Liquid chromatographic/electrospray ionization mass spectrometric identification of the oxidation end-products of metformin in aqueous solutions

Article abstract of DOI:10.1002/jms.656

Metformin is an antihyperglycemic drug that exhibits some antioxidant properties. HO^?-induced oxidation of metformin was studied in aqueous solution, in both aerated and deaerated conditions. Gamma radiolysis of water was used to generate HO^? free radicals, capable of initiating one-electron oxidation of metformin. Oxidation end-products were identified by direct infusion mass spectrometry (MS) and high-performance liquid chromatography/mass spectrometry (HPLC/MSⁿ): for every product, structure elucidation was based on its mass (simple mass spectra confirmed by HPLC/MS). In addition, fragmentation spectra (MS², MS³ and MS⁴) and the determination of deuterium-hydrogen exchange sites provided valuable information allowing the complete identification of some of the end-products. At low radiation dose, four products were identified as primary ones, since they result from the direct attack of HO^? radicals on metformin. These primary oxidation end-products were identified respectively as hydroperoxide of metformin, covalent dimer of metformin, methylbiguanide and 2-amino-4-imino-5-methyl-1,3,5-triazine. At high radiation dose, seven other products were identified as secondary ones, resulting from the HO^?-induced oxidation of the primary end-products. A reaction scheme was postulated for the interpretation of the results. Copyright

Full text of DOI:10.1002/jms.656

JOURNAL OF MASS SPECTROMETRY
J. Mass Spectrom. 2004; 39: 890–902
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jms.656
Liquid chromatographic/electrospray ionization mass
spectrometric identification of the oxidation
end-products of metformin in aqueous solutions
1
2
2
2
1
F. Collin,1∗ H. Khoury, D. Bonnefont-Rousselot, P. Therond, A. Legrand, D. Jore and
´
1
`
M. Gardes-Albert
1
Laboratoire de Chimie Physique, CNRS UMR 8601, Universite´ Paris 5, 75270 Paris, France
Laboratoire de Biochimie Me´ tabolique et Clinique, EA 3617, Universite´ Paris 5, 75270 Paris, France
2
Received 9 March 2004; Accepted 19 April 2004
Metformin is an antihyperglycemic drug that exhibits some antioxidant properties. HOž-induced oxidation
of metformin was studied in aqueous solution, in both aerated and deaerated conditions. Gamma radiolysis
of water was used to generate HOž free radicals, capable of initiating one-electron oxidation of metformin.
Oxidation end-products were identified by direct infusion mass spectrometry (MS) and high-performance
liquid chromatography/mass spectrometry (HPLC/MSn): for every product, structure elucidation was based
on its mass (simple mass spectra confirmed by HPLC/MS). In addition, fragmentation spectra (MS2, MS3
and MS4) and the determination of deuterium–hydrogen exchange sites provided valuable information
allowing the complete identification of some of the end-products. At low radiation dose, four products were
identified as primary ones, since they result from the direct attack of HOž radicals on metformin. These
primary oxidation end-products were identified respectively as hydroperoxide of metformin, covalent
dimer of metformin, methylbiguanide and 2-amino-4-imino-5-methyl-1,3,5-triazine. At high radiation
dose, seven other products were identified as secondary ones, resulting from the HOž-induced oxidation of
the primary end-products. A reaction scheme was postulated for the interpretation of the results. Copyright
 2004 John Wiley & Sons, Ltd.
KEYWORDS: metformin; oxidative stress; HO free radicals; liquid chromatography/electrospray ionization; mass
spectrometry
INTRODUCTION
NH
NH
CH3
H
Metformin, N,N-dimethylbiguanide (Fig. 1), is an oral anti-
hyperglycemic agent widely used for the management of
type 2 diabetes.1,2 Previous in vivo and in vitro studies sug-
gested that this molecule, which is an aminoguanidine-like
compound, has other beneficial effects that could be involved
in the prevention of diabetic complications. It has been shown
that metformin could contribute to the inhibition of the for-
mation of advanced glycation end-products (AGEs),3,4 to
lower systemic methylglyoxal concentrations5 and to react
with ˛-dicarbonyl compounds.6 In addition, it has been
hypothesized that metformin could directly scavenge reac-
tive oxygen species or indirectly act by modulating the
intracellular production of superoxide anions,7 exhibiting
antioxidant properties.
N
N
H
N
CH3
H
Figure 1. Structure of metformin (N,N-dimethylbiguanide).
stress in diabetes, i.e. an imbalance between the production
of oxidant species, particularly radical species, and the
antioxidant defences.10,11 This might partly explain the
elevated risk factors towards cardiovascular complications
in diabetic patients.12,13
Gamma radiolysis is a powerful tool for the modeling
of oxidative stress because it allows ROS to be gener-
ated in solution, with known radiolytic yields for every
species14 (the radiolytic yield is the number of free radi-
cals produced per unit of energy absorbed, i.e. per joule).
This method can be used to initiate the one-electron oxi-
dation (via HOž free radicals) of bio-organic substrates
(see, for example, Refs 15 and 16). Among the ROS gen-
Reactive oxygen species (ROS) play an important role in
the formation of AGE8 and, conversely, AGE are involved
in the enhancement of cellular oxidative stress.9 It has been
demonstrated that elevated glucose levels induce oxidative
erated by water radiolysis (i.e. HOž, O2 and H2O2), the
hydroxyl radical is the most reactive species. Indeed, it
žꢀ
ŁCorrespondence to: F. Collin, Laboratoire de Chimie Physique,
CNRS UMR 8601, Universite´ Paris 5, 45 rue des Saints Pe`res, 75270
Paris cedex 06, France. E-mail: fabrice.collin@univ-paris5.fr
Contract/grant sponsor: Merck Laboratories.
ž
0
°
possesses a high oxidation potential of E ꢀHO /H2Oꢁ D
2.3 V at pH 7.014 and reacts on numerous biomolecules
Copyright  2004 John Wiley & Sons, Ltd.

