Collision-induced dissociation studies of caffeine in positive electrospray ionisation mass spectrometry using six deuterated isotopomers and one N1-ethylated homologue

Article abstract of DOI:10.1002/rcm.6520

Rationale: In order to deepen the understanding of electrospray ionisation collision-induced dissociation (ESI-CID) fragmentation reactions of xanthine derivatives for the identification of metabolites using low-resolution liquid chromatography/mass spectrometry (LC/MS) analysis, basic experiments using caffeine (1,3,7-trimethylxanthine) as model compound have been performed. Methods: Six deuterium isotopomers and one N1-ethylated homologue of caffeine have been synthesized and their ESI fragmentation spectra have been obtained by using LC/MS in combination with either standard or perdeuterated eluent mixtures. RESULTS One result of these studies is the finding that the positive charges of the ESI-CID caffeine fragments are caused by the addition of protons. Furthermore, the performed experiments allow the determination of all molecular formulae of each ESI-CID caffeine fragment. Conclusions: As basic CID reactions of caffeine have been elucidated in this work, the developed fragmentation scheme may serve as a valuable tool for the interpretation of ESI-CID fragmentation spectra of more complex xanthine derivatives and their respective metabolites. Copyright

Full text of DOI:10.1002/rcm.6520

Research Article
Received: 22 October 2012
Revised: 19 January 2013
Accepted: 21 January 2013
Published online in Wiley Online Library
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
(wileyonlinelibrary.com) DOI: 10.1002/rcm.6520
Collision-induced dissociation studies of caffeine in positive
electrospray ionisation mass spectrometry using six deuterated
isotopomers and one N1-ethylated homologue
1
2
1
*
Dirk Bier , Rudolf Hartmann and Marcus Holschbach
1Institut für Neurowissenschaften und Medizin (INM-5), Forschungszentrum Jülich GmbH, 52425 Jülich, Germany
2Institute for Complex Systems (ICS-6), Forschungszentrum Jülich GmbH, 52425 Jülich, Germany
RATIONALE: In order to deepen the understanding of electrospray ionisation collision-induced dissociation (ESI-CID)
fragmentation reactions of xanthine derivatives for the identification of metabolites using low-resolution liquid chroma-
tography/mass spectrometry (LC/MS) analysis, basic experiments using caffeine (1,3,7-trimethylxanthine) as model
compound have been performed.
METHODS: Six deuterium isotopomers and one N1-ethylated homologue of caffeine have been synthesized and their
ESI fragmentation spectra have been obtained by using LC/MS in combination with either standard or perdeuterated
eluent mixtures.
RESULTS: One result of these studies is the finding that the positive charges of the ESI-CID caffeine fragments are caused
by the addition of protons. Furthermore, the performed experiments allow the determination of all molecular formulae of
each ESI-CID caffeine fragment.
CONCLUSIONS: As basic CID reactions of caffeine have been elucidated in this work, the developed fragmentation
scheme may serve as a valuable tool for the interpretation of ESI-CID fragmentation spectra of more complex xanthine
derivatives and their respective metabolites. Copyright © 2013 John Wiley & Sons, Ltd.
The xanthine (3,7-dihydro-purine-2,6-dione) structure is an ubi-
quitous heterocyclic scaffold and has, inter alia, been used
extensively as a lead template for the design and development
of adenosine receptor ligands.[1,2] Our main research is focused
on the development of new radiolabelled small molecules for
the non-invasive in vivo imaging of adenosine receptors in the
human brain using positron emission tomography (PET).[3]
Some years ago we reported on the synthesis, radiosynthesis,
and preclinical evaluation of the potent and selective A1
adenosine receptor (A1AR) antagonist 8-cyclopentyl-3-(3-
fluoropropyl)-1-propylxanthine (CPFPX), and its fluorine-18
radiolabelled isotopomer [18F]CPFPX.[4] CPFPX is a fluorinated
bioisistere of the well-established A1AR antagonist 1,3-
dipropyl-8-cyclopentylxanthine DPCPX.[5]
The clinical routine use of this radioligand is limited by
some unfavorable physiological properties, one being its
liability to fast enzymatic degradation. Consequently, new
projects aim at the development of ’second-generation’
radioligands which should overcome the disadvantages of
[18F]CPFPX without concomitantly impairing the essential
binding properties such as affinity and selectivity for the
A1AR subtype. Thus, the main improvement of follow-up
radioligands will be a refinement of metabolic in vivo stability
which should enhance radiotracer availability in plasma and
hence brain uptake. In order to achieve this goal it is crucial
to elucidate the metabolism of CPFPX and to unequivocally
identify the main metabolite(s).
Because the amount of radiotracer administered to humans
is so small, typically 0.5–5 nmol, or even less,[6] it is extremely
difficult or even impossible to detect those compounds with
the most advanced mass spectrometric (MS) methods. In
order to identify metabolites of such compounds in human
studies, additional steps are required.
Those radiolabelled – nearly massless – compounds are
analytically accessible only by radio-chromatographic meth-
ods. Chromatographic comparison of radioactive metabolites
from human blood with compounds generated by a suitable
model system (e.g. in vitro metabolites by human liver micro-
somes (HLM)) give a first indication of their chemical identity.
Compounds obtained from a suitable model system can be
further analysed applying high-performance liquid chroma-
tography (HPLC)/MS techniques.
Our previous work has demonstrated that spectra of
simple collision-induced dissociation (CID) fragmentation
reactions principally allow to obtain reliable information
about the identity of metabolites without using a collision
chamber.[7] The fragmentation scheme used to obtain
structural information about the metabolites of CPFPX was
derived from MS spectra of reference compounds CPFPX
and DPCPX. An elaborated study could reveal that the
chemical structure of the main metabolite of CPFPX in
* Correspondence to: D. Bier, Institut für Neurowissenschaften
und Medizin (INM-5), Forschungszentrum Jülich GmbH,
52425 Jülich, Germany.
E-mail: d.bier@fz-juelich.de
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
Copyright © 2013 John Wiley & Sons, Ltd.

D. Bier, R. Hartmann and M. Holschbach
human blood, tentatively identified by HPLC/MS, is in full
agreement with that of a reference compound obtained via a
multistep organic synthesis.[8]
our point of view it was most likely that the results obtained
with caffeine might help to elucidate and understand the frag-
mentation pathways of other synthetic xanthines.
