High-performance isotope labeling for profiling carboxylic acid-containing metabolites in biofluids by mass spectrometry

Article abstract of DOI:10.1021/ac102146g

We have developed a new isotope labeling method, based on the use of isotope-coded p-dimethylaminophenacyl (DmPA) bromide as a reagent, combined with liquid chromatography-mass spectrometry (LC-MS) for high-performance metabolome analysis with a focus on

Full text of DOI:10.1021/ac102146g

Anal. Chem. 2010, 82, 8789–8793
Letters to Analytical Chemistry
High-Performance Isotope Labeling for Profiling
Carboxylic Acid-Containing Metabolites in Biofluids
by Mass Spectrometry
Kevin Guo and Liang Li*
Department of Chemistry, University of Alberta, Edmonton, Alberta T6G 2G2 Canada
We have developed a new isotope labeling method, based
on the use of isotope-coded p-dimethylaminophenacyl
(DmPA) bromide as a reagent, combined with liquid
chromatography-mass spectrometry (LC-MS) for high-
performance metabolome analysis with a focus on profil-
ing carboxylic acid-containing metabolites. Derivatization
is simple, fast (1 h plus 30 min for quenching the
reaction), and applicable to a wide range of carboxylic
acids with a high yield and little or no side reaction
products. This labeling method is demonstrated to be not
only effective in introducing an isotope tag for accurate
metabolite quantification but also improving the chro-
matographic retention of the metabolites in reversed-
phase (RP) LC, enhancing ESI efficiency by 2-4 orders
of magnitude, and facilitating the identification of metabo-
lite peaks in LC-MS. In triplicate experiments of a 1:1
ratio of 13C-/12C-DmPA labeled human urine, we were
able to detect 2671, 2546, and 2820 ion pairs from
metabolites containing one or more carboxylic acid
groups.
combination with liquid chromatography (LC)-MS, for profiling
the metabolites containing carboxylic acid moieties in a biological
sample including body fluids such as urine.
Isotope labeling of chemicals has been widely used for
quantitative analysis of targeted molecules such as drug metabo-
lites by LC-MS.4 By overcoming matrix and ion suppression
effects, the use of an isotope internal standard for the analyte of
interest often provides high accuracy of quantitative measurement.
However, at present, it is not practical to generate individual
isotope standards for all the metabolites of a biological system or
the metabolome. An alternative approach is to use a chemical
reaction to introduce a mass tag to all the metabolites with a
common reactive moiety in a sample and apply the same
derivatization reaction to the standards with a differential mass
tag (e.g., 12C-labeling for the sample and 13C-labeling for the
standards).5 After the labeled sample and the mass-differentially
labeled standards are mixed, LC-MS is carried out and, based
on the accurate mass, retention time, and relative peak
intensities of ion pairs detected in the mass spectra, tentative
identification and absolute quantification of metabolites in the
sample can be attained. With the use of additional information
such as MS/MS spectral matches, positive metabolite identi-
fication can be made. Without the availability of standards,
information on the relative quantification of metabolites in
different samples can still be obtained.
Metabolomics is a rapidly evolving field for studying biological
systems and discovering potential disease biomarkers.1 For any
metabolomics application, metabolome analysis with adequate
sensitivity and specificity is essential in defining the metabolome.
Ideally, all metabolites present in a biological system are quali-
tatively and quantitatively profiled. Unfortunately, because of
technical limitations, only a fraction of metabolites are currently
analyzed by using techniques such as NMR and mass spectrom-
etry (MS).2,3 Because of limited metabolome coverage, many
important metabolome networks and some subtle changes in the
metabolome may not be revealed with current techniques. Herein
we report a high-performance isotope labeling strategy, in
While chemical derivatization is effective in introducing a mass
tag to the metabolites for LC-MS,5-13 it can also provide an
opportunity to enhance the performance of the overall LC-MS
(4) Ma, S. G.; Chowdhury, S. K.; Alton, K. B. Curr. Drug Metab. 2006, 7,
503–523
(5) Guo, K.; Li, L. Anal. Chem. 2009, 81, 3919–3932
(6) Ji, C. J.; Li, W. L.; Ren, X. D.; El-Kattan, A. F.; Kozak, R.; Fountain, S.;
Lepsy, C. Anal. Chem. 2008, 80, 9195–9203
.