LC/ESI-MS of oxidation end-products of metformin
891
with high rate constants (usually 109 –1010 l molꢀ1 sꢀ1
(Ref. 17)).
oxidation products are primary, i.e. result from a direct
HOž-induced oxidation of metformin (MTF) in solution. At
high radiation dose (e.g. at 300 Gy), other products can be
detected because they are derived from the disappearance
of the primary end-products, which are attacked by HOž
free radicals. Hence the oxidation of the primary oxidation
end-products gives rise to secondary end-products. By
analyzing solutions of MTF irradiated at increasing radiation
doses, identified oxidation end-products can be classified as
primary or secondary.
For irradiation under aerated conditions, 5 ml of solution
were introduced into a test-tube in order to be directly
irradiated. To remove oxygen in solution completely, and
thus to work under deaerated conditions, 5 ml of MTF
solution were introduced into a special test-tube, closed
at the top and equipped with two thin glass tubes allowing
bubbling of the solution with any gas. N2O gas was used
and the solution was bubbled for 1 h at a flow-rate of
¾1 ml minꢀ1. The test-tube was then hermetically closed
until analysis, in order to avoid any contact of the solution
with the oxygen present in the air before or after irradiation.
Volumes of 5 ml of a non-irradiated solution were
systematically taken as a control for each experiment. Prior
to each set of experiments, glassware was carefully washed
with TFD4 soap (Franklab, France), rinsed with ultra-pure
water (resistivity 18 Mꢂ, Maxima Ultra Pure Water, ELGA)
Since metformin has been shown to increase the antiox-
idant status in patients treated for type 2 diabetes, it could
act as an antioxidant, that is a reducing agent, by react-
ing with ROS and leading to metformin-oxidized products.
Hence this work was aimed at studying how metformin
could be oxidized in vitro by HOž free radicals produced by
water radiolysis and at identifying the oxidation products
by using high-performance liquid chromatography coupled
with mass spectrometry (HPLC/MSn).
EXPERIMENTAL
Aqueous solutions
Metformin hydrochloride was purchased from Sigma-
Aldrich (St. Louis, MO, USA). All aqueous solutions of
metformin (M D 129 g molꢀ1) were prepared at a concentra-
tion of 200 µmol lꢀ1 (pH 6.8) with ultra-pure water (resistivity
18 Mꢂ, Maxima Ultra Pure Water, ELGA). Methylbiguanide
(M D 115 g molꢀ1) was synthesized by Merck-Lipha (Lyon,
France) and aqueous solutions (50 µmol lꢀ1) were prepared
as described previously for metformin.
Generation of HOž free radicals by water gamma
radiolysis
Gamma irradiations were performed with an IBL 637
irradiator (CIS Biointernational, Gif-Sur-Yvette, France),
using a cesium 137 ꢃ-ray source whose activity was ¾222
TBq (6000 Ci). The main advantage of gamma radiolysis is
that it allows the modulation of the total amount of reactive
oxygen species produced by increasing or decreasing the
radiation dose (expressed in Gy; 1 Gy D 1 J kgꢀ1), i.e. by
selecting the time during which the sample is exposed to the
137Cs gamma ray source: the longer the exposure, the higher
is the radiation dose. It can be noted that for dilute solutions,
no direct interaction of ꢃ-radiation with the molecule under
focus occurs and the radiolytic effect is due only to the radical
°
and finally heated at 400 C for 4 h to avoid any pollution by
organic compounds.
Analyses
Mass spectrometry was performed on an ion-trap mass
spectrometer (LCQ Advantage, Thermofinnigan, Les Ulis,
France) equipped with an electrospray ionization (ESI)
source. Irradiated aqueous solutions of metformin were
diluted by 10 : 1 in methanol prior to being infused con-
tinuously into the ESI source with an SGE 250 µl syringe at a
flow-rate of 12.5 µl minꢀ1. Experiments performed at lower
pHs by adding formic acid solution (1% of the total volume)
did not provide any gain in peak intensity. The capillary
species produced by water radiolysis (i.e. HOž, eꢀ and
aq
Hž).14 The dosimetry was determined by Fricke and Morse’s
method18 and the dose rate was found to be 9.76 Gy minꢀ1
in our experiments.
°
temperature was held at 250 C and the relative sheath and
auxiliary gas flow-rates were set at 20 and 5, respectively
(sheath gas, 0–100 units corresponds to 0–1.5 l minꢀ1; aux-
iliary gas, 0–60 units corresponds to 0–18 l minꢀ1). Other
parameters, such as lens or capillary voltages, were tuned
systematically to obtain the best signal intensities for each ion
of interest. All experiments were performed in the positive-
ion mode, and each spectrum was typically an average of 15
acquired scans. For MS/MS experiments, a typical isolation
width of 1 Da was used.
In addition, some experiments were conducted by cou-
pling a high-performance liquid chromatograph (Surveyor,
Thermoquest, Les Ulis, France) to the mass spectrometer.
Volumes of 50 µl of the sample were injected in the col-
umn, which was a 250 ð 4.6 mm i.d. Atlantis dC18 (5 µm)
reversed-phase column (Waters, St. Quentin en Yvelines,
In the presence of oxygen, i.e. for aerated solutions, water
radiolysis leads to the generation of both hydroxyl (HOž)
žꢀ
and superoxide (O2žꢀ) free radicals (O2 coming from the
scavenging of eꢀ and Hž by O2), with formation yields of
aq
2.8 ð 10ꢀ7 and 3.4 ð 10ꢀ7 mol Jꢀ1, respectively.14 Bonnefont-
Rousselot et al. have shown that metformin is not able to
scavenge superoxide free radicals in vitro.7 As a consequence,
hydroxyl radicals are the only radical species initiating the
oxidation process. In the absence of oxygen, i.e. for deaerated
solutions saturated with N2O gas, hydroxyl radicals HOž
are generated with a formation yield of 5.6 ð 10ꢀ7 mol Jꢀ1
,
that is, twice the generation yield obtained under aerated
conditions, because of the transformation of eꢀaq free radicals
into hydroxyl radicals.14 In this work, both aerated and
deaerated solutions were irradiated and analyzed in order
to compare the oxidation end-products created.
°
France) maintained at 40 C, and eluted with 100% ammo-
nium acetate buffer (10 mmol lꢀ1), pumped at a flow-rate
of 500 µl minꢀ1. Prior to the mass spectrometer, the eluate
was detected with a photodiode-array UV–visible detector
Depending on the radiation dose, the oxidation is called
primary or secondary: at a radiation dose below 50 Gy,
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

892
F. Collin et al.
(Thermoquest Surveyor, PDA detector), working in both the
scan mode (between 200 and 400 nm) and single-wavelength
mode at 232 nm, previously identified as a typical wave-
length for metformin.19
In order to complete these experiments, samples of
oxidized MTF were evaporated to dryness, then rediluted in
the labelled medium D2O–MeOD (1 : 2, v/v); deuterium
exchanges were observed for both MTF and oxidation
products. Deuterium oxide and deuterated methanol have
minimum isotopic purities of 99.96 and 99.5% D, respectively,
and were purchased from Aldrich Chemical (Milwaukee, WI,
USA).
An experiment was also carried out in order to character-
ize the presence of hydroperoxide in solution. The method is
based on HPLC analysis with both spectrometric and spec-
trofluorimetric detection as developed by The´rond et al.20
The operating conditions were as following: isocratic run-
ning with an eluent consisting of MeOH and 10 mmol lꢀ1
ammonium acetate (96 : 4, v/v), flow-rate 1.2 ml minꢀ1 and
injection volume 200 µl. Two columns were used: a 25 cm
reversed-phase C18 followed by a 15 cm reversed-phase C8
and 66.1 (Fig. 2(b)). These ions could come from the frag-
mentation of MTF by the breaking of every carbon–nitrogen
single bond (except those of the terminal N-methyl groups).
An ion at m/z 71, previously found by Heinig and Bucheli to
be a specific fragment of MTF and used for its quantitation in
plasma samples,22 was detected, probably owing to different
experimental conditions.
CID spectra of MTF were also acquired at increasing
collision energies, from 15 to 27% (units as given by the
manufacturer; 0–100% relative collision energy corresponds
to 0 to ꢀ100 V multiple offset voltage for positive ions), and
the abundances of each detected fragment were plotted as
a function of the collision energy (data not shown). When
the collision energy increases, the abundance of m/z 130.2
(protonated MTF) decreases and the fragments increase.
Based on these experimental data, a fragmentation scheme
is proposed in Fig. 3, which includes detected ions for MTF
fragmentation in both H2O–MeOH and D2O–MeOD media
(ions in parentheses).
The most abundant ion is produced at low collision
energy and detected at m/z 60.1 (data not shown). Another
fragment starts to appear at low collision energy and is
detected at m/z 85.1. The two other fragments appear
at higher relative collision energies, above 20% for m/z
113.1 and 25% for m/z 88.3. These ions result from the
fragmentation of a carbon–nitrogen bond located in the ˛-
position of one of the imino groups of MTF (Fig. 3, pathways
a and b). The ions at m/z 60.1 and 85.1 are shifted up in
mass by 6 and 5 units respectively, when the CID spectrum
is acquired in D2O–MeOD medium, indicating the number
of H–D exchange sites for each ion and thus confirming the
structures postulated for these two ions (H–D exchanges
only occur on acidic hydrogens, i.e. those of amino or imino
groups).
°
(both of 4.6 mm i.d.), maintained at 40 C. Hydroperoxides
were specifically quantified by chemical luminescence, after
reaction with a microperoxidase.
RESULTS AND DISCUSSION
ESI-MS study of non-oxidized metformin
To obtain the mass spectra of non-oxidized MTF, under
ESI in the positive mode (ESIC), its fragmentation path-
way was determined. Collision-induced dissociation (CID)
spectra were acquired in both H2O–MeOH (Fig. 2(a)) and
D2O–MeOD (Fig. 2(b)) media. The parent ion of MTF was
detected at m/z 130.2, as protonated MTF. MTF has two pKa
values at 2.8 and 11.5,21 and therefore occurs as a monopro-
tonated ion at pH 7, which facilitates its ionization into the
ESI source. The fragmentation of m/z 130.2 shows four ions
detected at m/z 113.1, 88.3, 85.1 and 60.1 (Fig. 2(a)), along
with their equivalent labelled ions at m/z 116.0, 92.3, 90.0
The two other detected fragments, at m/z 88.3 and 113.1,
result from the same fragmentation pattern applied to the
second imino group (Fig. 3, pathways c and d). The structures
proposed for these two ions are confirmed by the shift in
Figure 2. (a) CID spectrum of MTF (200 µmol lꢀ1) in H2O–MeOH (1 : 10) at 33% collision energy and (b) CID spectrum of MTF
(200 µmol lꢀ1) in D2O–MeOD (1 : 1) at 28% collision energy. Collision energies are as given by the manufacturer. Ordinate, relative
abundance; abscissa, mass/charge ratio (m/z).
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

