Stable-isotope dimethylation labeling combined with LC-ESI MS for quantification of amine-containing metabolites in biological samples

Article abstract of DOI:10.1021/ac0704356

One of the challenges associated with metabolome profiling in complex biological samples is to generate quantitative information on the metabolites of interest. In this work, a targeted metabolome analysis strategy is presented for the quantification of amine-containing metabolites. A dimethylation reaction is used to introduce a stable isotopic tag onto amine-containing metabolites followed by LC-ESI MS analysis. This labeling reaction employs a common reagent, formaldehyde, to label globally the amine groups through reductive animation. The performance of this strategy was investigated in the analysis of 20 amino acids and 15 amines by LC-ESI MS. It is shown that the labeling chemistry is simple, fast (<10-min reaction time), specific, and provides high yields under mild reaction conditions. The issue of isotopic effects of the labeled amines on reversed-phase (RP) and hydrophilic interaction (HILIC) LC separations was examined. It was found that deuterium labeling causes an isotope effect on the elution of labeled amines on RPLC but has no effect on HILIC LC. However, ¹³C-dimethylation does not show any isotope effect on either RPLC or HILIC LC, indicating that ¹³C-labeling is a preferred approach for relative quantification of amine-containing metabolites in different samples. The isotopically labeled 35 amine-containing analogues were found to be stable and proved to be effective in overcoming matrix effects in both relative and absolute quantification of these analytes present in a complicated sample, human urine. Finally, the characteristic mass difference provides additional structural information that reveals the existence of primary or secondary amine functional groups in amine-containing metabolites. As an example, for a human urine sample, a total of 438 pairs of different amine-containing metabolites were detected, at signal-to-noise ratios of greater than 10, by using the labeling strategy in conjunction with RP LC-ESI Fourier-transform ion cyclotron resonance MS.

Full text of DOI:10.1021/ac0704356

Anal. Chem. 2007, 79, 8631-8638
Stable-Isotope Dimethylation Labeling Combined
with LC-ESI MS for Quantification of
Amine-Containing Metabolites in Biological
Samples
Kevin Guo, Chengjie Ji, and Liang Li*
Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada T6G 2G2
logues as internal standards for quantifying the metabolites of
interest. SIL internal standards are chemically and structurally
similar to the analytes with specific atoms in the analytes replaced
by their corresponding isotopes, such as deuterium for hydrogen,
13C for 12C, 15N for 14N, or 18O for 16O.1,2 The use of SIL internal
standards normalizes the MS intensity of analytes to their isotopic
analogues and, therefore, effectively compensates for the matrix
effect, ion suppression from other coeluting analytes, and varia-
tions caused by sample preparation, injection, and instrument
parameters.1-6 Unfortunately, only a limited number of SIL internal
standards are commercially available. In the absence of SIL
standards, structural analogues are used as the second best
choice; but the use of structural analogues may result in poor
quantification performance, particularly in analyzing metabolites
present in a complex matrix.7 Errors can be introduced during
sample processing (e.g., different recovery rates in the extraction
of a metabolite and its structural analogue) and the LC-MS
analysis step. Structural analogues are often not coeluted with the
analyte of interest, and therefore, they would experience levels
of matrix effect different from that of analyte. As a result, the
relative signal intensities of a metabolite and its analogue may
not reflect their concentration ratio in the sample.
One of the challenges associated with metabolome profil-
ing in complex biological samples is to generate quantita-
tive information on the metabolites of interest. In this
work, a targeted metabolome analysis strategy is pre-
sented for the quantification of amine-containing metabo-
lites. A dimethylation reaction is used to introduce a stable
isotopic tag onto amine-containing metabolites followed
by LC-ESI MS analysis. This labeling reaction employs a
common reagent, formaldehyde, to label globally the
amine groups through reductive amination. The perfor-
mance of this strategy was investigated in the analysis of
20 amino acids and 15 amines by LC-ESI MS. It is
shown that the labeling chemistry is simple, fast (<10-
min reaction time), specific, and provides high yields
under mild reaction conditions. The issue of isotopic
effects of the labeled amines on reversed-phase (RP) and
hydrophilic interaction (HILIC) LC separations was ex-
amined. It was found that deuterium labeling causes an
isotope effect on the elution of labeled amines on RPLC
but has no effect on HILIC LC. However, 13C-dimethyla-
tion does not show any isotope effect on either RPLC or
HILIC LC, indicating that 13C-labeling is a preferred
approach for relative quantification of amine-containing
metabolites in different samples. The isotopically labeled
35 amine-containing analogues were found to be stable
and proved to be effective in overcoming matrix effects in
both relative and absolute quantification of these analytes
present in a complicated sample, human urine. Finally,
the characteristic mass difference provides additional
structural information that reveals the existence of pri-
mary or secondary amine functional groups in amine-
containing metabolites. As an example, for a human urine
sample, a total of 438 pairs of different amine-containing
metabolites were detected, at signal-to-noise ratios of
greater than 10, by using the labeling strategy in conjunc-
tion with RP LC-ESI Fourier-transform ion cyclotron
resonance MS.
Another practical strategy for quantification in LC-MS is
chemical labeling, which has been widely applied to obtain relative
quantitative information of proteomes.8,9 For metabolome analysis,
there are a limited number of reports of using chemical deriva-
tization to introduce a stable isotopic tag to metabolites to facilitate
their quantitative analysis by LC-electrospray ionization (ESI)
MS.10-12 Development of facile derivatization methods for stable-
isotope dilution-based quantification of metabolomes is needed.