.
.
(7) O’Maille, G.; Go, E. P.; Hoang, L.; Want, E. J.; Smith, C.; O’Maille, P.;
Nordstrom, A.; Morita, H.; Qin, C.; Uritboonthai, W.; Apon, J.; Moore, R.;
* Corresponding author. E-mail: Liang.Li@ualberta.ca.
(1) Vinayavekhin, N.; Homan, E. A.; Saghatelian, A. ACS Chem. Biol. 2010,
Garrett, J.; Siuzdak, G. Spectrosc.: Int. J. 2008, 22, 327–343
(8) Shortreed, M. R.; Lamos, S. M.; Frey, B. L.; Phillips, M. F.; Patel, M.;
Belshaw, P. J.; Smith, L. M. Anal. Chem. 2006, 78, 6398–6403
.
5, 91–103
.
.
(2) Holmes, E.; Loo, R. L.; Stamler, J.; Bictash, M.; Yap, I. K. S.; Chan, Q.;
Ebbels, T.; De Iorio, M.; Brown, I. J.; Veselkov, K. A.; Daviglus, M. L.;
Kesteloot, H.; Ueshima, H.; Zhao, L. C.; Nicholson, J. K.; Elliott, P. Nature
(9) Abello, N.; Geurink, P. P.; Van Der Toorn, M.; Van Oosterhout, A. J. M.;
Lugtenburg, J.; Van Der Marel, G. A.; Kerstjens, H. A. M.; Postma, D. S.;
Overkleeft, H. S.; Bischoff, R. Anal. Chem. 2008, 80, 9171–9180
(10) Eggink, M.; Wijtmans, M.; Ekkebus, R.; Lingeman, H.; De Esch, I. J. P.;
Kool, J.; Niessen, W. M. A.; Irth, H. Anal. Chem. 2008, 80, 9042–9051
(11) Carlson, E. E.; Cravatt, B. F. J. Am. Chem. Soc. 2007, 129, 15780
(12) Johnson, D. W. Clin. Biochem. 2005, 38, 351–361
.
2008, 453, 396–U350
(3) Scalbert, A.; Brennan, L.; Fiehn, O.; Hankemeier, T.; Kristal, B. S.; Van
.
.
Ommen, B.; Pujos-Guillot, E.; Verheij, E.; Wishart, D.; Wopereis, S.
.
Metabolomics 2009, 5, 435–458
.
.
10.1021/ac102146g  2010 American Chemical Society
Published on Web 10/14/2010
Analytical Chemistry, Vol. 82, No. 21, November 1, 2010 8789

DmPABr. For labeling human urine samples, urine in 50% (v:
v) of acetonitrile was centrifuged for 10 min at 12 000 rpm.
About 50 µL of human urine in acetonitrile (v:v) or 50 µL of
carboxylic acid standards in acetonitrile (1.2 mM each) were
mixed with an equal volume of 750 mM of triethanolamine
(TEOA) in a reaction vial. The solutions were vortexed, spun
down, and then mixed with 50 µL of freshly prepared 13C-DmPA
(20 mg/mL) (for heavy labeling) or 12C-DmPA (20 mg/mL)
(for light labeling). The derivatization reaction proceeded for
60 min at 90 °C in a water bath. After 60 min, the mixtures
were vortexed, spun down, and 100 µL of triphenylacetic acid
(30 mg/mL) was added to consume the excess labeling
reagent, 12C-/13C-DmPA, at 90 °C in TEOA for 30 min. The
solutions were vortexed and spun down. The labeling reactions
were carried out in sealed glass vials with Teflon lined caps.
After labeling, the 13C-labeled mixtures were combined with
their 12C-labeled mixtures for LC-MS analysis. We note that
the labeled sample was usually analyzed within 2 weeks. No
degradation products for the labeled standards were observed
after storing the labeled samples in -20 °C for up to 2 weeks.
LC-MS. The LC-MS setup for analyzing the labeled samples
is similar to the one described elsewhere5 and is briefly described
in Supplemental Note N1 in the Supporting Information.
