Comparison of triple quadrupole, hybrid linear ion trap triple quadrupole, time-of-flight and LTQ-orbitrap mass spectrometers in drug discovery phase metabolite screening and identification in vitro - amitriptyline and verapamil as model compounds

Article abstract of DOI:10.1002/rcm.4465

Liquid chromatography in combination with mass spectrometry (LC/MS) is a superior analytical technique for metabolite profiling and identification studies performed in drug discovery and development laboratories. In the early phase of drug discovery the analytical approach should be both time- and cost-effective, thus providing as much data as possible with only one visit to the laboratory, without the need for further experiments. Recent developments in mass spectrometers have created a situation where many different mass spectrometers are available for the task, each with their specific strengths and drawbacks. We compared the metabolite screening properties of four main types of mass spectrometers used in analytical laboratories, considering both the ability to detect the metabolites and provide structural information, as well as the issues related to time consumption in laboratory and thereafter in data processing. Human liver microsomal incubations with amitriptyline and verapamil were used as test samples, and early-phase 'one lab visit only' approaches were used with all instruments. In total, 28 amitriptyline and 69 verapamil metabolites were found and tentatively identified. Time-of-flight mass spectrometry (TOFMS) was the only approach detecting all of them, shown to be the most suitable instrument for elucidating as comprehensive metabolite profile as possible leading also to lowest overall time consumption together with the LTQ-Orbitrap approach. The latter however suffered from lower detection sensitivity and false negatives, and due to slow data acquisition rate required slower chromatography. Approaches with triple quadrupole mass spectrometry (QqQ) and hybrid linear ion trap triple quadrupole mass spectrometry (Q-Trap) provided the highest amount of fragment ion data for structural elucidation, but, in addition to being unable to produce very high-important accurate mass data, they suffered from many false negatives, and especially with the QqQ, from very high overall time consumption.

Full text of DOI:10.1002/rcm.4465

RAPID COMMUNICATIONS IN MASS SPECTROMETRY
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/rcm.4465
Comparison of triple quadrupole, hybrid linear ion trap
triple quadrupole, time-of-flight and LTQ-Orbitrap mass
spectrometers in drug discovery phase metabolite
screening and identification in vitro – amitriptyline and
verapamil as model compounds
Timo Rousu1,2, Jukka Herttuainen3 and Ari Tolonen1
*
1Novamass Ltd., Medipolis Center, Kiviharjuntie 11, 90220 Oulu, Finland
2Department of Chemistry, University of Oulu, P.O. Box 3000, 90014 Oulu, Finland
3Orion Pharma, P.O. Box 65, 02101 Espoo, Finland
Received 18 December 2009; Revised 14 January 2010; Accepted 15 January 2010
Liquid chromatography in combination with mass spectrometry (LC/MS) is a superior analytical
technique for metabolite profiling and identification studies performed in drug discovery and
development laboratories. In the early phase of drug discovery the analytical approach should be
both time- and cost-effective, thus providing as much data as possible with only one visit to the
laboratory, without the need for further experiments. Recent developments in mass spectrometers
have created a situation where many different mass spectrometers are available for the task, each with
their specific strengths and drawbacks. We compared the metabolite screening properties of four
main types of mass spectrometers used in analytical laboratories, considering both the ability to
detect the metabolites and provide structural information, as well as the issues related to time
consumption in laboratory and thereafter in data processing. Human liver microsomal incubations
with amitriptyline and verapamil were used as test samples, and early-phase ‘one lab visit only’
approaches were used with all instruments. In total, 28 amitriptyline and 69 verapamil metabolites
were found and tentatively identified. Time-of-flight mass spectrometry (TOFMS) was the only
approach detecting all of them, shown to be the most suitable instrument for elucidating as
comprehensive metabolite profile as possible leading also to lowest overall time consumption
together with the LTQ-Orbitrap approach. The latter however suffered from lower detection
sensitivity and false negatives, and due to slow data acquisition rate required slower chromatog-
raphy. Approaches with triple quadrupole mass spectrometry (QqQ) and hybrid linear ion trap triple
quadrupole mass spectrometry (Q-Trap) provided the highest amount of fragment ion data for
structural elucidation, but, in addition to being unable to produce very high-important accurate mass
data, they suffered from many false negatives, and especially with the QqQ, from very high overall
time consumption. Copyright # 2010 John Wiley & Sons, Ltd.
Most drugs are eliminated from the human body by
oxidative and conjugative (phase I & II) biotransformation
reactions catalyzed mainly by cytochrome P450 and several
conjugative enzymes, of which uridinediphosphate-glucur-
onosyltransferase (UGT) plays the most important role. The
pharmacokinetic properties of drugs are determined to a
great extent by metabolic reactions and, therefore, metab-
olism is often behind bioavailability problems, interindivi-
dual variations, metabolic interactions and idiosyncrasies.1,2
Recently the importance of metabolite profiling studies in the
very early phase of drug discovery and development has
been stressed, especially by focusing on human metabolism,
but profiling metabolites in different toxicity species as
well.3,4 Also the latest FDA guidance encourages that early
metabolite profiling is carried out as soon as possible, both
in vitro and in vivo.5
Over the past ten years, liquid chromatography in
combination with mass spectrometry (LC/MS) has been
the preferred tool for the analysis of drugs and their
metabolites in both in vivo and in vitro studies. However,
there are several types of mass spectrometers available, each
with their specific strengths and drawbacks. Therefore, a
decision which one to use for which task also becomes more
important, so that as high quality data as possible can be
obtained as time and cost effectively as possible.6 Both data
quality and rapid delivery of results play a crucial role in
*Correspondence to: A. Tolonen, Novamass Ltd., Medipolis
Center, Kiviharjuntie 11, 90220 Oulu, Finland.
E-mail: ari.tolonen@novamass.net
Copyright # 2010 John Wiley & Sons, Ltd.

940 T. Rousu, J. Herttuainen and A. Tolonen
drug discovery and development because the information of
metabolic stability, metabolic routes, metabolic interactions
and metabolising enzymes and their kinetics is needed for
selecting leads and candidates.1
Triple quadrupole (QqQ) mass spectrometers have
traditionally been the main work horses of most analytical
laboratories. Excellent MS/MS capability of the triple
quadrupole instrument is widely used for further elucidation
of biotransformation sites in conjunction with accurate mass
measurements acquired with different instruments. Even
though very high detection sensitivity is obtained in various
biological matrices for known analytes in multiple reaction
monitoring (MRM) mode, thus being very applicable for
quantitative work, the detection sensitivity of QqQ is
however rather poor compared to other types of MS
instruments when full scan acquisition is used, limiting also
the applicability of full scan MS/MS product ion measure-
ment. In addition to product ion scanning, also other full scan
data acquisition modes, i.e. neutral loss (NL) and/or
precursor ion (PI) scanning functions, have been used for
the screening of drug metabolites with QqQ.14,15 These
structure-specific acquisitions result in true positives only if
metabolites undergo similar MS/MS fragmentation beha-
viour as the parent drug, thus many unexpected metabolites
may be missed, as well as metabolites with more than one
biotransformation site at opposite ends of the molecule. The
most commonly used screening method with QqQ is to scan
for the neutral loss of phase II conjugated metabolite, e.g. the
NL of 176 Da of a glucuronide.14,16 Moreover, the use of these
precursor ion related fragment ion patterns for screening
metabolites requires laborious preadjustment of the instru-
ment and usually several LC/MS runs to cover all desired
NL and PI acquisition types.
To facilitate rapid and time-effective high-quality analysis
in drug discovery and early phase development, the
analytical approach should provide as much data as possible
with only one visit to the laboratory, so that all crucial
information would be collected at once, without the need for
further experiments. This in turn means that the analytical
method should be able to collect data for both molecular ions
and fragment ions, at least for the main metabolites, enabling
the identification of biotransformations and their approxi-
mate sites (‘soft-spot’) in the parent compound structure.
Moreover, this should be provided from test samples with
very low parent drug concentrations, as in many cases the
new chemical entities (NCEs) under development have low
aqueous solubility due to high lipophilicity, limiting the test
assay concentrations. In addition to this also information
regarding the metabolic stability of the parent compound is
often desired from the same experiments, meaning that
in vitro assays with initial concentration around 1 mM or even
less is preferred.
