Phenylalanine 93 of the human UGT1A10 plays a major role in the interactions of the enzyme with estrogens

Article abstract of DOI:10.1016/j.steroids.2011.07.017

Little is currently known about the substrate binding site of the human UDP-glucuronosyltransferases (UGTs) and the structural elements that affect their complex substrate selectivity. In order to further understand and extend our earlier findings with phenylalanines 90 and 93 of UGT1A10, we have replaced each of them with Gly, Ala, Val, Leu, Ile or Tyr, and tested the activity of the resulting 12 mutants toward eight different substrates. Apart from scopoletin glucuronidation, the F90 mutants other than F90L were nearly inactive, while the F93 mutants' activity was strongly substrate dependent. Hence, F93L displayed high entacapone and 1-naphthol glucuronidation rates, whereas F93G, which was nearly inactive in entacapone glucuronidation, was highly active toward estradiol, estriol and even ethinylestradiol, a synthetic estrogen that is a poor substrate for the wild-type UGT1A10. Kinetic analyses of 4-nitrophenol, estradiol and ethinylestradiol glucuronidation by the mutants that catalyzed the respective reactions at considerable rates, revealed increased K_m values for 4-nitrophenol and estradiol in all the mutants, whilst the K _m values of F93G and F93A for ethinylestradiol were lower than in control UGT1A10. Based on the activity results and a new molecular model of UGT1A10, it is suggested that both F90 and F93 are located in a surface helix at the far end of the substrate binding site. Nevertheless, only F93 directly affects the selectivity of UGT1A10 toward large and rigid estrogens, particularly those with substitutions at the D ring. The effects of F93 mutations on the glucuronidation of smaller or less rigid substrates are indirect, however.

Full text of DOI:10.1016/j.steroids.2011.07.017

Steroids 76 (2011) 1465–1473
Contents lists available at ScienceDirect
Steroids
journal homepage: www.elsevier.com/locate/steroids
Phenylalanine 93 of the human UGT1A10 plays a major role in the interactions
of the enzyme with estrogens
Camilla Höglund a,b,1, Nina Sneitz a,b,1, Anna Radominska-Pandya c, Liisa Laakonen a, Moshe Finel a,
⇑
a Centre for Drug Research, Faculty of Pharmacy, P.O. Box 56 (Viikinkaari 5), University of Helsinki, FI-00014 Helsinki, Finland
b Division of Pharmaceutical Chemistry, Faculty of Pharmacy, P.O. Box 56 (Viikinkaari 5), University of Helsinki, FI-00014 Helsinki, Finland
c Department of Biochemistry and Molecular Biology, University of Arkansas for Medical Sciences, Little Rock, AR, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 24 June 2011
Received in revised form 28 July 2011
Accepted 29 July 2011
Available online 9 August 2011
Little is currently known about the substrate binding site of the human UDP-glucuronosyltransferases
(UGTs) and the structural elements that affect their complex substrate selectivity. In order to further
understand and extend our earlier findings with phenylalanines 90 and 93 of UGT1A10, we have replaced
each of them with Gly, Ala, Val, Leu, Ile or Tyr, and tested the activity of the resulting 12 mutants toward
eight different substrates. Apart from scopoletin glucuronidation, the F90 mutants other than F90L were
nearly inactive, while the F93 mutants’ activity was strongly substrate dependent. Hence, F93L displayed
high entacapone and 1-naphthol glucuronidation rates, whereas F93G, which was nearly inactive in ent-
acapone glucuronidation, was highly active toward estradiol, estriol and even ethinylestradiol, a syn-
thetic estrogen that is a poor substrate for the wild-type UGT1A10. Kinetic analyses of 4-nitrophenol,
estradiol and ethinylestradiol glucuronidation by the mutants that catalyzed the respective reactions
at considerable rates, revealed increased Km values for 4-nitrophenol and estradiol in all the mutants,
whilst the Km values of F93G and F93A for ethinylestradiol were lower than in control UGT1A10. Based
on the activity results and a new molecular model of UGT1A10, it is suggested that both F90 and F93 are
located in a surface helix at the far end of the substrate binding site. Nevertheless, only F93 directly
affects the selectivity of UGT1A10 toward large and rigid estrogens, particularly those with substitutions
at the D ring. The effects of F93 mutations on the glucuronidation of smaller or less rigid substrates are
indirect, however.
Keywords:
Estradiol
Estriol
Ethinylestradiol
Glucuronidation
Ó 2011 Elsevier Inc. All rights reserved.
1. Introduction
properties. This promiscuity leads to partial overlaps in the
substrate selectivity of individual UGTs, although closer inspection
The UDP-glucuronosyltransferases (UGTs) are membrane bound
enzymes of the endoplasmic reticulum that catalyze the conjugation
of many compounds with glucuronic acid from UDP-glucuronic acid
(UDPGA) [27]. This glucuronidation reaction increases the
water-solubility of mostly lipophilic substrates and stimulates their
excretion from the body through bile or urine. The 19 functional
human UGTs of subfamilies 1A, 2A and 2B [16] are expressed in a
tissue-specific manner. Many of them are expressed in the liver,
but some are only or mainly expressed in extrahepatic tissues.
UGT1A10, the subject of this study, belongs to the latter group and
it is mainly expressed in the intestine [18,5–7].
of their kinetic properties, as well as regio- and stereo-selectivity,
reveals that there are clear differences even among highly homolo-
gous individual UGTs such as 1A7–1A10 [4,6,7]. The crystal struc-
ture of C-domain of UGT2B7, the domain that contains the
UDPGA (co-substrate) binding site, has been resolved [17], but it
does not include the binding site for the aglycone substrate. Based
on the homology between UGTs and other glycosyltransferases,
UGTs are predicted to be comprised of two large pseudosymmetric
a–b sandwich domains, the N and C domains, and two additional
small domains, the envelope helices connecting the globular
domains, and a transmembrane helix with a short cytoplasmic tail
[13]. In mature UGT1A10, the N-terminal domain, the C-terminal
domain, the envelope helices and the transmembrane domain are
predicted to correspond to residues 26–277, 278–430, 431–468
and 469–530, respectively. In analogy to resolved crystal structures
of homologous glycosyltransferases from plant and bacteria, UDP-
GA very likely binds to the cleft between the N and the C domains,
and is stabilized by several conserved, specific interactions to the C
domain. In UGT1A10, UDPGA is probably bound by C-domain
Most human UGTs can catalyze the glucuronidation of several
compounds that vary greatly in structure, size and physicochemical
Abbreviations: 4-MU, 4-methylumbelliferone; pNP, 4-nitrophenol; EED, ethiny-
lestradiol; UDPGA, UDP-glucuronic acid; UGT, UDP-glucuronosyltransferase.
⇑
Corresponding author. Fax: +358 9 1915 9556.
E-mail address: moshe.finel@helsinki.fi (M. Finel).
These authors contributed equally to this study.
1
0039-128X/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.steroids.2011.07.017