Figure 1. Reaction scheme for labeling carboxylic acid-containing
metabolites using isotope coded p-dimethylaminophenacyl (DmPA)
bromide (light chain, x ) 12; heavy chain, x ) 13).
analysis process. Within this context, a rational design of the tag
to be attached to the metabolites becomes important. Considering
that the LC-MS metabolome profiling work involves several
sample handling and analysis steps, an ideal tag would provide
concurrent improvement in the analytical performance of each
step. In this work, we report a new isotope labeling method to
tag carboxylic acid-containing metabolites (CAMs) and demon-
strate its application for LC-MS metabolome profiling of complex
samples. Global profiling of these metabolites is significant in
metabolomics, as a large portion of the metabolome including a
vast number of fatty acids belong to this class. For example, about
65% of the ∼5000 known endogenous human metabolites contain
at least one carboxylic acid group in a chemical structure.14
RESULTS AND DISCUSSION
Derivatization of carboxylic acids can be done with a variety
of chemical reactions for analytical applications and, among them,
phenacyl bromide (PBr) has been used to label the acids to
improve the performance of HPLC and UV detection.18 Our
labeling chemistry is based on this reaction. However, to tailor
our needs, we designed a new reagent that allows the introduction
of a mass tag and concurrent improvement in LC-MS analysis.
Figure 1 shows the structure of the reagent, p-dimethylami-
nophenacyl (DmPA) bromide, and the reaction scheme for
labeling the carboxylic acid to form isotope mass-coded deriva-
tives. Triethanolamine (TEOA) was used as a base catalyst for
the reaction. The mass difference of the 13C-/12C-labeled products
with one tag has a nominal mass of 2 Da.
We have also constructed a standard library of 113 carboxylic
acid-containing metabolites (CAMs) by labeling individual com-
pounds with DmPABr one-by-one (see Supplemental Table T1 in
the Supporting Information). The reaction was found to be
complete within 60 min, and the yield was in the range of 95-99%.
Supplemental Table T2 in the Supporting Information shows the
reproducibility and reaction yield of 10 standards; the average CV
was 4.3% with a range from 2.2% to 7.2% for four repeated
experiments. This labeling reaction has high specificity toward
the acids. We tested a variety of compounds with different
functional groups, such as alcohols, thiols, amides, amines,
ketones, and aldehydes, and did not see reaction products. The
reaction can be performed in an aprotic solvent such as acetoni-
trile, acetone, or N,N-dimethylformamide (DMF). A small amount
of water (up to 20%) does not significantly affect the reaction yield.
To quench the reaction, various acids were tested and it was found
that triphenylacetic acid could be used to effectively consume the
excess amount of DmPABr and the labeled product eluted often
as the last peak in reversed-phase (RP) LC, thereby avoiding the
EXPERIMENTAL SECTION
Synthesis of p-Dimethylaminophenacyl (DmPA) Bromide.
Chemical reagent information is provided in Supplemental Note
N1 in the Supporting Information. The synthesis of DmPABr is
based on a two-step procedure modified from those described in
the literature (see Supplemental Scheme S1 and Note N1 in the
Supporting Information).15-17 The first step is the dimethylation
reaction,15,16 where an isotope tag is introduced to the reagent
using 13C2-dimethyl sulfate for the heavy chain reagent or 12C2-
dimethyl sulfate for the light chain reagent. The second step
involves bromation and debromation.17 The synthesis proce-
dures were optimized for reactions with ∼1 g of starting
material. Two semipreparative reversed-phase separation and
normal phase flush chromatography were used to obtain the
pure labeling reagents. The purity of the labeling reagent
synthesized was >99.5% determined by high-performance liquid
chromatography ultraviolet detection (HPLC UV) and MS. The
structure and purity were further confirmed by NMR.
Labeling Reaction. Figure 1 shows the reaction scheme for
labeling carboxylic acids using the isotope reagents, 13C- or 12C-
(13) Lamos, S. M.; Shortreed, M. R.; Frey, B. L.; Belshaw, P. J.; Smith, L. M.
Anal. Chem. 2007, 79, 5143–5149
.