LC/ESI-MS of oxidation end-products of metformin
893
NH3+
NH3+
131.1, and four others at m/z 116.1, 126.1, 144.1 and 162.0.
No peaks were detected above 180 Da.
ND
+
ND
+
3
3
DN
D
N
ND
N(CH
)
H2N
NH
m/z 60
2
3 2
HN
N(CH3)2
Experiments were also carried out on deaerated oxidized
(400 Gy) aqueous solutions of MTF (200 µmol lꢀ1) and the
mass spectrum between 50 and 400 Da allowed the detection
of many ions (Fig. 4(b)). Among them, some are of interest a
priori: the ion at m/z 384.1, along with its equivalent doubly
protonated ion at m/z 192.5, whose mass is close to the mass
of three molecules of MTF (3 ð 129 D 387); the ion detected
at m/z 257.1, along with its equivalent doubly protonated
ion at m/z 129.1, whose mass is close to the mass of two
molecules of MTF (2 ð 129 D 258); the ion at m/z 116.1 that
has been identified under aerated conditions; and the ion at
m/z 130.1, which is the protonated MTF. Above 400 Da, no
peaks were detected.
m/z 66
m/z 88
m/z 92
a
c
+
NH
NH
b
a
d
+H
N
H
H2N
m/z 130
N(CH3)2
c
b
d
ND
ND
DN+
ND+
NH
N(CH3)2
NH
HN+
NH+
C
C
ND
N(CH3)2
D2N ND
C
C
N
H
N
H
H2N
m/z 116
m/z 90
These results were taken as a base in the identification of
the oxidation products by HPLC/MS. In addition, all masses
from 50 to 400 Da were sought step by step (1 Da). Figure 5
shows trace chromatograms obtained for an aerated (Fig. 5(a)
and (b)) and a deaerated (Fig. 5(c) and (d)) solution of MTF
(200 µmol lꢀ1), oxidized at 50 and 300 Gy, for every mass of
detected ions.
m/z 113
m/z 85
Figure 3. Proposed scheme for the fragmentation of the
protonated MTF at m/z 130 (ions in parentheses are labelled
equivalent fragments detected in D2O–MeOD).
mass of 3 and 4 units for m/z 113.1 and m/z 88.3, respectively,
when the CID spectrum is acquired in D2O–MeOD medium.
For a solution of MTF oxidized at 300 Gy under aerated
conditions (Fig. 5(b)), previous identifications are confirmed
(m/z 162.1, 126.1 and 116.1), except for m/z 144.1. New ions
are also detected at m/z 148.1, 112.1 and 102.1. It appears
that two products have the same mass (m/z 126.1) but
have significantly different retention times (5.4 and 8.6 min),
indicating a possible different molecular structure. The same
results are observed for m/z 112.1, detected at two retention
times (7.1 and 13.5 min). When MTF is oxidized at 50 Gy (low
radiation dose), the chromatogram (Fig. 5(a)) shows peaks
corresponding only to the ions at m/z 130.1, 162.1, 126.1 and
116.1: they correspond to the primary oxidation products
of MTF (i.e. resulting from a direct attack of HOž radicals
on MTF), since they are detected at low radiation dose
(50 Gy). For the other products (not detected below 50 Gy,
but present at 300 Gy), they are secondary ones because they
are produced by the HOž-induced oxidation of the primary
MS and HPLC/MS study of oxidized metformin
Hydroxyl radicals can initiate three types of oxidizing
reactions: abstraction of an H-atom, abstraction of an electron
(charge transfer) or addition to double bonds.23 MTF has
many reaction sites towards the oxidation of HOž radical:
two carbon–nitrogen double bonds, five electronic doublets
and six H-atoms on two methyl groups. Hence the potential
structures of oxidation products are difficult to predict. As
a consequence, the first approach in this work was to study
the MS of an aerated oxidized (300 Gy) aqueous solution
of MTF (200 µmol lꢀ1) by infusing directly a dilute sample
(H2O–MeOD, 1 : 10, v/v) into the ESI source and acquiring a
full mass spectrum from 50 to 180 Da (Fig. 4(a)). Six ions are
present in this spectrum: the [M C H]C protonated MTF at
m/z 130.1, along with the isotopic contribution of 13C at m/z
Figure 4. Full mass spectra of an aqueous solution of MTF (200 µmol lꢀ1), oxidized (a) at 300 Gy under aerated conditions and (b) at
400 Gy under deaerated conditions. Abscissa, mass/charge ratio (m/z); ordinate, relative abundance.
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