From a retrospective point of view the identification of
metabolites originating from the synthetic xanthine CPFPX
should theoretically have been possible with the knowledge
extracted from existing publications which deal with various
MS-fragmentation mechanisms of caffeine.[9–13] However, the
published experiments were performed either by using
advanced spectrometers in combination with additional
ionisation modes like collision-activated dissociation (CAD)
or by a priori using different ionisation methods like electron
ionisation (EI). It seemed not reasonable to us to directly
translate the published results into simple CID fragmentation.
It cannot be excluded that, depending on the particular
fragmentation mode (CID or collision chamber), fragments
having identical m/z values but belonging to different
chemical species (radicals or ions), may be obtained. Thus, the
aim of the present study was to investigate the ESI fragmenta-
tion of caffeine as a reference xanthine in deeper detail. From
Regardless of the successful results of the studies mentioned
above a more general fragmentation scheme for chemically
different xanthines would be desirable. Such a tool could be
helpful to minimise time-consuming and cost-intensive
preparative organic syntheses of putative metabolites.
In order to expand the scope in the present work six differ-
ently deuterated isotopomers (2–7) and one N-ethylated deri-
vative (8) of caffeine (1) have been synthesized with two
derivatives, namely 6 and 7, being new compounds (Fig. 1).
By performing HPLC/MS experiments using those deriva-
tives 2–7 in combination with deuterated eluents and compu-
ter-based analyses it has been possible to clearly determine
the molecular formulae of all caffeine fragments found by
low-resolution CID. One recently published study used a col-
lision fragmentation cell and a deuterated caffeine isotopo-
mer for the clarification of possible fragmentation reactions
and was the motivation for the present work.[12]
Figure 1. Chemical syntheses of caffeine isotopomers 1–7, and N3-ethyl caffeine derivative 8.
wileyonlinelibrary.com/journal/rcm
Copyright © 2013 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2013, 27, 885–895

Collision-induced dissociation studies of caffeine
EXPERIMENTAL
For HPLC/MS studies a Kromasil 100-5 C 18 column
(250 ꢁ 4.6 mm; CS-Chromatographie Service GmbH,
Langerwehe, Germany) and either the eluent specified above
or the respective perdeuterated eluent (D2O/methanol-d4/acetic
acid-d4, 30/70/0.2 (v/v/v)) at a flow rate of 0.8 mL/min
was used.
General
Methanol, water and caffeine (1) were purchased from Merck
(Darmstadt, Germany) and all other solvents and reagents
were obtained from Sigma-Aldrich (Taufkirchen, Germany).
Solvents and reagents in the highest state of purity were used
as supplied by the vendors.
Mass spectrometry
Melting points were measured on a B 545 melting point
apparatus (Büchi Labortechnik, Flawil, Switzerland) and are
uncorrected. 1H-NMR and proton-decoupled 13C-NMR spec-
tra (given in Tables 1 and 2, respectively) were obtained in
CDCl3 at 200.13 and 50.32 MHz, respectively, by means of a
Bruker DPX-200 Avance instrument (Bruker Bio Spin GmbH,
Rheinstetten, Germany) in 5% solution at 25ꢀC. All shifts are
given in d ppm and are referenced to tetramethylsilane
(TMS) as an internal standard (d = 0 ppm). Coupling con-
stants J are given in Hertz. Protons were assigned with the
aid of COSY, and carbons with the aid of DEPT and HMQC
experiments. The multiplicity symbols s and d refer to singlet
and doublet, respectively.
For HPLC/MS measurements the outlet of the UV detector
was coupled to a mass spectrometer (Surveyor MSQ; Thermo
Fisher Scientific GmbH, Dreieich, Germany) with an electro-
spray interface. Nebuliser gas pressure was 4 bar and
desolvation temperature was 575ꢀC. The sprayer voltage
and the cone voltage were 3000 and 80–100 V, respectively.
For the breakdown graph, cone voltages of 20, 40, 60, 70, 80,
90, 100 and 110 V were used. Positive ion spectra were
recorded over a m/z range of 1–298 at a scan time of 2 s. The
XcaliburW software (version 1.4) provided with the instru-
ment permitted scans of the chromatograms for ions of a
desired m/z over the range of 50–250.
For analyses of the synthesized compounds the same
parameters were used but the cone voltage was set to 20 V.
High-performance liquid chromatography
Analysis of caffeine analogues employed a Kromasil 100-5 C 18
column (250 ꢁ 4.6 mm; CS-Chromatographie Service GmbH,
Langerwehe, Germany) in a HPLC system consisting of a
WellChrom K-1001 pump (Knauer, Berlin, Germany), a K-2001
UV detector (Knauer) and a manual sample injector (type
7125; Rheodyne, Bensheim, Germany) fitted with a 500 mL
sample loop. Isocratic elution with water/methanol/acetic acid
(30/70/0.2 v/v/v) was at a flow rate of 1 mL/min and UV
monitoring at 275 nm.