The use of time-of-flight (TOF) or quadrupole time-of-
flight (Q-TOF) mass spectrometers in drug metabolite
screening is in many case preferred for their high full scan
detection sensitivity and mass accuracy coupled to their
compatibility for modern fast chromatography by very high
data acquisition rate. The detection of all expected and
unexpected metabolites in a single run, without the need for
preadjustment of detection for certain predicted metabolites
and with the possibility of various post-acquisition data
filtering and processing options, makes the screening of
metabolites both straightforward and cost effective.7–9
Although some TOF instruments are capable of producing
tandem mass spectrometric (MS/MS) information via in-
source fragmentation, the advantage of Q-TOF over TOF
instruments is in the possibility to obtain more detailed
information about the site for the detected biotransformation.
The use of both high and low collision energy data
acquisition functions in a single LC/MS run provides so
called MSE data, where fragment ion data is acquired for all
detected compounds and linked by their retention times.10
This functionality can be also applied with some limitations
to some TOF instruments without a real collision cell, by
using two parallel acquisition functions with high and low
cone/aperture voltages.
A quite recent major advancement for triple quadrupole
technology is the hybrid linear ion trap triple quadrupole (Q-
Trap) in which the last quadrupole is replaced by a linear ion
trap.17–19 The ion trap is capable of MSn as well as high
sensitivity scanning, whereas the instrument is also capable
of triple quadrupole like collision-induced dissociation (CID)
fragment ion production without low mass cut-off and
of MRM mode sensitivity. For drug metabolite screening
purposes the Q-Trap offers both structure-specific and data-
dependent data acquisition modes, of which selected/
multiple reaction monitoring (SRM/MRM) triggered
enhanced product ion (EPI) MS/MS scanning is most widely
used for metabolite screening.20–22 In addition, approaches
utilizing a wide-range scan with the ion trap as a survey scan
(EMS, enhanced mass scan), followed by EPI MS/MS (EMS/
EPI), have been described.23
Depending on what type of mass spectrometer has been
used for metabolite screening, a number of different software
programs exist for prediction and post-acquisition proces-
sing of data.8,9,24–27 For post-acquisition data processing,
the comparison of sample and control chromatograms
with these software programs is typically straightforward,
revealing the differences in sample and its negative control.
All software designed to ease the screening procedure and to
increase the throughput must be subjected to careful setup,
as with dozens of different parameters and filters the
possibility of false negatives is significantly increased.
In this study, we compared the metabolite screening
properties of various mass spectrometers widely used in
analytical laboratories. Amitriptyline and verapamil human
liver microsomal (HLM) incubations in vitro were screened
In a LTQ-Orbitrap mass spectrometer, where the high-
resolution mass analyzer is usually combined with a linear
ion trap, accurate mass measurements are combined with the
high trapping capacity and MSn scan function of the linear
ion trap. Therefore, the LTQ-Orbitrap is capable of MS/MS
and produces very high accuracy mass data which makes it a
useful instrument in identification of biotransformations and
their sites as well.11–13 As a down side, the relatively slow
data acquisition rate of the LTQ-Orbitrap makes it incompa-
tible with very fast liquid chromatography, at least if the aim
is to analyze several types of data within a single LC/MS run,
which is essential for fast high-throughput metabolite
screening.
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 941
for metabolites with TOF, QqQ and Q-Trap mass spec-
trometers coupled to ultra-performance liquid chromato-
graphic (UPLC), and amitriptyline samples were also
screened with a LTQ-Orbitrap. Amitriptyline and verapamil
were selected as test compounds in this study as both of them
are known to undergo extensive oxidative metabolism and
have been widely studied by LC/MS.28–31 These two model
compounds also have distinctive differences: (a) amitripty-
line forms through very comprehensive fragmentation ten
major fragment ions in CID, whereas verapamil forms only
four major fragment ions; and (b) amitriptyline serves as a
model compound with quite simple expected metabolism,
whereas verapamil has several expected biotransformation
sites, producing a very high number of metabolites formed
via combinations of several reactions in various sites of the
parent compound structure. The number of major fragment
ions for the parent compound has naturally a large impact on
the number of NL/PI transition reactions or ions that can be
utilized in targeted metabolite analysis. The aim was to use
an approach suitable for tentative discovery/early develop-
ment phase metabolite screening meaning as few runs per
sample as possible, and that all data, including precursor and
fragment ion data, should be acquired with only one visit to
the laboratory without the need for additional experiments.
10 min at 16100 g (Eppendorf 5415D, Eppendorf AG,
Hamburg, Germany) and pipetted to Maximum Recovery
vials (Waters Corp., Milford, MA, USA).
Liquid chromatography
In combination with triple quadrupole, hybrid linear ion trap
triple quadrupole and time-of-flight mass spectrometry, a
Waters Acquity ultra-performance liquid chromatographic
(UPLC) system (Waters Corp.) with an autosampler, a
vacuum degasser and a column oven was used. A Waters
BEH ShieldRP18 column (2.1 ꢁ 50 mm, 1.7 mm) was used
together with an on-line filter. The eluents were 0.1% acetic
acid (A, pH 3.2) and acetonitrile (B). A linear gradient elution
with profile 5% – 20% – 85% – 85% B in 0 – 5 – 6 – 7 min was
applied, followed by column equilibration. The flow rate was
0.5 mL/min and the column oven temperature was 358C. A
4 mL injection volume was used. The first 0.6 min of the run
was directed to waste by using a divert valve to decrease the
ion source contamination from early eluting matrix con-
stituents. The UPLC system was operated under MassLynx
4.1 software (with TOFMS and QqQ) or Analyst 1.4.2
software (with Q-Trap).
With LTQ-Orbitrap mass spectrometry the chromato-
graphic system consisted of
a Thermo Accela liquid
chromatograph (Thermo-Fischer, San Jose, CA, USA) with
a Thermo Hypersil Gold C18 column (2.1 ꢁ 50 mm, 3.0 mm).
The same eluents as described above were used at a flow rate
of 0.2 mL/min and a gradient profile of 5% – 20% – 85% – 85%
B in 0 – 9 – 11 – 13 min, followed by column equilibration. An
ambient column oven temperature was used, and the
injection volume was 4 mL. The LC system was operated
under Xcalibur 2.0.7 software.
EXPERIMENTAL
Reagents and materials
HPLC grade acetonitrile (LiChrosolv GG) and dimethyl
sulfoxide (DMSO) (SeccoSolv) were purchased from Merck
(Darmstadt, Germany). Acetic acid was purchased from
BDH Laboratory Supplies (Poole, UK). Laboratory water was
distilled and purified with a Direct-Q water purifier
(Millipore, Molsheim, France). Both of the test compounds,
NADPH and UDPGA were all purchased from Sigma-
Aldrich (Helsinki, Finland).
TOF experiments
Data was acquired with a Waters LCT Premier XE time-of-
flight (TOF) mass spectrometer (Waters Corp.) equipped
with a LockSpray electrospray ionization (ESI) source. The
positive ionization mode of ESI was used with a capillary
voltage of 2800 V and a cone voltage of 40 V. Two parallel
data acquisition functions were used, with aperture voltages
of 5 V (for molecular ions) and 50 V (for fragmentation). An
acquisition time (scan time) of 100 ms per data point was
used leading to a total cycle time of 200 ms. Lock mass data
was acquired in every 10th scan. The mass range of m/z 150–
750 was acquired, using a W-mode flight tube optics at
resolution of 12 000 (FWHM) and dynamic range enhance-
ment (DRE) option. Desolvation temperature used was 3508C
and source temperature was 1508C. Nitrogen was used both
as desolvation and as nebulizer gas. Leucine enkephalin was
used as a lock mass compound ([MþH]þ m/z 556.2771) for
accurate mass measurements and was infused into the
LockSpray ion source via a separate ionization probe using a
syringe pump. The mass spectrometer was operated under
MassLynx 4.1 software. Metabolites were mined from the
data by using the MetaboLynx XS subroutine of the
MassLynx software, employing the dealkylation tool and
‘chemically intelligent’ (structure based) mass defect filtering
(MDF) with a 40 mDa tolerance window. The real positives
(metabolites) and their identifications were confirmed from
the data manually. Metabolites with signal-to-noise (S/N)
Microsomal incubations and sample
preparation
Pooled liver microsomes containing 20 mg protein/mL were
obtained from BD Biosciences Discovery Labware (Woburn,
MA, USA). Human microsomal pool (Lot#99268) consisted
of liver samples from 30 donors of both genders. The basic
incubation mixture of 250 mL in volume consisted of the
following components: 0.5 mg of microsomal protein per mL,
substrate in DMSO, 1 mM NADPH and 1 mM UDPGA. The
substrate concentration used was 10 mM. Two parallel
incubations, one with cofactors and one without, were
employed. The final amount of DMSO in the incubation was
1% (v/v). Each reaction mixture was preincubated for 2 min
at 378C in a shaking incubator block (Eppendorf Thermo-
mixer 5436, Hamburg, Germany). The incubation reactions
were started by addition of NADPH and UDPGA. After an
incubation period of 0 or 60 min, a 100 mL sample was
collected and the reaction was terminated by adding an
equal volume of ice-cold acetonitrile. Control incubations
were performed without the test compounds. Samples were
subsequently cooled in an ice bath and the tubes were stored
at ꢀ188C until analysis. The incubation samples were
thawed at room temperature, shaken and centrifuged for
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

942 T. Rousu, J. Herttuainen and A. Tolonen
ratio >3 in the ion chromatogram, measured accurate mass
within 3 mDa from the calculated mass and signals missing
from negative controls were considered as real positives.
