Ion trap MS/MS of intact testosterone and epitestosterone conjugates-Adducts, fragile ions and the advantages of derivatisation

Article abstract of DOI:10.1016/j.steroids.2008.01.026

In ion trap mass spectrometry, fragile ions may fragment under the application of resonance ejection during precursor mass isolation, reducing MS/MS spectral intensity. In this study the steroidal epimers testosterone glucuronide (TG) and epitestosterone

Full text of DOI:10.1016/j.steroids.2008.01.026

steroids 7 3 ( 2 0 0 8 ) 621–628
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/steroids
Ion trap MS/MS of intact testosterone and epitestosterone
conjugates—Adducts, fragile ions and the advantages of
derivatisation
David A. Cowana,∗, Andrew T. Kicmana, Carlos Kubli-Garfiasb, Helen J. Welchmana
a
Drug Control Centre, King’s College London, 150 Stamford Street, London SE1 9NH, United Kingdom
Instituto de Investigaciones Biomedicas, Universidad Nacional Autonoma de Mexico, Coyoacan, Mexico
b
a r t i c l e i n f o
a b s t r a c t
Article history:
In ion trap mass spectrometry, fragile ions may fragment under the application of reso-
nance ejection during precursor mass isolation, reducing MS/MS spectral intensity. In this
study the steroidal epimers testosterone glucuronide (TG) and epitestosterone glucuronide
(EG) have been chosen as a model for exploring whether compound structure is linked
to ion trap fragility. Both compounds form multiple adducts by ESI-MS, namely protona-
tion, ammonium and sodium, however, the mass spectrum of EG displays a more intense
ammonium adduct peak than TG. [TG + NH4]+, [EG + NH4]+ and [EG + H]+ were found to be
fragile ions. To explain the differences in adduct formation and fragility, molecular mod-
elling was employed. Ammonium adduction was localised to the glucuronide ring oxygens
and while EG has eight possible adduction sites, only seven were located for TG explaining
the increased ammonium adduct abundance with EG. In EG the bond between the steroid
and the glucuronide was slightly longer and the oxygen in this bond was more basic than
TG. This shows that the EG bond is weaker which may contribute to the fact that [EG + H]+ but
not [TG + H]+ is fragile. To investigate whether stability could be restored by chemical means,
EG was derivatised with tris(trimethoxyphenyl)phosphonium chloride or methylated on the
carboxylic acid and Girard P or methoxylamine on the 3-keto group. Derivatisation of the
steroid rather than the glucuronide eliminated fragility and using a charged derivative elim-
inated adduct formation. This work demonstrates the importance of carefully considering
the nature of the derivative and the site of derivatisation.
Received 31 August 2007
Received in revised form
20 January 2008
Accepted 22 January 2008
Published on line 9 February 2008
Keywords:
Ion trap mass spectrometry
Testosterone glucuronide
Epitestosterone glucuronide
Adducts
Fragile ions
Derivatisation
© 2008 Elsevier Inc. All rights reserved.
1.
Introduction
this has been performed using GC/MS, however, due to the
polarity and thermal lability of such compounds a lengthy
Chromatography coupled to mass spectrometry is particu-
lar valuable for the study of steroid metabolites, which are
mainly excreted as phase II glucuronide and sulphate conju-
gates in urine. These steroid conjugates are excellent markers
of endocrine perturbation and are routinely monitored in
clinical, forensic and drug control laboratories. Historically
sample preparation procedure is required including enzy-
matic hydrolysis of the glucuronide and derivatisation. More
recently, the introduction of LC/MS has enabled the analysis of
intact steroid glucuronides and sulphates, reducing the need
for sample preparation. Quadrupole ion trap (QIT) mass spec-
trometers are particularly valuable when used for the analysis
∗
Corresponding author.
E-mail address: david.cowan@kcl.ac.uk (D.A. Cowan).
0039-128X/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.steroids.2008.01.026

622
steroids 7 3 ( 2 0 0 8 ) 621–628
can be used to predict the chemical behaviour, taking into
account nuclei and all electrons. This type of modelling has
previously been used to explain negative ion [2] and alkali
metal adduct [3] formation in ESI-MS. Previously published ab
initio modelling of epitestosterone and testosterone showed
the qualitatively different intermolecular forces present at
C17 [4]. Building on this knowledge, molecular modelling was
applied to the glucuronide conjugates of these steroids to
explain adduct formation.
The mass spectrometer used was a 3D QIT, which is
more sensitive for obtaining full scan data than typical linear
quadrupole mass spectrometers. Multiple mass spectrometry
(MS) experiments above MS [2] can only readily be performed
by separation in time, making QITs particularly valuable for
structural analysis and enhancing the signal-to-noise ratio.