General procedure for the synthesis of mono-trideuteromethyl
caffeine isotopomers 2–4 and N1-ethyl caffeine 8 (Fig. 1)
Deuteromethylation of 3,7-dimethylxanthine (theobromine, 1a),
1,7-dimethylxanthine (paraxanthine, 1b), 1,3-dimethylxanthine
(theophylline, 1c), and xanthine (1c) was performed according
to a classical alkylation reaction described using an alkyl halide
in the presence of a base.[12,14]
Table 1. 1H-NMR spectral data of compounds 1–7 (d-values in ppm)
cpd
N1-Me
N3-Me
N7-Me
H8
4
1
2
3
4
5
6
7
3.41 (s)
absent
3.43 (s)
3.43 (s)
absent
absent
3.42 (s)
3.59 (s)
3.61 (s)
absent
3.61 (s)
absent
absent
3.60 (s)
4.00 (d, JH7-H8 = 0.6 Hz)
7.53 (d, 4JH8-H7 = 0.6 Hz)
4.02 (s)
7.53 (s)
7.54 (s)
7.55 (s)
7.55 (s)
absent
4.01 (d, 4JH7-H8 = 0.59 Hz)
absent
absent
absent
4.01 (s)
absent
Table 2. Proton decoupled 13C-NMR spectral data of compounds 1–7 (d-values in ppm)
cpd
N1-Me
N3-Me
N7-Me
C2
C4
C5
C6
C8
1
2
3
4
5
6
7
28.32
30.15
30.16
29.18*
30.21
29.19*
29.16*
30.17
34.00
34.01
34.04
32.97*
32.98*
32.96*
34.00
152.11
152.13
152.12
152.12
152.13
152.12
152.13
149.07
149.09
149.00
149.03
149.03
149.06
149.07
107.99
108.02
108.01
108.00
108.01
108.00
107.97
155.81
155.85
155.83
155.83
155.85
155.84
155.84
141.78
141.76
141.72
141.71
141.76
141.19**
141.17**
27.26 *
28.34
28.35
27.28 *
27.34 *
28.34
1
*septet, JC-D = 21.2 Hz,
1
**triplet, JC-D = 33.6 Hz
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/rcm

D. Bier, R. Hartmann and M. Holschbach
Under the exclusion of moisture the respective xanthine was
suspended in dry acetonitrile (7 mL/mmol), dry potassium car-
bonate (K2CO3, 1.2 eq.) was added, and the suspension stirred
at 80ꢀC for 30 min. After the addition of iodomethane-d3
(CD3I, 1.2 eq.) the mixture was refluxed for the time specified,
cooled to ambient temperature, filtered through a 0.2 m filter,
and the filtrate was evaporated to dryness in vacuo. The residue
was dissolved in an aqueous solution of sodium chloride (1 g)
and K2CO3 (100 mg) in H2O (100 mL)[12] and the aqueous solu-
tion was extracted three times with chloroform. The combined
organic extracts were evaporated to dryness in vacuo and the
obtained product recrystallized from ethanol.
Evaporation of the combined organic extracts to dryness gave
a solid residue (243 mg) which was recrystallized from
ethanol to furnish 155 mg (20%) of product, mp 236.1ꢀC.
HPLC: k0 2.22, purity 99.05%. MS (ESI) calc. for C8HD9N4O2,
exact mass 203.14, m/z 204.16 [M+H]+.
General procedure for the deuteration of carbon-8 in caffeines
via palladium-catalyzed H–D exchange reaction[15,16]
8-Deutero-1,3,7-tri-trideuteromethylxanthine, caffeine-d10, 6
At ambient temperature caffeine-d9 5 (240 mg, 1.17 mmol) was
dissolved in D2O (15 mL) and 10% Pd/C (24 mg) was added. A
gentle hydrogen flow was bubbled for 60 s through the solu-
tion. The flask was tightly stoppered and the contents stirred
at 80ꢀC for 70 h. The reaction mixture was filtered through a
0.2 m filter and the filtrate evaporated to dryness in vacuo to
yield 218 mg (90.8%) of product, mp 237.8ꢀC. HPLC: k0 2.18,
purity 99.5%.
1-Trideuteromethyl-3,7-dimethylxanthine, N1-d3-caffeine, 2
3,7-Dimethylxanthine 1a (532 mg, 2.94 mmol) in CH3CN
(20 mL), K2CO3 (490 mg, 3.53 mmol and CD3I (511 mg, 219 mL,
3.53 mmol), reaction time 3 h, crude yield 835 mg, dry CHCl3
extract yield 434 mg, recrystallized yield 303 mg (52.4%),
mp 235.4ꢀC. HPLC: k’ 2.278, purity 99.3%. MS (ESI) calc. for
C8H7D3N4O2, exact mass 197.10, m/z 198.1 [M+H]+.
8-Deutero-1,3,7-trimethylxanthine, caffeine-d1, 7
At ambient temperature 10% Pd/C (26 mg) was added to a
stirred solution of caffeine 1 (264 mg, 1.36 mmol) in D2O
(15 mL). At that temperature a gentle flow of hydrogen was
bubbled through the solution for 60 s. The flask was tightly
stoppered and the contents stirred at 90ꢀC for 1.5 h. After
cooling to ambient temperature the reaction mixture was fil-
tered through a 0.2 m filter and the filtrate was evaporated
to dryness in vacuo giving 241 mg (86.9%) of analytically pure
product as colourless crystals, mp 236.2ꢀC. HPLC: k’ 2.27,
purity 98.7%. MS (ESI) calc. for C8H9D1N4O2, exact mass
195.09, m/z 196.09 [M+H]+. MS (ESI) calc. for C8D10N4O2,
exact mass 204.17, m/z 205.18 [M+H]+.
3-Trideuteromethyl-1,7-dimethylxanthine, N3-d3-caffeine, 3
1,7-Dimethylxanthine 1b (485 mg, 2.7 mmol) in CH3CN (20 mL),
K2CO3 (442 mg, 3.2 mmol) and CD3I (463 mg, 199 mL, 3.2 mmol),
reaction time 3 h, crude yield 767 mg, dry CHCl3 extract yield
362 mg, recrystallized yield 268 mg (50.5%), mp 236.3ꢀC. HPLC:
k’ 2.264, purity 99.2%. MS (ESI) calc. for C8H7D3N4O2, exact
mass 197.10, m/z 198.0 [M+H]+.
7-Trideuteromethyl-1,3-dimethylxanthine, N7-d3-caffeine, 4
1,3-Dimethylxanthine 1c (563 mg, 3.1 mmol) in CH3CN
(20 mL), K2CO3 (514 mg, 3.7 mmol) and CD3I (540 mg, 232
mL, 3.7 mmol), reaction time 5.5 h, crude yield 913 mg, dry
CHCl3-extract yield 466 mg, recrystallized yield 376 mg
(61.6%), mp 236.4ꢀC. HPLC: k’ 2.256, purity 98.4%. MS (ESI)
calc. for C8H7D3N4O2, exact mass 197.10, m/z 198.1 [M+H]+.
NMR spectra of compounds 1–7
The 1H and 13C NMR spectra of compounds 2 to 7 were identi-
cal to an authentic standard of unlabeled caffeine 1, differing
only as italicized due to the incorporation of deuterium. The
1H-NMR and proton decoupled 13C-NMR spectral data of com-
pounds 1–7 are given in Tables 1 and 2, respectively.
3,7-Dimethyl-1-ethyl-xanthine, N1-ethyl-caffeine, 8
3,7-Dimethylxanthine 1a (400 mg, 2.2 mmol) in CH3CN
(20 mL), K2CO3 (400 mg, 2.9 mmol) and C2H5I (530 mg,
274 mL, 3.4 mmol), reaction time 15 h, dry CHCl3-extract
yield 130 mg, recrystallised yield 15 mg (3.2%), mp 164.7ꢀC.