Information-dependent acquisition (IDA) methods link the
survey scan to ER (Enhanced Resolution Scan) and EPI
(Enhanced Product Ion Scan) utilizing the linear ion trap
mode of the instrument. Instrument parameters for SRM
mode were optimized by infusing each analyte dissolved in
methanol/water/formic acid (50:50:0.1). Parameters for
amitriptyline were declustering potential (DP) 50, EP 10,
collision energy (CE) 25, CXP 18 and for verapamil DP 70, EP
10, CE 45, CXP 20. SRM parameters optimized for amitripty-
line and verapamil were used as a basis when generating the
SRM transitions for the metabolites, assuming fragmentation
behaviour of the metabolites and the parent drug to
be similar. Predicted SRM transitions used for the metabolite
screening of amitriptyline and verapamil samples are
presented in Table 1. The scan time for each SRM transition
was 5 ms. ER and EPI acquisitions were triggered when ions
detected with SRM exceeded 500 counts per second (cps). ER
scanning was performed at step size 0.03 amu, scan rate
250 amu/s, 2 ms pause between mass ranges and Q0
trapping on. EPI scanning was performed at step size
0.08 amu, scan rate 1000 amu/s and 2 ms pause between
mass ranges. Dynamic fill time of the trap was applied in
both ER and EPI. In the EPI scan, DP and CE were as follows:
amitriptyline 50 and 30; verapamil 70 and 40. Total cycle time
of the amitriptyline method was 1.19 s and of the verapamil
method 1.44 s. Ion source conditions for both amitriptyline
and verapamil were set as follows: CUR ¼ 20, CAD ¼ 5,
IS ¼ 5000, TEM ¼ 4508C, GS1 ¼ 50 and GS2 ¼ 50. Samples
were also analyzed with a full scan method (EMS-ER-EPI)
using mass range 100–600 amu for amitriptyline and 200–
700 amu for verapamil. EMS scanning was performed at step
size 0.08 amu, scan rate 1000 amu/s, 2 ms pause between
mass ranges and with dynamic fill time on. ER and EPI
parameters in the full scan methods were identical to the
SRM-ER-EPI method described above. Total cycle time of the
amitriptyline method was 1.71 s and of the verapamil
method 1.80 s. Nitrogen was used as the nebulizer and
auxiliary gas. The mass spectrometer was operated under
Analyst1 (version 1.4.2) software. Metabolites were mined
from the data manually, by looking directly for visually
detectable peaks in LC/MS/MS chromatograms. Metab-
olites with S/N ratio >3 in SRM chromatograms, ER
spectrum confirming the metabolite found with SRM, EPI
fragment data fitting to study compound related structures,
Triple quadrupole experiments
Data was acquired with a Waters Quattro Premier triple
quadrupole mass spectrometer (Waters Corp.) equipped
with a Z-spray ESI source. The positive ionization mode of
ESI was used with a capillary, extractor and RF lens voltages
of 2800 V, 3 V and 0.2 V, respectively. With amitriptyline a
cone voltage of 32 V was used and with verapamil the cone
voltage was set to 40 V. With amitriptyline, the ten most
abundant fragment ions were selected for precursor ion (PI)
scan/neutral loss (NL) scan functions assuming similar
fragmentation behaviour for the metabolites as for the parent
drug, whereas for verapamil the four most abundant
fragment ions were selected. The used PI/NL scan functions
with corresponding collision energies for both of the
compounds are summarized in Table 1. All NL/PI
experiments were run using only one PI or NL scan/
transition at a time, i.e. in total 21 runs/sample for amitripty-
line and 9 runs/sample for verapamil were acquired. Scan
times used were 0.3 s. Collision gas was argon with a CID gas
cell pressure of 3.90 ꢁ 10ꢀ3 mbar. Desolvation temperature
was 3508C and source temperature was 1508C. Nitrogen was
used both as desolvation and as nebulizer gas. The mass
spectrometer was operated under MassLynx 4.1 software.
Metabolites were mined from the data manually, by looking
directly for visually detectable peaks in LC/MS/MS
chromatograms, and by searching ion chromatograms for
the common predicted metabolites. Metabolites with S/N
ratio >3 in ion chromatograms extracted from PI or NL data
and signals missing from negative controls were considered
as real positives.
Hybrid linear ion trap triple quadrupole
experiments
LC/MS/MS data was acquired with a Applied Biosystems/
MDS SCIEX 4000 QTrap1 triple quadrupole/linear ion trap
(QqQLIT) mass spectrometer (Applied Biosystems, Foster
City, CA, USA) equipped with a TurboIonSpray1 interface
operating in the positive ESI mode. A SRM-ER-EPI approach
was used in which SRM functions as survey scan and
exploits the triple quadrupole mode of the instrument.
Table 1. The NL/PI and SRM reactions used with the triple quadrupole and Q-Trap instruments
PI/NL functions for amitriptyline and metabolites; parents for
233/45 (16), 218/60 (24), 205/73 (24), 191/87 (22), 155/123 (24), 129/149
m/z/neutral loss of Da (collision energy eV)
(24), 117/161 (22), 105/173 (24), 91/187 (24), 84/194 (24), NL of 176 (25)
for glucuronides
PI/NL functions for verapamil and metabolites; parents for
m/z/neutral loss of Da (collision energy eV)
303/152 (24), 260/195 (32), 165/290 (26), 150/305 (38), NL of 176 (25) for
glucuronides
SRM transitions for amitriptyline and metabolites; m/z>m/z
278>191, 294>207, 294>191, 294>100, 310>223, 310>207, 310>116,
264>191, 264>70, 250>191, 250>56, 270>189, 276>82, 454>278,
452>276, 470>294, 292>205, 292>98, 296>209, 298>225, 298>70,
312>225, 468>292, 280>207, 280>89, 440>264
SRM transitions for verapamil and metabolites; m/z>m/z
455>303, 441>151, 441>165, 441>289, 441>303, 617>441, 457>165,
457>167, 457>303, 457>305, 633>457, 427>137, 427>151, 427>165,
427>275, 427>289, 427>303, 603>427, 443>151, 443>289, 471>165,
471>181, 471>303, 471>319, 631>455, 469>165, 469>303, 647>471,
291>248, 277>234, 307>264, 307>307, 453>277
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 943
and signals missing from negative controls were considered
as real positives.
chromatograms with a 20 mDa window, and by using
Thermo MetWorks software. The same expected metabolites
list as with MetaboLynx (see above) was used and MDF was
used with 80 mDa tolerance window.
LTQ-Orbitrap experiments
Data was acquired using a LTQ-Orbitrap mass spectrometer
(Thermo Fisher Scientific, Bremen, Germany). An approach
with high-resolution (15 000 FWHM) scan at mass range m/z
100–750 as a survey scan was utilized, and parallel low-
resolution MS/MS data using the ion trap function was
acquired data-dependently for the two most intense ions in
each high-resolution survey scan (unit resolution). CID was
conducted with an isolation width of 5 Da, normalized
collision energy of 35, activation q of 0.25 and an activation
time of 30 ms. Dynamic exclusion was conducted by utilizing
a repeat count of one prior to exclusion. Ions with unassigned
or multiple charge states were excluded from CID, as well as
known background ions. Total scan cycle time was about 1 s,
formed from 0.4 s high resolution scan and two 0.25 s data-
dependent MS/MS scans. The source parameters of 5000 V
for capillary voltage, 2758C of capillary temperature, sheath
gas (nitrogen) flow of 8 units and tube lens voltage of 90 V
were used. External calibration of the instrument was
conducted just before experiments. The instrument was
controlled by Xcalibur 2.0.7 software. The data was
processed both manually for expected metabolites, using
ion chromatograms with 20 mDa window, and by using
Thermo MetWorks software with MDF with 80 mDa
tolerance window.