(1) Stokvis, E.; Rosing, H.; Beijnan, J. Rapid Commun. Mass Spectrom. 2005,
19, 401-407.
(2) Liang, H.; Foltz, R.; Meng, M.; Bennet, P. Rapid Commun. Mass Spectrom.
2003, 17, 2815-2821.
(3) Annesley, T. M. Clin. Chem. 2003, 49, 1041-1044.
(4) Magni, F.; Pereira, S.; Leoni, M.; Grisenti, G.; Kienle, G. J. Mass Spectrom.
2001, 36, 670-676.
(5) Jacob, P.; Wilson, M.; Yu, L.; Mendelson, J.; Jones, R. T. Anal. Chem. 2002,
74, 5290-5296.
(6) Kuklenyik, Z.; Ashley, D. L.; Calafat, A. M. Anal. Chem. 2002, 74, 2058-
2063.
LC-MS has become an important tool for metabolome
analysis. However, generation of accurate quantitative information
by LC-MS is not straightforward.1 One of the most reliable
quantitative methods is to use stable-isotope-labeled (SIL) ana-
(7) Stokvis, E.; Rosing, H.; Lopez-Lazaro, L.; Schellens, J.; Beijnan, J. Biomed.
Chromatogr. 2004, 18, 400-402.
(8) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R.
Nat. Biotechnol. 1999, 17, 994-999.
* To whom correspondence should be addressed. E-mail: Liang.Li@ualberta.ca.
(9) Aebersold, R.; Mann, M. Nature 2003, 422, 198-207.
10.1021/ac0704356 CCC: $37.00 © 2007 American Chemical Society
Published on Web 10/10/2007
Analytical Chemistry, Vol. 79, No. 22, November 15, 2007 8631

For an ideal derivatization protocol, the derivatization reagent
should be specific for the target functional group, and the analytes
of interest need to be rapidly and quantitatively derivatized with
minimum byproducts (i.e., high yield). The reaction should be
performed under mild conditions with minimum manipulation. For
quantitative applications, the resulting products need to be stable.
Differential isotopic dimethyl labeling of N-terminal peptides
with d(0) and d(2) or d(0), 12C and d(2), 13C-formaldehyde
combined with LC-ESI or LC-MALDI have been successfully used
for relative proteome quantification.13-17 The labeling is carried
out by using reductive amination chemistry.18 In this work, we
report our studies of using reductive amination to introduce
isotopic tags to amine-containing metabolites and applying this
strategy to the quantification by LC-ESI MS of both primary and
secondary amine metabolites in human urine. Amine-containing
metabolites play essential roles in biological functions. For
example, amino acids and their derivatives are common biomar-
kers for human physiological process.19 Their identification and
quantification in human fluids provide significant insights related
to human health. The polycationic polyamines are essential for
eukaryotic cellular growth and viability; rapid tumor growth was
associated with polyamine biosynthesis and accumulation.20 Many
studies indicate that significantly higher levels of polyamines and
their metabolites were present in the biological fluids and the
affected tissues of cancer patients and other hyperproliferative
diseases.20-22 Some therapeutic polyamine analogues are showing
exciting potentials to treat cancer and other hyperproliferative
disorders.20 Thus, quantitative profiling of amine-containing me-
tabolites could potentially be applied for the discovery of new
disease biomarkers as well as for the monitoring of tumor growth
and regression in cancer study.
sodium acetate, LC-MS grade formic acid, and acetic acid were
also obtained from Sigma-Aldrich. LC-MS grade of water,
methanol, and acetonitrile were purchased from Fisher Scientific
Canada (Edmonton, AB, Canada). Formaldehyde-13C (20 wt %
solution in water, >99% isotope purity) solution, and d(2)-
formaldehyde (20 wt % solution in deuterated water, >98% isotope
purity) were the products of Cambridge Isotope Laboratories, Inc.
(Andover, MA).
Dimethylation Labeling Reaction. The freshly collected
human urine was centrifuged for 10 min at 12 000 rpm. A total of
500 µL of urine supernatant, 20 amino acid solutions, or 15 amine
standard solutions were mixed with an equal volume of am-
monium acetate buffer (0.2 M, pH 5.3) in a reaction vial. The
solutions were vortexed, centrifuged, and mixed with 125 µL of
freshly prepared sodium cyanoborohydride (1.0 M). After further
mixing, centrifugation, and the addition of 100 µL of 4% formal-
dehyde, or formaldehyde-13C, or d(2)-formaldehyde solution, the
mixtures were vortexed and centrifuged again, and the reaction
was allowed to proceed for 10 min at 37 °C and 200 rpm in an
Innova 4000 benchtop incubator shaker. The pH of mixtures was
adjusted to pH 2-3 by adding ∼25 µL of formic acid. The solutions
were then centrifuged or filtered before being injected onto an
LC column. The samples for hydrophilic interaction liquid chro-
matography (HILIC) were diluted with acetonitrile to obtain
optimal separation efficiency. Because sodium cyanoborohydride
is a highly toxic chemical that will produce hydrogen cyanide gas
when exposed to acid, and formaldehyde is a known carcinogen
on inhalation exposure, the dimethylation labeling reaction was
carried out in a fume hood.