HPLC: purity 98.32%. 1H-NMR (CDCl3): 1.25–1.32 (t, 3H,
N3CH2CH3), 3.60 (s, 3H, N7CH3), 4.02 (s, 3H, N7CH3),
4.06–4.17 (m, 2H, N1CH2CH4), 7.52 (s, 1H, C8H). MS (ESI)
calc. for C9H12N4O2, exact mass 208.1 m/z 209.02 [M+H]+.
Computer calculations
An in-house developed computer program (Listing 1) written
in Python[17] (version 2.6). Python runs on Windows, Linux/
Unix, Mac OS X, and has been ported to the Java and.
NET virtual machines. It is free to use, even for commercial
products, because of its open source license. The script calcu-
lated possible compositions of fragment ions [F+1]+ of given
masses taking into account only the elements in the molecu-
lar formula of caffeine. The number of ‘hits’ per fragment are
summarized in Table 3. If a composition of elements agrees
with the mass of a calculated fragment (Fc) the ‘double-bond
equivalents’ (DBE) parameter was computed.
1,3,7-Tri-trideuteromethylxanthine, caffeine-d9, 5
To a stirred suspension of xanthine 1d (517 mg, 3.83 mmol) in
CH3CN (25 mL) was added K2CO3 (700 mg, 5.07 mmol) and
the mixture was stirred at 70ꢀC for 1.5 h. After the addition of
CD3I (1.5 g, 10.3 mmol) the mixture was refluxed for 4 h and
an additional amount of CD3I (1.5 g, 10.3 mmol) was added.
After 4 h the temperature was set to 40ꢀC and stirring was
continued for 12 h. The mixture was filtered through a 0.2 m
filter and the filtrate was evaporated to dryness in vacuo.
The solid residue (889 mg) was dissolved in a solution of
sodium chloride (1 g) and K2CO3 (100 mg) in H2O (100 mL),
and the solution was extracted three times with chloroform.
RESULTS AND DISCUSSION
The use of isotopomers for the structural analysis of
fragments obtained in mass spectrometry has been estab-
lished for
a
long time.[18] Thus, specifically for this
wileyonlinelibrary.com/journal/rcm
Copyright © 2013 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2013, 27, 885–895

Collision-induced dissociation studies of caffeine
purpose, 15N-isotopomers of caffeine have already been
synthesized for the structural elucidation of its fragments
obtained in EI-MS experiments.[19]
MS provide the possibility to obtain fragments exclusively
by increasing the cone voltage to achieve a so-called ‘cone
voltage collision-induced dissociation’ (CID).
For the investigation of the fragmentation chemistry of
xanthine derivatives in the ESI-MS mode, in addition to other
unlabeled xanthines, a deuterated isotopomer of caffeine
has been described and examined.[12] For fragmentation
experiments the authors of this work have used a collision
cell; however, such a cell is not available in small and simple
mass spectrometers. Commonly used machines for routine
Some comparative studies show that many similarities
exist between CID fragments and those which are formed
in a collision chamber, particularly when comparing fragmen-
tations of pure protonated molecules.[20] Spectra obtained
from preliminary CID fragmentation experiments with caf-
feine revealed a very close similarity to a published spectrum
whose fragments had been generated by means of a collision
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/rcm

D. Bier, R. Hartmann and M. Holschbach
Table 3. Approach for the determination of the molecular formulae of fragments a–g using information from Table 1 and the
Python script. For structures of a–g, see Fig. 8
Step 1
Step 2
Step 3
Number
of H in
fragment
Number of
solutions with
steps 1 and 2
Number of
solutions
with step 3
Type
Number
of hits
Number
of N
Elemental
composition
F = [M+H]+
.
Fragment
m/z
R. = [F+H]+
a
b
c
195
181
138
123
110
83
1
2
10
8
-
-
-
C8H10N4O2
C7H8N4O2
C6H7N3O
C5H4N3O
C5H7N3
F
F
1
2
2
2
2
2
N ≥ 3
N ≥ 3
N ≥ 3
N ≥ 2
N ≥ 2
N ≥ 1
1
1
1
1
1
1
9
7
F
d
e
f
13
12
14
12
4
R.
F
F
F
7
6
C4H6N2
g
69
4
C3H4N2
chamber.[12] Intrigued by this study, it was our aim to gain a
deeper insight into the pure CID fragmentation chemistry of
xanthine derivatives. Using unlabelled caffeine as a model
substance we have additionally synthesized a number of
deuterated caffeine isotopomers and one caffeine derivative
bearing an ethyl instead of a methyl substituent at position
3 of the xanthine scaffold.
Syntheses
Several synthetic routes for the preparation of deuterated caf-
feine derivatives have been described in the literature.[12,14,21,22]
The syntheses of the 1-, 3-, and 7-mono-trideuteromethyl
isotopomers 2, 3, and 4, the 1,3,7-tris-trideuteromethyl
isotopomer 5 of caffeine and the N-3-ethyl derivative 7 (Fig. 1)
were carried out using modified literature procedures. In order
to obtain the C-8 deuterated caffeine isotopomer 7, various
described synthetic routes were investigated.[23,24]
Figure 2. ESI-CID ion spectrum of caffeine 1 at a cone voltage
of 90 V.
It appeared, however, that high reaction temperatures and
long reaction times, prerequisite conditions for a sufficiently
high degree of C-H/C-D isotopic exchange, resulted in poor
deuterium incorporation and led to products of low purity.
Moreover, uncatalysed reactions in D2O at 100ꢀC for
more than 10 h did not afford the target compound 9 in a
satisfactory product purity.[25]
Breakdown graphs are established tools for the interpreta-
tion of CID fragmentation processes.[26,27] They display the
percentage of a certain fragment population as a function of
the cone voltage and allow the interpretation of the genesis
and formation of the respective fragments. For caffeine no
fragments could be observed up to 40 V but raising the cone
voltage to about 70 V led to fragmentation. A further increase
of the cone voltage to 80 V caused the formation of all frag-
ments (Fig. 2); the respective breakdown graph of caffeine is
shown in Fig. 3.
