Regioselectivity of human UDP-glucuronosyltransferase isozymes in flavonoid biotransformation by metal complexation and tandem mass spectrometry

Article abstract of DOI:10.1016/j.bcp.2011.08.015

Based on reactions with five flavonoids, the regioselectivities of twelve human UDP-glucuronosyltransferase (UGT) isozymes were elucidated. The various flavonoid glucuronides were differentiated based on LC-MS/MS fragmentation patterns of [Co(II)(flavonoid-H)(4,7-diphenyl-1,10-phenanthroline) ₂]⁺ complexes generated upon post-column complexation. Glucuronide distributions were evaluated to allow a systematic assessment of the regioselectivity of each isozyme. The various UGT enzymes, including eight UGT1A and four UGT2B, displayed a remarkable range of selectivities, both in terms of the positions of glucuronidation and relative reactivity with flavanones versus flavonols.

Full text of DOI:10.1016/j.bcp.2011.08.015

Biochemical Pharmacology 82 (2011) 1764–1770
Contents lists available at SciVerse ScienceDirect
Biochemical Pharmacology
journal homepage: www.elsevier.com/locate/biochempharm
Regioselectivity of human UDP-glucuronosyltransferase isozymes in flavonoid
biotransformation by metal complexation and tandem mass spectrometry
*
Scott A. Robotham, Jennifer S. Brodbelt
Department of Chemistry and Biochemistry, University of Texas at Austin, 1 University Station A5300, Austin, TX 78712, USA
A R T I C L E I N F O
A B S T R A C T
Article history:
Based on reactions with five flavonoids, the regioselectivities of twelve human UDP-glucuronosyl-
transferase (UGT) isozymes were elucidated. The various flavonoid glucuronides were differentiated
based on LC–MS/MS fragmentation patterns of [Co(II)(flavonoid-H)(4,7-diphenyl-1,10-phenanthro-
line)2]+ complexes generated upon post-column complexation. Glucuronide distributions were
evaluated to allow a systematic assessment of the regioselectivity of each isozyme. The various UGT
enzymes, including eight UGT1A and four UGT2B, displayed a remarkable range of selectivities, both in
terms of the positions of glucuronidation and relative reactivity with flavanones versus flavonols.
ß 2011 Elsevier Inc. All rights reserved.
Received 21 June 2011
Accepted 15 August 2011
Available online 30 August 2011
Keywords:
Human UDP-glucuronosyltransferase
Flavonoid
Regioselectivity
Mass spectrometry
Metal complexation
Glucuronidation
1. Introduction
compound metabolized or unmodified that is not absorbed prior to
reaching the large intestine may be absorbed by microflora, a
Flavonoids,
a
class of polyphenols found in fruits and
process leading to decomposition of the flavonoid by ring fission
and causing the release of small phenolic acids that are excreted in
the urine [10]. Since most of the flavonoids ingested are conjugated
and consequently absorbed as conjugates, there has been
increasing interest in understanding the formation, uptake,
distribution, and chemopreventive properties of the conjugates.
To facilitate such investigations, the development of sensitive
analytical methods to characterize, identify and track the flavonoid
conjugates is paramount.
Glucuronidation of flavonoids is carried out in the body by the
UDP-glucuronosyltransferase (UGT) family of enzymes. These
enzymes have been found in every major organ involved in
digestion, as well as the kidneys and liver [11]. To date, nineteen
different isomers of the UGT enzyme have been identified [12],
which are categorized into three different subgroups (UGT1As,
UGT2As and UG2Bs). There are nine isoforms of the UGT1A group
and seven in the UGT2B group, and together they play a major role
in Phase II metabolism. The role of UGT2A isoforms remains
unknown [12]. UGT enzymes catalyze the addition of glucuronic
acid at a hydroxyl group, carboxylic acid, sulfide group, amine, or in
rare cases a methyl group [11]. Flavonoids possessing one to
multiple hydroxyl groups may undergo O-glucuronidation at
various positions when metabolized by UGT enzymes. However,
the specific positions which are glucuronidated by each enzyme
are still not fully established.
vegetables, have been shown not only to have anti-inflammatory
properties but also to exhibit promising chemopreventive
properties against cancer and cardiovascular disease [1,2]. The
positive bioactivities of flavonoids have been demonstrated in a
variety of in vitro, in vivo, and case control studies [2–4]. In recent
years the presumed chemopreventive properties have been under
closer scrutiny due to the poor bioavailability of most unmodified
flavonoid aglycones in the body [5], coupled with the growing
documentation that flavonoids undergo extensive biotransforma-
tion [6–8]. The metabolism of flavonoids has a great impact on
their absorption and distribution, and importantly biotransforma-
tion can substantially alter the chemical properties of the
flavonoids, such as altering the bioactivities [9].
Most flavonoids are found naturally as glycosylated forms in
fruits and vegetables. When ingested the flavonoid glycosides
undergo deglycosylation by
hydrolase enzymes primarily found in the small intestine [10].
After loss of their sugar side-chains, flavonoids are rapidly
metabolized by mainly Phase II enzymes found in small intestine,
kidneys, and most importantly the liver [10]. This process results in
glucuronidation, sulfation, methylation, or hydroxylation depend-
ing on the nature of the interacting enzyme [11]. Any flavonoid
b-glucosidase or lactose phloridzin
The structural characterization of flavonoids and their
metabolites has proven to be a challenging task. All flavonoids
* Corresponding author. Tel.: +1 512 471 0028.
E-mail address: jbrodbelt@mail.utexas.edu (J.S. Brodbelt).