(14) Wishart, D. S.; Knox, C.; Guo, A. C.; Eisner, R.; Young, N.; Gautam, B.;
Hau, D. D.; Psychogios, N.; Dong, E.; Bouatra, S.; Mandal, R.; Sinelnikov,
I.; Xia, J. G.; Jia, L.; Cruz, J. A.; Lim, E.; Sobsey, C. A.; Shrivastava, S.; Huang,
P.; Liu, P.; Fang, L.; Peng, J.; Fradette, R.; Cheng, D.; Tzur, D.; Clements,
M.; Lewis, A.; De Souza, A.; Zuniga, A.; Dawe, M.; Xiong, Y. P.; Clive, D.;
Greiner, R.; Nazyrova, A.; Shaykhutdinov, R.; Li, L.; Vogel, H. J.; Forsythe,
I. Nucleic Acids Res. 2009, 37, D603–D610
(15) Szylhabelgodala, A.; Madhavan, S.; Rudzinski, J.; Oleary, M. H.; Paneth, P.
J. Phys. Org. Chem. 1996, 9, 35–40
(16) Sekiya, M.; Tomie, M.; Leonard, N. J. J. Org. Chem. 1968, 33, 318
(17) Diwu, Z. J.; Beachdel, C.; Klaubert, D. H. Tetrahedron Lett. 1998, 39, 4987–
4990
.
.
.
.
(18) Toyo’oka, T. Anal. Chim. Acta 2002, 465, 111–130.
8790 Analytical Chemistry, Vol. 82, No. 21, November 1, 2010

Figure 2. Base-peak ion chromatogram of a mixture of 10 carboxylic acid standards (6 pmol each injected). The peaks labeled with masses
from left to right are (1) acetic acid, (2) hippuric acid, (3) 4-hydroxybenzoic acid, (4) 3-hydroxybenzoic acid, (5) malic acid, (6) butyric acid, (7)
malonic acid, (8) phenylacetic acid, (9) hydrocinnamic acid, (10) octanoic acid, and (11) triphenylacetic acid (triphenylacetic acid was added to
consume the remaining labeling reagent and quench the reaction). The standard mixture was labeled with either 13C-DmPABr or 12C-DmPABr,
and a 1:1 mixture of the labeled compounds was analyzed by RPLC quadrupole time-of-flight mass spectrometry (QTOFMS).
interference of the analyte peaks from the quenching reagent. We
note that the reaction product of the hydrophobic quenching
reagent, triphenylacetic acid, has low solubility. The majority of
the product remained as solid and would not be extracted to the
sample and thus not injected into the column. This could avoid
column overloading. Furthermore, derivatized triphenylacetic acid
eluted when there was around 90-95% ACN in the mobile phase.
Under this condition, the ionization efficiency of the derivatized
triphenylacetic acid was low, and thus their signal intensity was
low.
Figure 2 shows a base-peak ion chromatogram (BPC) of a
mixture of 10 carboxylic acids labeled with 13C-/12C-DmPABr
obtained by LC-MS using the QSTAR mass spectrometry (6
pmol each injected). Figure 3 shows the chromatograms of a
human urine sample labeled with 13C-/12C-DmPABr (Figure 3A)
and the same sample without labeling (Figure 3B). In each case,
a sample amount of equivalent to 30 nL of urine was injected into
a 2.1 mm RPLC column combined with the Bruker Fourier
transform ion cyclotron resonance (FTICR) mass spectrometer.
The molecular region of a representative mass spectrum obtained
from the labeled urine sample is shown in Figure 3C.
There are several important features shown in Figures 2 and
3 that demonstrate the merits of performing DmPA derivatization
for acid analysis. First of all, DmPA derivatization improves the
Figure 3. Base-peak ion chromatograms of (A) a labeled human
urine sample obtained in positive ion mode and (B) a urine sample
metabolite separation by RPLC. While many unlabeled acids are
not retained on a RP column, the labeled analytes have sufficient
interaction with the column to be separated by RPLC. For the
standard mixture shown in Figure 2, except the last two acids,
the unlabeled acids eluted out in the void or in a short retention
time of less than 5 min (data not shown). For the unlabeled urine
sample, most peaks were observed at or near the void volume
(Figure 3B); many other acids expected to retain on the column
were not observed due to low signal intensities (see below).