Finally, a recently published isotopic exchange method was
adopted, namely the heterogeneous Pd-catalysed reaction of
a solution of caffeine in a D2O medium previously saturated
with hydrogen.[15] The Pd-catalysed exchange method was
also used to obtain the perdeuterated caffeine isotopomer 6
from compound 5.
Ion species
All resulting products were analysed by 1H-NMR, 13C-
NMR, ESI-MS and HPLC and contained no or maximal 2%
measurable impurities as determined by HPLC.
In order to obtain information about the nature and identity
of the individual ion species formed in the fragmentation
experiments, a perdeuterated eluent with a deuteration grade
of 99.98% was used. In the past, deuterated eluents have been
used in mass spectrometric experiments to detect labile
protons such as hydrogen atoms bound to heteroatoms, for
example in amino alcohols.[28,29]
Although there are no labile hydrogens in the parent
compound 1, this technique was used in the present study
in order to obtain more detailed information about the nature
of the detected ions. As shown in Fig. 4, using perdeuterated
eluents, the m/z values of fragments a, c, d, e, f, and g
increased by one m/z value compared to those (Fig. 2)
ESI–MS experiments
General experiments
Breakdown graphs. Figure 2 shows a typical ESI mass spectrum
of caffeine 1 which was obtained at a cone voltage of 80 V.
The fragmentation pattern is consistent with that of earlier
published spectral data.[9,12] Considerations about the forma-
tion, the composition, and the chemical structures of the
obtained fragments are the subject of the following discussion
(vide infra).
wileyonlinelibrary.com/journal/rcm
Copyright © 2013 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2013, 27, 885–895

Collision-induced dissociation studies of caffeine
Figure 3. Breakdown graph of caffeine 1.
Figure 5. ESI-CID ion spectrum of caffeine-d10 6 at a cone
voltage of 90 V using a normal eluent system. Delta values
indicate the differences in m/z values in comparison to those
in Fig. 1.
Figure 4. ESI-CID ion spectrum of caffeine 1 at a cone voltage
of 90 V using a deuterated eluent system.
obtained using ’normal’ hydrogen-containing eluents. This
finding suggests that the charge of these fragments (F) results
from addition of a deuteron and that these species can
formally be written as protonated molecules such as [F+D]+.
Figure 6. ESI-CID ion spectra of N3-ethyl caffeine 8 at a cone
voltage of 80 V using (A) a normal eluent system and (B) a
deuterated eluent system.
Number of hydrogen atoms
(Fig. 2). It can be supposed that a nitrogen atom is still present
in a fragment if the corresponding (bound) methyl group can
be detected. The minimum number of nitrogen atoms in a given
fragment should therefore at least equal the number of methyl
groups still present in the molecule.
To obtain further information about the chemical nature of
the observed fragments, a caffeine isotopomer was synthe-
sized in which all hydrogen atoms are substituted by deuter-
ium atoms (caffeine-d10, 6). The difference between the m/z
values of the perdeuterated analogue 6 (Fig. 5) and the m/z
values of the corresponding fragments of ‘normal’ caffeine 1
(Fig. 2) furnishes the number of hydrogen atoms in a given
fragment. In Fig. 5 these values are tagged with a delta.
Molecular formulae
With this information about the individual fragments
(summarized in Table 3) it is possible to calculate the molecu-
lar formula of a particular fragment and to clearly determine
whether this species is a radical or a saturated protonated
molecule.
For this purpose, a simple Python (2.7) script was written.
Python has the advantages of being open-source, is available
for all relevant computer platforms and forces a human-
readable source code. The execution speed of the interpreter
language is sufficiently fast on modern personal computers
and has already been used in the past for mass spectrum
analysis.[30]
Remaining nitrogen atoms
Further fragmentation analysis of the synthesized isotopomers
gives evidence about the remaining nitrogen atoms depending
on the methyl groups detectable in a fragment. This informa-
tion can easily be obtained, since the three possible mono-
trideuteromethyl derivatives 2, 3, and 4 (Fig. 1) had been
synthesized for the present work. The results of these experi-
ments are summarized in Table 4. The delta values are the
differences between the m/z values of deuterated fragments
and those of the corresponding values of native caffeine 1
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/rcm

D. Bier, R. Hartmann and M. Holschbach
Table 4. Peaks of fragments resulting from ESI CID spectra of caffeine isotopomeres. Delta values indicate the difference
between m/z values of one given fragment and the m/z value of the corresponding caffeine peak in column 1. For structures
of compounds 1–7, see Fig. 1
N1,3,7-tris-
deuteromethyl
(5)
Caffeine with
perdeuterated
eluent
Caffeine (1)
N1d3 (2)
N3d3 (3)
N7d3 (4)
C8d1 (7)
D10 (6)
a
b
c
195
181
138
123
110
83
204
Δ 9
187 (vw)
Δ 6
144
Δ 6
126
Δ 3
116
Δ 6
89
198
Δ 3
184 (vw)
Δ 3
138
198
Δ 3
181 (vw))
Δ 0
198
Δ 3
181 (vw)
Δ 0
196
Δ 1
182 (vw)
Δ 1
139
205
Δ 10
n.d.
196
Δ 1
n.d.
141
Δ 3
141
Δ 3
145
Δ 7
127
Δ 4
117
Δ 7
89
Δ 6
73
Δ 4
139
Δ 1
124
Δ 1
111
Δ 1
84
Δ 1
70
Δ 1
Δ 0
123
Δ 0
110
Δ 0
83
Δ 0
69
Δ 0
Δ 1
124
Δ 1
111
Δ 1
83
Δ 0
70
Δ 1
d
e
f
123 (126)15%
Δ 0 (Δ 3)
113
126 (123)15%
Δ 3 (Δ 0)
113
Δ 3
Δ 3
86
86
Δ 6
72
Δ 3
Δ 3
Δ 3
g
69
72
69
Δ 3
Δ 0
In a nested loop construction starting with the maximum
number of available atoms of each element all possible
compilations of molecular compositions are calculated. The
results are summarized in Table 3 (step 1) which shows the
number of hits for each fragment a–g.
The next step to reduce this number is to consider the
number of hydrogen atoms in the fragment, which is
obtained by comparing the spectra of unlabelled caffeine 1
and caffeine-d10 6. These results are listed in Table 4.