0006-2952/$ – see front matter ß 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.bcp.2011.08.015

S.A. Robotham, J.S. Brodbelt / Biochemical Pharmacology 82 (2011) 1764–1770
1765
R '
3
2.2. Synthesis of flavonoid glucuronides by UGT enzymes
R '
3
OH
R '
4
The procedure for the glucuronidation reactions was modified
from the protocol reported in Davis et al. [34]. Each enzyme was
B
B
HO
O
7
HO
O
7
divided into 25
m
L aliquots and stored at ꢀ80 8C until use. The
A
C
A
C
3
following reaction was set up for each combination of UGT enzyme
(UGT1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, 2B15,
2B17) and flavonoid (hesperetin, isorhamnetin, kaempferol,
naringenin, quercetin). All volumes were delivered using appro-
priate micropipettes. The synthesis was carried out by adding
3
OH
5
5
OH
O
OH
O
Flavanones (MW) R3’ R4’
Hesperetin (302) OH OCH3
Naringenin (272) OH
Flavonols (MW)
R3’
Isorhamnetin (316) OCH3
2 mM aqueous UDPGA (65
buffer pH 7.0 (378.75 L), and 10 mM methanolic solution of
flavonoid (6.25 L) to a microcentrifuge tube. The reaction was
started by addition of 25 L of a UGT enzyme (5 mg/mL). This
mL), 20 mM potassium phosphate
H
Kaempferol (286)
Quercetin (302)
H
OH
m
m
Fig. 1. Structures of flavonoids.
m
concentration of enzyme was used based on experiments reported
by Plumb et al. [41]. The mixture was incubated at 37 8C overnight.
To stop the reaction, 1.5 mL of acetone was added. The tubes were
centrifuged for 10 min at 16,000 ꢁ g. The supernatant was
removed and the acetone was evaporated by placing the tubes
in a Savant DNA120 SpeedVac Concentrator (Thermo Electron,
Waltham, MA, USA) on low heat for 1 h and 40 min. The remaining
mixture was refrigerated until analysis. In a study by Easterbrook
et al. [42], the activities of UGT enzymes were evaluated in the
presence of organic solvents at various concentrations, and it was
reported that there were no apparent effects on enzyme activities
for solutions containing up to 2% methanol content, so a significant
effect on enzymatic activity was not anticipated for the present
study. A low concentration of methanol was used in the present
study to enhance solubility of the flavonoids and assure more
accurate concentrations in solution.
share the same basic three-ring structure and may differ by the
position of a single functional group, making their positive
identification difficult by many analytical methods. Mass spec-
trometry has proven to be one of the most effective tools for
identification of flavonoids, in large part due to the informative
fragmentation patterns generated by collision induced dissocia-
tion (CID) upon application of MS/MS strategies [13], especially
when coupled with HPLC to allow separation of complex mixtures
of flavonoids [14–20]. We have extended the capabilities of MS/
MS methods for differentiation of flavonoids by formation of
complexes containing a flavonoid, a metal, and an auxiliary
organic ligand [21]. These complexes, upon CID, give unique
fragmentation patterns that allow confident identification and
differentiation of flavonoids, even for isomers. We have evaluated
a number of metal complexation approaches and shown their
versatility [22–29], including the adaptation of the methods for
identification of metabolites in urine and plasma [30–33]. More
recently, we applied the metal complexation/MS/MS methodolo-
gy to gain insight into the regioselectivity of the UGT1A1 enzyme
with various flavonoids [34]. In this prior study, the products were
confidently identified and the distributions of various glucur-
onidated products were quantified.
2.3. LC–MS/MS analysis
LC–MS analysis of the flavonoid glucuronides was undertaken
using a Waters Alliance 2695 HPLC system (Milford, MA, USA) or a
Hitachi L-7000 HPLC system (Hitachi High Technologies America,
Pleasanton, CA, USA) and LCQ Duo quadrupole ion trap mass
spectrometer (Thermo Electron, Waltham, MA, USA) with electro-
spray ionization (ESI). The column was a Waters Symmetry C18
In this present study, we have expanded our investigation of the
selectivity of glucuronidation of the twelve most common UGT
enzymes (1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, 2B15,
2B17) for five of the most commonly consumed flavonoids
(hesperetin, isorhamnetin, kaempferol, naringenin, quercetin).
(Fig. 1) While biotransformation of flavonoids has been an area
of much interest [12,35–40], this is the first time the isomeric
flavonoid glucuronide products of such a large array of enzymatic
syntheses have been determined, thus providing detailed insight
into the selectivities of the UGT isoenzymes. This systematic study
provides benchmark data for assessment of UGT enzymatic
regioselectivity and establishes predictive correlations of biotrans-
formation upon consumption of flavonoids.
column, 2.1 mm ꢁ 50 mm, 3.5
m
m particle size, with a guard
L. The mobile phase was
column. The injection volume was 30
m
0.33% formic acid in water (A) and 0.33% formic acid in acetonitrile
(B). The gradient used began at 15% B and increased to 40% over
30 min.
Samples were first analyzed in the negative ESI mode in order to
search for flavonoid glucuronides. The spray voltage was set at
4.5 kV, the heated capillary temperature was 200 8C, and the
automatic gain control was set to 5 ꢁ 107 ions with a maximum
injection time of 500 ms and 5 microscans averaging. All other
parameters were set to obtain optimal signal. The positive ESI
mode was used for MS/MS analysis of the flavonoid/metal
complexes. The metal complexes were formed by post-column
addition of a methanolic solution of 5
mM CoBr2 and 4,7-dpphen,
which was infused at a rate of 20 L/min controlled by a syringe
m
2. Methods and materials
pump. The spray voltage for the positive ion mode was set to 5 kV,
and the heated capillary temperature was 200 8C. The automatic
gain control for MS/MS was set to 2 ꢁ 107 ions with a maximum
injection time of 500 ms and 5 microscan averaging, the isolation
width was set to 4 Da, and a collision energy of 35% normalized
collision energy was used for collision induced dissociation.