With data from all instrument types, a detectable peak (S/
N ratio >3) in the sample (60 min incubation) that was not
present in the control samples (incubation without the study
compound and the 0 min sample with the study compound)
was considered as a positive hit. In addition, an accurate
mass criteria with <3 mDa difference to the calculated mass
value was set for TOFMS and the LTQ-Orbitrap. Also
adequate fragment ion data confirming the metabolite to be
in a similar structural family as the parent compound was an
additional criterion for both types of triple quadrupole
instruments. All positive hits were also evaluated manually
to exclude them from false positives.
Data from TOFMS and both types of triple quadrupole
instrument approaches were processed and mined for
metabolites as individual studies so that no information of
the found metabolites with one instrument was used with the
other instruments, thus providing the comparison of
metabolite screening possibilities of these instruments to
be more informative and directly comparable. When mining
the LTQ-Orbitrap data, the results from TOFMS were used as
a basis, to avoid the features and suitability of the used
software (MetaboLynx or MetWorks) having too large an
impact on the results between these two instrument types.
RESULTS AND DISCUSSION
Data processing methods and set criteria for
metabolite hits
Metabolite hits and detection sensitivity with
different mass spectrometers
The actual samples were compared with the control samples
without the study compound. Positive hits were also
compared manually with the 0 min samples to distinguish
them from possible impurities present with the study
compound. TOFMS data was mined using a MetaboLynx
XS subroutine of the MassLynx-software, the expected
metabolite list included the expected phase I and phase II
biotransformations for amitriptyline and verapamil to
undergo in a human liver microsomal incubation with the
used cofactors, i.e. hydroxylation, dihydroxylation, demethy-
lation,didemethylation,hydrogenation,dehydrogenationand
glucuronide conjugation, as well as all possible combinations
of the abovementioned reactions (combinations created by the
software itself). The dealkylation tool of the software was
enabled with a mass cut-off of 70Da allowing it to break two
bonds. The mass defect filter was enabled with a tolerance of
40 mDa. Expected mass chromatograms were created with a
mass window of 0.05 Da. For unexpected metabolites full
acquisition mass range chromatograms were created with a
1 Da window. The peak detection threshold for absolute peak
area was set to slightly overcome the normal noise level with
the instrument.
In total, 28 confirmed amitriptyline metabolites were found
and tentatively identified from 60 min in vitro HLM
incubations. All of these were found by the TOFMS
approach. The same samples analyzed with Q-Trap, QqQ
or LTQ-Orbitrap resulted in 11, 10 and 13 metabolite hits,
respectively. Ten of the Q-Trap metabolites were detected
using the SRM-ER-EPI-approach, and one additional metab-
olite was found by the acquisition of EMS-ER-EPI data.
Verapamil incubations resulted in 69, 21 and 24 metabolite
hits when analyzed with TOF, Q-Trap or QqQ, respectively.
No data with the LTQ-Orbitrap was acquired for verapamil.
Twenty of the verapamil Q-Trap metabolites were detected
using the SRM-ER-EPI-approach, and again one additional
metabolite was found by the acquisition of EMS-ER-EPI data.
Metabolite hits with various instruments for both of the
compounds are collected in Tables 2 and 3, including
identifications of biotransformations (and their sites where
applicable). Detailed LC/MS data for all detected metab-
olites are presented in Tables 4 and 5, including accurate
mass (in-source) fragment ion data from TOFMS and all low-
resolution fragment ion data from the other MS instruments.
Metabolic schemes with structures of the metabolites for both
compounds are presented in Figs. 1 and 2.
The precursor ion scan and neutral loss scan data and the
SRM/EPI-MS data obtained from triple quadrupole instru-
ments were processed manually. In the case of NL and PI
data, 20 different predefined extracted ion chromatograms
with amitriptyline and 42 with verapamil were used to cover
possible biotransformations. The LTQ-Orbitrap data was
processed both manually for expected metabolites, using ion
The number of hits correlated well with the assumption
that the detection sensitivity is clearly higher when TOFMS is
used instead of QqQ in PI or NL scanning mode, and that
certain unexpected metabolites are missed by the PI/NL or
SRM-EPI approach, as those are limited to only predefined
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

944 T. Rousu, J. Herttuainen and A. Tolonen
Table 2. The number of metabolite hits with different MS instruments for amitriptyline
Hits TOF
Hits QTrap Hits QqQ
Hits Orbitrap
Metabolite Biotransformation/Data processing
MetaboLynx þ manual
manual
manual
MetWorks þ manual
Parent
M1
M2
Unchanged
N-Demethylation
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
2 ꢁ N-Demethylation
M3
M4
Hydroxylation to ethylene bridge in central ring
Hydroxylation
M5
M6
Hydroxylation to ethylene bridge in central ring
Aromatic hydroxylation
Aromatic hydroxylation
Aromatic hydroxylation
Hydroxylation
x
x
x
x
x
x
x
x
x
x
x
x
x
M7
M8
x
M9
M10
M11
M12
M13
M14
M15
M16
M17
M18
M19
M20
M21
M22
M23
M24
Hydroxylation to area between N and central ring
Methyl hydroxylation
x
x
x
x
2 ꢁ Hydroxylation/N-Oxidation
2 ꢁ Hydroxylation/N-Oxidation
2 ꢁ Hydroxylation/N-Oxidation
2 ꢁ Hydroxylation/N-Oxidation
2 ꢁ Hydroxylation/N-Oxidation
Hydroxylation þ N-Demethylation
Hydroxylation þ N-Demethylation
Hydroxylation þ N-Demethylation
Hydroxylation þ N-Demethylation
Hydroxylation þ N-Demethylation
Hydroxylation þ N-Demethylation
2 ꢁ Hydroxylation þ N-Demethylation
Hydroxylation þ Dehydrogenation
(oxo to ethylene bridge)
x
x
x
x
x
x
M25
M26
Hydroxylation þ Dehydrogenation
Hydroxylation þ Dehydrogenation þ
N-Demethylation
N-Glucuronide conjugation
Hydroxylation þ Glucuronide conjugation
x
x
M27
M28
x
x
x
Table 3. The number of metabolite hits with different MS instruments for verapamil
Hits TOF
MetaboLynx þ manual
Hits QTrap
manual
Hits QqQ
manual
Metabolite
Biotransformation/Data processing
Parent
M1
M2
Unchanged
O-Demethylation (to ring A)
O-Demethylation (to ring A)
O-Demethylation (to ring B)
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
M3
M4
M5
N-Demethylation
2 ꢁ Demethylation (both to ring A or one to ring A
and N-Demethylation)
M6
2 ꢁ Demethylation (both to ring A or one to ring A
and N-Demethylation)
x
x
x
M7
M8
2 ꢁ Demethylation
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
3 ꢁ Demethylation
M9
Hydroxylation
M10
M11
M12
M13
M14
M15
M16
M17
M18
M19
M20
M21
Hydroxylation/N-oxidation to C17H25N2O2 area
Hydroxylation/N-oxidation to C17H25N2O2 area
Hydroxylation/N-oxidation
Hydroxylation/N-oxidation
Hydroxylation (to ring B or carbon next to it)
Hydroxylation/N-oxidation
Hydroxylation (to ring B or carbon next to it)
2 ꢁ Hydroxylation (to C10H12O2 area)
3 ꢁ Hydroxylation/N-oxidation
3 ꢁ Hydroxylation/N-oxidation
3 ꢁ Hydroxylation/N-oxidation
Hydroxylation to isopropyl þ N-Demethylation
x
x
x
x
x
x
x
x
x
x
(Continues)
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 945
Table 3. (Continued)
Hits TOF
MetaboLynx þ manual
Hits QTrap
manual
Hits QqQ
manual
Metabolite
Biotransformation/Data processing
M22
M23
M24
M25
M26
M27
M28
M29
M30
M31
M32
M33
M34
M35
M36
M37
M38
M39
M40
M41
M42
M43
M44
M45
M46
M47
M48
M49
M50
Methyl hydroxylation þ Demethylation
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Methyl hydroxylation þ Demethylation (not in ring B)
Hydroxylation þ Demethylation to C17H25N2O2 area
Hydroxylation þ Demethylation
x
x
x
x
Hydroxylation þ Demethylation
Hydroxylation to isopropyl þ O-Demethylation (to ring B)
Hydroxylation to C10H12O2 area þ N-Demethylation
Hydroxylation þ 2 ꢁ Demethylation
x
Hydroxylation þ 2 ꢁ Demethylation
Hydroxylation þ 2 ꢁ Demethylation
Hydroxylation þ 2 ꢁ Demethylation
