174212. Sauvage FL, Picard N, Saint-Marcoux F, Gaulier JM, Lachatre G,

Marquet P (2009) J Sep Sci 32:3074–308313. Weinmann W, Wiedemann A, Eppinger B, Renz M, Svoboda M

(1999) J Am Soc Mass Spectrom 10:1028–1037

116 S. Broecker et al.

14. Dresen S, Kempf J, WeinmannW (2006) Forensic Sci Int 161:86–9115. Mueller CA, Weinmann W, Dresen S, Schreiber A, Gergov M

(2005) Rapid Commun Mass Spectrom 19:1332–133816. Dresen S, Gergov M, Politi L, Halter C, Weinmann W (2009)

Anal Bioanal Chem 395:2521–252617. Dresen S, Ferreiros N, Gnann H, Zimmermann R, Weinmann W

(2010) Anal Bioanal Chem 396:2425–243418. Gergov M, Ojanpera I, Vuori E (2003) J Chromatogr B 795:41–5319. Gergov M, Boucher B, Ojanperä I, Vuori E (2001) Rapid

Commun Mass Spectrom 15:521–52620. Pelander A, Ojanperä I, Laks S, Rasanen I, Vuori E (2003) Anal

Chem 75:5710–571821. Ojanperä I, Pelander A, Laks S, Gergov M, Vuori E, Witt M

(2005) J Anal Toxicol 29:34–4022. Ojanperä S, Pelander A, Pelzing M, Krebs I, Vouri E, Ojanperä I

(2006) Rapid Commun Mass Spectrom 20:1161–116723. Kolmonen M, Leinonen A, Pelander A, Ojanperä I (2007) Anal

Chim Acta 585:94–10224. Pelander A, Ristimaa J, Rasanen I, Vuori E, Ojanperä I (2008)

Ther Drug Monit 30:717–72425. Pelander A, Ristimaa J, Ojanperä I (2010) J Anal Toxicol 34:312–31826. Tyrkkö E, Pelander A, Ojanperä I (2010) Drug Test Anal 2:259–27027. Ristimaa J, Gergov M, Pelander A, Halmesmäki E, Ojanperä I

(2010) Anal Bioanal Chem 398:925–93528. Pavlic M, Libiseller K, Oberacher H (2006) Anal Bioanal Chem

386:69–8229. Decaestecker TN, Clauwaert KM, Van Bocxlaer JF, Lambert WE,

Van den Eeckhout EG, Van Peteghem CH, De Leenheer AP(2000) Rapid Commun Mass Spectrom 14:1787–1792

30. Decaestecker TN, Vande Casteele SR, Wallemacq PE, VanPeteghem CH, Defore DL, Van Bocxlaer JF (2004) Anal Chem76:6365–6373

31. Polettini A, Gottardo R, Pascali JP, Tagliaro F (2008) Anal Chem80:3050–3057

32. Liotta E, Gottardo R, Bertaso A, Polettini A (2010) J MassSpectrom 45:261–271

33. Lee HK, Ho CS, Iu YP, Lai PS, Shek CC, Lo YC, Klinke HB,Wood M (2009) Anal Chim Acta 649:80–90

34. Stroh JG, Petucci CJ, Brecker SJ, Nogle LM (2008) J Sep Sci31:3698–3703

35. Kaufmann A, Butcher P, Maden K, Widmer M (2007) Anal ChimActa 586:13–21

36. Nielsen MK, Johansen SS, Dalsgaard PW, Linnet K (2010)Forensic Sci Int 196:85–92

37. http://molgen.de (molecular structure generation), assessed 20August 2010

38. www.chemspider.com, assessed 20 August 201039. Pragst F, Herzler M, Herre S, Erxleben B-T, Rothe M (2001)

UV-Spectra of toxic compounds. Database of photodiode arrayUV spectra of illegal and therapeutic drugs, pesticides, ecotoxicsubstances and other poisons. Verlag Dieter Helm, Heppen-heim; Suppl Vol (2008) Edition Toxicological Chemistry,Berlin

40. Stone P, Zweigenbaum J (2009) An application kit for thescreening of samples for analytes of forensic and toxicologicalinterest using TOF or Q-TOF LC/MS with a personal forensics/toxicology database. Agilent application note 5990-4252EN,Agilent Technologies, Inc

41. Broecker S, Herre S, Pragst F, Kuhlman F, Wüst B,Zweigenbaum J (2010) Toxicological screening with theagilent LC/MS-QTOF and the personal compound databaseand the Broecker, Herre and Pragst accurate mass spectrallibrary. Agilent Application Note 5990-6419EN Agilent Tech-nologies, Inc

Development and practical application of a library of CID accurat 117