2.1. Reagents
All UDP-glucuronosyltransferase isozymes were purchased
from BD Biosciences (Woburn, MA, USA). UDP-Glucuronic acid
(UDPGA) trisodium salt, 4,7-diphenyl-1,10-phenanthroline (4,7-
dpphen), cobalt(II) bromide, hesperetin, naringenin, isorhamnetin,
kaempferol, and quercetin were purchased from Sigma–Aldrich
(St. Louis, MO, USA). HPLC grade acetonitrile, HPLC grade water,
potassium phosphate, and methanol were purchased from Thermo
Fisher Scientific, Inc. (Waltham, MA, USA).
2.4. LC-UV monitoring of reaction products
In order to estimate the relative product distributions of
different flavonoid glucuronides for each enzymatic reaction
leading to multiple products, the peak area for each resulting

1766
S.A. Robotham, J.S. Brodbelt / Biochemical Pharmacology 82 (2011) 1764–1770
product was integrated based on its LC-UV chromatographic
profile at 360 nm. The area of each product peak was divided by the
total area of all product peaks in order to calculate the product
distributions as percentages. This percentage was rounded to the
nearest 5%.
onidated products (aglycon + 176 + 176) based on the m/z values of
the anticipated products. A typical total ion chromatogram obtained
forthe hesperetin product mixtureis shownin Fig. 2a withextracted
ion chromatograms shown in Fig. 2b and c for deprotonated
hesperetin (m/z 301) and the monoglucuronidated products (m/z
477). No diglucuronidated products were detected. After screening
the enzymatic incubates, then the metal complexes were generated
upon LC–MS by post-column complexation and subjected to CID for
structural characterization. Examples of the resulting MS/MS
spectra are shown in Fig. 3 for hesperetin glucuronides produced
from UGT1A6 and in Fig. 4 for kaempferol glucuronides produced
from UGT1A1. All products identified are summarized for the two
classes of flavonoids (flavanones and flavanols) in Table 1 along with
the distribution of products based on integration of the chro-
matographic peak areas of each product and unreacted flavonoid.
Inclusion of the unreacted flavonoids gives an indication of the
relative reactivity promoted by each enzyme for a particular
flavonoid.
3. Results
Our objective was to map the formation of various flavonoid
glucuronides for each glucuronosyltransferase enzyme. In order to
differentiate the flavonoid glucuronide isomers, MS/MS spectra of
the metal complexes [Co(II) (FG–H) (4,7-dpphen)2]+ were analyzed
along with HPLC retention times (where FG represents a flavonoid
glucuronide). These metal complexes were produced via post-
column complexation in LC–MS runs of the product mixtures
obtained for each flavonoid/glucuronosyltransferase combination.
Whereas the fragmentation patterns of deprotonated flavonoid
glucuronides are typically not distinctive for isomers, the MS/MS
spectra of the metal complexes allows differentiation and
confident elucidation of the isomers. As described previously
[30,34], the metal complexes containing 7-O-glucuronides show
losses of the auxiliary ligand (-Aux) and the glucuronic acid moiety
(-GlcA), both individually or together (-(GlcA + Aux)), upon CID.
The loss of the flavonoid aglycon (-Agl) is also a characteristic
fragment of 7-O-glucuronides. The metal complexes of the 5-O-
glucuronides and 3-O-glucuronides both show a prominent loss of
the glucuronic acid moiety in conjunction with the auxiliary ligand
(-(GlcA + Aux)); moreover, the 5-O product has been shown to
elute prior to the 7-O product [34] whereas the 3-O-glucuronide
elutes after. For B-ring-glucuronides, the characteristic fragments
include the loss of the auxiliary ligand (-Aux) and the losses of both
the auxiliary ligand and glucuronide moiety together (-
(GlcA + Aux)) [30]. 40-O-glucuronides elute before 30-O glucur-
onides and after the 3-O-glucuronides. These systematic MS/MS
and elution patterns were utilized for assignment of the
glucuronide products in the present study. A table listing all the
fragments found for the products is shown in Supplement Table 1.
Once the enzymatic reactions were quenched, each of the
incubates was centrifuged and the supernatant was screened by LC–
MS in the negative ESI mode. The resulting total ion chromatograms
were searched both for the unreacted flavonoid aglycon, mono-
glucuronidated flavonoid products (aglycon + 176), and diglucur-
3.1. Flavanone glucuronides
Hesperetin formed monoglucuronidated products upon expo-
sure to all the glucuronosyltransferases except for UGT1A4 (which
in fact produced no glucuronides for any of the flavonoids). These
products were analyzed by the metal complexation/MS/MS
method described in the methods and materials section and
assigned based on their characteristic fragmentation patterns and
retention times. For UGT1A1, 1A3, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7,
2B15, and 2B17, two products were found (Table 1). The MS/MS
spectrum of the first glucuronide product showed losses of GlcA,
Agl, and Aux, which is characteristic of 7-O-glucuronide products.
The second eluting product yielded the prominent loss of
(GlcA + Aux) as well as the loss of Aux, which as described above
is characteristic of 30-O-glucuronidation. For UGT1A6, hesperetin
showed a total of three products (see corresponding MS/MS
patterns in Fig. 3). Two of the products are identical to those
described above for the other glucuronosyltransferase reactions.
The third product eluted prior to the 7-O-glucuronide, and the MS/
MS spectrum shows one prominent loss of (GlcA + Aux), thus
confirming this product as a 5-O-glucuronide.
For naringenin all of the glucuronosyltransferases resulted in
only a single product except for UGT1A4, which promoted no
Hesperetin 7-O-
Glucuronide
Hesperetin
100
a
monosaccharide
80
(m/z 463)
60
40
20
Hesperetin 3’-
O-Glucuronide
Hesperetin
100
80
Hesperetin
b
60
40
20
100
80
Hesperetin 7-O-
Glucuronide
c
60
40
20
Hesperetin 3’-
O-Glucuronide
6
10
14
18
22
26
30
Time (min)
Fig. 2. (a) Total ion chromatogram for UGT1A10 reaction with hesperetin, including background subtraction. (b) Selected ion chromatogram for deprotonated hesperetin (m/z
301). (c) Selected ion chromatogram for deprotonated hesperetin monoglucuronide (m/z 477).