However, the labeled acids were separated over the gradient
elution time. The addition of the DmPA tag containing a hydro-
phobic aromatic ring to a polar and hydrophilic acid increases
hydrophobicity, thereby enhancing its retention on a RP column.
without labeling obtained in negative ion mode that was optimized
for detecting carboxylic acids (the expanded view of the chromato-
gram at the earlier elution is shown in the inset). The urine sample
was labeled with either 13C-DmPABr or 12C-DmPABr, and the labeled
urine samples were mixed in 1:1. A sample with an amount equivalent
to about 30 nL of the original urine was injected into RPLC FTMS for
analysis. (C) The molecular ion region of an ESI mass spectrum
displaying a pair of 13C-/12C-DmPABr labeled metabolite ions. Some
of the metabolites tentatively identified and their retention times from
part A are provided in Supplemental Table T3 in the Supporting
Information.
The major advantage of using RPLC for separating the metabolites
is that it has superior separation efficiency to other modes of
separation such as hydrophilic interaction (HILIC) chromatogra-
Analytical Chemistry, Vol. 82, No. 21, November 1, 2010 8791

phy. In addition, the use of one column, i.e., RPLC, instead of
using different columns to handle the different polarity of the
metabolites of the same sample, saves the analysis time.
Second, DmPA derivatization enhances the ESI efficiency
significantly. As Figure 2 shows, 10 chromatographic peaks from
the standards are observed with similar responses, despite large
differences in chemical structures of these acids. The injection of
the same amount of the mixture without derivatization did not
produce many detectable signals, even in the optimized negative
ion mode. For the 113 carboxylic acid standards, DmPA deriva-
tization was found to enhance the detection sensitivity by about
2-4 orders of magnitude, depending on the compound structure
(as an example, Supplemental Figure S1 in the Supporting
Information shows the calibration curves of two standards before
and after derivatization obtained by LC-FTICR-MS). The signal
enhancement can be attributed to several factors. One is related
to the increased propensity of being charged for the labeled acid
due to the presence of the dimethyamine moiety attached to the
aromatic ring of the tag where a tertiary amine can be readily
formed. The labeled acid has higher hydrophobicity than its
unlabeled counterpart, making it easier to stay on the surface of
the droplets during ESI. An elution solvent with higher organic
solvent content where a labeled acid is eluted out, as opposed to
the unlabeled one eluted at void or high water content solvent,
also enhances the ionization efficiency. Note that the first large
peak eluted at or near the void volume in Figure 2 was mainly
from the base catalyst, TEOA. The small peaks shown were most
likely from the unknown impurities present in the standards that
were also labeled with DmPA. The ESI signal enhancement is
quite dramatic in the analysis of biological samples such as urine.
As Figure 3A,B illustrates, only a few peaks are observed in the
analysis of unlabeled urine (Figure 3B and the expanded chro-
matogram in the inset) while many peaks are detected from the
labeled urine sample (Figure 3A).
The third advantage of DmPA derivatization is that a proper
isotope mass tag can be readily attached to a carboxylic acid-
containing metabolite, and the labeled metabolite does not display
any isotope effect on RPLC. Co-elution of the analyte and its
internal standard is a key to achieve accurate quantification by
LC-MS. Otherwise, ion suppression can cause significant varia-
tion in detectability of analyte and standard, resulting in quantita-
tive errors. The 13C-/12C-labeled metabolites coelute perfectly.
The relative amount of two comparative samples, one labeled
with 13C-DmPA and another with 12C-DmPA, can be determined
directly from the peak ratio of the ion pair in the mass spectra
(e.g., Figure 3C), not in the chromatogram. For example, a linear
relation (R2 > 0.995) was found in the LC-MS analysis of
differential isotope labeled carboxylic acid standards, such as
malic acid, where the amount ratios of 13C-/12C-labeling were
5:1, 1:1, 1:5, 1:10, and 1:20. The relative standard derivation
(RSD) on the ratios of the ion pairs in replicate analysis was
less than 6%, indicating that good reproducibility could also
be achieved.
over 15 000 features could be seen), and many of these peaks
are from impurities present in the solvents, column, tubing, and
interface, salts and adducts, electric noises, column coatings, etc.
Identification of the true metabolite peaks can be done by
analyzing a mixture of an equal amount of a sample differentially
labeled with 13C- and 12C-DmPABr, such as the urine sample
results shown in Figure 3A and Supporting Table T3 in the
Supporting Information. Perfect coelution of the ion pairs and their
characteristic mass difference of 2 × n Da for metabolites with n
tags or acidic groups with <2 ppm errors in mass difference can
be used to identify the metabolite peaks. Other nonmetabolite
peaks and unlabeled metabolites would show up as singlet peaks.