Taking into account the minimum number of available
nitrogen atoms for a given fragment, it can be deduced
that only one possible elemental composition exists for the
respective fragment (Table 3, step 3).
Fragmentation mechanisms
Fragment b
Although fragment b does not provide an unambiguous
signal (Fig. 2), it is most likely formed by a demethylation
reaction followed by subsequent addition of hydrogen
[M–CH3+2H]+. In order to verify this hypothesis, compound
8, an N1-ethylated homologue of caffeine, was synthesised
and used for further MS investigations.
In contrast to peak b in the caffeine spectrum (Fig. 2) the
corresponding peak b’ in the mass spectrum of 8 at m/z 181
(Fig. 6(A)) appears as a prominent peak. In the presence of a
deuterated eluent, peak b’ of homologue 8 appears at m/z
182 (Fig. 6(B)). This indicates that the fragmentation process
of the ethyl group takes place via an ethyl/hydrogen
exchange reaction and that addition of a positively charged
deuteron (D+), absorbed from the eluent, is responsible for
the positive charge of the resulting ion.
The observed m/z value of 182 raises the question on the
origin of the hydrogen atom since exclusive homolytic
cleavage of the ethyl group and subsequent addition of a
positive deuteron would lead to a fragment with an m/z
value of 181. Obviously, addition of a hydrogen atom or
radical to the initially formed desethyl radical must occur
after the homolytic cleavage of the ethyl group. However,
the addition of a small charged particle, e.g. a proton, would
decrease the m/z value to 180. The source of the hydrogen
atom is not the deuterated eluent, because these experiments
exclusively led to a fragment having an m/z value of 182. The
above findings and considerations lead to the conclusion
that the parent molecule 8 itself acts as a hydrogen donor
during the fragmentation.
As shown in Fig. 7, an increase in the cone voltage leads
to a relatively increased appearance of ions of the type [M
+H]+1 – n (n = 1 . . . 4). Assuming that the molecules were
completely isolated in a vacuum, the hydrogen transfer could
only take place by an intramolecular process that would give
rise to a m/z value of 180 ([M+H-H-CH2CH3]+1).
Double-bond equivalents
Finally, further consideration of possible ‘double-bond
equivalents’ is helpful for the decision whether a given frag-
ment is an unsaturated radical or a saturated protonated
molecule. This parameter is also calculated by the Python
script.
The total number of rings and sites of unsaturation can be
calculated from any elemental composition by using the fol-
lowing formula[31]
:
D ¼ 1 þ 0:5 ꢂ i Ni ꢂ ðVi ꢃ 2Þ
max
X
i
where D is the unsaturation, imax is the total number of
different elements in the composition, Ni the number of atoms
of element i, and Vi is the valence of atom i.
The conclusions that can be drawn from the experiments of
the present work compared to the ESI-CID fragmentation
processes of caffeine are summarized in Fig. 8. Although the
chemical structures of the fragments are only postulated, both
their molecular formulae and the origin of the atoms from
the parent caffeine molecule are strongly supported by the
presented arguments.
wileyonlinelibrary.com/journal/rcm
Copyright © 2013 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2013, 27, 885–895

Collision-induced dissociation studies of caffeine
Therefore, it seems to be reasonable to postulate that the
hydrogen transfer occurs in an intermolecular fashion, but this
transfer is unlikely to take place between spatially separated
and hence isolated structures. This type of fragments will
hereinafter be referred to as ‘wet fragments’, because the
particle transfer must take place in a matrix.
isolated so that there are actually no more solvent molecules
which can act as a hydrogen transfer matrix. In the present
work these fragments are referred to as ‘dry fragments’.
Fragment e
During the formation of all other fragments, as shown by c in
Fig. 8, ESI fragmentation processes may be explained by bond
scission of carbon–heteroatom bonds; only the reaction of
fragment c to form fragment e involves a carbon–carbon
bond scission with the elimination of carbon monoxide. One
possible explanation for this exception is the existence of a
very unstable synthon.
Indeed it is known from the literature that the parent
gamma-delta unsaturated b-lactam of fragment c, namely
N-methylazetinone cb, has an extremely limited thermal
stability.[33,34] Those authors claim that lactam cb exists in
two tautomeric forms, in the cyclic b-lactam form ca and in
the (preferred) open-chain imino ketene form cb. The relief
of ring strain together with a low activation energy associated
with such a transformation are sufficient to account for the
conversion of ca into cb.[34]
Fragment c
This fragment arises directly from the parent substance a by a
neutral loss of methyl isocyanate (CH3NCO). In the literature,
the formation of this fragment is explained by a retro-Diels-
Alder reaction and the resulting molecule is consequently
presented as an open-chain structure.[10,12,19] Taking into
account the calculated empirical formula C6H7N3O (Table 3)
and the calculated number of ‘double-bond equivalents’ a
cyclic b-lactam structure for fragment c seems to be feasible.[9]
Except for fragment b all fragments observed directly or
indirectly arise from fragment c.
The first fragmentation reaction (a to c) leads to the
elimination of the N1 substituent of the xanthine core; the
same is true for more complex substituted xanthines.
The structure of fragment e shown in Fig. 8 is consistent
with the results shown in Table 3. Fragment e contains the
two methyl groups stemming from nitrogens N3 and N7
and the hydrogen atom from C8 of the original molecule a.
Fragment d
All fragment peaks found are isomerically pure except peak d
in Table 4 (m/z 123) which is a mixture of two isomers. This
peak splits into two different fractions which are visible in
the spectra of the isotopomers N3-d3-caffeine, 3, and N7-d3-
caffeine, 7. Obviously, these fragments differ in the position
of one methyl group.
The results clearly show (Table 3) that only peak d with m/z
123 is a protonated odd radical, which can be observed at cone
voltages higher than 70 V and therefore it is an example for an
exception of the ‘Even-Electron Rule’.[32] Looking at the postu-
lated fragmentation pathway for caffeine (Fig. 8) it becomes
clear that the radical species d could arise either from the frag-
ment b or fragment c. During the formation of fragment d no
addition of hydrogen is observable as is the case in the reactions
leading to the fragment with m/z 182 in Fig. 6(B) (vide supra).