S.A. Robotham, J.S. Brodbelt / Biochemical Pharmacology 82 (2011) 1764–1770
1767
-(GlcA + Aux)
692
a
b
c
100
80
60
40
20
0
-GlcA
1024
-(GlcA + Aux)
692
-GlcA
1024
100
80
60
40
20
0
-Aux
868
-Agl
898
-(Aux + Alg)
-H2O
1182
566
-(GlcA + Aux)
692
100
80
60
40
20
0
-Aux
868
-Agl
-GlcA
1024
-(Aux + Agl)
566
898
400
500
600
700
800
900
1000
1100
1200
m/z
Fig. 3. CID mass spectra of [Co(II) (FG-H) (4,7-dpphen)2]+ for all UGT1A6/hesperetin products: (a) 5-O glucuronide, m/z 1200, (b) 7-O glucuronide, m/z 1200, (c) 30-O-
glucuronide, m/z 1200. -Aux (loss of auxiliary ligand); -GlcA (loss of glucuronic acid moiety); -Agl (loss of flavonoid aglycon).
-(GlcA + Aux)
676
100
a
80
60
40
20
0
-GlcA
1008
-Aux
-Aux
100
80
60
40
20
0
852
b
c
-(GlcA + Aux)
676
-(Aux + Agl)
566
-GlcA
1008
-Agl
898
100
80
60
40
20
0
852
-(GlcA + Aux)
-(Aux +Agl)
-GlcA
1008
676
566
400
500
600
700
800
900
1000
1100
1200
m/z
Fig. 4. CID mass spectra of [Co(II) (FG-H) (4,7-dpphen)2]+ for all UGT1A1 kaempferol products: (a) 5-O-glucuronide, (m/z 1184), (b) 7-O-glucuronide, (m/z 1184), (c) 40-O-
glucuronide, (m/z 1184). -Aux (loss of auxiliary ligand); -GlcA (loss of glucuronic acid moiety); -Agl (loss of flavonoid aglycon).
products at all and UGT1A6 which led to the formation of two
products. The single dominant product dissociated by pathways
characteristic of a 7-O-glucuronide (losses of GlcA, Agl, and Aux).
For the UGT1A6 reaction the second product showed a prominent
loss of (GlcA + Aux) which is assigned as the 5-O-glucuronide
owing to its elution prior to the 7-O-glucuronide product (see CID
mass spectra in Supplementary Fig. 1).
very low abundance. This product could either be a 3-O
glucuronide or a 40-O glucuronide because its elution after the
7-O glucuronide ruled it out as a 5-O glucuronide. Moreover the 30
position has a methoxy group, and the absence of a methyl group
loss allowed a 30-glucuronide structure to be ruled out. The minor
loss of Aux in the MS/MS spectrum suggested that this was the 40-O
glucuronide, but the evidence was somewhat inconclusive due to
the lack of other diagnostic ions.
3.2. Flavonol glucuronides
The UGT1A1 enzyme promoted the formation of one product for
isorhamnetin that dissociated by a major loss of (GlcA + Aux) and a
minor loss of Aux which had a very low abundance. Since there are
no other products in this reaction, confident identification of this
product was not possible; however comparing the elution time of
this product with confirmed products of other isorhamnetin/
glucuronidation reactions shows that this product has a retention
time matching that of the second product of the UGT1A3, which
suggest that this product also having the minor loss of Aux in the CID
spectrum appears to be a 40-O-glucuronide. However this product
Upon reactions with most of the glucuronosyltransferases,
isorhamnetin displayed no reactivity in the presence of the
UGT1A4, 1A6, 1A8, 2B4, 2B7, 2B15, and 2B17 enzymes. Reaction
with the UGT1A3 enzymes led to the formation of two products.
The first product showed losses of GlucA, Agl and Aux upon CID,
allowing ready assignment as the 7-O-glucuronide. The second
product demonstrated loss of (GlcA + Aux) as the major product,
and also showed a loss of Aux though this fragment occurred in

1768
S.A. Robotham, J.S. Brodbelt / Biochemical Pharmacology 82 (2011) 1764–1770
Table 1
Glucuronide product distributions relative to unreacted flavonoid.
1A1
1A3
1A4
1A6
1A7
1A8
1A9
1A10
2B4
2B7
2B15
2B17
Hesperetin
Hesperetin
60
–
15
–
100
–
10
45
–
70
–
30
–
75
–
80
60
–
100
–
100
–
5-O-glucuronide
7-O-glucuronide
30-O-glucuronide
Naringenin
Trace
90
–
30
10
10
75
–
5
10
20
30
40
15
10
20
30
10
Trace
Trace
Trace
Trace
–
Trace
50
Trace
Naringenin
40
–
50
–
100
–
30
80
–
65
–
20
–
90
–
70
–
80
–
95
–
95
–
5-O-glucuronide
7-O-glucuronide
40-O-glucuronide
Isorhamnetin
Trace
70
50
–
50
–
–
20
–
35
–
80
–
10
–
30
–
20
–
5
5
–
–
–
–
Isorhamnetin
80
–
35
100
–
100
–
90
10
Trace
–
100
–
95
5
95
100
–
100
–
100
–
100
–
5-O-glucuronide
7-O-glucuronide
3-O-glucuronide
40-O-glucuronide
Kaempferol
–
Trace
Trace
(5)
–
50
–
–
–
–
–
–
–
–
(20)
(20)
(15)
(15)
–
–
–
–
–
–
–
–
–
–
–
–
–
(5)
–
–
–
–
Kaempferol
40
5
90
100
–
60
75
10
15
–
90
75
100
95
-
100
100
–
100
–
5-O-glucuronide
7-O-glucuronide
3-O-glucuronide
40-O-glucuronide
Quercetin
Trace
10
30
Trace
10
Trace
Trace
–
Trace
Trace
–
50
–
–
10
5
–
5
15
5
–
–
–
–
–
–
–
–
–
5
Trace
–
Trace
–
Trace
Trace
–
Trace
–
–
Quercetin
25
–
55
–
100
–
95
100
–
100
–
70
–
95
100
–
90
Trace
–
100
–
100
–
5-O-glucuronide
7-O-glucuronide
3-O-glucuronide
40-O-glucuronide
30-O-glucuronide
Trace
–
5
25
–
–
5
–
–
–
–
Trace
–
30
–
5
–
–
–
–
–
–
–
–
–
–
–
10
60
5
–
–
Trace
Trace
–
Trace
Trace
–
–
–
–
15
–
–
–
–
10
–
–
All values are percentages of total product distribution. Values in parenthesis indicate products with unconfirmed structures, meaning that two alternatives are possible. A
dash is used to indicate the absence of a product. All values rounded to the nearest 5%. The average standard deviation is ꢂ5%.