LC-ESI MS. The HPLC system used in conjunction with the
mass spectrometer was an Agilent 1100 series binary system, and
it was modified to reduce extra column system volume according
to an Agilent protocol (Agilent Publication Number, 5988-2682EN).
A reversed-phase (RP) Agilent Zorbax XDB C18 column (1.0 ×
150 mm, 3.5-µm particle size, 80-Å pore size) and a Zorbax XDB
EXPERIMENTAL SECTION
Chemicals and Reagents. Twenty amino acids,
-arginine, -aspartic acid, -asparagine, -cysteine, -glutamic acid,
-glutamine, glycine, -histidine, -isoleucine, -leucine, -lysine,
-methionine, -phenylalanine, -proline, -serine, -threonine,
-tyrosine, -tryptophan, and -valine; and 15 amines, 1-ephedrine,
L-alanine,
L
L
L
L
L
L
L
L
C18 rapid resolution high-throughput cartridge column (2.1 × 15
L
L
L
L
mm, 1.8 µm, 80 Å) were purchased from Agilent Technologies,
Inc. (Palo Alto, CA). For RP chromatography, solvent A was 0.1%
formic acid, 5% methanol in water, and solvent B was 0.1% formic
acid in methanol. All the formic acid, methanol, and water used
were LC-MS grade. The 64-min binary gradient elution profile
was as follows: t ) 0, 0% B; t ) 6 min, 0% B; t ) 21 min, 30% B;
t ) 54 min, 90% B; t ) 64 min, 90% B. The flow rate was 50 µL/
min, and sample injection volumes were 10 µL.
L
L
L
L
L
L
1,4-diaminobutane, (-)-epinephrine, 2-methylbenzylamine, 3-meth-
yl- -histidine, aniline, benzylamine, cysteamine, dopamine, hista-
mine, -4-hydroxyproline, p-aminohippuric acid, pyridoxamine,
L
L
γ-aminobutyric acid, and tyramine, were purchased from Sigma
Aldrich (Oakville, ON, Canada). Formaldehyde (37 wt % solution
in water), sodium cyanoborohydride (95%), ammonium acetate,
A TSKgel Amide-80 HILIC column (1.0 × 250 mm, 5 µm) was
obtained from Tosoh Bioscience LLC (Montgomeryville, PA). For
HILIC, solvent A was 10% 15 mM ammonium acetate (pH 5.5) in
LC-MS grade acetonitrile, and solvent B was 40% 15 mM
ammonium acetate (pH 5.5) in LC-MS grade acetonitrile. The
45-min binary gradient elution profile was as follows: t ) 0, 10%
B; t ) 30 min, 30% B; t ) 37 min, 45% B; t ) 42 min, 70% B; t )
45 min, 70% B. The flow rate was 55 µL/min, and sample injection
volumes were 3 µL. The flow from RP or HILIC columns was
directed to the ESI source of a Bruker Esquire-LC ion trap LC-
MS system or a Bruker 9.4-T Fourier transform (FT) ion cyclotron
resonance (ICR) mass spectrometer. All MS spectra were obtained
in the positive ion mode. Negative ion detection was found to be
not as sensitive as the positive ion detection for the labeled amines
tested in these instruments.
(10) Fukusaki, E.; Harada, K.; Bamba, T.; Kobayashi, A. J. Biosci. Bioeng. 2005,
99, 75-77.
(11) Yang, W. C.; Mirzaei, H.; Liu, X.; Regnier, F. E. Anal. Chem. 2006, 78,
4702-4708.
(12) 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.
(13) Hsu, J. L.; Huang, S. Y.; Chow, N. H.; Chen, S. H. Anal. Chem. 2003, 75,
6843-6852.
(14) Ji, C.; Li, L. J. Proteome Res. 2005, 4, 734-742.
(15) Ji, C.; Li, L. J.; Gebre, M.; Pasdar, M.; Li, L. J. Proteome Res. 2005, 4, 1419-
1426.
(16) Ji, C.; Guo, N.; Li, L. J. Proteome Res. 2005, 4, 2099-2108.
(17) Ji, C.; Lo, A.; Marcus, S.; Li, L. J. Proteome Res. 2006, 5, 2567-2576.
(18) Means, G. E.; Feeney, R. E. Biochemistry 1968, 7, 2192-2201.
(19) Husek, P.; Simek, P. LC-GC 2001, 19, 986-999.
(20) Qureshi, G.; Khuhawar, M. J. Chromatogr., B 2001, 764, 385-407.
(21) Marton, L.; Casero, R., Jr. Nat. Rev. 2007, 6, 373-390.
(22) Wallace, H.; Saunder, F. Biochem. Soc. Trans. 2007, 35, 364-368.
8632 Analytical Chemistry, Vol. 79, No. 22, November 15, 2007

analyte corresponds to the number of primary or secondary amino
groups present. This is another benefit of using this derivatization
chemistry; i.e., one can deduce information about the number of
amino groups present in an analyte, which may be useful for
unknown metabolite identification or structural analysis. We also
observed some ESI signal enhancement (∼1-10-fold depending
on the compound structures) for the products, compared to the
unlabeled analytes. This can be attributed to the fact that
dimethylation converts primary or secondary amine into a more
easily protonated tertiary amine. In addition, in the case of amino
acids, the labeled ones are slightly more hydrophobic than the
unlabeled ones (e.g., they elute out at a longer retention time in
RP LC). The increased hydrophobicity for some of the labeled
polar amino acids may enhance the ESI signals.