QIT mass spectrometers operate by applying isolation and
resonance ejection frequencies to ions in the mass analyser.
These frequencies increase the internal energies of the ions
and can cause some ‘fragile’ ions to fragment. Previously iden-
tified fragile ions include orlistat [5], and some explosive,
acylcarnitine and macrolide antibiotic molecules [6].
Derivatisation was used as a complementary approach
to investigate whether fragile ions could be made stable by
chemical means. The derivatised analyte broadly retains its
structure but gains new chemical properties that can be used
to gain insights into which structural features account for
fragility. In this study, either the 3-oxo (ketone) group of the
steroid or the carboxylic acid group of the glucuronide moiety
were reacted with uncharged or permanently positive charged
reagents to form the structures displayed in Fig. 2.
Fig. 1 – (Top) TG and EG with oxygen numbers and the
electrostatic potential. (Bottom) Total energy curves of the
steroid–ammonium adducts studied. The oxygen atoms
linked to the ammonium moiety are in parentheses.
of unknown compounds in complex mixtures due to their abil-
ity to provide data dependent MSn full scan data. Metabolite
screening is one area where ion traps have been effectively
used in this respect. However, routine use of LC/MS for the
quantitative analysis of steroid glucuronides has been ham-
pered by the formation of multiple adducts in the ion source.
In this study, the steroidal epimers testosterone glu-
curonide (TG) and epitestosterone glucuronide (EG) were
chosen as model compounds (Fig. 1) for studying adduct for-
mation with steroid glucuronides. Attempts to detect the
intact glucuronides by LC/MS have shown that there are
remarkable differences in the ammonium adduct affinities
of TG and EG leading to the hypothesis that there may be
important structural differences in the molecules which affect
adduct formation. Kuuranne et al. [1] reported that the inten-
sity of the ammonium adduct is 80% of the base peak [EG + H]+
intensity by ESI-triple quadrupole. By contrast the epimer TG
gave [TG + H]+ as the base peak with a [TG + NH4]+ intensity of
only 20%. Sodium adducts were also prominent. Competing
2.
Materials and methods
2.1.
Materials
TG, EG, testosterone, epitestosterone, testosterone sulphate
and epitestosterone sulphate were obtained from Promochem
(Teddington, UK). Testosterone benzoate, epitestosterone
benzoate, testosterone hemisuccinate and epitestosterone
hemisuccinate were obtained from Steraloids (Newport, USA).
Ammonium acetate, sodium hydroxide and diethyl ether (all
analytical reagent grade) were obtained from Fisher Scientific
(Loughborough, UK). Acetonitrile and methanol (HPLC grade)
were obtained from Fisher Scientific (Loughborough, UK). Pro-
pranolol, 1-methyl-3-nitrosoguanidine (MNNG), 2-chloro-1-
methyl-pyridinium iodide (CMPI), triethylamine (TEA), Girard
P reagent, methoxylamine hydrochloride, pyridine, LiCl, NaCl,
KCl and RbCl were obtained from Sigma–Aldrich (Poole, UK).
Acetic acid and ammonia (both Analar) was obtained from
BDH (Poole, UK). Tris(trimethoxyphenyl)phosphonium (TMPP)
were prepared in house as described elsewhere [7,8]. Solid
phase extraction cartridges (C8, Isolute) were obtained from
Kinesis (St. Neots, UK).
ionisation processes such as addition of H+, Na+ or NH4 ions
+
to a single pure substrate are disadvantageous to quantitative
analysis because they reduce sensitivity by dividing the ana-
lyte signal between several discrete m/z values. Also irregular
adduct formation increases analytical variability.
The objective of the current investigation was to use struc-
tural elucidation to understand firstly, why EG has a higher
affinity for ammonium than TG. In order to obtain precise
three-dimensional structures and charge localisations on each
atom a combination of X-ray crystallography and ab initio cal-
culations have been used. Quantum mechanics ab initio theory
2.2.