could not be confidently distinguished between the 3-O-glucuro-
nide and 40-O-glucuronide structures. The UGT1A7 enzyme resulted
in formation of two products upon reaction with isorhamnetin. The
first product exhibited the dominant loss of (GlucA + Aux), and the
second product displayed all the characteristic losses of a 7-O-
glucuronide species (-GlucA, -Agl, -Aux). The first product corre-
sponded to the 5-O-glucuronide due to its elution prior to the 7-O
product. The UGT1A9 enzyme promoted formation of one product
that dissociated by loss of (GlcA + Aux) upon CID, indicating that this
could be a 3-O or 5-O glucuronide. With only a single product
eluting, the only way this product can be identified is based on
comparison of retention times to products from those of the 1A7
reaction. The elution time of the UGT1A9 product matched that of
the 5-O-glucuronide identified from the 1A7 reaction mixture, thus
confirming the UGT1A9 product as 5-O-glucuronide (see CID mass
spectra in Supplemental Fig. 2).
The reaction of isorhamnetin with UGT1A10 yielded three
products. The first two eluting products were identified as the 5-O
glucuronide and 7-O glucuronide, respectively; this is because the
second product showed losses of GlucA, Agl, and Aux upon CID,
allowing it to be identified as the 7-O glucuronide. The first eluting
product showed a dominant loss of (GlucA + Aux), which because it
eluted before the 7-O glucuronide can be readily assigned as the 5-
O glucuronide. The third product showed a loss of (GlcA + Aux) as
the major product, and also exhibited a loss of Aux although this
latter fragment occurred in very low abundance similar to that of
the second product of the UGT1A3 reaction. This product could not
be confidently distinguished between the 3-O glucuronide and 40-
O glucuronide structures. NMR spectroscopy would arguably allow
more confident identification and differentiation of these products,
albeit requiring far greater quantities of products as well as
isolation of individual products or use of LC-NMR.
yielded only two products, and UGT2B4 which led to only one
glucuronide. As described above, the identities of the products
were assigned based on their MS/MS patterns, as illustrated in
Fig. 4, and their relative retention times. Upon analysis of the
UGT1A1 glucuronides, the first eluting product dissociated by
losses of GlcA and (GlucA + Aux), the second product demonstrated
losses of GlcA, Agl, and Aux, some occurring together, and the third
product showed prominent losses of Aux and (GlcA + Aux). The
first product was readily identified as kaempferol-5-O-glucuronide
because of the characteristic sole major loss of (GlucA + Aux), as
well as the fact that this product eluted prior to the second product,
the latter which is confidently assigned as the 7-O-glucuronide
based on its characteristic fragmentation pattern. The third
product could conceivably be a 3-O or 40-O glucuronide, both
which should elute after the 7-O-glucuronide. However, the fact
that the loss of Aux was observed along with the combined loss of
(GlcA + Aux) ruled out the 3-O-glucuronide. For the reactions with
the UGT1A7 enzyme, the two products formed included the 5-O-
glucuronide and 7-O-glucuronide but not the 40-O-glucuronide.
The one product formed for the UGT2B4 reactions was identified as
the 7-O- glucuronide product, while neither the 5-O-glucuronide
nor the 40-O-glucuronide were produced.
Quercetin has five possible glucuronidation sites, and it
generated three different products upon reaction with the
UGT1A1, 1A3, 1A8, and 1A10 glucuronosyltransferases. Upon
CID, the first eluting product showed fragment pathways
characteristic of 7-O-glucuronide, including losses of GlcA, Agl,
and Aux. The next two products both showed losses of Aux and
Aux + GlcA. The first of these two products is assigned as the 40-O-
glucuronide based on its elution prior to the other 30-O-
glucuronide. Reaction with the UGT1A9 enzyme resulted in only
one product, the 7-O-glucuronide. The UGT1A6 enzyme predomi-
nantly led to formation of the 7-O-glucuronide as well as the 5-O
product which was not observed for any of the enzymatic reactions
of quercetin. This 5-O product was confirmed because it showed
Kaempferol formed three glucuronides (5-O, 7-O, and 40-O)
with the exception of reactions promoted by UGT1A4, UGT2B15,
and UGT2B17 which resulted in no products, UGT1A7 which

S.A. Robotham, J.S. Brodbelt / Biochemical Pharmacology 82 (2011) 1764–1770
1769
only one characteristic loss (-(GlcA + Aux)) and eluted prior to the
7-O-glucuronide. The UGT2B7 enzyme led to formation of two
products, the first of which showed the loss of (GlcA + Aux) as the
only major fragmentation pathway while the second product
showed losses of Aux and (GlcA + Aux). The fragmentation pattern
tin, kaempferol, and quercetin with UGT1A3 and UGT1A9 [35], but
specific product identities were not assigned. In this past study
three products were found for isorhamnetin, two for kaempferol,
and four for quercetin [35]. For each of isorhamnetin and quercetin,
this represents one additional product then reported in the present
study but for kaempferol it is one less for the UGT1A3 reactions.