Hydroxylation þ 2 ꢁ Demethylation
N-Dealkylation by loss of C10H12O2
x
x
x
x
x
N-Dealkylation þ O-Demethylation
N-Dealkylation þ O-Demethylation
N-Dealkylation þ N-Demethylation
N-Dealkylation þ 2 ꢁ Demethylation
N-Dealkylation þ Demethylation þ Hydroxylation
N-Demethylation þ Dehydrogenation
Demethylation þ Dehydrogenation
Demethylation þ Dehydrogenation
Demethylation þ 2 ꢁ Dehydrogenation
2 ꢁ Demethylation þ Dehydrogenation
2 ꢁ Demethylation þ Dehydrogenation
2 ꢁ Demethylation þ Dehydrogenation
2 ꢁ Demethylation þ Dehydrogenation
Hydroxylation þ Demethylation þ Dehydrogenation
Hydroxylation þ Demethylation þ Dehydrogenation
Hydroxylation þ Demethylation þ Dehydrogenation to
C17H25N2O2 area
x
M51
M52
M53
M54
M55
M56
M57
M58
M59
Hydroxylation þ Demethylation þ Hydrogenation
Cleavage of C17H25N2O2 þ Hydroxylation
Glucuronide conjugation
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
O-Demethylation (to ring A) þ Glucuronide conjugation
O-Demethylation (to ring B) þ Glucuronide conjugation
Demethylation þ Glucuronide conjugation
2 ꢁ Demethylation þ Glucuronide conjugation
2 ꢁ Demethylation þ Glucuronide conjugation
2 ꢁ Demethylation (both to ring A or one to ring A and
N-Demethylation) þ Glucuronide conjugation
Hydroxylation þ Glucuronide conjugation
Hydroxylation þ Demethylation þ Glucuronide conjugation
Hydroxylation þ Demethylation þ Glucuronide conjugation
Hydroxylation þ Demethylation þ Glucuronide conjugation
Hydroxylation þ Demethylation þ Glucuronide conjugation
Hydroxylation þ Demethylation þ Glucuronide conjugation
Hydroxylation þ Demethylation þ Glucuronide conjugation
Hydroxylation þ 2 ꢁ Demethylation þ Glucuronide conjugation
2 ꢁ Demethylation þ Dehydrogenation þ Glucuronide conjugation
N-Dealkylation þ Demethylation þ Glucuronide conjugation
x
x
x
M60
M61
M62
M63
M64
M65
M66
M67
M68
M69
x
x
x
x
x
x
x
x
x
x
x
expected metabolites, or metabolites that undergo similar
predefined fragmentation reactions. The main metabolites
were detected with all MS approaches. It is also worth
mentioning that as the Q-Trap is also a triple quadrupole
instrument with fully functional NL and PI scan modes, it
must be assumed that all metabolites detected here with the
traditional triple quadrupole instrument would have been
detected also with the Q-Trap by using the same NL/PI scan
approach. When comparing the hit rate with the triple
quadrupole and the Q-Trap instruments, it is also worth
noticing that a higher number of PI/NL reactions were used
with the triple quadrupole than fragmentation reactions for
predicting the used SRM transitions with the Q-Trap. Thus, if
the number of PI/NL reactions with triple quadrupole
instruments had been decreased so that only one LC/MS run
had been acquired (as with Q-Trap), the hit rate would have
been probably clearly decreased to be less than that with the
Q-Trap.
An example of the detection sensitivities of the four
different mass spectrometers used in this study is illustrated
in Fig. 3, in which extracted ion chromatograms (XICs) of the
didemethylated metabolite of amitriptyline (M2) are shown.
On the top is a XIC from LTQ-Orbitrap data (Fig. 3(A)), the
next XIC being from Q-Trap data (Fig. 3(B)) and the following
two XICs being from QqQ (Fig. 3(C)) and TOF (Fig. 3(D))
data, respectively. The figure shows that S/N ratio with
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

946 T. Rousu, J. Herttuainen and A. Tolonen
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 947
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

948 T. Rousu, J. Herttuainen and A. Tolonen
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 949
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

950 T. Rousu, J. Herttuainen and A. Tolonen
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 951
Figure 1. Proposed amitriptyline biotransformations and fragmentation.
TOFMS is about 15-fold in comparison to the other three
instruments, with which an approximately similar S/N ratio
was observed. It is however worth noticing that the signal
from the QqQ instrument is strongly skewed and defined
only by a couple of chromatographic data points, so that
probably even a very slight decrease in concentration would
have led to missing this metabolite. Moreover, the XIC from
LTQ-Orbitrap data is created using the same 50 mDa
window as that from TOFMS data, to observe some noise and
to be able to calculate the S/N ratio. Use of a smaller XIC
window size would have cut down the noise from both
TOFMS and LTQ-Orbitrap data practically completely,
making it more difficult to evaluate the S/N ratio. Thus,
the sensitivity difference between TOFMS and the LTQ-
Orbitrap is somewhat overestimated in this figure, but the
data however shows a better sensitivity with TOFMS with
respect to other tested MS instruments.
Identification of metabolites and proposed
fragmentation pathways
Biotransformations for all found metabolites were tentatively
identified with accurate mass obtained from TOFMS/
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

952 T. Rousu, J. Herttuainen and A. Tolonen
Figure 2. Proposed verapamil biotransformations and fragmentation.
LTQ-Orbitrap data, and accurate mass in-source fragment
ion data from TOF experiments were used together with low-
resolution fragment ion data obtained from QqQ, Q-Trap
and LTQ-Orbitrap (for amitriptyline only) experiments to
further elucidate the metabolite structures. All accurate mass
and fragment ion data obtained for amitriptyline and
verapamil are collected in Tables 4 and 5, respectively. Very
good mass accuracies were obtained for all molecular ions
and their in-source fragments detected with TOFMS, as in
general the differences between calculated and measured m/z
values were below 2 mDa.
(m/z 276) points to the hydroxylation site being either in the
ethylene bridge or in the C2H4 part of the side chain (non-
aromatic or non-double-bond carbon), and the fragment ions
at m/z 84 point to an unmodified side chain in both M3 and
M5. Also the ions at m/z 216/215 and 153, being 2 or 3 mass
units smaller than the corresponding ions with the parent,
after loss of water from the hydroxylated fragment, point to
the hydroxylation site being in this ethylene bridge area.
These two ethylene bridge hydroxylations are indeed well
known in amitriptyline metabolism.29 The contradiction in
biotransformation site elucidation is however raised by the
M3 and M5 fragment ions at m/z 205, 191, and 203, which are
the same as the parent, and, if formed via an identical
fragmentation pathway (see Fig. 1), this would point to the
biotransformation site being the side chain. If figuring out
the fragmentation map for the first two of these, it can be
suggested that cleavages of water and unmodified side chain
in M3 and M5 can lead to fragment ions at m/z 205 and 191
that are isobaric to amitriptyline fragments, even though
having slightly different structure. The ion at m/z 203 is
however more problematic, as the only identification we
could come up for that is loss of H2 from m/z 205, meaning
that in the case of ethylene bridge-hydroxylated metabolites
the ion at m/z 203 should be preceded by an ion at m/z 221
(m/z 205þ16), that was not detected here. The ion at m/z 231
may be formed from both possible abovementioned hydroxyl-
ation sites as well, and do not elucidate the biotransformation
site more exactly. One additional suggestion to the ethylene
For amitriptyline, the clear main metabolic route was N-
demethylation (M1), which had about 65% share of the total
combined metabolite LC/MS peak area (in TOFMS data). In
addition, nine hydroxylated (or N-oxidized) metabolites
(M3–M11) were the only metabolites formed via single
biotransformation reaction only. After this, all other detected
metabolites were formed as a combination of several
biotransformation reactions, i.e. hydroxylations/N-oxida-
tions, one or two N-demethylations, dehydrogenation, along
with N- or O-glucuronide conjugations. Both mono- and di-
demethylated M1 and M2 produced the same fragment ions
as the parent molecule.