Sample preparation
For infusion the steroid glucuronides (100 ␮g/mL, 50 ␮L) were
dried under nitrogen (60 ◦C). They were then redissolved in the
positive ion mode infusion solvent (1 mL) consisting of 50:50

steroids 7 3 ( 2 0 0 8 ) 621–628
623
Fig. 2 – The structure of epitestosterone glucuronide (EG, centre) with the ketone and carboxylic acid functional groups
labelled. The structures of the EG ketone derivatives (left); methoxyamine (top) and Girard P (bottom) and carboxylic acid
derivatives (right); diazomethane (top) and TMPP (bottom) are shown. Theoretical masses shown.
automatic gain control using 3 microscans per scan and a
maximum injection time of 500 ms, which are the default set-
tings for this instrument. The automatic gain control target
values were full MS 2e7, SIM 2e7, MSn 2e7 and ZoomScan 1e7.
methanol:water containing ammonium acetate (7.5 mM) and
acetic acid (0.1%) to give a final concentration of 5 ␮g/mL. The
negative ion mode infusion solvent was 50:50 methanol:water
containing ammonia (0.01%). To assess the fragility of lithium,
sodium, potassium and rubidium adducts, solutions of EG
(5 ␮g/mL) in 1:1 water:methanol were prepared containing
either LiCl, NaCl, KCl or RbCl (0.5 mM).
2.5.
Derivatisation procedures
2.5.1. Derivatisation of the carboxylic acid on the
glucuronide moiety
2.3.
Instrumentation
2.5.1.1. Methylation. A diazomethane generator (Sigma–
Aldrich, Poole, UK) was used to prepare the diazomethane.
Ether (3 mL) was placed in the outer tube and MNNG (147 mg)
and water (0.5 mL) were placed in the inner tube. Sodium
hydroxide (5 M, 0.6 mL) was added dropwise to the inner tube.
The diazomethane co-distills with the ether ready for use.
To methylate the samples, diazomethane in ether (0.5 mL)
was added to EG or TG (10 ␮g/mL, 0.5 mL). The solutions were
dried under nitrogen (30 ◦C) and diluted with the positive ion
mode infusion solvent (1 mL) [9].
The LCQ DECA XP (Thermo Electron, Hemel Hempstead) QIT
mass spectrometer equipped with an orthogonal electrospray
source was used with the Xcalibur Version 1.3 software. The
instrument was tuned using the non-fragile [M + H]+ ion from
propranolol for mass, resolution and abundance.
2.4.
Infusion experiments
Sample solutions were introduced into the electrospray source
at a flow rate of 5 ␮L/min via a metal capillary. The built
in syringe pump was used for sample delivery. The sheath
gas and auxillary gas flows were set to 10 and 0 ‘arbitrary
units’, respectively. The capillary temperature was 260 ◦C and
the spray and capillary voltages were 4 kV and 4 V respec-
tively. Both MS and MS/MS experiments were analysed with
2.5.1.2. TMPP. The steroid glucuronides (100 ␮g/mL, 50 ␮L)
were dried under nitrogen (60 ◦C). Solutions of the catalysts
CMPI (15 mM in acetonitrile, 10 ␮L) and TEA (30 mM in ace-
tonitrile, 10 ␮L) were added and the samples were sonicated
(20 ◦C, 15 min). TMPP solution was then added (20 mM in ace-

624
steroids 7 3 ( 2 0 0 8 ) 621–628
tonitrile, 20 ␮L) and the samples again sonicated (20 ◦C, 15 min)
and then made up to 1 mL with the positive ion mode infusion
solvent [8].
K [13] which is co-ordinated with the oxygen atoms of the
steroid–glucuronide molecule.
+
The NH4 ion was placed close to the oxygen atoms look-
+
ing for spontaneous hydrogen bond formation. Thus the NH4
ion was linked to the glucuronide by one or two hydrogen
bonds forming the adduct. For each adduct after the spon-
taneous hydrogen bond formation, a molecular mechanics
2.5.2. Derivatisation of the ketone on the steroid moiety
2.5.2.1. Girard P. The steroid glucuronides (100 ␮g/mL, 50 ␮L)
were dried under nitrogen (60 ◦C).
A solution of Girard
+
energy minimization was carried out allowing the NH4 ion
P (10 mg/mL, 1 mL) prepared in methanol:water:acetic acid
(7:2:1) was added to each sample. The samples were incubated
(70 ◦C, 30 min), dried and the steroid hydrazone derivative
reconstituted in the positive ion mode infusion solvent (1 mL)
[10].
to find its best place. However, in some cases some con-
straints were used (e.g., fixing a neighbouring torsional angle)
to obtain the conformer. After that, each steroid–ion com-
plex was submitted to single point energy calculations with
ab initio levels similar to those applied to TG and EG. The fol-
lowing properties were measured: total energy, free energy
of the steroid and ion components, dipole moment, electro-
static charges, highest occupied molecular orbital (HOMO)
and lowest unoccupied molecular orbital (LUMO) location and
energies, electron density encoded electrostatic potential and
hydrogen bond distances. Calculations were carried out using
Spartan software (Wavefunction Inc., Irvine, CA, 2004).