Due to the lack of product identities [35], further comparisons to
the present results were not possible.
of this second product was consistent with either
a
30-O
glucuronide or a 40-O glucuronide. The similar MS/MS patterns
and elution times of 30-O and 40-O glucuronides did not allow
conclusive assignment of this particular quercetin product.
However, upon LC–MS separation and analysis of a mixture
containing the products from the UGT1A3 reaction and the
products from this UGT2B7 reaction, the formation of the 40-O
glucuronide product was ruled out, and the product was identified
as a 30-O modification. The first product was identified as a 5-O
glucuronide because of its similar retention time to the 5-O
product confirmed from the UGT1A6 reactions (see CID mass
spectra in Supplemental Fig. 3).
Several other notable features emerged upon inspection of the
glucuronidation results for the other enzymes. UGT1A6 proved to
be unusual in that it is the only glucuronidase that is able to
promote 5-O-glucuronidation of the two flavanones, hesperetin
and naringenin, and also the flavonol quercetin. UGT1A7 and
UGT1A9 exhibited a preference for modifying the 30 position of the
hesperetin, but favored the 7-O position of naringenin and the
flavonols. In particular, the 30-O glucuronides were most abundant
for hesperetin, whereas kaempferol and quercetin showed no or
low abundance B-ring products. Instead the most abundant
products for kaempferol were the 7-O and 5-O (A-ring) glucur-
onides. The UGT1A8 enzyme displayed a high preference for
modification of flavonoids on the B-ring. For all the flavonoids that
yielded multiple products, the most abundant glucuronide
involved the B-ring. UGT1A10 preferentially glucuronidated the
7-O position, except for isorhamnetin which showed the 3-O or 40-
O position as the more highly favored sites for glucuronidation.
The UGT2B enzymes (UGT2B4, UGT2B7, UGT2B15, and
UGT2B17) all shared one similar characteristic: they all showed
poor to no reactivity with the flavonols, suggesting that the general
flavanol structure restricted interaction with the UGT2B enzymes
or otherwise deactivated the glucuronidation reactions.
There are also notable differences in the glucuronidation
selectivities based on subtle structural differences of the flavo-
noids. For example, naringenin and kaempferol are analogs, with
the former being a flavanone and the latter being a flavanol with its
characteristic double bond on the C ring and hydroxyl at the 3-O
position. The flavanol structure enhances the ability of the UGT
enzymes to modify the 40-O position. In no cases does naringenin
undergo glucuronidation at the 40-O position, whereas 40-O
glucuronide products are produced in nearly every reaction for
kaempferol.
Kaempferol and quercetin also share similar structures except
for the additional hydroxyl group at the 30 position of quercetin.
The 30-OH group proved to be a very reactive site for glucuronida-
tion for quercetin (as well as the flavanone hesperetin). Another
point of interest when comparing kaempferol and quercetin is that
the presence of the hydroxyl at the 30 position seems to render the
5-O position almost completely inactive, except for in the cases of
the UGT1A6 and the UGT2B7 enzyme, which both lead to 5-O
glucuronide as a minor product.
Quercetin also is similar in structure to isorhamnetin with the
only difference being that the 30-O position is methylated for
isorhamnetin. The methoxy at the 30 position appears to deactivate
glucuronidation in general by the UGT enzymes since products are
only detected for five of the UGT enzymes (UGT1A1, UGT1A3,
UGT1A7, UGT1A9, and UGT1A10). Interestingly, isorhamnetin
exhibited greater reactivity with UGT1A7 than was observed for
quercetin (no products).
4. 4.Discussion
4.1. Selectivity trends
EachUGTenzymeexhibitedselectivitywithrespecttothesitesof
glucuronidation of flavonoids. To establish
a benchmark for
evaluating the array of UGT glucuronosyltransferases, the glucur-
onidation trends for 1A1 in the present study were first compared to
a previouslimitedset ofresultsobtained using thesameenzyme and
LC–MS/MS analysis [34]. It was previously reported that UGT1A1
selectivelymodifies only the hydroxylgroup at the 7 position, unless
there is a hydroxyl at the 30 position in which case UGT1A1 also
modifies hydroxyl groups on the B ring [34]. The relative quantities
of the hesperetin and quercetin glucuronidation products in this
current study were in good agreement with those found in the
previous study [34]. The sole glucuronidation of naringenin at the 7-
O position noted in the past study also matches the same finding in
the present study. The glucuronidation of kaempferol, however,
shows a discrepancy based on our present results relative to past
results [34] and another recent study [40]. In one study, only the 7-
O-glucuronide product was reported [34], unlike the three distinc-
tive products (5-O, 7-O, and 40-O) found in the present study. In a
more recent study by Singh et al., the interaction of kaempferol with
the UGT1A1, 1A3, 1A6, 1A7, 1A8, 1A9, 1A10, and 2B7 enzymes was
evaluatedbased ona UVshiftmethodforglucuronidedifferentiation
[40]. Singh et al. reported modification of kaempferol at the 3-O
position for several oftheUGTenzymes, aswellasa low likelihood of
5-O product formation, in contrast to our findings in which
glucuronidation was found to occur at the 5-O position but not at
the 3-O position based on the MS/MS patterns of the metal
complexes. This difference may require the synthetic production of
individual flavonoid glucuronides and NMR spectral analysis for
confident resolution. Isorhamnetin is the only flavonoid whose
glucuronidation in the presence of UGT1A1 has not been reported
previously, and it in fact yields just one product.