Of the hydroxyl metabolites M3–M11, in M3 and M5 the
biotransformation sites were identified to the ethylene bridge
between two aromatic rings, even though this will lead to
some inconsistencies with regard to the fragment ion
identification. The presence of very intense loss of water
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 953
Figure 3. Extracted ion chromatograms of di-demethylated M2 of amitriptyline from LTQ-
Orbitrap (A), Q-Trap (B), QqQ (C), and TOF (D) data illustrating the detection sensitivity of
different MS instruments in this study.
bridge as a hydroxylation site is M24 (see below), with clearer
of the precursor. In M7 the hydroxylation was located to the
three-ring area based on fragment ions m/z 207 (m/z 191 þ 16)
and 84, and the fragment ion at m/z 121 (m/z 105 þ 16)
narrows the hydroxylation site to an aromatic ring. In M8 the
fragment ions at m/z 207 (m/z 191 þ 16), 107 (m/z 91 þ 16) and
91 suggest also aromatic hydroxylation. In M10 the fragment
ion at m/z 249, by loss of C2H7N, and the fragment ion at m/z
205, pinpoint the hydroxylation site to the three-carbon area
between the nitrogen atom and the central ring, and the lack
of a fragment ion by loss of water suggests that the site may
identification of the formation of an oxo to ethylene bridge
area (possibly by further dehydrogenation of M3 or M5).
Despite this somewhat ambiguous fragment ion identifi-
cation, more experiments were not conducted, as the main
issue of the work was to compare the instrumentation in one
lab visit only, and not the actual comprehensive identifi-
cation of amitriptyline metabolism.
Aromatic hydroxylation in M6 was identified with
fragment ions þ16 mass units compared to fragment ions
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

954 T. Rousu, J. Herttuainen and A. Tolonen
most probably be the carbon next to the ring. This however
requires that the proton/hydroxyl attached to this carbon is
cleaved in the reaction forming the ion at m/z 205. The
cleavage of CH2O to produce a fragment ion at m/z 264 was
used to identify M11 as methylhydroxylation instead of N-
oxidation. For metabolites M12–M23 and M25–M26, no
fragment ion data (or only a few less informative fragments)
was obtained due to their low abundance, and they
were identified based only on the accurate mass data for
the molecular ion. In M24, the formation of an oxo group to
the ethylene bridge after hydroxylation and dehydrogena-
tion reactions were suggested by the presence of fragment
ions þ14 mass units compared to the precursor (m/z 232, 219
and 205 with respect to m/z 218, 205 and 191). Also the
fragment ion at m/z 84 suggests the cleavage of an
unmodified side chain in M24. The fragment ion at m/z
191 of M24 was then suggested to be formed via cleavage of
the side chain and loss of the keto-oxygen as H2O. M27 was
identified as an N-glucuronide conjugate, as no other site for
conjugation is available. M28 was formed by hydroxylation
and further glucuronide conjugation. The observed fragment
ions did not reveal the site for the hydroxylation nor the
conjugation. The proposed biotransformation reactions and
their sites (for amitriptyline metabolites to which fragment
ion data was obtained) are shown in Fig. 1.
the fragment ion at m/z 165. In M27 the fragment ions at m/z
396 and 245 locate the hydroxylation to isopropyl and the O-
demethylation to ring B. In M28, fragment ions at m/z 181 and
260 suggest a hydroxylation to C10H12O2 area and an N-
demethylation. Metabolites M35–M39 were further metab-
olites of N-dealkylated M34. The fragment ion at m/z 246 for
M35 and at m/z 260 for M37 differentiate the O- and N-
demethylations. M40 and M50 were the only dehydroge-
nated verapamil metabolites for which fragment ion data
was obtained. The fragment ion at m/z 260 along with the ion
at m/z 289 (m/z 303 – 14) resolved the N-demethylation in
M40. For M50, the fragment ion at m/z 165 shows that the ring
B area is intact. Metabolites from M53 to M69 were identified
as glucuronide conjugates. For M54 the fragment ions at m/z
165 (unmodified ring B), 289 (m/z 303 – 14) and 246 (m/z 260 –
14) identify the O-demethylation to ring A. For M55, the
fragment ions at m/z 291 and 260, together with an ion at m/z
151 (m/z 165 – 14) locates the O-demethylation to ring B. For
M59, the two demethylations were either both in ring A, or
the other is N-demethylation. For M60, the fragment ions
at m/z 260 and 165 suggest the additional oxygen atom to
be either N-methylhydroxylation or N-oxidation, which is
followed by further glucuronide conjugation. The proposed
biotransformation reactions and their sites (for the verapamil
metabolites for which fragment ion data was obtained) are
shown in Fig. 2.
For verapamil, the main metabolic routes were N-
demethylation (M4) and N-dealkylation by loss
of C10H12O2 (M34), having about 40% and 33% share of
the total combined metabolite peak area (from LC/TOFMS
data), respectively. Metabolites M1–M8 were formed via one
or more N- and/or O-demethylations, M9–M20 via one to
three hydroxylations/N-oxidations, M34 via N-dealkylation,
and all the remaining metabolites were formed via
combinations of these presented reactions or additional
dehydrogenation or hydrogenation reactions, along with N-
or O-glucuronide conjugations. For the demethylated
metabolites, M1, M2 and M4, the presence of fragment ions
at m/z 165 and 150, corresponding to unmodified ring B,
suggests that the biotransformations have occurred as O-
demethylations in ring A and/or as N-demethylation. Of
these M4 with the fragment ion at m/z 260 was identified as
N-demethylation. Similarly, the fragment ion at m/z 165
observed with M5 and M6 suggests that ring B is unmodified,
and thus that demethylations have occurred either both in
ring A or one in ring A and the other being N-demethylation.
Of the hydroxy/N-oxide metabolites M9–M16, M10 and M11
had fragment ions at m/z 165 and 150, pinpointing the
biotransformation sites to the C17H25N2O2 area. The frag-
ment ions at m/z 181 and 166 for M14 and M16 show that the
hydroxylation site is in either ring B or the carbon atom next
to it. In M17, dihydroxylation is suggested to be located in
the C10H12O2 area, based on the fragment ions at m/z 303 and
260. In M21 the cleavage of C3H7O, resulting in the fragment
ion at m/z 399, pinpoints the hydroxylation to the isopropyl
group, and the fragment ion at m/z 220 shows that the
demethylation is located on the nitrogen atom. In M22 and
M23 the fragment ion by cleavage of methanol suggests that
the hydroxylations have occurred to the methyl group. In
M24 the hydroxylation or N-oxidation and demethylation
were pinpointed to the C17H25N2O2 area by the presence of
As the main objective of this comparison is to test the
feasibility of different MS instruments in metabolite screen-
ing and evaluate the data obtained with the instruments,
the identification of the metabolites and biotransformation
sites was not complete. Thus, we did not try to confirm the
biotransformation sites by other available methods, nor were
any of the identified metabolites purchased or synthesized.
Aspects for obtained structural data
Accurate mass data was obtained for all of the identified
metabolites with TOF and LTQ-Orbitrap instruments, and in
general the measured masses differed from the calculated
monoisotopic masses by less than 2 mDa. For some of the
minor metabolites or fragment ions with very low abun-
dances slightly poorer mass accuracies were obtained. As
expected, the overall mass accuracy for molecular ions was
slightly better in LTQ-Orbitrap data (survey scan) in
comparison to TOFMS data, but also the TOFMS mass
accuracy was good enough and did not leave more than
one possible identification of molecular formula in any case,
i.e. no real meaningful difference was observed between the
instruments. Moreover, it is worth keeping in mind that
recent advances in TOF mass spectrometers, i.e. at least two
instrument vendors now market Q-TOFs capable of mass
resolution of 40 000 (FWHM) and very high mass accuracy,
have narrowed the gap in comparison to the LTQ-Orbitrap.