It should be noted that the dimethylation labeling reaction is
pH dependent. Ammonium acetate buffer (pH 5.3) was found to
provide the optimal condition resulting in the highest yield
possible for all the analytes. The pH of reaction mixtures was
carefully controlled and checked when necessary. The labeling
reaction is fast. No or very little byproducts were observed from
a 10-min incubation at 37 °C. Some byproducts could be observed
if the reaction time was too long or the incubation temperature
was too high. The stability of the labeled products was also
studied. LC-MS chromatograms show that the products were
stable at room temperature over a period of at least two weeks.
The labeled products can be stored at -20 °C for long-term
storage.
In summary, reductive amination provides a simple means of
labeling amine-containing compounds, such as amino acids and
various amines. The labeling chemistry produces a high yield,
requires a short reaction time, and is very specific under mild
conditions. Due to its long-term stability, an isotope-labeled
analogue can serve as the internal standard for absolute quanti-
fication, and differential labeling of comparative samples can be
used for relative quantification of amine-containing compounds
(see below).
Evaluation of Isotopic Effects. For quantitative analysis using
the isotope dilution method, analytes and their isotopic analogues
should coelute on LC; i.e., there should be no isotopic effect on
analyte retention. In this way, a pair of isotope labeled analytes
would experience the same degree of ion suppression or matrix
effect in MS analysis. Thus, the overall analytical efficiencies for
the pair would be the same, which is important for accurate
relative quantification. Figure 2 shows the results of the isotopic
effect studies. In Figure 2A, the heavy deuterium dimethylated
isoleucine (Ile) and leucine (Leu) eluted noticeably faster than
light hydrogen dimethylated Ile and Leu on RP chromatography.
This result indicates that there was less interaction between the
stationary phase and the deuterium-labeled compounds than there
was between the stationary phase and the hydrogen-labeled
counterpart. The origin of the deuterium isotopic effect has been
attributed primarily to the differences in the lengths of the C-D
and C-H bonds;25-27 the smaller amplitude of vibrations, and
therefore lower average volumes and polarizabilities for C-D
bonds. Therefore, van der Waals interactions between the hydro-
Figure 1. Overview of dimethylation labeling strategy for quantitative
analysis of amine-containing metabolites. (A) Scheme of reductive
amination labeling. X ) 13C or 12C; in the case of D(2) labeling, two
deuterium atoms replace two hydrogen atoms in formaldehyde. (B)
Workflow for demonstrating the feasibility and performance of relative
quantitative analysis of primary and secondary amine metabolites in
urine samples by LC-ESI MS.
RESULTS AND DISCUSSION
Dimethyl Isotope Labeling. When amine-containing mol-
ecules are treated with low concentrations of simple aliphatic
aldehydes and a small amount of sodium borohydride, amine
groups can be converted in high yield into their corresponding
mono- or dialkylamine derivatives.18 We investigated the reaction
of 20 amino acids and 15 amines with formaldehyde and observed
the formation of predominantly dimethylated derivatives (28-Da
difference for each labeled site) from primary amines, and
monomethylated derivatives (14-Da difference) from secondary
amines, such as proline. Judging from the LC-MS results of the
labeled products, the conversion yield is better than 97%. This
observation is similar to those found in other reported studies.23,24
The use of dimethylation labeling is attractive for several reasons.
The formaldehyde used as the labeling reagent is inexpensive and
13C- or deuterium-labeled formaldehyde is commercially available.
The experimental conditions for reductive amination are extremely
mild, and the reaction is easily completed without any special
reagents or reaction equipment. As Figure 1A shows, an inter-
mediate Schiff base is formed in reductive amination. The high
yield of methylated products observed in our experiment suggests
that the Schiff base intermediate was reduced at a reaction rate
much greater than that of formaldehyde. Consequently, the
reactive intermediate readily reacts with formaldehyde and is
reduced to dimethylated products. The dimethylation labeling
reaction appears highly specific for amine groups for the 20 amino
acids and the 15 amines we studied.18
We also tested the dimethylation reaction for each amino acid
and amine, one by one, to examine any possible byproducts from
the reaction. In general, no significant amount of side reaction
products were observed, at least for the 20 amino acids and 15
amines tested. For Lys, there were two dimethylated tags
introduced to the molecule. The number of tags introduced to an
(25) Turowski, M.; Yamakawa, N.; Meller, J.; Kimata, K.; Ikegami, T.; Hosoya,
K.; Tanaka, N.; Thornton, E. R. J. Am. Chem. Soc. 2003, 125, 13836-13849.
(26) Wade, D. Chem.-Biol. Interact. 1999, 117, 191-217.
(23) Bowman, R. E.; Stroud, H. H. J. Chem. Soc. 1950, 1342-1345.
(24) Ingram, V. M. J. Biol. Chem. 1953, 202, 193-201.
(27) Thornton, E. R. Annu. Rev. Phys. Chem. 1966, 17, 349-372.
Analytical Chemistry, Vol. 79, No. 22, November 15, 2007 8633

dimethylated standards did not show any isotopic effect in HILIC
separation either (data not shown). Thus, 13C-dimethylation is a
preferred labeling method for stable-isotope quantitative analysis.