1466
C. Höglund et al. / Steroids 76 (2011) 1465–1473
Fig. 1. Molecular structures of the eight substrates used in the screening assays. The compounds are oriented so that the reactive hydroxyls point down.
residues between W351 to Q394, and at the same time situated
close to His37 of the N-terminal domain, the ‘‘catalytic His’’, that
directly contributes to O-glucuronidation reactions by stabilizing
the deprotonated form of the hydroxyl group on the substrate
[17,21,22]. Contrary to the binding site of UDPGA, the exact location
and architecture of the acceptor–substrate binding site is currently
unknown. The available data suggests that two variable segments
in the N-domain, L83–H123 and I172–F224 (UGT1A10 numbering),
contribute to the substrate binding site. We hypothesize that the
actual interaction residues will vary from case to case, even if
provided by these segments. The complex substrate selectivity of
the UGTs – promiscuous and specific at the same time – suggests
that part of the binding site is rigidly refined and part is flexible.
The latter process may be affected by the substrate itself in an
induced fit manner.
UGT1A10 is highly active in the glucuronidation of estrogens
and several other compounds [28,25,3,30,5–7]. A photoaffinity
labeling study, using 4-azido-2-hydroxybenzoic acid as a probe,
suggested that F90 and F93 of UGT1A10 are important for interac-
tion with phenolic substrates [28]. We have now undertaken to ex-
plore this site in more detail, using site directed mutagenesis,
activity analyses of steroids, coumarins and aromatic alcohols,
and homology modeling. The results shed new light on glucuroni-
dation of several substrates by UGT1A10, as well as raise many
new questions.
other samples were: UGT1A10, 2.0; F90A, 7.3; F90G, 6.4; F90V,
11.5; F90I, 7.8; F90Y, 1.4; F93A, 7.6; F93L, 4.3; F93G, 5.6; F93V,
10.2; F93I, 10.8; F93Y, 2.6. Protein concentrations were determined
by the BCA method (Pierce Biotechnology, Rockford, IL, USA).
2.2. Glucuronidation activity analyses
Glucuronidation rates at a single protein and substrate concen-
trations (‘‘screening assays’’) were measured in duplicates (at most
15% difference between duplicates), in a final volume of 100 ll. In
addition, the kinetics of pNP, estradiol and EED were determined,
and in these assays triplicates were used. The protein concentra-
tions for the screenings and the kinetics were 20–200 lg/ml,
depending on the tested substrate and enzyme variant. The
screening reaction mixtures contained 50 mM phosphate buffer,
pH 7.4, 5 mM MgCl2, 5 mM saccharolactone, 5 mM UDPGA and
the following concentrations of the tested substrates: 500
the cases of scopoletin, pNP, entacapone and estriol, 200
l
M in
l
M of
4-MU, or 100 lM for the screening assays with estradiol, EED and
1-naphthol. Estradiol, EED and estriol were dissolved in DMSO
and the final concentration of the solvent in assays with these
substrates was either 5% (estradiol and estriol), or 10% (EED). These
concentrations of DMSO do not inhibit the activity of recombinant
UGTs [29].
Saccharolactone was included in the first screening analyses,
but once it became clear that this inhibitor of beta glucuronidase
activity does not improve the activity results [19], it was left out
of subsequent assays. Hence, the estradiol and EED screenings
and all the kinetic analyses were carried out in the absence of sac-
charolactone. The reactions were started by the addition of UDPGA,
followed by mixing and transfer to 37 °C. The reactions were car-
ried out for 60 min in the cases of pNP and EED, 30 min for estra-
diol, estriol, 4MU, 1-naphthol and entacapone, or 20 min in the
case of scopoletin. The incubations were terminated by adding
2. Materials and methods
2.1. Materials, recombinant UGT1A10, and mutants
Scopoletin, 1-naphthol, 17b-estradiol, 4-nitrophenol (pNP),
16a,17b-estriol, ethinylestradiol (EED), 4-methylumbelliferone,
(4-MU), UDPGA (triammonium salt) and
D-saccharic acid 1,4-lac-
tone were purchased from Sigma–Aldrich (St. Louis, MO, USA). So-
dium dihydrogen phosphate dihydrate was from Fluka (Buchs,
Switzerland) and perchloric acid was from Merck (Darmstadt, Ger-
many). Entacapone was a generous gift from Orion Pharma (Espoo,
Finland).
The mutations in F90 and F93 of UGT1A10 were generated by
the QuikChange system (Stratagene, La Jolla, CA, USA) and the se-
quence of the entire cloned UGT1A10 was verified by DNA
sequencing. Recombinant human UGT1A10 and the mutants were
expressed as His-tagged proteins in baculovirus-infected Sf9 insect
cells as previously described [9,12]. The relative expression level of
the UGT in each membrane batch was determined using the mono-
clonal anti-His antibodies tetra-His (Qiagen, Hilden, Germany) as
detailed elsewhere [11]. The expression level of the variant with
the lowest expression per mg protein, F90L, was taken as the refer-
ence whose expression level is 1.0. The corresponding levels of the
10 ll perchloric acid, cooling on ice for 10 min, and subsequent
centrifugation at 16,000g for 10 min. The supernatants from the
latter centrifugation were subjected to HPLC analysis.
The HPLC analyses were carried out on either an Agilent (1100
series, Agilent Technologies, Palo Alto, CA, USA), or a Shimadzu
HPLC (LC-10 model, Shimadzu Corporation, Kyoto, Japan). These
two HPLCs were equipped with both UV absorbance and fluores-
cence detectors. The estradiol-3-glucuronide was separated from
the estradiol 17-glucuronide using Chromolith SpeedRod column
(Merck, Darmstadt, Germany). The column temperature was
30 °C. The mobile phase consisted of 60% 25 mM phosphate buffer,
pH 3, and 40% methanol. The flow rates were as follow: 0–9.5 min,
1 ml/min; 9.5–19 min, 2 ml/min; 19–20 min, 1 ml/min. The glucu-
ronides were detected by fluorescence (excitation 216 nm, emis-
sion 316 nm) and the retention time for estradiol-3-glucuronide