894
F. Collin et al.
Figure 5. Aqueous solution of MTF (200 µmol lꢀ1), oxidized under aerated conditions at (a) 50 and (b) 300, and under deaerated
conditions at (c) 50 and (d) 300 Gy. Trace chromatograms for each detected ion. —Abscissa, retention time; ordinate,
relative intensity.
oxidation end-products of MTF when they are sufficiently
accumulated.
covalent bond between these two radicals). The second
ion at m/z 162.1 corresponds to a formal addition of
molecular oxygen to MTF (162 D 130 C 32). As oxygen is
known to react rapidly with a carbon-centered free radical
leading to peroxyl ꢀfr1ee radicals and finally to hydroperoxides
(k ³ 108 –109 l mol sꢀ1 (Ref. 24)), this ion could be assumed
to be a protonated hydroperoxide of MTF. The ion detected
at m/z 126.1 corresponds to a loss of 4 Da in relation to MTF.
There are not many possibilities for such a very small loss,
and, in a first approach, it could be the result of two losses of
molecular dihydrogen. The last primary product, detected
at m/z 116.1, could come from a loss of a formal CH2 group
from MTF.
Under deaerated conditions, for a solution of MTF
oxidized at 300 Gy (Fig. 5(d)), some of the ions previously
detected by full MS infusion are confirmed (m/z 257.1, 243.0
and 116.1) and new ions are found at m/z 253.1, 229.1,
126.1, 112.1 and 102.1. By comparing the results obtained at
high (Fig. 5(d)) and low radiation doses (Fig. 5(c)), as done
previously, three products are found to be primary (m/z
257.1, 126.1 and 116.1) and five secondary (m/z 253.3, 243.0,
229.1, 112.1 and 102.1).
Table 1 summarizes all the masses of the ions detected
for a solution of MTF oxidized at 300 Gy, along with their
retention times and a possible identification for each one.
The first ion at m/z 257.1 could result from a covalent
dimerization of MTF (H-abstraction on two molecules of
MTF, leading to two MTF radicals, and formation of a
As previously described, the oxidation of MTF seems
to lead to four primary oxidation end-products, generated
by four different reactions: a covalent dimerization, a per-
oxidation, a demethylation and a double dehydrogenation.
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

LC/ESI-MS of oxidation end-products of metformin
895
Table 1. Oxidation products of MTF identified by HPLC/MS, in both aerated and deaerated conditions: products are sorted
according to whether they result from a primary (low radiation dose) or a secondary oxidation (high radiation dose), and based on a
comparison of their masses with a reference parent mass, a possible identification is proposed
Parent
ion (Da) (D)eaerated
(A)erated or
(P)rimary or Difference in
(S)econdary mass/MTFa
tr (min)
Possible identification
257.1
162.0
126.1
116.1
384.1
253.1
243.1
229.1
148.1
112.1
102.1
D
A
Both
Both
D
D
D
D
A
16,4
9,5
5,4 and 8,6
8,1
P
P
P
P
S
S
S
S
S
S
S
[2 ð ꢀMTF ꢀ 1ꢁ C 1]
[MTF ꢀ MTF C H]C covalent dimer of MTF
[MTFOOH C H]C hydroperoxide of MTF
[MTF ꢀ 2H2 C H]C
[MTF C 1] C 32
[MTF C 1] ꢀ 4
[MTF C 1] ꢀ 14
[MTF ꢀ CH2 C H]C methylbiguanide (MBG)
???
24,7
12,7
9,8
[3 ð ꢀMTF ꢀ 1ꢁ ꢀ 1 C 1] [MTF ꢀ MTF ꢀ MTF C H]C covalent trimer of MTF
257 ꢀ 4
[MTF ꢀ MTF ꢀ 2H2 C H]C
257 ꢀ 14
[MTF ꢀ MTF ꢀ CH2 C H]C
[2 ð ꢀ115 ꢀ 1ꢁ C 1]
116 C 32
[MBG ꢀ MBG C H]C covalent dimer of MBG
[MBGOOH C H]C hydroperoxide of MBG
[MBG ꢀ 2H2 C H]C
7,0
Both
Both
7,1 and 13,5
6,5
116 ꢀ 4
116 ꢀ 14
[MBG ꢀ CH2 C H]C biguanide (BGN)
a MTF: 129 g molꢀ1; protonated molecule [M C H]C detected at m/z 130.2. MBG: 115 g molꢀ1; protonated molecule [M C H]C detected
at m/z 116.1.
The formation of the secondary oxidation products that
have been detected in this study (Table 1, bottom) can be
explained by a similar reaction scheme, applied to two pri-
mary oxidation products (m/z 116.1 and 257.1). Indeed, ions
at m/z 229.1, 148.1, 112.1 and 102.1 could be generated
by covalent dimerization, peroxidation, double dehydro-
genation and demethylation of methylbiguanide (MBG, m/z
116.1), respectively. Likewise, ions detected at m/z 253.3 and
243.0 could come from covalent dimers of MTF (m/z 257.1)
via a double dehydrogenation and a demethylation, respec-
tively. These results give a first approach to the behavior of
metformin when oxidized by HOž free radicals.
(aqueous solution saturated with N2O). The full mass
spectrum (Fig. 6(a)) exhibits two major peaks at m/z 257.1
and 129.2. If the first one could be assumed to correspond
to the monoprotonated diMTF, the second one could be
the doubly protonated ion (ꢀ257 C 2ꢁ/2 D 129). The other
ions present in this spectrum (m/z 116.1, 126.1, 243.0 and
253.3) have been previously identified as potential oxidation
products of MTF under deaerated conditions (see Fig. 5).
Full mass spectrum of oxidized MTF acquired in the labeled
medium, D2O–MeOD (50 : 50, v/v), shows a single peak
detected at m/z 268.1 (Fig. 6(a), inset), indicating that diMTF
has exchanged 11 times its acidic hydrogen atoms with
deuterium, meaning that diMTF contains 11 exchange sites.
There are only three possibilities for diMTF to be
generated from two molecules of MTF by abstracting
two hydrogen atoms and creating a covalent bond: the
Identification of m/z 257.1
Figure 6 shows the mass spectra of an aqueous solution
of MTF oxidized at 300 Gy under deaerated conditions
Figure 6. Mass spectra of an aqueous solution of MTF (200 µmol lꢀ1), oxidized at 300 Gy, under deaerated conditions (N2O
saturated): (a) full mass spectrum in H2O–MeOH and (inset) D2O–MeOD, and (b) CID spectra of (b1) m/z 257 (MS2 at 25% collision
energy), (b2) m/z 198.1 (MS3 at 30% collision energy) and (b3) m/z 181.1 (MS4 at 30% collision energy). Abscissa, mass/charge ratio
(m/z); ordinate, relative abundance. diMTF D covalent dimer of metformin.
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