This reaction produces at least a 2-Da mass difference between
the light- and heavy-labeled analytes except for a secondary amine
where the mass difference would be 1 Da. However, the more
expensive 13C-labeled analogues may not be needed if a HILIC
separation is used for LC-ESI MS. In this case, deuterium labeling
is sufficient and provides at least a 4-Da mass difference, which
can be advantageous in avoiding the overlaps of isotope envelopes,
particularly for high-mass analytes.
Quantitative Response. To study the feasibility of our
labeling strategy for quantitative analysis, 13C- and 12C-dimethylated
amino acid and amine standards were mixed in ratios of 1:8, 1:4,
1:1, 4:1, and 8:1 in aqueous solution. The standard mixtures were
then injected onto a RP column followed by ESI-MS. The resulting
extracted ion chromatograms were obtained from the correspond-
ing mass for heavy 13C- and light 12C-dimethylated products. The
ratios of chromatographic peak areas were calculated for each
pair of 20 amino acids and 15 amines from their corresponding
extracted ion chromatograms. The linear regression plots of
tryptophan, phenylalanine, and methionine, as examples, are
shown in Figure 3A. An R2 value of above 0.99 was obtained for
all the retained species, indicating good correlation of the
experimental data with the theoretical ratios. Thus, the amine-
containing compounds were quantitatively and reproducibly de-
rivatized by this protocol. The D- and H-dimethylated standards
in water with the same mixing ratios were also examined using
HILIC LC-ESI MS. Similar R2 values for all the polar species were
obtained (data not shown).
Figure 2. Evaluation of isotopic effects. (A) Deuterium dimethyl-
labeled isoleucine and leucine eluted earlier than those of hydrogen
counterparts in RP chromatography. (B) Deuterium/hydrogen di-
methyl-labeled isoleucine and leucine coeluted perfectly in HILIC
separation. (C) 1:4 ratio of 13C-/12C-dimethyl-labeled isoleucine and
leucine coeluted perfectly in RP separation.
To demonstrate the feasibility of our labeling strategy to
overcome the matrix effect in complex biological samples, 13C-
and 12C-dimethylated standards were mixed in ratios of 1:8, 1:4,
1:1, 4:1, and 8:1 in water containing 20% unlabeled human urine
to mimic a complex matrix. The unlabeled urine would not
contribute to the signals of the labeled ion pairs of interest, but
would introduce chemical noise and coeluting components to the
analysis. With the same procedure described above, linear
regression plots were performed and three examples are shown
in Figure 3B. Similarly, the D- and H-dimethylated standards in
20% unlabeled urine with the same mixing ratios were also
examined for polar species in HILIC separation. In general, the
R2 values for those in 20% unlabeled urine were over 0.99 (Figure
3C and D), about the same as those in pure water. These results
of good linearity indicate that the use of dimethylated analogues
as internal standards can effectively overcome the matrix effect.
In general, the performance of stable-isotope dilution quanti-
fication largely depends on the control of the two labeling
reactions, i.e., heavy and light dimethylation, and also the quality
of the mass spectra obtained that display the isotope peak pairs.
These spectra can be affected by the ionization processes of the
analytes. Reductive amination occurs with a minimum change in
hydrophobicity of amino acids and amines. For some of the less
polar species, which can be retained on C18 reversed-phase
columns, good-quality mass spectra of the dimethylated products
with high signal-to-noise ratios can be obtained. However, many
of the amino acids and amines studied in this work are very polar.
They are poorly retained on the hydrophobic C18 stationary phase
phobic stationary phase and deuterium-labeled species or disper-
sion forces that result in attraction binding forces of deuterium
analogues within the hydrophobic stationary phase are less than
that of hydrogen-labeled counterparts. Thus, the deuterium-labeled
internal standard is not the best choice for stable-isotope dilution
quantitative analysis due to the isotopic effect in RP LC.
In contrast, as shown in Figure 2B, the deuterium dimethylated
Ile and Leu perfectly coeluted with their hydrogen counterparts
in HILIC mode separation. There appears to be no difference in
the magnitude of interactions with the stationary phase of the
deuterium-labeled species and that of their hydrogen counterparts.
In HILIC LC, the combination of the hydrophilic interaction and
ion-exchange mechanisms results in the enhanced polar retention,
and the hydrophilic interaction is dominated when the mobile
phase is above 70% organic solvent. The hydrophilic interaction,
the partitioning of polar analytes into and out of the adsorbed water
layer for deuterium and the corresponding hydrogen species, are
approximately the same. Thus, no isotopic effect was observed
in HILIC LC. Thornton et al. reported that much less isotopic effect
was observed when a less hydrophobic stationary phase (com-
pared to C18) was used, even in reversed-phase LC.25 Our results
are consistent with the notion that increasing hydrophilic interac-
tions decreases the isotopic effect of deuterium-labeled com-
pounds.
As expected, the heavy 13C-dimethylated Ile and Leu eluted
with exactly the same retention time as the light 12C-dimethylated
Ile and Leu in RP LC (see Figure 2C). Not surprisingly, 13C-
8634 Analytical Chemistry, Vol. 79, No. 22, November 15, 2007

Table 1. LC-Ion Trap-MS Results of Four Replicate
Measurements of 1:1 Mixtures of 20 Isotope-Labeled
Amino Acids Using Dimethylation
measured ratio
of the isotope-
metabolite
labeled pair
RSD %
glycine
1.03
0.96
1.02
1.08
0.97
1.07
1.10
1.01
1.02
1.09
1.10
1.05
1.09
1.00
1.07
1.00
1.08
1.05
1.07
1.02
8.0
4.3
6.6
7.7
4.5
8.1
8.9
4.5
5.6
3.9
4.4
2.1
9.3
3.1
6.2
2.7
7.0
8.5
5.2
2.3
alanine
proline
serine
valine
threonine
cysteine
leucine
isoleucine
asparagine
aspartic acid
glutamine
glutamic acid
methionine
histidine
phenylalanine
arginine
lysine
tyrosine
tryptophan
background levels. For those highly polar species, HILIC separa-
tion that utilizes high organic mobile phases (>70%) provides
excellent complementary selectivity to that of RP chromatography.