The quality and quantity of the obtained fragment ion data
varied between different MS instruments, as expected, and
all of them had strengths and weaknesses. The lowest
quantity of fragment ions was obtained with TOF, being the
only instrument lacking a collision cell. Even though quite
good elucidation of the biotransformation sites for the main
metabolites was observed also from TOFMS data, a number
of fragment ions important for more accurate structural
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 955
elucidation were missed in the TOFMS approach with
respect especially to the Q-Trap. This disadvantage would
have naturally been overcome by using the Q-TOF instru-
ment instead of plain TOF, but that was not available for this
study. Moreover, this might have decreased the TOF
detection sensitivity due to the lower overall ion trans-
mission to the detector in the Q-TOF with respect to TOF. The
obvious strength of TOFMS is that accurate mass data was
obtained also for fragment ions, which was not the case with
the LTQ-Orbitrap, with which the use of high-resolution
detection mode for fragment ions would have led to
absolutely too long scan cycle times. The accurate mass
information for fragment ions is much emphasized in
structural elucidation, as often the fragment ion identifi-
cations may be easily erroneous without this data.32 With
QqQ and the Q-Trap we obtained pretty much equal
amounts of fragment ion data. However, we used a high
number of PI and NL functions with the QqQ, leading to
several LC/MS runs, whereas the Q-Trap was used only in
one run per sample approaches, meaning that if more runs
had been used to utilize the ion trap feature of the Q-Trap, the
number of fragment ions obtained could have easily been
even higher. This is naturally the case also if the ion trap
feature of the LTQ-Orbitrap would have been used more
with additional LC/MS runs. Even though providing high
amounts of fragment ion data, the serious drawback of QqQ
as well as the Q-Trap instrument was naturally the inability
to produce accurate mass data, which has a very important
role in confirming the changes in molecular formulae due to
biotransformation reactions.
from the first injection, to overcome the interference of matrix
constituents in fragment ion data acquisition. Moreover, they
showed that with proper settings also the data acquisition
speed of the LTQ-Orbitrap could be improved, even though
their data still contained 0.3 min broad chromatographic
peaks with long chromatographic run times.
Aspects for the workload of different
approaches
The time consumption for the described analyses was clearly
the lowest when using the TOF instrument. With this
approach only one injection/sample is needed and all of
the expected as well as all of the unexpected metabolites are
detected within a single run without any need for metabolite
prediction for certain special metabolites, or elucidation of
parent compound fragmentation pathways before analysis,
thus being very applicable for high-throughput analysis.
Based on the high amount of data from a single run, it is
laborious to mine metabolites from the total ion current data,
but fortunately very good software exists to ease the process.
In most cases, and to guarantee the results – practically
always, there is a need for manual expert evaluation and
confirmation of the software-produced results, and therefore
the whole process cannot be totally automated, especially
with regard to the unexpected metabolites. With this
approach the total workload, after finding suitable LC/MS
conditions (gradient strength and pH in chromatography), is
typically about 30 min of instrument time and 1–4 h of
data processing per compound/species, depending on
the number of metabolites and on the complicity of the
interpretation of the fragment ion data. However, some of
the recently described software may also take care of the
fragment ion data interpretation, especially if a Q-TOF
instrument is used.27
With LTQ-Orbitrap experiments high-resolution full scan
was used as a survey scan in conjunction with data-
dependent acquisition of low-resolution MS/MS spectra
for the two most abundant ions. With this kind of setup a
total cycle time of around 1 s was obtained, which produced
adequate structural information for the main metabolites in
this study. The amount and quality of the MS/MS data
obtained with the LTQ-Orbitrap for the minor metabolites
was however rather poor, as the measured MS/MS data was
mainly produced by ions from matrix constituents, despite
the use of a comprehensive background ion exclusion list.
This disadvantage could have been overcome by a more
thorough setup of the inclusion list for expected metabolite
ions, so that those ions are chosen for CID when present, even
if more abundant ions were detected in the same survey scan.
Also an approach similar to the one described by Ruan
et al.,13 i.e. using mass resolution of 7500 for MS/MS events,
was tested but this resulted in a scan cycle exceeding 2 s that
was clearly too long and led to missing metabolites due
to inadequately defined chromatographic peak shapes and
decreased sensitivity, even though a slower chromato-
graphic method with broader peak widths was used in
LTQ-Orbitrap experiments with respect to other instru-
ments. Despite this disadvantage some approaches with high
mass resolution for fragment ion data acquisition, suffering
from slow data acquisition rates, have been described. For
example, a total duty cycle of even 5.6 s has been reported.12
Recently, Li et al.33 described an two-injection approach,
where the other injection with precursor-specific MS/MS
experiments were conducted after interpretation of the data
The experiments with the triple quadrupole consumed a
lot of instrument time and also a high volume of sample.
Amitriptyline produced ten abundant fragment ions under
CID and therefore ten different PI scan and 11 different NL
scan functions were acquired, thus 21 injections for each of
blank, 0 min sample and 60 min sample were conducted.
Twenty different traces were processed so that over 1000
extracted ion chromatograms were evaluated manually in
order to mine the data for metabolites. The processing of
verapamil data also involved a manual evaluation of over
1000 ion chromatograms as 42 traces were checked from nine
scan functions within three samples. In total the acquisition
of the data was conducted overnight and the metabolites
were mined manually in a time frame of a few working days.
The workload of the Q-Trap experiments lies somewhere
in between those of TOF and QqQ experiments. With the
used SRM-ER-EPI and EMS-ER-EPI approaches the most
laborious part is the prediction of possible biotransform-
ations and thus the generation of predicted SRM/MRM
reactions, even though the construction of compound-
specific predicted SRM-IDA methods can also be automated
by using Applied Biosystems/MDS SCIEX LightSight soft-
ware.
In the case of the LTQ-Orbitrap experiments with the
approach where only low-resolution MS/MS was per-
formed, the workload of the acquisitions themselves is also
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

956 T. Rousu, J. Herttuainen and A. Tolonen
low – as only one injection per sample was used. Due to slow
data acquisition rate about twice as many runs were needed
in comparison to other instruments. In data processing, times
quite similar to TOFMS would be expected, as a similar type
of software-aided data processing was used. Because of our
greater experience in using the MetaboLynx software
(Waters, for TOF) with respect to MetWorks (Thermo-
Scientific, for LTQ-Orbitrap), we wanted to avoid differences
in results due to this, and used the TOFMS-detected
metabolites as a list of expected metabolites when going
through the LTQ-Orbitrap data.
use of the ion mobility feature must be mentioned, as
described in metabolite profiling recently by Chan et al.34
In addition to the used approaches in Q-Trap experiments,
the use of PI/NL scan functions can be used as a survey scan
to trigger the acquisition of EPI spectra, as mentioned earlier.
Also recently, yet one more alternative Q-Trap approach was
introduced and applied to drug metabolite screening,
utilizing monitoring of multiple predicted metabolite ions
(MIM) in both quadrupoles with minimized fragmentation,
and using it to trigger the acquisition of EPI spectra.22,35
Selectivity of this methodology was similar to that of the SRM
approach, and the sensitivity to obtain MS/MS data was
improved in comparison to full scan approaches.
Comparison with literature data and possible
alternative screening methods
CONCLUSIONS
For amitriptyline, all the known main metabolites, and also
most of the minor known metabolites,29 were detected and
tentatively identified by their accurate masses and fragment
ion data. Many of the metabolites were of very low
abundance and no fragment ion data was obtained for them
with any of the used MS approaches. Therefore, the
comparison of metabolites detected here with earlier
reported amitriptyline metabolites is quite rough, e.g. with
respect to biotransformation sites that remained unclear in
many cases. Recently, a study of verapamil metabolism
conducted with a newer and more sensitive version of the Q-
Trap mass spectrometer (AB SCIEX QTRAP 5500) was
reported, leading to a total of 35 detected verapamil
metabolites from a fresh rat hepatocyte incubation spiked
into rat bile.30 That study used a two-injection methodology
(predicted SRM using LightSight software and dual
precursor and neutral loss survey scan) with 158 predicted
SRM reactions generated from one parent compound
fragmentation reaction (m/z 455 > 165). This shows that even
the use of a very high number of predicted SRM reactions
with the Q-Trap did not lead to as high a hit rate as in the
TOFMS-based approach used here, but naturally it must
be noted that also the used enzyme source was different (rat
hepatocytes). In our own other recent study with a clearly
less sensitive older generation TOF instrument than the one
used in this study, the number of detected metabolites was
higher with rat liver microsomal incubation in comparison to
human liver microsomal incubation (8 vs. 7).31
The time-of-flight mass spectrometer with high scanning
sensitivity and rapid data acquisition speed was shown to
have a superior performance over the other instrument types
used here in early phase ’one lab visit only’ metabolite
profiling. The very high wide-mass range detection sensi-
tivity allows easy simultaneous detection of both expected
and unexpected metabolites in a single run, acquiring also
fragment ion data at the same time. Moreover, the approach
does not require any metabolite prediction or pre-adjusted
detection reactions or changes in settings between different
test compounds (except parameters with respect to ionis-
ation properties), which was the case also when the LTQ-
Orbitrap mass spectrometer was used in the analyses.