For example, the ion pair at 298.144 09 and 300.150 97 shown in
Figure 3C has a mass difference error of -0.53 ppm. The mass
difference of 2.007 88 Da indicates this metabolite has one
carboxylic acid. With the comparison of the retention time and
accurate mass of the standards, the ion pair was identified to be
from phenylacetic acid. While the use of ion pairing of differentially
labeled samples to assist in distinguishing the metabolite peaks
from the others is not unique to DmPA labeling, this labeling
method does provide two additional benefits. One is related to
the reduction of interference from low mass ions in LC-MS. The
addition of a tag (162 Da) to a metabolite increases the molecular
ion mass, effectively shifting away from the low mass region of
the mass spectra where high background signals are usually
observed at m/z < 200. Another benefit is related to the increased
stability of the labeled metabolites where, in general, only the
molecular ions are observed in the mass spectra. Dissociation of
molecular ions of unlabeled metabolites in the interface and during
ion translation to the mass analyzer can cause ambiguity in
assigning the mass spectra peaks; a number of peaks observed
may be from the fragment ions of the metabolites, not the intact
metabolite ions. In triplicate experiments of the 1:1 ratio of 13C-/
12C-DmPA labeled human urine, 2671, 2546, and 2820 ion pairs
were detected (see the Supporting Information, Table T3, for
the list). Using the 113 carboxylic acid standard library, we
identified 51, 43, and 48 metabolites (see Table T3 in the
Supporting Information). This example illustrates that a large
number of metabolites can be profiled from human urine using
the DmPA labeling LC-FTICR-MS method. Of course, positive
identification of these ion pairs remains to be challenging. Future
work on MS/MS analysis of these labeled ions may result in
identifying more metabolites. We also note that in the case of the
urine sample, we could not determine if there were any side
reactions contributing to the overall number of peaks observed,
as the identities of most of the peaks were unknown.
In summary, we have developed a new isotope labeling method
for high-performance metabolome analysis with a focus on global
profiling of carboxylic acid-containing metabolites. This labeling
method is demonstrated to be not only effective in introducing
an isotope tag for accurate metabolite quantification but also
improving the chromatographic retention of the metabolites in
RPLC, enhancing ESI efficiency by 2-4 orders of magnitude, and
facilitating the identification of metabolite peaks in LC-MS. This
method along with other high-performance labeling methods, such
as dansylation, that are targeted at amine-,5 phenol-,5 and ketone/
aldehyde-containing metabolites should cover the majority of the
metabolome. We believe that high-performance isotope labeling
Finally, DmPA isotope labeling facilitates the identification of
metabolite peaks among many spectral features observed in
LC-MS (a feature refers to a mass spectral peak with a specific
m/z at a given retention time). There are usually many peaks
detected in an LC-MS run (e.g., for the labeled urine samples,
8792 Analytical Chemistry, Vol. 82, No. 21, November 1, 2010

LC-MS should open the possibility of carrying out comprehensive
metabolome profiling experiments on any biological sample,
thereby increasing the power of metabolomics for investigating
subtle changes in the metabolome for biological studies and
disease biomarker discovery. Finally, we note that design of
multiplex isotope labeling reagents based on a similar reaction
scheme of DmPABr is possible and will be reported in the future.
SUPPORTING INFORMATION AVAILABLE
Synthesis of the isotope-coded DmPABr reagents (Scheme S1),
information on chemical reagents, reagent synthesis, and LC-MS
conditions (Note N1), list of standard acids (Table T1), list of 10
standards showing labeling reaction reproducibility and yield
(Table T2), list of ion pairs detected from the human urine samples
and names of the positively identified metabolites by LC-FTMS
(Table T3), and calibration curves of two standards (Figure S1).
This material is available free of charge via the Internet at
http://pubs.acs.org.
ACKNOWLEDGMENT
This work was supported by the Genome Canada and Genome
Alberta through the Human Metabolomics Toolbox (HMDB)
project and the Canada Research Chairs program. We thank
Professor David Wishart for providing some of the standards used
in this work.
Received for review August 15, 2010. Accepted October 1,
2010.
AC102146G
Analytical Chemistry, Vol. 82, No. 21, November 1, 2010 8793