2.5.2.2. Methoximation. The steroid glucuronides (100 ␮g/mL,
50 ␮L) were dried under nitrogen. Methoxylamine hydrochlo-
ride (2% in pyridine, 50 ␮L) was added to each sample and the
samples were incubated at 70 ◦C for 1 h [11]. The samples were
then dried under nitrogen (60 ◦C) and the residue was resus-
pended in water (0.5 mL). The steroid methyloxime derivative
was extracted from the excess methoxylamine hydrochloride
by passing the solution through a C8 solid phase extraction
cartridge pre-conditioned with methanol (3 mL) and water
(3 mL). The sample was washed with water (3 mL) and the
derivatised steroid glucuronides were eluted with methanol
(1 mL). The solution was dried under nitrogen (60 ◦C) and
reconstituted in the positive ion mode infusion solvent (1 mL).
3.
Results and discussion
3.1.
Analysis of EG and TG by mass spectrometry
Comparing the results from a triple quadrupole [1] to those
from a QIT, similar abundances of protonation to ammonium
adduction were obtained for TG and EG when infused into the
mass spectrometer via an ESI source in positive ion mode. The
same solvents were used in the two experiments. In the QIT,
the ratio of protonation to ammonium adduction was 5:1 for
TG and 3:1 for EG. This indicates that the pattern of adduction
is unaffected by the type of mass analyser used, being deter-
mined mainly in the ion source. Varying the concentration of
ammonium acetate (0, 7.5, and 15 mM) caused a proportionate
increase in ammonium adducts. Of course, chromatography
is also affected by this change in concentration of the buffer.
Furthermore, attempts to control adduct formation were ham-
pered by sodium contamination in solvents and the LC/MS
equipment. This contamination could not be eradicated.
ESI of unconjugated testosterone (T) and epitestosterone (E)
and sulphated T and E using the same infusion solvent did not
generate ammonium adducts, nor did hemisuccinate or ben-
zoate moieties, substituents with structural features similar to
those of glucuronic acid. Hence, ammonium adduct formation
appears to be specific to glucuronide conjugates.
The MS/MS spectra of each of the TG and EG adducts
were determined using a QIT instrument. It was observed
that an isolation width >2.5 u was required to achieve sen-
sitive MS/MS spectra. This occurs with “fragile ions” or ions
which (unusually) fragment under the application of the iso-
lation waveform, leaving little or no precursor ion available for
CID and hence generating weak MS/MS spectra [6]. Increasing
the isolation width, reduces the amount of energy imparted
to the isolated precursor and hence reduces fragmentation.
McClellan et al. proposed criteria for fragile ions, those requir-
ing an isolation width greater than 2.5 u and giving a peak
width greater than 0.31 u in ZoomScan mode. Using these cri-
teria, the [M + H]+, [M + NH4]+ and [M + Na]+ ions of EG and
2.5.2.3. Fragile ion experiments. Using the standard tune con-
ditions the resolution of the instrument was fixed with the
non-fragile [M + H]+ ion from propranolol. The test solutions
were infused into the QIT mass spectrometer (5 ␮L/min) using
the operating parameters described above. ZoomScan (10 u
mass range) data was acquired for 30 s centred on the mass
of interest ([M + H]+, [M + NH4]+, and [M + Na]+ in MS mode)
and the data averaged so that the width at 10% (W10%) could
be determined. The instrument was then set to MS/MS mode
using each appropriate centroid mass (to one decimal place).
To optimise isolation of ions, rather than fragmentation, no
collision energy was applied. The isolation width was then
increased from 1 to 10 u in increments of 0.25 u. Each isolation
width was monitored for 20 s and the average spectrum over
this period was used to determine the peak intensity.
2.5.2.4. TG and EG modelling from X-ray conformers and com-
putational chemistry procedures. TG was modelled from the
testosterone ab initio conformer reported previously [4] and the
glucuronide moiety was added from data of the X-ray crystal
of estriol 17␤-monoglucuronide [12]. Co-ordinates of EG were
acquired from its crystal co-ordinated with potassium (TG-K)
[13]. Regardless of the intrinsic lower energy conformation of
the crystals and ab initio conformers, both molecules were sub-
mitted to further energy minimization. An initial geometry
was obtained by the semi-empirical method MP3, and refined
at ab initio Hartree–Fock theory using sequentially the 3-21G
and 6-31G(*) basis set levels, to unveil the electronic properties
of the models.