Both TOF and LTQ-Orbitrap instruments showed ade-
quate mass accuracies to elucidate the change in molecular
formula by metabolism, even though the mass accuracy of
the LTQ-Orbitrap was slightly better. The drawback of the
LTQ-Orbitrap with respect to TOFMS was its low data
acquisition rate, because much longer LC/MS run times with
poorer chromatographic resolution had to be used. In
addition, the LTQ-Orbitrap was shown to have lower
detection sensitivity in full scan-type analysis, which led
to false negatives for many metabolites that the TOFMS
approach was able to find. Considering the fragment ion
data, the LTQ-Orbitrap was able to produce a higher number
of different fragment ions in comparison to the TOFMS
approach, but due to its low data acquisition speed the low-
resolution mode had to be used to acquire the fragment ion
data, and thus accurate masses were not obtained for them.
This in turn suggests that the more evident strengths of the
LTQ-Orbitrap instrument lie in the later phases of drug
development studies, where the more accurate structural
elucidation has an increased value, and time used per sample
is less critical, meaning that the LTQ-Orbitrap can be used to
acquire several LC/MS runs for one sample (or even one
metabolite) with high-resolution MS/MS settings and even
MSn data.
With TOFMS we also tested a different approach where the
fragment ion data was obtained with two injections of the
sample, separating the acquisition of low and high aperture
voltage data with 150 ms scan time into two separate runs.
This data was then evaluated against a single injection
protocol with two parallel scan functions with 100 ms scan
time. The peak areas for abundant main metabolites in
extracted ion chromatograms with two parallel functions
were about 50–80% of the peak areas when using only one
function with longer scan time. For peaks of very low
abundance (S/N ratio around 10 or less) the relative size of
peaks in the two data function run were about 80–100% of
those in separate one function runs, leading thus to an equal
number of metabolite hits regardless of the approach
used (data not shown). If yet alternative TOFMS-based
approaches are searched, in addition to earlier mentioned
latest generation very high resolution Q-TOF instruments, a
The approaches used here with the Q-Trap and triple
quadrupole mass spectrometers produced the highest
amount of fragment ion data for structural elucidation, but
the low sensitivity and very high time consumption with
the triple quadrupole instrument made the approach very
laborious, being neither time- nor cost-effective, and also
provided the lowest number of detected metabolites. Also
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm

Comparison of MS instruments in metabolite screening 957
the SRM/EPI approach with the Q-Trap instrument required
laborious prediction of metabolites and their SRM reactions,
and even with careful selection of these reactions a number of
unexpected metabolites were missed.
12. Lim H-K, Chen J, Sensenhauser C, Cook K, Subrahmanyam
V. Rapid Commun. Mass Spectrom. 2007; 21: 1821.
13. Ruan Q, Peterman S, Szewc MA, Ma L, Cui D, Humphreys
WG, Zhu M. J. Mass Spectrom. 2008; 43: 251.
14. Liu DQ, Hop CECA. J. Pharm. Biomed. Anal. 2005; 37: 1.
15. Hakala KS, Kostiainen R, Ketola RA. Rapid Commun. Mass
Spectrom. 2006; 20: 2081.
16. Prakash C, Soliman V. Drug Metab. Dispos. 1997; 25: 1288.
17. Hager JW. Rapid Commun. Mass Spectrom. 2002; 16: 512.
18. Hopfgartner G, Husser C, Zell M. J. Mass Spectrom. 2003; 38:
138.
Acknowledgements
The expert technical assistance of Dr. Yue Xuan from Ther-
moFisher Scientific is gratefully acknowledged for the LTQ-
Orbitrap measurements. We are also grateful to Mrs Katri
Sunnari and Mrs Birgitta Paldanius for technical assistance.
19. Hopfgartner G, Varesio E, Tschappat V, Grivet C, Bourgogne
¨
¨
E, Leuthold LA. J. Mass Spectrom. 2004; 39: 845.
20. Li AC, Alton D, Bryant MS, Shou WZ. Rapid Commun. Mass
Spectrom. 2005; 19: 1943.
21. Li AC, Shou WZ, Mai TT, Jiang X. Rapid Commun. Mass
Spectrom. 2007; 21: 4001.
REFERENCES
1. Pelkonen O, Turpeinen M, Uusitalo J, Rautio A, Raunio H.
Basic Clin. Pharmacol. Toxicol. 2005; 96: 167.
22. Yao M, Ma L, Duchoslav E, Zhu M. Rapid Commun. Mass
Spectrom. 2009; 23: 1683.
2. Coecke S, Ahr H, Blaauboer BJ, Bremer S, Casati S, Castell J,
Combes R, Corvi R, Crespi CL, Cunningham ML, Elaut G,
Eletti B, Freidig A, Gennari A, Ghersi-Egea JF, Guillouzo A,
Hartung T, Hoet P, Ingelman-Sundberg M, Munn S, Janssens
W, Ladstetter B, Leahy D, Long A, Meneguz A, Monshouwer
M, Morath S, Nagelkerke F, Pelkonen O, Ponti J, Prieto P,
Richert L, Sabbioni E, Schaack B, Steiling W, Testai E, Vericat
JA, Worth A. Altern. Lab Anim. 2006; 34: 49.
23. Xia Y-Q, Miller JD, Bakhtiar R, Franklin RB, Liu DQ. Rapid
Commun. Mass Spectrom. 2003; 17: 1137.
24. Anari MR, Sanchez RI, Bakhtiar R, Franklin RB, Baillie TA.
Anal. Chem. 2004; 76: 823.
25. Zhang H, Zhang D, Ray K, Zhu M. J. Mass Spectrom. 2009; 44:
999.
26. Pelander A, Tyrkko E, Ojanpera I. Rapid Commun. Mass
¨
¨
Spectrom. 2009; 23: 506.
3. Leclercq L, Cuyckens F, Mannens GSJ, de Vries R, Timmer-
man P, Evans DC. Chem. Res. Toxicol. 2009; 22: 280.
4. Walker D, Brady J, Dalvie D, Davis J, Dowty M, Duncan JN,
Nedderman A, Obach RS, Wright P. Chem. Res. Toxicol. 2009;
22: 1653.
27. Leclercq L, Mortishire-Smith RJ, Huisman M, Cuyckens F,
Hartshorn MJ, Hill A. Rapid Commun. Mass Spectrom. 2009;
23: 39.
28. Walles M, Thum T, Levsen K, Borlak J. J. Chromatogr. B 2003;
798: 265.
5. FDA Guidance for Industry. Safety Testing of Drug Metab-
olites, 2008. Available: http://www.fda.gov/downloads/
Drugs/GuidanceComplianceRegulatoryInformation/
GGuidance/ucm079266.pdf.
6. Tolonen A, Turpeinen M, Pelkonen O. Drug Discov. Today
2009; 14: 120.
29. Breyer-Pfaff U. Drug Metab. Rev. 2004; 36: 723.
30. Liu H-F, Jones E. Applied Biosystems Technical Note:
Breakthrough in vivo metabolite analysis using the
AB Sciex QTRAP 5500 System and Predictive MRM.
September 2009; 16.
31. Pelkonen O, Tolonen A, Rousu T, Tursas L, Turpeinen M,
Hokkanen J, Uusitalo J, Bouvier d’Yvoire M, Coecke S.
ALTEX 2009; 3: 214.
7. Tiller PR, Yu S, Castro-Perez J, Fillgrove KL, Baillie TA. Rapid
Commun. Mass Spectrom. 2008; 22: 1053.
8. Mortishire-Smith RJ, O’Connor D, Castro-Perez J, Kirby J.
Rapid Commun. Mass Spectrom. 2005; 19: 2659.
32. Hill AW, Mortishire-Smith RJ. Rapid Commun. Mass Spec-
trom. 2005; 19: 3111.
9. Cuyckens F, Hurkmans R, Castro-Perez J, Leclercq L, Morti-
shire-Smith RJ. Rapid Commun. Mass Spectrom. 2009; 23: 327.
10. Wrona M, Mauriala T, Bateman KP, Mortishire-Smith RJ,
O’Connor D. Rapid Commun. Mass Spectrom. 2005; 19: 2597.
11. Hu Q, Noll RJ, Li H, Makarov A, Hardman M, Cooks RG.
J. Mass Spectrom. 2005; 40: 430.
33. Li AC, Ding J, Jiang X, Denissen J. Rapid Commun. Mass
Spectrom. 2009; 23: 3003.
34. Chan ECY, New LS, Yap CW, Goh LT. Rapid Commun. Mass
Spectrom. 2009; 23: 384.
35. Yao M, Ma L, Humphreys WG, Zhu M. J. Mass Spectrom.
2008; 43: 1364.
Copyright # 2010 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2010; 24: 939–957
DOI: 10.1002/rcm