A polar species generally elutes after a less polar one, and at a
high percentage of organic solvent. As a consequence, much
higher sensitivity can be obtained from HILIC LC-ESI MS for
the polar compounds. This is important for achieving quantitative
results from the stable-isotope dilution method. Thus, in this work,
all the data for the polar species were obtained from the HILIC
mode, and that of less polar species were from RP LC. Dimethyl-
ated products of some less polar species, such as leucine and
isoleucine, were retained in both modes and could be used for
comparison studies. As illustrated above, the linear regression R2
values of both modes were similar, indicating that both separation
modes can be used for quantitative analysis with the dimethylation
labeling strategy.
Relative Quantification of Amine-Containing Metabolites
in Human Urine. Relative quantification of different metabolome
samples is important for many biological studies, including the
search for potential biomarkers of diseases. Our labeling strategy
combined with LC-ESI MS can potentially be used to determine
relative quantities of all the primary and secondary amines in two
biological samples. The reproducibility of our method was evalu-
ated by running four replicate experiments of mixtures containing
1:1 12C- and 13C-dimethylated 20 amino acids in aqueous solution.
Relative standard deviation (RSD) of the measured intensity ratio
for tryptophan, phenylalanine, leucine, isoleucine, tyrosine, argi-
nine, methionine, valine, and lysine was calculated from extracted
ion chromatograms of RP LC-MS runs. RSD of other dimethyl-
ated amino acids were calculated from extracted ion chromato-
grams of HILIC LC-MS runs. The results are summarized in
Table 1. As Table 1 shows, the measured relative intensity ratio
of an amino acid pair varies from 0.96 to 1.10 with RSD ranging
from 2.1 to 9.3 (average RSD ) 5.6%).
Figure 3. Linear regression plot of (A) RP chromatography of (i)
13C-/12C-dimethyl-labeled tryptophan in water, (ii) 13C-/12C-dimethyl-
labeled phenylalanine in water, and (iii) 13C-/12C-dimethyl-labeled
methionine in water. Linear regression plot of (B) HILIC separation
of (i) 13C-/12C-dimethyl-labeled leucine in water, (ii) 13C-/12C-dimethyl-
labeled tyrosine in water, and (iii) 13C-/12C-dimethyl-labeled aspar-
agine in water. Linear regression plot of (C) RP chromatography of
(i) 13C-/12C-dimethyl-labeled tryptophan in 20% urine and (ii) 13C-/
12C-dimethyl-labeled phenylalanine in 20% urine. Linear regression
plot of (D) HILIC separation of (i) 13C-/12C-dimethyl-labeled tyrosine
in 20% urine and (ii) 13C-/12C-dimethyl-labeled asparagine in 20%
urine.
and elute in the initial void volume, resulting in low sensitivity of
ESI-MS detection due to analyte ion suppression and high
To demonstrate the feasibility of comparative quantification by
this labeling strategy, we divided a biological sample into two equal
Analytical Chemistry, Vol. 79, No. 22, November 15, 2007 8635

pair (m/z 194.117 55 and 196.124 24, see Figure 4C) were 1.00
and 1.02. For the 20 amino acids, the measured ratio of an amino
acid pair ranges from 0.96 to 1.10 with an average ratio of 1.04
and a coefficients of variation (CV) of 4.1%. These LC-MS
experimental values are in excellent agreement with the expected
ratios of 1.00.
Note that, as pointed out earlier, the characteristic mass
differences between heavy and light dimethylated compounds in
mass spectra indicate the existence of primary or secondary amine
functional group in a metabolite, and this could potentially be used
for unknown amine identification, in addition to the information
gained from retention time, MS/MS, accurate mass, and isotopic
pattern by high-resolution MS. In our FT-ICR-MS, the mass
measurement accuracy for the metabolites is typically less than
2 ppm. This high mass accuracy facilitates the identification of
ion pairs of amine-containing metabolites in the urine sample.
Supporting Information Table S1 lists 33 amine-containing me-
tabolites detected by either RP or HILIC LC-ESI FT-ICR-MS with
each pair having signal-to-noise ratios of greater than 10. The 20
amino acids plus 13 amines were positively identified based on
their accurate mass and retention time in comparison to those of
labeled standards. Among them, 16 were found in both RP and
HILIC LC-MS runs, 2 was found in RP LC-MS alone, and 15
were found in HILIC LC-MS alone.
In the RP LC-ESI FT-ICR-MS analysis of the labeled urine
sample, besides the 18 metabolites identified and listed in Table
S1, we actually detected additional 420 ion pairs (see Supplorting
Information Table S2). All these pairs displayed a signal-to-noise
ratio of greater than 10. Over 100 ion pairs showed signal-to-noise
ratios of greater than 80, indicating they are high-abundance
metabolites. Based on accurate mass and retention time informa-
tion, we conclude that these ion pairs belong to different
metabolites. However, these metabolites remain to be identified.