896
F. Collin et al.
first is to create a carbon–carbon bond, the second a
nitrogen–nitrogen bond and the third a carbon–nitrogen
bond. Since deuterium only exchanges with acid hydrogen,
i.e. hydrogens of the amino and imino functional groups
of MTF, this result (Fig. 6(a) and inset) demonstrates
that the formation of diMTF involves the creation of a
carbon–carbon single bond. Indeed, MTF possesses five
exchange sites, diMTF has 10 potential exchange sites (plus
one for the protonation) and is detected at m/z 268.1
(268.1 D 257.1 C 10 C 1). Hence the molecular structure
shown in Fig. 7 is proposed for diMTF.
CID spectra of m/z 257.1 are presented in Fig. 6(b)
and confirm this proposal. First, on fragmenting the ion
of m/z 257.1 (MS2, Fig. 6(b1)), two losses of 17 and 42 Da
are observed, corresponding to the transitions 257.1 !
240.2 and 240.2 ! 198.1 respectively. They come from
the fragmentation of the non-methylated part of MTF,
as previously proposed in Fig. 3 (pathways c and d).
Fragmentation of m/z 198.1 (MS3, Fig. 6(b2)) and m/z
181.1 (MS4, Fig. 6(b3)) leads to the same losses of 17 Da
(198.1 ! 181.1) and 42 Da (181.1 ! 139.1). Hence the
fragmentation pathway proposed for MTF can be applied
twice to the ion of m/z 257.1, confirming that this ion
contains two molecules of MTF. However, the loss of 45 Da,
corresponding to the neutral fragment ꢀCH3ꢁ2NH and taking
place for MTF (Fig. 3, pathway b), is not observed on
fragmenting m/z 257.1. This shows that diMTF cannot be
fragmented to lose the same neutral fragment ꢀCH3ꢁ2NH as
MTF because the dimethylamino group is involved in the
formation of diMTF, since a carbon–carbon bond has been
created.
Identification of m/z 162.1
MS of the HPLC peak detected at a retention time
tr D 9.54 min is presented in Fig. 8. The base peak
of this spectrum is detected at m/z 162.1 and could
correspond to the ion [M C H]C of the protonated molecule,
previously proposed (in Table 1), i.e. a hydroperoxide of
MTF (MTFOOH). The peak at m/z 163.0 ([M C H C 1]C) is
due to the natural contribution of 13C (1%) and 15N (0.4%)
and accounts for about 6% of the abundance of the parent
ion: MTFOOH contains four carbon atoms and five nitrogen
atoms (4 ð 1% C 5 ð 0.4% D 6%).
This spectrum shows that m/z 162.1 is easily fragmented
because three other peaks are simultaneously detected.
The two first are at m/z 144.1 and 126.1. The transitions
162.1 ! 144.1 and 144.1 ! 126.1 involve the same neutral
loss (18 Da), which could correspond to the mass of water.
Hence this double loss of water implies the presence of two
atoms of oxygen for the ion of m/z 162.1. The last peak of
this spectrum (at m/z 60.0) has been previously detected for
the fragmentation of authentic MTF (Fig. 3, pathway a) and
has been shown to be present at low collision energies. The
other fragments involved in the fragmentation of MTF (see
Fig. 2(a)) are not detected for MTFOOH because they need
relatively high collision energies to be generated and the
experiment described in Fig. 8 present a full mass spectrum
where only in-source fragmentation could occur.
NH NH
CH3
The molecular structure shown in Fig. 9 is proposed for
the oxidation product detected at m/z 162.1.
H2N
N
H
N
H
N
To verify this assumption, an aqueous solution of MTF
(200 µmol lꢀ1), irradiated at 300 Gy in aerated medium, was
diluted twofold with D2O–MeOD (50 : 50, v/v) and intro-
duced directly into the mass spectrometer (infusion). The
full mass spectrum thus obtained (Fig. 8(b)) shows a statisti-
cal distribution of H–D exchanges on MTFOOH. It was not
N
H2N
CH3
NH NH
Figure 7. Proposed structure of the covalent dimer
of metformin.
Figure 8. Mass spectra of an aqueous solution of MTF (200 µmol lꢀ1), oxidized at 300 Gy, under aerated conditions. (a) Full mass
spectrum of the HPLC peak detected at tr D 9.5 min. Abscissa, mass/charge ratio (m/z); ordinate, absolute intensity (cps). (b) Full
mass spectrum of the solution diluted in D2O–MeOD (1 : 1) and analyzed with direct infusion. Abscissa, mass/charge ratio (m/z);
ordinate, relative abundance.
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

LC/ESI-MS of oxidation end-products of metformin
897
Identification of m/z 116.1
NH NH
Based on the mass of the ion detected at m/z 116.1, a
possible identification was proposed in Table 1: it could be
the protonated demethylmetformin, called methylbiguanide
(MBG). This product was analyzed by MS. Because its
molecular structure is very close to that of MTF, its
fragmentation should follow the same pathway as MTF
(see Fig. 3). Indeed, by subtracting 14 from the mass
of the different fragments of MTF (those containing a
dimethylamino group), the fragmentation spectrum of MBG
could be predicted: it should present peaks at m/z 60, 74,
85 and 90. This is exactly what is observed when the CID
spectrum of an aqueous solution of synthetic MBG (non-
oxidized) is acquired, as shown in Fig.10(a). This result
confirms that MBG exhibits the same mass spectrometric
behaviour as MTF. The structure in Fig. 11 is proposed.
A similar behaviour is also observed for MS performed
in D2O–MeOD. Figure 10(b) shows a CID spectrum of the
ion detected at m/z 123.1 for an aqueous solution of MTF
(200 µmol lꢀ1) irradiated at 300 Gy under aerated conditions,
then evaporated to dryness and rediluted in D2O–MeOD
(50 : 50, v/v). MBG shows six exchange sites plus one for the
protonation and is detected at m/z 123.1 (123 D 116 C 6 C 1).
The proposed structures of the ions detected, based on the
fragmentation pathway of MTF (Fig. 3), are all in agreement
with the shifts in mass due to H–D exchanges. For example,
the ion at m/z 99.1 (Fig. 10(a)) can exchange four times
to give an ion at m/z 103, detected when working in
D2O–MeOD (Fig. 10(b)). The peak detected at m/z 116.1
for a solution of oxidized MTF can exchange seven times
(one more time than MTF) and its fragmentation leads to
fragment ions that are all compatible with ions detected when
fragmenting synthetic MBG (non-oxidized). These results
demonstrate that methylbiguanide is one of the oxidation
products of MTF.
CH3
H2N
N
H
N
OOH
Figure 9. Structure of hydroperoxometformin (MTFOOH).
possible to obtain a single peak corresponding to the high-
est number of exchanges because the solution still contains
H2O: to remove it, and thus avoid possible back-exchanges
between deuterium and hydrogen, an evaporation should
have been performed by heating the solution. However, this
procedure could initiate the decomposition of hydroperox-
ides with temperature. The protonated MTFOOH is detected
at m/z 162.1 and the statistical distribution of exchanges
seems to end at m/z 169.2, showing seven possible exchange
sites on this molecule. In relation to MTF, which has five
hydrogen atoms susceptible to be exchanged plus one for the
protonation, a new exchange site is detected in MTFOOH: it
could correspond to the hydrogen atom of the hydroperoxide
functional group. This result allows the previous proposed
structure for MTFOOH to be confirmed.
It should be noted that no ion at m/z 162.1 was detected
for solutions of MTF irradiated under deaerated conditions
(Fig. 5). Indeed, the generation of hydroperoxides by water
radiolysis involves the reaction of a carbon-centered free
radical with molecular oxygen (O2) as an intermediate step.24
To finalize this identification, an aqueous solution of MTF
(200 µmol lꢀ1) was analyzed specifically for hydroperoxides
by HPLC/chemiluminescence, as described previously (see
Experimental section). The chromatogram obtained for the
previous solution irradiated at 200 and 400 Gy (data not
shown) contains a peak whose area depends on the radiation
dose (area of the peak: 53 075 at 400 Gy and 22 467 at 200 Gy),
thus showing that the oxidized solution of MTF contains a
hydroperoxide.
It can be noted that a peak at m/z 116.1 is also detected
for solutions of MTF irradiated under deaerated conditions
Figure 10. CID spectrum of the ion detected at m/z 116.1 for an aqueous solution of synthetic methylbiguanide (MBG) at
50 µmol lꢀ1, non-oxidized, diluted in H2O–MeOH (1 : 10) and analyzed with direct infusion (collision energy D 23%). Abscissa,
mass/charge ratio (m/z); ordinate, relative Abundance. (b) CID spectrum of the ion detected at m/z 123 for an aqueous solution of
MTF (200 µmol lꢀ1), oxidized at 300 Gy, under aerated conditions, diluted in D2O–MeOD (1 : 1) and analyzed with direct infusion
(collision energy D 26%).
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