The lower energy conformers of TG and EG were used
to study the steroid ammonium adduct formation and its
+
ion fragility. The NH4 moiety was joined to TG and EG
based on the co-ordinates of the potassium atom from TG-

steroids 7 3 ( 2 0 0 8 ) 621–628
625
respectively). The rubidium adduct and ammonium adducts
Table 1 – Minimum isolation width for the different
adducts of EG and TG
˚
whose ionic radii differ by just 0.05 A demonstrated the same
W10% providing strong evidence that the ammonium and
alkali metal cations adduct at the same site. The degree of
fragility of these adducts is likely therefore to be based on the
dissociation of the adduct ion from the adduction site dur-
ing the isolation process because cations of the alkali metals
with a higher charge density and therefore stronger interac-
tion with the adduction site were not fragile. However, the
limitation of this explanation is that metallic cations have
their own charge, which is completely available to accept elec-
trons from the oxygen lone pairs (ionic bonding) whereas the
Ion
m/z
Min i.w. (u)a
W10% (u)
[EG − H]−
[TG − H]−
[EG + H]+
463
463
465
465
482
482
487
487
1.5
1.5
7.0
2.5
6.5
8.0
2.5
2.5
0.22
0.25
0.56
0.25
0.61
0.89
0.17
0.25
[TG + H]+
[EG + NH4]+
[TG + NH4]+
[EG + Na]+
[TG + Na]+
a
The minimum isolation width (min i.w.), defined as the isolation
waveform notch width that preserves 90% of the maximum iso-
lated ion intensity, is presented for each ion. For comparison with
peak width data, the W10% data for each ion are presented.
+
NH4 ion has its charge delocalised among its atoms leaving
only sufficient charge to form weaker H-bonds.
In negative ion mode [EG − H]− and [TG − H]− are non-
fragile ions. The abstraction of a hydrogen atom is most likely
to occur at the carboxylic acid of the glucuronic moiety as this
is the most acidic hydrogen in the molecule. However, previ-
ous investigators have found that using negative ion mode for
the analysis of TG and EG is not as sensitive as positive ion
mode. MS/MS spectra from triple quadrupoles (which do not
suffer from a low mass cut-off) also show that fragments are
mainly derived from the glucuronide giving poorer specificity
and signal to noise which is especially important if quanti-
fying the analyte in a complex biological matrix [1]. MS/MS
analysis in the QIT gave very weak spectra as any fragments
derived from the glucuronide were below the low mass cut-off
and could not be stored in the trap.
Even though TG and EG are epimers, the [EG + H]+ was
fragile but [TG + H]+ was not. Since both the EG and TG ammo-
nium adducts were also fragile, complete conversion of these
analytes to ammonium adducts using a high ammonium
acetate concentration confers no benefit for reliable quan-
tification. Given the apparent structural similarity of the two
glucuronides it was possible that structural analysis of the
two compounds may uncover important but subtle differences
that would account for the fragility of one but not the other.
This would add to the scientific knowledge of fragile ions since
no publication was identified that explained which structural
features made an analyte more likely to be fragile.
TG in positive ion mode and the [M − H]− ion in negative
ion mode were tested for fragility in the QIT. The results are
shown in Table 1. TG was found to be stable in the proto-
nated form with a peak width of 0.25 u in ZoomScan mode
and requiring an isolation width in MS/MS mode of 2.5 u.
By contrast, the ammonium adduct of TG was very fragile
with a peak width of 0.89 u (i.e. ꢀ0.31 u) and requiring iso-
lation widths of 8.0 u (Table 1) because of the dissociation
of the adducts caused by energy imparted by the notch fre-
quency isolation waveform. Unlike TG, protonated EG was
fragile (W10% = 0.56 u, i.w. = 7.0 u) as was its ammonium adduct
(W10% = 0.61 u, i.w. = 6.5 u). Comparatively, [TG + NH4]+ is less
stable than [EG + NH4]+ as evidenced by their ZoomScan peak
widths and this is also indicated by the relative proportion
of protonation to ammonium adduction of TG (5:1) and EG
(3:1). This difference in stability indicates the importance of
the 17-configuration of these epimers causing their different
behaviour in the mass spectrometer. A comparison of the
extent of fragmentation in normal full MS mode shows that
EG fragments more than TG. Of note is the greater formation
of the aglycone (at m/z 289) suggesting that the fragility stems
from the glucuronide breaking away from the steroid molecule
as discussed later.
We tested whether the size of the adducting ion or
molecule might be important to determining its stability. All
the sodium adducts in positive ion mode were non-fragile.
Sodium ions exhibit strong affinity to oxygenated centres
(ether, ester, carboxylic or alcohol groups) and hence are also
likely to adduct at the same site as the ammonium. The
fact that the ammonium adducts are fragile but the sodium
adducts are not can be explained in terms of the specific inter-
3.2.