The results for the UTG1A3 enzyme show that its selectivity
differs slightly from that of UGT1A1, with greater preference for
glucuronidation of the 7-O position of the flavonols. The best
example of this is reflected by the quercetin and isorhamnetin
results, in which the production of 7-O glucuronides is enhanced
for the UGT1A3 reactions. This is particularly evident in the
reactions of isorhamnetin with UGT1A3 for which the 7-O-
glucuronide product was by far the most abundant, a product not
even detected upon reaction in the presence of UGT1A1. The
selectivity of the UGT1A3 enzyme for hesperetin is likewise very
different from that of the UGT1A1 reaction, in that the 30-O
glucuronide is the most abundant product, not the 7-O product. A
previous recent study reported the glucuronidation of isorhamne-
5. Conclusion
The regioselectivity of the reactions of twelve human UDP-
glucuronosyl-transferase (UGT) isozymes with five common flavo-
noids was evaluated by LC–MS/MS with post-column metal
complexation. Metal complexation results in the formation of
[Co(II) (FG-H) (4,7-dpphen)2]+ ions which are key for confident
identification of the modification site promoted by a given UGT

1770
S.A. Robotham, J.S. Brodbelt / Biochemical Pharmacology 82 (2011) 1764–1770
[17] Luo Y, Wu S, Li X, Li P. LC-ESI–MS–MS determination of rat plasma protein
binding of major flavonoids of flos lonicerae japonicae by centrifugal ultrafil-
tration. Chromatographia 2010;72:71–7.
isozyme due to the more diagnostic fragmentation patterns than
produced by conventional deprotonated flavonoid glucuronides.
The UGT1A enzyme selectivities are affected by the presence of a
hydroxyl group at the 3, 40, or 30 positions as well as by the presence
of a methoxy at the 30 position. The UGT2B enzymes all share one
similar trend: they all show poor to no reactivity with the flavonols.
This result implies that the greater planarity of the flavonols
compared to the non-planar structures of flavanones or the
additional hydroxyl group at the 3 position of the flavonols inhibits
interaction with the UGT2 enzymes. This study also shows the
effectiveness of metal complexation/tandem mass spectrometry in
conjunctionwithHPLCretentiontimesforidentificationofflavonoid
monoglucuronides. Of the 60 reactions reported in this study only
three resulted in a product that could not be differentiated.
[18] Stobiecki M, Staszko´w A, Piaskecka A, Garcia-Lopez PM, Zamora-Natera F,
Kachlicki P. LC–MS/MS profiling of flavonoid conjugates in wild Mexican
Lupine, Lupinus reflexus. Journal of Natural Products 2010;73:1254–60.
[19] Dehkharghanian M, Adenier H, Vijayalakshmi MA. Study of flavonoids in
aqueous spinach extract using positive electrospray ionisation tandem quad-
rupole mass spectrometry. Food Chemistry 2010;121:863–70.
[20] Rak G, Fodor P, Abrank L. Three-step HPLC-ESI–MS/MS procedure for screening
and identifying non-target flavonoid derivatives. International Journal of Mass
Spectrometry 2010;290:32–8.
[21] Satterfield M, Brodbelt JS. Enhanced detection of flavonoids by metal com-
plexation and electrospray ionization mass spectrometry. Analytical Chemis-
try 2000;72:5898–906.
[22] Zhang J, Wang J, Brodbelt JS. Characterization of flavonoids by aluminum
complexation and collisionally activated dissociation. Journal of Mass Spec-
trometry 2005;40:350–63.
[23] Pikulski M, Brodbelt JS. Differentiation of flavonoid glycoside isomers by using
metal complexation and electrospray ionization mass spectrometry. Journal of
the American Society for Mass Spectrometry 2003;14:1437–53.
[24] Zhang J, Brodbelt JS. Gas-phase hydrogen/deuterium exchange and conforma-
tions of deprotonated flavonoids and gas-phase acidities of flavonoids. Journal
of the American Chemical Society 2004;126:5906–19.
[25] Davis BD, Brodbelt JS. LC–MSn methods for saccharide characterization of
monoglycosyl flavonoids using postcolumn manganese complexation. Ana-
lytical Chemistry 2005;77:1883–90.
[26] Zhang J, Brodbelt JS. Silver complexation and tandem mass spectrometry for
differentiation of isomeric flavonoid diglycosides. Analytical Chemistry
2005;77:1761–70.
[27] Satterfield M, Brodbelt JS. Structural characterization of flavonoid glycosides
by collisionally activated dissociation of metal complexes. Journal of the
American Society for Mass Spectrometry 2001;12:537–49.
Acknowledgement
Funding from the NIH (R03 CA133924-02) and the Welch
Foundation (1155) is gratefully acknowledged.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bcp.2011.08.015.
[28] Zhang J, Brodbelt JS, Wang J. Threshold dissociation and molecular modeling of
transition metal complexes of flavonoids. Journal of the American Society for
Mass Spectrometry 2005;16:139–51.
References
[29] Pikulski M, Aguilar A, Brodbelt JS. Tunable transition metal–ligand complexa-
tion for enhanced elucidation of flavonoid diglycosides by electrospray ioni-
zation mass spectrometry. Journal of the American Society for Mass
Spectrometry 2007;18:422–31.
[30] Davis BD, Needs PW, Kroon PA, Brodbelt JS. Identification of isomeric flavonoid
glucuronides in urine and plasma by metal complexation and LC-ESI–MS/MS.
Journal of Mass Spectrometry 2006;41:911–20.
[31] Zhang J, Satterfield MB, Brodbelt JS, Britz SJ, Clevidence B, Novotny JA.
Structural characterization and detection of kale flavonoids by electrospray
ionization mass spectrometry. Analytical Chemistry 2003;75:6401–7.
[32] Zhang J, Brodbelt JS. Screening flavonoid metabolites of naringin and narirutin
in urine after human consumption of grapefruit juice by LC–MS and LC–MS/
MS—analyst (RSC Publishing). Analyst 2004;1227–33.
[33] Brett GM, Hollands W, Needs PW, Teucher B, Dainty JR, Davis BD, et al.