898
F. Collin et al.
(MTF)
NH NH
NH
H O
2
HO
NH NH
NH NH
2
CH
3
HN
C
N
H2N
N
H
NH
CH3
CH
CH
3
3
H N
N
N
N
H
N
2
HN
H
H N
N
CH
2
H C
2
3
H
biradicalar
reaction (-H )
Figure 11. Structure of methylbiguanide (MBG).
NH
NH
NH
+
CH
CH
CH
H
3
3
3
(see Fig. 5). As a consequence, the generation of MBG seems
not to be dependent on the presence of molecular oxygen
in solution; this point will be considered further in the
discussion of a possible reaction mechanism.
N
N
N
N
N
N
oxidation (H2O2)
- 2H2O
H N
N
2
N
H N
H N
2
N
H
2
(1)
Figure 12. Proposed mechanism of formation of the ion
detected at m/z 126.1 during HOž radical-induced oxidation of
metformin in solution.
Identification of m/z 126.1
The presence of ions detected at m/z 126.1, corresponding
to a mass of 125.1 Da for the oxidation products, implies
a chemical transformation of MTF leading to a loss of
4 Da. The simplest assumption that can be made is that
such a low loss should correspond to two molecular
dihydrogens (H2) (double dehydrogenation). According to
the molecular structure of MTF, there is only one mechanism
that would allow a double dehydrogenation: cyclization of
MTF followed by the formation of a double bond (oxidation).
Assuming that the reactive part of MTF towards HOž
radicals is the dimethylamino group (previously shown to be
involved in the formation of MBG, diMTF and MTFOOH), a
hypothetical mechanism could be considered, leading to an
oxidation product of mass 125.1 Da (Fig. 12).
water radiolysis) leads finally to an oxidation end-product
that has a molecular structure close to an aromatic 1,3,5-
triazine (compound 1 along with its mesomeric form for
which the aromatic feature appears more clearly). Once
again, molecular oxygen cannot be involved in this last step
because of the occurrence of m/z 126.1 under deaerated
conditions.
The aromatic feature of this product is confirmed by the
UV spectrum of the HPLC peak detected at tr D 8.6 min
(on the chromatogram presented in Fig. 5). The differential
spectrum, i.e. for which the UV spectrum of MTF is taken
as a reference, exhibits two main bands with maximum
absorptions at around 208 and 255 nm (data not shown). This
last wavelength is characteristic of the aromatic compounds.
The structure of m/z 126.1, parent ion of 4-amino-2-
imino-1-methyl-1,2-dihydro-1,3,5-triazine (4,2,1-AIMT (1)),
was confirmed by the CID spectra in the direct infusion
mode (Fig. 13). The fragmentation of the ion of m/z 126.0
is characterized by three main losses of 27, 42 and 56 Da,
which could correspond to each part of the molecule
previously suggested, as indicated in Fig. 13(a). This kind
of fragmentation was proposed previously by Baglio et al.
for triazine herbicides.26 In addition, for a solution of MTF
diluted in D2O–MeOD, the ion at m/z 126.0 seems to
The first step is the action of HOž radical, abstracting a
hydrogen atom from a terminal methyl group, leading to a
carbon-centered free radical. This radical is then susceptible
to react intramolecularly with an imine group by creating a
nitrogen–carbon single bond, giving a six-membered ring
radical: endo cyclization seems to occur, even though it
is usually unfavorable for other systems.25 No reaction of
this carbon-centered free radical with molecular oxygen is
involved because the end-product (m/z 126.1) is also detected
under deaerated conditions (see chromatogram in Fig. 5).
A biradical reaction followed by a two-electron oxidation
by H2O2 (one of the molecular species generated during
Figure 13. CID spectra of an aqueous solution of MTF (200 µmol lꢀ1) oxidized at 300 Gy, under aerated conditions, in direct infusion
mode. Abscissa, mass/charge ratio (m/z); ordinate, relative abundance (30 accumulated scans). (a) Fragmentation of the ion
detected at m/z 126.0 (collision energy D 29%), for the solution diluted in H2O–MeOH (1 : 10). (b) Fragmentation of the ion detected
at m/z 130.1 (collision energy D 26%), for the solution diluted in D2O–MeOD (1 : 1).
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

LC/ESI-MS of oxidation end-products of metformin
899
exchange four times to give an ion at m/z 130.1 (Fig. 13(b)),
in good agreement with the molecular structure of 4,2,1-
AIMT (three exchange sites plus one for protonation). Its
fragmentation leads to ions with masses compatible with
those of the fragments detected in H2O–MeOH (Fig. 13(a)):
fragments at m/z 70.4 and 84.0 can exchange three times
and twice, respectively, to give ions detected at m/z 72.9
and 86.1. These results demonstrate that 4,2,1-AIMT is one
of the primary oxidation end-products of MTF in aqueous
solution.
As shown in Fig. 5, another HPLC peak is detected at m/z
126.1 (tr D 5.4 min), which probably corresponds to a second
oxidation product that has the same mass as 4,2,1-AIMT. A
simple way to find a second molecular structure consists in
assuming the migration of a methyl group from 4,2,1-AIMT
(1), as represented in Fig. 14.
This rearrangement leads to an aromatic molecule, called
2-amino-4-methylamino-1,3,5-triazine (2,4-AMT (2)). The
molecular structure of this second compound is compatible
with the CID spectra presented in Fig. 13. In particular,
the three main losses of 27, 42 and 56 Da observed in
the CID spectrum acquired in the non-labeled medium
(H2O–MeOH) could correspond to each part of 2,4-AMT.
Moreover, the number of H–D exchange sites observed on
the CID spectrum acquired in D2O–MeOD (Fig. 13(b)) are
also compatible with the molecular structure of 2. These
results confirm the assumption that 2,4-AMT would be a
possible oxidation end-product of MTF in aqueous solution.
Ions at m/z 126.1 were difficult to observe as they could
arise from aromatic compounds (triazines), usually very
difficult to fragment. By accumulating a lot of scans when
working in MS infusion, a correct CID spectrum was finally
obtained. However, this CID spectrum corresponds to a
mixture of the two peaks detected at m/z 126.1 by HPLC/MS.
Indeed, the accumulation time, equal to the peak width,
was not sufficient to obtain valuable MS/MS information
(intensities are too low) when we tried to obtain the CID
spectra for the individual HPLC peaks.
Finally, other experiments using nuclear magnetic res-
onance (NMR) and high-resolution Fourier transform ion
cyclotron resonance (FT-ICR) methods in the search for
additional data for structure elucidation were carried out in
order to provide elements of confirmation for the proposed
structures of the products detected at m/z 126.1. However,
no valuable information was obtained since dilute aqueous
solution (radiolysis conditions) is not a good medium for
NMR work and the mass of metformin (129 g molꢀ1) is too
low to work correctly with FT-ICR.
Identification of secondary oxidation products
As previously explained (see Experimental section), sec-
ondary oxidation products are generated at high radia-
tion doses (above 50 Gy) and are the result of the attack
of HOž radicals on some of the primary oxidation end-
products. Based on MS identifications, a general reaction
mechanism has been postulated, that includes four differ-
ent reactions (covalent dimerization, peroxidation, double
dehydrogenation, demethylation). Since the molecular struc-
tures of primary oxidation end-products and the reactions
involved in their formation have been characterized, molec-
ular structures were deduced for each secondary oxidation
end-product and are presented in Table 2.
The heaviest secondary oxidation product is detected
at m/z 384.1 and could correspond to a covalent trimer of
MTF (called triMTF). Full mass and CID spectra confirm
this assumption. A doubly protonated ion is detected at m/z
192.6, indicating that this ion contains at least two potential
sites for protonation. The triply charged ion that could be
detected at m/z 128.7 is not clearly identified because of a lack
of resolution of the mass spectrometer: it should be observed
together with the peak at m/z 129.1, the doubly protonated
diMTF. The full mass spectrum for a solution of oxidized
MTF (300 Gy, deaerated conditions) diluted in D2O–MeOD
(1 : 1) shows a peak at m/z 400.3, due to 16 H–D exchanges
on the ion of m/z 384.1, which confirms the formation of
two carbon–carbon single bonds (triMTF contains three
molecules of MTF that can exchange 3 ð 5 D 15 times plus
one for protonation). Finally, the CID spectrum of m/z 384.1
exhibits three losses of 17 and 42 Da, as previously observed
for diMTF (see Fig. 6(b)). These observations confirm that
a covalent trimer of MTF is one the secondary oxidation
end-products.
The next two products, detected at m/z 253.1 and
243.1, could come from the double dehydrogenation
and the demethylation of diMTF, respectively, and are
called 4-amino-2-imino-1-[2-(1-methylbiguanid-1-yl)ethyl]-
1,2-dihydro-1,3,5-triazine (4,2,1-AIBT) and N-demethyldi-
metformin (DMdiMTF). Actually, DMdiMTF could also be
generated by a radical–radical reaction between an MTF rad-
ical and an MBG radical (both carbon-centered). Full mass
and CID data provide elements of confirmation for these
proposals. For 4,2,1-AIBT, no peak at m/z 127 is detected,
meaning that the double protonation of its parent ion is not
easy, probably because of the cyclic part of this molecule
(in relation to MTF). Concerning DMdiMTF, the doubly
protonated molecule is detected at m/z 122.1, and the CID
spectrum of this last ion exhibits fragment ions at m/z
113.6, 85.1 and 60.1, which are typical MTF fragments as
determined previously (see Fig. 3).
The last reaction that could have been applied to diMTF,
to find another product, could be peroxidation. However,
all of these products (diMTF, triMTF, 4,2,1-AIBT and
DMdiMTF) are generated only under deaerated conditions;
peroxidation was made impossible because of the absence of
molecular oxygen.
NHCH3
N
NH
CH3
N
N
N
N
H2N
N
H2N
(2)
(1)
Figure 14. Proposed mechanism involved in the formation of
the second ion detected at m/z 126.1 during HOž
radical-induced oxidation of metformin in solution.
The other secondary oxidation products could come
from reactions on MBG: covalent dimerization, peroxidation,
double dehydrogenation and demethylation that could
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