Theoretical assessment of EG and TG ammonium
adduct abundance and fragility
Structural features may be responsible for the different
behaviour of TG and EG in MS. To explore this further, struc-
tural analyses involving ab initio molecular modelling of TG
and EG were performed. The intention was to better under-
stand the structural features that favour adduct formation
and fragility so that strategies could be developed to overcome
these problems.
Although ammonium adduction to steroid glucuronides is
a well known phenomenon, a publication mapping the site of
adduction was not found. Therefore the adduction of ammo-
nium to EG and TG was studied using molecular models based
on X-ray crystallographic data. The EG X-ray crystallography
results contained a potassium ion and co-ordination of this
ion by oxygens of the glucuronide and the oxygen of the C3
keto group was observed. Based on that, and similar features
observed in potassium and ammonium picrate crystals [14],
actions of the cation with the glucuronide. Since Na+ and NH4
+
+
+
˚
˚
have different ionic radii (Na : 0.98 A, NH4 : 1.43 A) their co-
ordination with the glucuronide moiety depends on the spatial
availability which is determined by the constitution and con-
figuration of the glucuronic acid. To test this hypothesis the
fragility of adducts of increasing ionic radii were evaluated
˚
˚
˚
(lithium, 0.68 A; sodium, 0.98 A; potassium, 1.33 A; rubidium
˚
1.48 A) to encompass that of ammonium. Metal cation adduc-
tion compared with protonation was approximately four times
greater for lithium and sodium than for potassium and rubid-
ium. Rubidium was the only metal cation adduct that was
fragile (W10% values: Li+ 0.31; Na+ 0.28; K+ 0.28; Rb+ 0.56 u

626
steroids 7 3 ( 2 0 0 8 ) 621–628
it was clear that oxygen atoms readily accept the ammonium
ion through hydrogen bonds, forming the ammonium adduct.
+
Only one NH4 ion is allowed at once in both TG and EG
+
because the NH4 ion modifies the glucuronide electrostatic
environment changing it from negative to positive and gen-
erating a huge dipole moment. This concurs with the mass
spectral results where no di-ammonium adduct was observed.
Several sites of ammonium adduction were possible utilising
the oxygens of the glucuronide ring to co-ordinate the adduct
molecule. In total, molecular modelling identified seven pos-
sible ammonium adduct conformers for TG and eight for EG
(Fig. 1). Interestingly, the two ether oxygens labelled O2 and
O3 of TG do not partake in adduct formation together because
of steric hindrance by the steroidal C18 methyl group and the
two hydrogens of C16. These findings show that TG has some
+
restrictions in NH4 adduct formation giving fewer possible
conformers than EG and this may explain why ammonium
adduction is more favourable with EG than with TG.
Since the fragility of the ammonium adduct is likely to be
caused by displacement of the adduct molecule during the
isolation step of MS/MS in the QIT, the binding energy of the
ammonium to the glucuronide for each conformer was calcu-
lated. This showed that although EG ammonium adducts are
more abundant, they have higher energy than TG adducts as
shown in Fig. 1. The energy difference is more marked in four
of the eight EG adducts suggesting lower stability. Although
all these conformers are possible theoretically it is unknown
which ones actually form in the source. For example, one of
the most favourable of the theoretical conformers was ammo-
nium adducted to the C3 ketone group (O1 in Fig. 1). However,
this conformer could also form with unconjugated or sulpho
conjugated T and E by MS but ammonium adducts are not
observed with these molecules. This suggests that formation
of this conformer is not favourable in the ESI with T, E and
their conjugates presumably because protonation at this site
is much more energetically favourable.
The proposed adduction sites agree well with the finding
that only glucuronide and not sulphate, benzoate or hemisuc-
cinate conjugates of T or E form ammonium adducts. Also,
previous investigations have shown that steroid glucuronides
without the 4-ene-3-one structure form the ammonium
adduct almost exclusively [1]. These workers conclude that
Table 2 – Minimum isolation width for the derivatives of
EG and TG
Derivative
m/z
Min i.w. (u)a
W10% (u)
EG diazomethane
TG diazomethane
EG TMPP
TG TMPP
EG Girard P
TG Girard P
EG methoxyamine
TG methoxyamine
479.0
479.0
1036.7
1036.7
598.5
598.5
494.3
494.3
10.0
2.5
8.5
2.2
1.0
1.0
1.0
1.0
0.78
0.29
0.33
0.28
0.28
0.28
0.28
0.28
Fig. 3 – Spectra of EG (EG was derivatised and infused into
the ion trap at 5 ␮L/min, and the samples analysed in MS
mode): EG underivatised (A); EG Girard P (B); EG
methoxyamine (C); EG methylated (D); EG TMPP (E). For
structures see Fig. 2.
a
The minimum isolation width (min i.w.), defined as the isolation
waveform notch width that preserves 90% of the maximum iso-
lated ion intensity, is presented for each ion. For comparison with
peak width data, the W10% data for each ion are presented.