Absorption, metabolism and excretion of flavanones from single portions of
orange fruit and juice and effects of anthropometric variables and contracep-
tive pill use on flavanone excretion. British Journal of Nutrition 2009;
664–75.
[1] Ross JA, Kasum CM. Dietary flavonoids: bioavailability, metabolic effects, and
safety. Annual Review of Nutrition 2002;22:19–34.
[2] Kaur C, Kapoor HC. Antioxidants in fruits and vegetables—the millennium’s
health. International Journal of Food Science & Technology 2001;36:703–25.
[3] Kupeli E, Sahin FP, Yesilada E, Calis I, Ezer N. In vivo anti-inflammatory and
antinociceptive activity evaluation of phenolic compounds from Sideritis
stricta. Journal of Biosciences A 2007;62:519–25.
[4] Terao J. Dietary flavonoids as antioxidants. Forum of Nutrition 2009;61:87–94.
[5] Kroon PA, Clifford MN, Crozier A, Day AJ, Donovan JL, Manach C, et al. How
should we assess the effects of exposure to dietary polyphenols in vitro?
American Journal of Clinical Nutrition 2004;80:15–21.
[6] Manach C, Williamson G, Morand C, Scalbert A, Remesy C. Bioavailability and
bioefficacy of polyphenols in humans. I. Review of 97 bioavailability studies.
American Journal of Clinical Nutrition 2005;81:230S–42S.
[7] Rice-Evans C. Flavonoid antioxidants. Current Medicinal Chemistry
2001;8:797.
[8] Atmani D, Chaher N, Atmani D, Berboucha M, Debbache N, Boudaoud H.
Flavonoids in human health: from structure to biological activity. Current
Nutrition and Food Science 2009;5:225–37.
[9] Williamson G, Barron D, Shimoi K, Terao J. In vitro biological properties
of flavonoid conjugates found in vivo. Free Radical Research 2005;39:457–
69.
[10] Williamson G, Day AJ, Plumb GW, Couteau D. Human metabolic pathways of
dietary flavonoids and cinnamates. Biochemical Society Transactions
2000;28:16–22.
[11] King C, Rios G, Green M, Tephly T. UDP-glucuronosyltransferases. Current Drug
Metabolism 2000;1:143–61.
[12] Wong YC, Zhang L, Lin G, Zuo Z. Structure–activity relationships of the
glucuronidation of flavonoids by human glucuronosyltransferases. Expert
Opinion on Drug Metabolism & Toxicology 2009;5m:1399–419.
[13] Stobiecki M. Application of mass spectrometry for identification and structural
studies of flavonoid glycosides. Phytochemistry 2000;54:237–56.
[14] Gutzeit D, Winterhalter P, Jerz G. Application of preparative high-speed
counter-current chromatography/electrospray ionization mass spectrometry
for a fast screening and fractionation of polyphenols. Journal of Chromatogra-
phy A 2007;1172:40–6.
[15] Zettersten C, Co M, Wende S, Turner C, Nyholm L, Sjo¨berg PJR. Identification
and characterization of polyphenolic antioxidants using on-line liquid chro-
matography, electrochemistry, and electrospray ionization tandem mass
spectrometry. Analytical Chemistry 2009;81:8968–77.
[34] Davis BD, Brodbelt JS. Regioselectivity of human UDP-glucuronosyl-transfer-
ase 1A1 in the synthesis of flavonoid glucuronides determined by metal
complexation and tandem mass spectrometry. Journal of the American Society
for Mass Spectrometry 2008;19:246–56.
[35] Chen Y, Xie S, Chen S, Zeng S. Glucuronidation of flavonoids by recombinant
UGT1A3 and UGT1A9. Biochemical Pharmacology 2008;76:416–25.
[36] Xie S, You L, Zeng S. Studies on the flavonoid substrates of human UDP-
glucuronosyl transferase (UGT) 2B7. Pharmazie 2007;62:625–9.
[37] Chen YK, Chen SQ, Li X, Zeng S. Quantitative regioselectivity of glucuronidation
of quercetin by recombinant UDP-glucuronosyltransferases 1A9 and 1A3
using enzymatic kinetic parameters. Xenobiotica 2005;35:943–54.
[38] Xie S, Chen Y, Chen S, Zeng S. Structure-metabolism relationships for the
glucuronidation of flavonoids by UGT1A3 and UGT1A9. Journal of Pharmacy
and Pharmacology 2011;63:297–304.
[39] Xie S-G, You L-Y, Zeng S. Phase II metabolism of flavonoids mediated by human
glucuronosyltransferase: an advanced research. Zhongguo Yaolixue Yu
Dulixue Zazhi 2007;21:438–43.
[40] Singh R, Wu B, Tang L, Hu M. Uridine diphosphate glucuronosyltransferase
isoform-dependent regiospecificity of glucuronidation of flavonoids. Journal
of Agricultural and Food Chemistry 2011;59:7452–64.
[41] Plumb GW, O’Leary K, Day AJ, Williamson G. Enzymatic synthesis of quercetin
glucosides and glucuronides. In: Santos-Buelga C, Williamson G, editors.
Methods in polyphenol analysis. Cambridge: The Royal Society of Chemistry;
2003. p. 177–86.
[42] Easterbrook J, Lu C, Sakai Y, Li AP. Effects of organic solvents on the activities of
cytochrome P450 isoforms, UDP-dependent glucuronyl transferase, and phe-
nol sulfotransferase in human hepatocytes. Drug Metabolism and Disposition
2001;29:141–4.
[16] Schmidt S, Zietz M, Schreiner M, Rohn S, Kroth LW, Krumbein A. Identification
of complex, naturally occurring flavonoid glycosides in kale (Brassica oleracea
var. sabellica) by high-performance liquid chromatography diode-array de-
tection/electrospray ionization multi-stage mass spectrometry. Rapid Com-
munications in Mass Spectrometry 2010;24:2009–22.