900
F. Collin et al.
Table 2. Proposed structures for the secondary oxidation end-products of MTF, along with the reaction type involved in their
formation and any experimental results allowing confirming the proposals
Parent ion (Da)
384
Postulated structure (name)
Reaction involved
Analytical resultsa
Covalent
dimerization of
diMTF and MTF
m/z 192.7 (CC) detected 16
exchange sites CID of m/z
384.1 D losses of 17 and
42 Da (as for diMTF)
CH3
N
H
N
NH NH
NH2
H2N
N
H
N
NH NH
H
N
H2N
N
CH3
NH NH
(triMTF )
253
243
Dehydrogenation
of diMTF
m/z 127 (CC) not detected
NH
H
N
CH3
N
NH2
N
N
NH NH
H2N
N
4-Amino-2-imino-1-[2-(1-methylbiguanid-1-yl)
ethyl]-1, 2-dihydro-1, 3, 5-triazine (4, 2, 1-AIBT)
Demethylation of
diMTF
m/z 122.1 (CC) detected
CID of m/z 122.1 D 113.6,
85.1, 60.1
NH NH
CH3
N
H
N
NH2
H2N
N
H
N
H
NH NH
N-Demethyldimetformin (DMdiMTF)
229
148
Covalent
dimerization of
MBG
Signal too weakb
NH NH
H
N
H
N
NH2
H2N
N
H
N
H
NH NH
(diMBG)
Peroxidation of
MBG
CID of m/z 148 D double
loss of water (as for
MTFOOH)
NH NH
H2N
N
H
N
H
OOH
Hydroperoxomethylbiguanide
(MBGOOH)
112
102
Dehydrogenation
of MBG
Signal too weakb
Signal too weakb
NH2
N
N
H2N
N
2, 4-Diamino-1, 3, 5-triazine (2, 4-DAT)
NH NH
Demethylation of
MBG
H2N
N
H
NH2
Biguanide (BGN)
a Data not shown but available on request.
b The signal was too weak to perform any MS/MS experiment.
generate a covalent dimer of MBG (diMBG at m/z 229.1),
hydroperoxomethylbiguanide (MBGOOH at m/z 148.1), 2,4-
diamino-1,3,5-triazine (2,4-DAT at m/z 112.1) and biguanide
(BGN at m/z 102.1), respectively. In the case of MBGOOH, the
full mass spectrum of the HPLC peak detected at tr D 7.0 min
shows three main peaks at m/z 148.1, 130.1 and 112.1,
indicating a possible double loss of water (148.1 ! 130.1
and 130.1 ! 112.1) occurring easily without any CID of the
ion. A similar result was observed for MTFOOH, showing
the presence of two oxygen atoms in the molecule (see
Identification of m/z 162.1). For the other detected ions (at
m/z 229.1, 122.1 and 102.1, Table 2), the intensity of the signal
was not high enough to perform MS/MS experiments and
full mass spectra of the corresponding HPLC peaks do not
provide any information (except the mass of the protonated
molecule) that confirms the postulated structures.
CONCLUSION
MTF was oxidised by HOž free radicals, the most reac-
tive species involved in the oxidative stress phenomena
in vivo. Towards this entity, MTF acts as an antioxidant, i.e. a
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