steroids 7 3 ( 2 0 0 8 ) 621–628
627
protonation occurs on the 4-ene-3-one structure but do
not comment on the location of the ammonium adduct.
Thus, since they are epimers both with 4-ene-3-one struc-
tures, the fact that [EG + H]+ was fragile but [TG + H]+ was
not is paradoxical. Molecular modelling was employed to
investigate whether any particular structural characteristic
could explain this. Firstly the adduction sites were differ-
entiated by their position on the analyte. There appears to
be two major adduction sites, one located at the A-ring
and one located around the glucuronide oxygens. Ammo-
nium adduct formation elicits a large dipole moment which
makes the glucuronide moiety quite positive and the A–B ring
region very negative. At the A–B ring region the highly res-
onant 4-ene-3-one system is very important to this polarity
because it is the site of both the highest occupied molec-
ular orbital (HOMO) and the lowest unoccupied molecular
orbital (LUMO) in TG and EG. This concurs with previously
published molecular models of unconjugated T and E [4].
This negative system in the A-ring, is the site of proto-
nation while the cation adducts such as sodium adducts
attach in a similar way to the ammonium adducts at the glu-
curonide.
release E (with associated proton at O2) which would explain
the observed peak of m/z 289 in the MS spectrum.
3.3.
Derivatisation
Since structural features could account for some of the differ-
ences in the behaviour of TG and EG by MS, it was possible
that by derivatising these molecules fragility and multiple
adduct formation would be eliminated. Derivatives have been
used in mass spectrometry to improve sensitivity [8]. The
analyte broadly retains its chemical structure and gains new
chemical properties that can be used for quantification and
separation such as a positive charge to enhance ionisation.
The ab initio calculations demonstrated that the addition of
a second ammonium adduct to an already charged analyte
is highly unfavourable. Therefore it was likely that the use
of charged derivatives on these molecules would eliminate
multiple adduct formation. Positively charged and neutral
derivatives of the keto function on the steroid moiety and
the carboxylic acid of the glucuronide moiety were tested for
fragility in the QIT to investigate whether stability could be
restored by chemical means.
The stability of Li+, Na+ and K+ adducts in the ion trap mass
spectrometer and the fragility of Rb+ demonstrates that the
size of the adduct is important in terms of the strength of
its interactions with the analyte. By modelling these adducts
with EG it was possible to determine the bond distances
between the oxygens of the analyte and the cation. The aver-
age bond distance increases steadily as we descend the alkali
The keto structure was derivatised using Girard P hydrazine
or methoxylamine hydrochloride. These derivatives convert
the steroid ketone group to a hydrazone or oxime, respectively,
thus removing the 4-ene-3-one functionality. The carboxylic
acid was derivatised using diazomethane or TMPP. Dia-
zomethane methylates the carboxylic acid leaving it less
acidic whilst TMPP is a bulky group with a permanent posi-
tive charge. The structures of these molecules are shown in
Fig. 2.
When EG was derivatised using either of the charged
derivatives (Girard P or TMPP) no adduct was observed in
the MS spectra. However, when non-charged derivatives were
used (methoxylamine and diazomethane) adducts were still
formed, for example in the spectrum of methylated EG,
intense ions for [EG + CH2 + NH4]+, [EG + (CH2)2 + NH4]+ and
[EG + (CH2)3 + NH4]+ were observed and in the methoxyamine
(MO) spectrum [EG MO + Na]+ was observed (Fig. 3). This was
predicted since it is unfavourable for a charged adduct to asso-
ciate with a previously permanently charged analyte.
Derivatisation of the carboxylic acid on the glucuronide
with TMPP led to partial stabilisation of EG but not enough
to bring it within the non-fragile range. The fragility of EG
remained after methylation of the carboxylic acid group
with diazomethane. The fragility experiments showed that
derivatisation of the ‘4-ene-3-one’ structure totally stabilised
both protonated EG and TG so that they could be analysed by
ion trap MS/MS with an isolation width of 1 u (Table 2).
metal adducts: Li+ (1.793 A), Na (2.149 A), K (2.470 A) and Rb
+
+
+
˚
˚
˚
+
(2.626 A). Since K adducts are stable but Rb+ adducts are not,
the critical oxygen to cation bond length required to retain the
adduct during isolation in the ion trap is likely to be between
˚
˚
2.470 and 2.626 A. This explains why the association between
the Rb+ and the analyte is weaker than the other cations
resulting in this adduct being a fragile ion.