Accurate mass measurement alone does not lead to unambiguous
metabolite identification, considering there are many metabolites
potentially present in body fluids. Nevertheless, this work dem-
onstrates that, using the labeling strategy combined with RP LC-
ESI FT-ICR-MS, we can potentially profile over 438 amine-
containing metabolites in human urine. One would expect that
multiple dimensional separation of the labeled metabolites followed
by LC-MS analysis should further expand the metabolome
coverage.
Absolute Quantification of 20 Amino Acids and 15
Amines in Human Urine. Absolute quantification of analytes of
interest can provide direct information on the expression of a given
metabolite in relation to other metabolites present in a sample.
Absolute quantification by stable-isotope dilution LC-MS can be
done by using this dimethylation labeling strategy, as long as the
targeted amines are commercially available. The availability of the
13C- or deuterium isotope-labeled analogue of an amine of interest
is no longer an issue, because stable isotopic tags will be
introduced onto the parent compounds using the dimethylation
labeling reaction to produce the isotope analogue, while the same
reaction will be done on the sample to be analyzed.
Figure 4. Feasibility of relative quantification of metabolites in
human urine. (A) Reversed-phase base peak ion chromatogram of
equivalent mixing of 13C- and 12C-dimethyl-labeled human urine
obtained by using RPLC FT-ICR-MS; (B) mass spectrum of 13C-/12C-
dimethyl-labeled tryptophan in human urine at the retention time of
22.35 min; (C) mass spectrum of 13C-/12C-dimethyl-labeled phenyl-
alanine in human urine at the retention time of 16.52 min.
fractions (see Figure 1B), followed by labeling with heavy and
light dimethyl isotopes, and then examined whether the ratio of
heavy to light labeled pair was close to 1:1. For this demonstration,
a human urine sample was split into four fractions which were
13C- or 12C- and D(2)- or H-dimethyl labeled under the same
reaction conditions. Then the fractions from the 13C- and 12C-
labeled samples or D(2)- and H-labeled samples, respectively, were
well mixed by vortexing. The aliquots of mixtures were injected
onto RP or HILIC columns, followed by ESI MS detection. Note
that, for the treatment of the urine sample, only centrifugation
was applied to remove possible particles in urine. Molecular
weight cutoff filters were initially used to remove proteins, but
found to be not necessary. The amount of sample injected was
very small, and the protein concentration in urine should be very
low. We have not observed any adverse effect on the lifetime of
a column or the chromatographic performance of the urine
metabolites from injecting the urine samples without removal of
the proteins. As an example of the urine analysis, Figure 4A shows
the resulting base peak ion chromatogram from RP LC on the
Bruker 9.4-T FT-ICR mass spectrometer. The ratio of integrated
peak areas or peak heights for the 13C- and 12C-dimethylated
tryptophan pair (m/z 233.128 62 and 235.135 30, see Figure 4B)
were 1.02 and 1.04, respectively. The ratio of the phenylalanine
As an example of absolute quantification, aliquots of human
urine were heavy dimethyl labeled with 13C- or D(2)-formaldehyde.
The 20 amino acid and 15 amine standard solutions of known
concentration were light dimethyl labeled with normal formalde-
8636 Analytical Chemistry, Vol. 79, No. 22, November 15, 2007

Figure 6. Base peak ion chromatograms of hydrophilic interaction
chromatography of (A) 13C-/12C dimethyl-labeled 20 amino acids, (B)
13C-/12C-dimethyl-labeled 15 amines, and (C) 13C-dimethyl-labeled
human urine mixed with 12C-dimethyl-labeled standards. The 15
amines in (B): 1, aniline; 2, 2-methylbenzylamine; 3, p-aminohippuric
acid; 4, benzylamine; 5, 1-ephedrine; 6, pyridoxamine; 7, tyramine;
8, histamine; 9, dopamine; 10, cysteamine; 11, (-)-epinephrine; 12,
4-hydroxyproline; 13, diaminobutane; 14, γ-aminobutyric acid; 15,
3-methylhistidine.
Figure 5. Base peak ion chromatograms of RP chromatography.
(A) 13C-/12C-dimethyl-labeled 20 amino acids; the components that
eluted in the void volume are A, C, E, Q, G, S, and T, and (B) 13C-/
12C-dimethyl-labeled 15 amines. Histamine, diaminobutane, 4-hy-
droxproline, 3-methylhistidine, and γ-aminobutyric acid were eluted
in the column void volume. 1, pyridoxamine and cysteamine; 2, (-)-
epinephrine; 3, dopamine; 4 and 5, unknown; 6, tyramine; 7,
benzylamine; 8, aniline; 9, 2-methylbenzylamine; 10, 1-ephedrine; 11,
p-aminohippuric acid; (C) 13C-dimethyl-labeled human urine mixed
with 12C-dimethyl-labeled standards.