Obviously, these radical-forming reactions take place when a
molecule has already travelled a longer distance and is spatially
Fragments f and g
Fragments f and g can result from the decay of the reactive
fragment e after neutral loss of HCN (f) or methyl cyanide
(g), respectively. It is clearly shown in Table 4 that in fragment
f the N1 and N7 methyl groups of parent compound a are
retained, whereas, in fragment g, only the N3 methyl group
of parent compound a is conserved. The structure in curly
brackets in Fig. 8 shows a little thicker drawn the origin from
the parent compound.The most probable structural embed-
ding of these two fragments into the parent molecule a is
represented in the structures shown in curly brackets in Fig. 8.
The characterized fragments obtained using a simple CID
fragmentation protocol nicely correspond to fragments
described earlier in the literature.[9–13,35] Although more
Figure 7. Details of ESI-CID spectra of caffeine 1 recorded at different cone
voltages: (A) at 20 V, (B) at 60 V, and (C) at 90 V.
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/rcm

D. Bier, R. Hartmann and M. Holschbach
Figure 8. Proposed fragmentation pathway of the protonated caffeine molecule after CID.
sophisticated experimental MS paradigms have been used,
the resulting information is very similar to that obtained in
the present study.
accordance with a homolytic bond cleavage followed by the
addition of a hydrogen atom so as to form a neutral molecule.
Since the first step in the fragmentation is a homolytic
bond scission the addition of hydrogen is formally that of a
hydrogen radical. It is likely that this hydrogen transfer takes
place in a (partially) solvated state (‘wet’ fragments) but it is
implausible that these hydrogen atoms stem from the
perdeuterated eluent. Consequently it is most likely that the
hydrogen atoms originate from the parent compound
caffeine. In the second fragment type no addition of hydrogen
takes place after bond scission and accordingly these
fragments are radicals which arise from originally unsolvated
molecules (‘dry’ fragments).
CONCLUSIONS
The present study clearly reveals that the positive charge
found on every caffeine fragment (F), formed by CID during
ESI-MS investigations in the positive ion mode, originates
from proton additions and results in the formation of species
such as [F+H]+ and [R.+H]+. This could unambiguously be
shown by experiments using perdeuterated eluents.
Additionally, the fragmentation spectrum of perdeuterated
caffeine allows to precisely specify the exact number of
hydrogen atoms per fragment ion equalling the m/z value dif-
ferences between the m/z values of the deuterated fragments
and the m/z values of the corresponding peaks in fragmenta-
tion spectra obtained using hydrogenated (‘normal’) caffeine.
With regard to the number of hydrogen atoms in a given
fragment in combination with a computer-aided method for
the calculation of the elemental composition and following
the rule ‘fragment parameter is even’ (vide supra), both the
ion type of the fragment ([F+H]+ or [R.+H]+) and its molecu-
lar formula can unequivocally be defined.
Through knowledge of the molecular formulae of the
fragments combined with conclusions drawn from some
well-known fragmentation reactions for alkylxanthines (e.g.
N-dealkylations), the deduced chemical structures presented
in this work seem to be reasonable. Knowledge of the
described CID fragmentation reactions allows the assembly
of fragmentation patterns which may prove to be very useful
for the structural elucidation of xanthine metabolites in
future studies.
Caffeine fragments which have been identified as radical
species most likely result from homolytic bond cleavage and
are exceptions to the ‘Even-Electron Rule’.
The formation of all fragments except one may be explained
by carbon–heteroatom bond scission. Only the formation of one
fragment involves a carbon–carbon bond scission, most likely
due to the existence of a very unstable species.
A closer examination of the formed species suggests that two
fragment types can be distinguished from N-dealkylation
reactions. They are apparently generated depending on the
chemical environment: mechanistically the first type is in
REFERENCES
[1] U. Schwabe, D. Ukena, M.J. Lohse. Xanthine derivatives
as antagonists at A1 and A2 adenosine receptors.
Naunyn-Schmiedeberg’s Arch. Pharmacol. 1985, 330, 212.
[2] K.N. Klotz. Adenosine receptors and their ligands.
Naunyn-Schmiedeberg’s Arch. Pharmacol. 2000, 362, 382.
[3] P.T. Meyer, D. Bier, M. H. Holschbach, C. Boy, R. A. Olsson,
H. H. Coenen, K. Zilles, A. Bauer. Quantification of cerebral
A1 adenosine receptors in humans using [18F]CPFPX and
PET. J. Cerebr. Blood F. Met. 2004, 24, 323.
wileyonlinelibrary.com/journal/rcm
Copyright © 2013 John Wiley & Sons, Ltd.
Rapid Commun. Mass Spectrom. 2013, 27, 885–895

Collision-induced dissociation studies of caffeine
[4] M.H. Holschbach, R.A. Olsson, D. Bier, W. Wutz, W. Sihver,
M. Schüller, B. Palm, H.H. Coenen. Synthesis and evalua-
tion of no-carrier-added 8-cyclopentyl-3-(3-[18F] fluoropro-
pyl)-1-propylxanthine ([18F]CPFPX): a potent and selective
A1-adenosine receptor antagonist for in vivo imaging.
J. Med. Chem. 2002, 45, 5150.
[5] M.J. Klotz, K.N. Lindenborn-Fotinos, J. Reddington,
M. Schwabe, U. Olsson, R. A. Lohse. 8-Cyclopentyl-1,3-
dipropylxanthine (DPCPX) – a selective high affinity antagonist
radioligand for A1 adenosine receptors. Naunyn-Schmiedeberg’s
Arch. Pharmacol. 1987, 336, 204.
aromatic compounds using heterogeneous Pd/C in D2O.
Org. Lett. 2004, 6, 1485.
[17] Python Programming Language
Available: http://www.python.org.
–
official website.
[18] C. Djerassi. Isotope labelling and mass spectrometry of
natural products. Pure Appl. Chem. 1964, 9, 159.
[19] A. Kenani, J. Bernier, J. Henichart. Synthesis and EIMS
fragmentation analysis of [1,3-15N2]xanthine and [1,3-15N2]
caffeine. J. Labelled Compd. Rad. 1995, 36, 187.
[20] C. Bure, C. Lange. Comparison of dissociation of ions in an
electrospray source, or a collision cell in tandem mass
spectrometry. Curr. Org. Chem. 2003, 7, 1613.
[21] F. Balssa, Y. Bonnaire. Easy preparative scale syntheses of
labelled xanthines: caffeine, theophylline and theobromine.