Analysis of the two molecules shows that the A–B ring
is stable and very similar in EG and TG, therefore the main
differences exist at the glucuronide end of the molecule.
Examination of the MS spectrum of EG showed that the agly-
cone was also present in the spectrum indicating that the bond
between the C17 oxygen and the glucuronide is very labile. EG
appears to be more labile than TG since the aglycone is not
observed in the MS spectrum of TG. Given this information,
the bonds C17 to O2 and O2 to C1ꢁ of the glucuronide were
examined in more detail since these connect the glucuronide
moiety to the steroid. The lengths of these bonds were found to
˚
be 1.422 and 1.369 A respectively in EG compared to the slightly
˚
shorter bonds (1.411 and 1.360 A) in TG. In this respect the
shorter bonds of TG may be considered to be slightly stronger.
However, greater differences were observed when the point
charges were calculated on each atom of these bonds. Both
the O2 atoms were slightly basic but the charge on the O2
atom of EG is −0.708 eV compared to −0.497 eV with TG. Hence
protonation of the O2 atom of EG is more favourable than in
TG. Protonation is a common precursor to hydrolysis reactions
in chemistry by stabilising the charge on the leaving group.
In a similar way, protonation at the O2 of EG may promote
bond rupture under the stresses of the ion trap environment
explaining its observed fragile ion status. This scenario would
4.
Conclusions
Adduct formation occurs in the source of ESI mass spectrome-
ters and is therefore a problem with all types of analyser such
as triple quadrupole and QIT. In the current study the sites
of ammonium and sodium adduction in steroid glucuronides
were mapped which can help to develop strategies to min-
imise adduct formation in the future. One such strategy is to
derivatise the molecules.

628
steroids 7 3 ( 2 0 0 8 ) 621–628
[2] Cai J, Henion J. J Chromatogr B: Biomed Sci Appl
Molecular modelling has proved a useful tool to assist with
1997;691:357–70.
[3] Sudha R, Panda M, Chandrasekhar J, Balaram P. Chemistry
2002;8:4980–91.
[4] Kubli-Garfias C. Theochem-J Mol Struct 1998;422:167–77.
[5] Wieboldt R, Campbell DA, Henion J. J Chromatogr B: Biomed
Sci Appl 1998;708:121–9.
[6] McClellan JE, Murphy JP, Mulholland JJ, Yost RA. Anal Chem
2002;74:402–12.
[7] Leavens WJ, Lane SJ, Carr RM, Lockie AM, Waterhouse I.
Rapid Commun Mass Spectrom 2002;16:433–41.
[8] Barry SJ, Carr RM, Lane SJ, Leavens WJ, Manning CO, Monte
S, et al. Rapid Commun Mass Spectrom 2003;17:484–97.
[9] Huskova R, Bartak P, Cap L, Friedecky D, Adam T. J
Chromatogr B: Anal Technol Biomed Life Sci 2004;799:
303–9.
[10] Griffiths WJ, Wang Y, Alvelius G, Liu S, Bodin K, Sjovall J. J
Am Soc Mass Spectrom 2006;17:341–62.
[11] Chen BG, Wang SM, Liu RH. J Mass Spectrom
2007;42:1012–23.
the design of chemistries to avoid adduction and fragility of
labile steroids. Our results show that derivatisation can elim-
inate both adduct formation and fragility from TG and EG. To
eliminate adduct formation the derivative should be charged
thus making the addition of a second charge by adduction
highly unfavourable. However, to eliminate fragility, the site
of derivatisation should be on the steroid rather than the
glucuronide. Hence in the current example Girard P was the
optimum derivative to use. Nevertheless, for other analytes, it
is important to consider carefully which derivative to use and
the site of derivatisation.
Acknowledgement
The authors would like to thank Ricardo Vazquez-Ramirez for
his technical assistance.
[12] Wu YQ, Waters JM, Blackwell LF. J Chem Soc-Perkin Trans
1996;2:1449–53.
[13] Kubli-Garfias C, Welchman H, Soriano-Garcia M. Anal Sci:
X-ray Structure Analysis 2006;22:x297–8.
[14] Maartman-Moe K. Acta Crystal 1969;B25:1452–60.
r e f e r e n c e s
[1] Kuuranne T, Vahermo M, Leinonen A, Kostianen R. J Am Soc
Mass Spectrom 2000;11:722–30.