Table 2. Absolute Quantificiation of 20 Amino Acids
and 15 Amines in Human Urine
concn
concn
hyde under exactly the same reaction conditions. Heavy-labeled
urine was mixed with light-labeled standards and then injected
onto RP and HILIC columns followed by ESI-MS analysis. Figures
5 and 6 show the base peak ion chromatograms of the 13C- and
12C-dimethyl-labeled amino acid standards, amine standards, and
the heavy-labeled human urine sample mixed with light-labeled
standards obtained by RP LC-ESI MS and HILIC LC-ESI MS,
respectively. The calculated ratios of heavy- to light-labeled pairs
of amines from their extracted ion chromatograms were used to
perform absolute quantification, since the concentrations of the
standards are known. The absolute concentrations of 20 amino
acids and 15 amines in human urine were determined and are
listed in Table 2. As Table 2 shows, the metabolites detected in
the urine sample have concentrations ranging from 0.05 (arginine
or tyramine) to 21.2 mg/L (glutamine) with a concentration
dynamic range of at least 424-fold.
metabolites
(mg/L)
metabolites
(mg/L)
glycine
10.1
0.97
0.14
1.40
0.51
2.20
0.20
0.37
0.27
0.96
1.40
21.2
0.17
0.22
16.2
1.00
0.05
1.60
tyrosine
3.00
4.80
<DL
<DL
0.09
1.90
1.40
0.05
0.22
0.26
0.29
0.12
0.14
0.06
0.25
0.38
0.08
alanine
tryptophan
proline
aniline
serine
2-methylbenzylamine
benzylamine
valine
threonine
cysteine
1-ephedrine
p-aminohippuric acid
tyramine
leucine
isoleucine
asparagine
aspartic acid
glutamine
glutamic acid
methionine
histidine
phenylalanine
arginine
γ-aminobutyric acid
histamine
L,4-hydroxyproline
cysteamine
1,4-diaminobutane
(-)-epinephrine
pyridoxamine
dopamine
3-methyl-L-histidine
lysine
While the preliminary results shown in Table 2 illustrate the
possibility of carrying out absolute metabolite quantification using
the isotope dilution method with dimethylation. Additional work
is required to validate this method. To this end, we are planning
to study the analyte recovery issue during sample workup with
spiked samples, compare this method with other widely used
techniques such as LC-UV method, and investigate the run-to-
run and day-to-day reproducibility of the method. We envisage
that, once the method is validated with the amino acid standards,
we should be able to extend this method for absolute quantification
of many other amine-containing metabolites.
It should be noted that, in addition to allowing absolute
quantification, spiking a complex sample with isotopically labeled
Analytical Chemistry, Vol. 79, No. 22, November 15, 2007 8637

standards can facilitate the identification of targeted amines. This
can be done by looking at the extracted ion chromatogram of the
labeled standards: the analytes of interest in urine appear in the
same mass spectrum with a characteristic mass difference.
not been identified, they appear to be different metabolites based
on the accurate mass and retention time information. Future work
will involve expanding the list of 35 standards to many more
biologically relevant amines and apply this isotope dilution strategy
to generate quantitative profiles of amine-containing metabolites
in different natures of complex samples, such as urine samples
collected from disease and control objects and cells of different
states. We note that, while the method can potentially be used
for high-throughput quantitative profiling of amine-containing
metabolome, the major challenge is for the identification of
unknown metabolites. To this end, we are in the process of
developing a human metabolome MS/MS database, which should
facilitate the identification of unknown metabolites.28
CONCLUSIONS
In this work, we examined the use of reductive amination to
provide isotope tags for amine-containing metabolites. The simple
and rapid labeling reaction is carried out under very mild
conditions with minimum sample and reagent manipulation. The
targeted amine-containing metabolites could be quantitatively
derivatized under controlled reaction conditions. The dimethylated
products were stable at room temperature, and the ionization
efficiency of the products was not compromised. This labeling
strategy essentially produces inexpensive 13C- or deuterium-labeled
analogues of targeted amines for stable-isotope dilution analysis
in RP and HILIC LC-ESI MS. This method was proved to be
effective to overcome matrix effects. It was shown to be feasible
to quantify amine-containing metabolites in the complex biofluid,
human urine. Absolute quantification can be performed using this
strategy as long as the parent standards are available for di-
methylation labeling. Relative quantification can be easily per-
formed on biological samples by differential isotope labeling.
Another advantage of labeling is that the characteristic mass
difference between the heavy and light dimethylamines provides
additional information to facilitate peak and compound identifica-
tion. Using FT-ICR-MS combined with one-dimensional RP LC
separation, we detected a total of 438 ion pairs of amine-containing
metabolites in a human urine sample. Although most of them have
ACKNOWLEDGMENT
This work was funded by Genome Canada through Genome
Alberta’s Human Metabolomics Toolbox project and the Canada
Research Chairs Program. We thank the Canada Foundation for
Innovation and the Alberta Science and Research Investments
Program for providing the equipment fund for purchasing the FT-
ICR-MS instrument.
SUPPORTING INFORMATION AVAILABLE
A list of 33 ion pairs detected and identified by RP and HILIC
LC-ESI FT-ICR-MS from a mixture of 1:1 human urine samples
labeled by 12C- and 13C-formaldehyde, respectively (Table S1), and
a list of additional 420 ion pairs detected by RP LC-ESI FT-ICR-
MS (Table S2). This material is available free of charge via the
Internet at http://pubs.acs.org.
Received for review March 3, 2007. Accepted August 16,
2007.
(28) Wishart, D. S.; Tzur, D.; Knox, C.; et al. Nucleic Acids Res. 2007, 35, D521-
D526.
AC0704356
8638 Analytical Chemistry, Vol. 79, No. 22, November 15, 2007