J. Labelled Compd. Rad. 2007, 50, 33.
[22] W.M. Schlager, J.J. Madden, R.J. Hurst, H.E. Pierce Jr. A
simple, rapid synthesis of caffeine-1, 7-13CH3. J. Labelled
Compd. Rad. 1984, 21, 187.
[23] W. Traube. Der synthetische Aufbau der Harnsäure, des
Xanthins, Theobromins, Theophyllins und Caffeins aus der
Cyanessigsäure. Ber. Dtsch. Chem. Ges. 1900, 33, 3035.
[24] H. Biltz, F. Max. Alkylierung des Theobromins. Liebigs Ann.
Chem. 1921, 423, 318.
[6] D. Bier, K. Dutschka, E.J. Knust. Radiochemical synthesis of
[
123I]2-iodo-lisuride for dopamine D2-receptor studies. Nucl.
Med. Biol. 1996, 23, 373.
[7] D. Bier, M.H. Wutz, W. Olsson, R.A. Coenen, H.H.
Holschbach. Metabolism of the A1 adenosine receptor
positron emission tomography ligand [18F]8-cyclopentyl-3-
(3-fluoropropyl)-1-propylxanthine ([18F]CPFPX) in rodents
and humans. Drug Metab. Dispos. 2006, 34, 570.
[8] M.H. Holschbach, D. Bier, W. Wutz, S. Willbold, R.A.
Olsson. Synthesis of the main metabolite in human blood
of the A(1) adenosine receptor ligand [18F]CPFPX. Org. Lett.
2009, 11, 4266.
[9] F. Beaudry, J.P. Vachon, P. Lavoie. Development of an
electrospray ionization mass spectrometric method for the
quantification of theophylline in horse serum. Biomed.
Chromatogr. 2005, 19, 643.
[10] G. Bianco, S. Abate, C. Labella, T.R.I. Cataldi. Identifica-
tion and fragmentation pathways of caffeine metabolites
in urine samples via liquid chromatography with
[25] M. Maeda, M. Saneyoshi, Y. Kawazone. Studies on hydro-
gen exchange. XII. Reaction mechanism for hydrogen
exchange of C-8 hydrogen of purin ribosides. Chem. Pharm.
Bull. 1971, 19, 1641.
[26] N.N. Dookeran, T. Yalcin, A. G. Harrison. Fragmentation
reactions of protonated alpha-amino acids. J. Mass Spectrom.
1996, 31, 500.
positive electrospray ionization coupled to
a
hybrid
[27] A.G. Harrison. Energy-resolved mass spectrometry: a com-
parison of quadrupole cell and cone-voltage collision-induced
dissociation. Rapid Commun. Mass Spectrom. 1999, 13, 1663.
[28] K.E. Karlsson. Deuterium oxide as a reagent for the modifica-
tion of mass spectra in electrospray microcolumn liquid chro-
matography-mass spectrometry. J. Chromatogr. A 1993, 647, 31.
[29] M.A. Cummings, P.G. Kennedy-Gabb, S. Wagner, B.M.
Nicol, G.R. Munson, B. Olsen. The use of deuterium oxide
as a mobile phase for structural elucidation by HPLC/
UV/ESI/MS. Anal. Chem. 2000, 72, 5070.
[30] R. Winkler. ESIprot: a universal tool for charge state
determination and molecular weight calculation of proteins
from electrospray ionization mass spectrometry data. Rapid
Commun. Mass Spectrom. 2010, 24, 285.
[31] V. Pellegrin. Molecular formulas of organic compounds:
the nitrogen rule and degree of unsaturation. J. Chem. Educ.
1983, 60, 626.
[32] M. Karni, A. Mandelbaum. The ‘even-electron rule’. Org.
Mass. Spectrom. 1980, 15, 53.
[33] G. Ege. Photolyse von 3.4-Dihydro-4-oxo-benzo-1.2.3-triazinen,
die in 3-Stellung substituiert sind. Angew. Chem. 1965, 77, 723.
[34] G. Kretschmer, R.N. Warrener. A photochemical route to the
unsaturated b-lactam, N-methyl-azetinone: A thermally labile
ring-system. Tetrahedron Lett. 1975, 16, 1335.
[35] T. Tuomi, T. Johnsson, K. Reijula. Analysis of nicotine, 3-
hydroxycotinine, cotinine, and caffeine in urine of passive
smokers by HPLC-tandem mass spectrometry. Clin. Chem.
1999, 45, 2164.
quadrupole linear ion trap (LTQ) and Fourier transform
ion cyclotron resonance mass spectrometry and tandem
mass spectrometry. Rapid Commun. Mass Spectrom. 2009,
23, 1065.
[11] A. Weimann, M. Sabroe, H.E. Poulsen. Measurement of
caffeine and five of the major metabolites in urine by high-
performance liquid chromatography/tandem mass spectro-
metry. J. Mass Spectrom. 2005, 40, 307.
[12] M. Thevis, G. Opfermann, O. Krug, W. Schänzer. Electro-
spray ionization mass spectrometric characterization and
quantitation of xanthine derivatives using isotopically
labelled analogues: an application for equine doping
control analysis. Rapid Commun. Mass Spectrom. 2004,
18, 1553.
[13] M.S. Caubet, B. Comte, J.L. Brazier. Determination of
urinary 13C-caffeine metabolites by liquid chromatogra-
phy-mass spectrometry: the use of metabolic ratios to assess
CYP1A2 activity. J. Pharmaceut. Biomed. Anal. 2004, 34, 379.
[14] J.B. Brazier, J.L. Desage, M. Falconnet. Synthesis of seven
deuteromethyl-caffeine analogues: observation of deuter-
ium isotope effects on CMR analysis. J. Labelled Compd.
Rad. 1986, 23, 267.
[15] H. Sajiki, H. Esaki, F. Aoki, T. Maegawa, K. Hirota.
Palladium-catalyzed base-selective HD exchange reaction
of nucleosides in deuterium oxide. Synlett 2005, 1385.
[16] H. Sajiki, F. Aoki, H. Esaki, T. Maegawa, K. Hirota. Efficient
CH/CD exchange reaction on the alkyl side chain of
Rapid Commun. Mass Spectrom. 2013, 27, 885–895
Copyright © 2013 John Wiley & Sons, Ltd.
wileyonlinelibrary.com/journal/rcm