LC/ESI-MS of oxidation end-products of metformin
901
Figure 15. Proposed scheme of the general reaction mechanism for the oxidation of MTF by HO radicals. MTF D metformin;
MTFOOH D hydroperoxometformin; diMTF D covalent dimer of metformin; 4,2,1-AIMT D 4-amino-2-imino-1-methyl-1,2-
dihydro-1,3,5-triazine; MBG D methylbiguanide; triMTF D covalent trimer of metformin; 4,2,1-AIBT D 4-amino-2-imino-1-
[2-(1-methylbiguanid-1-yl)ethyl]-1,2-dihydro-1,3,5-triazine; DMdiMTF D N-demethyldimetformin; MBGOOH D hydroperoxo-
methylbiguanide; diMBG D covalent dimer of methylbiguanide; 2,4-DAT D 2,4-diamino-1,3,5-triazine; BGN D biguanide.
reducing agent that can be oxidized, leading to various prod-
ucts. The primary and secondary oxidation end-products
were characterised by HPLC/MSn.
same properties as MTF in terms of increasing the antioxi-
dant status.
Acknowledgements
As discussed previously, a hypothetical scheme of the
reaction mechanism could be deduced from the identification
of the primary and secondary oxidation end-products of
MTF, and is presented in Fig. 15. The first step is the attack
of hydroxyl radicals on MTF, leading to an MTF radical.
Two pathways are then postulated, since differences were
observed in the identifications under aerated and deaerated
conditions. The first pathway (1) is O2-dependent because
it leads to MTFOOH under aerated conditions and diMTF
under deaerated conditions. The second pathway (2) does
not depend on the presence of O2 and leads to 4,2,1-AIMT
and MBG. These two latter oxidation products were detected
under both aerated and deaerated conditions. To summarize,
the four compounds MTFOOH, diMTF, MBG and 4,2,1-
AIMT characterize the primary HOž-induced oxidation of
MTF observed in aqueous solution at low radiation doses.
The secondary oxidation of MTF at high radiation dose
corresponds to the attack of HOž radicals on the primary
oxidation products diMTF and MBG. Pathways (1) and (2)
applied to diMTF give triMTF, 4,2,1-AIBT and DMdiMTF,
whereas when applied to MBG they lead to 2,4-DAT, BGN,
MBGOOH and diMBG (depending on the experimental
conditions, i.e. aerated or deaerated). Such a reaction scheme
(Fig. 15) clearly shows the similar reactivities of MTF, diMTF
and MBG towards HOž radical-induced oxidation. However,
kinetic studies should be carried out in order to confirm each
elementary step of this mechanism.
The authors gratefully thank Y. Hoppilliard and J. Chamot-Rooke
(Laboratoire des Me´canismes Re´actionnels, CNRS UMR 7651, Ecole
Polytechnique, Palaiseau, France) for their advices on mass spec-
trometry, and R. Azerad (CNRS UMR 8601, Universite´ Rene´
Descartes, Paris, France) for the use of the HPLC/MS facilities. We
are indebted to Jamila Mekhloufi for the HPLC/chemiluminescence
experiments. We acknowledge N. Wiernsperger and Merck Lab-
oratories (Lyon, France) for financial support and for providing
methylbiguanide.
REFERENCES
1. Wiernsperger NF, Bailey CJ. The antihyperglycaemic effet of
metformin. Therapeutic and cellular mechanism. Drugs 1999; 58:
31.
2. Bailey CJ, Turner RC. Drug therapy: metformin. N. Engl. J. Med.
1996; 334: 574.
3. Tanaka Y, Iwamoto H, Onuma T. Inhibitory effect of metformin
on formation of advanced glycation end products. Curr. Ther.
Res. 1997; 58: 693.
4. Tanaka Y, Uchino H, Shimizu T, Yoshii H, Njwa M, Ohmura C,
Mitsuhashi N, Onuma T, Kawamori R. Effect of metformin on
advanced glycation end product formation and peripheral nerve
function in streptozocin-induced diabetic rats. Eur. J. Pharmacol.
1999; 376: 17.
5. Beisswenger PJ, Howell SK, Touchette AD, Lal S, Szwergold BS.
Metformin reduces systemic methylglyoxal levels in type 2
diabetes. Diabetes 1999; 48: 198.
6. Ruggiero-Lopez D, Lecomte M, Moinet G, Patereau G, Lag-
arde M, Wiernsperger N. Reaction of metformin with dicarbonyl
compounds. Possible implication in the inhibition of advanced
glycation end product formation. Biochem. Pharmacol. 1999; 58:
1765.
These experimental results could open the way to the
development of new drugs, that could potentially have the
7. Bonnefont-Rousselot D, Raji B, Walrand S, Garde`s-Albert M,
Jore D, Legrand A, Peynet J, Vasson MP. An intracellular
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902

902
F. Collin et al.
modulation of free radical production could contribute to
the beneficial effects of metformin towards oxidative stress.
Metabolism 2003; 52: 586.
recent results on formation of hydroperoxides. Can. J. Physiol.
Pharmacol. 2001; 79: 176.
17. Buxton GV, Greenstock CL, Helman WP, Ross AB. Critical
review of rate constants for reaction of hydrated electrons,
hydrogen atoms and hydroxyl radicals in aqueous solutions. J.
Phys. Chem. Ref. Data. 1988; 17: 513.
18. Fricke H, Morse S. The chemical action of Ro¨ntgen rays on dilute
ferrosulfate solutions as a measure of dose. Am. J. Roentgenol.
Radium Ther. 1927; 18: 430.
19. Beckmann R. The fate of biguanides in man. Ann. N.Y. Acad. Sci.
1968; 148: 820.
20. The´rond P, Couturier M, Demelier JF, Lemonier F. Simultane-
ous determination of the main molecular species of soybean
phosphatidylcholine or phosphatidylethanolamine and their
corresponding hydroperoxides obtained by lipogenase treat-
ment. Lipids 1993; 28: 245.
8. Giardino I, Edelstein D, Brownlee M. BCL-2 expression or
antioxidant prevent hyperglycemia-induced formation of
intracellular advanced glycation endproducts in bovine
endothelial cells. J. Clin. Invest. 1996; 97: 1422.
9. Yan SD, Schmidt AM, Anderson GM, Zhang J, Brett J, Zou YS,
Pinsky D, Stern D. Enhanced cellular oxidant stress by the
interaction of advanced glycation end products with their
receptor/binding proteins. J. Biol. Chem. 1994; 269: 9889.
10. Betteridge DJ. What is oxidative stress? Metabolism 2002; 49: 3.
11. Ceriello A. Oxidative stress and glycemic regulation. Metabolism
2000; 49: 27.
12. Chu NV, Kong APS, Kim DP, Armstrong D, Baxi S, Deutch R,
Caulfield M, Mudaliar SR, Reitz R, Henry RR, Reaven PD.
Differential effects of metformin and troglitazone on
cardiovascular risk factors in patients with type 2 diabetes.
Diabetes Care 2002; 25: 542.
13. Rosen P, Nawroth PP, King G, Moller W, Tritchler HJ, Packer L.
The role of oxidative stress in the onset and progression
of diabetes and its complications: a summary of a congress
series sponsored by UNESCO–MCBN, the American Diabetes
Association and the German Diabetes Society. Diabetes Metab.
Res. Rev. 2001; 17: 189.
21. Elpern B. Chemistry of the biguanides. Ann. N.Y. Acad. Sci. 1968;
148: 577.
22. Heinig K, Bucheli F. Fast liquid chromatographic–tandem mass
spectrometric (LC–MS–MS) determination of metformin in
plasma samples. J. Pharm. Biomed. Anal. 2004; 34: 1005.
23. Steeken S. Addition–elimination paths in electron-transfer
reactions between radicals and molecules. J. Chem. Soc., Faraday.
Trans. 1 1987; 83: 113.
24. Ross AB, Neta P. Rate Constants for Reactions of Aliphatic Carbon-
centeredRadicals in Aqueous Solution. National StandardReference
Data System. US Department of Commerce, National Bureau of
Standards: Washington, DC, 1982.
14. Spinks JWT, Woods RJ. Water and Inorganic Aqueous Systems.
Introduction to the radiation chemistry (3rd edn). Wiley: New York,
1990; 243.
15. Bonnefont-Rousselot D, Garde`s-Albert M, Delattre J, Ferra-
dini C. Oxidation of low-density lipoprotein by HOž and
O2žꢀ/HOž free radicals produced by gamma radiolysis. Radiat.
Res. 1993; 134: 271.
16. Hauville C, Remita S, The´rond P, Jore D, Garde`s-Albert M.
Radiation induced peroxidation of polyunsaturated fatty acids:
25. Fossey J, Lefort D, Sorba J. Free Radicals in Organic Chemistry.
Wiley: New York, 1995; 151.
26. Baglio D, Kotzias D, Larsen BR. Atmospheric pressure ionisation
multiple mass spectrometric analysis of pesticides. J. Chromatogr.
A 1999; 854: 207.
Copyright  2004 John Wiley & Sons, Ltd.
J. Mass Spectrom. 2004; 39: 890–902