C. Höglund et al. / Steroids 76 (2011) 1465–1473
1467
was 5.0 min. Estriol and its 3-glucuronide were separated using
Hypersil BDS-C18, 5
m, 4.6 ꢀ 150 mm (Agilent Technologies, Palo
hydroxybenzoic acid upon photolabeling [28]. Replacing these
two phenylalanine residues with either Ala or Leu was then shown
to reduce glucuronidation activity toward 4-nitrophenol (pNP) and
4-methylumbelliferone (4-MU) [28], and to variably affect the glu-
curonidation of native and hydroxylated estrogens [25]. This sug-
gested that F90 and F93 may form part of the substrate binding
site of UGT1A10 and in order to gain further insight into the active
site properties, we have now examined the effect of other amino
acids at the same positions. Both these Phe residues were sepa-
rately replaced with 6 different residues, either the aromatic Tyr,
or amino acids with increasing side chain size and hydrophobicity,
namely Gly, Ala, Leu, Val and Ile. The 12 resulting single mutants
were expressed in baculovirus-infected insect cells and their activ-
ities, corrected for relative expression level, were compared to wild
type UGT1A10. The aglycone substrates in this study included the
simple aromatics pNP and 1-naphthol, the coumarins scopoletin
and 4-MU, the estrogens estradiol, estriol and EED, and a less com-
monly tested drug, entacapone, an inhibitor of catechol O-methyl-
transferase (COMT) and a very good substrate of UGT1A9 [14]
(Fig. 1).
l
Alto, CA) and recorded with fluorescence detection (excitation
280 nm, emission 305 nm). The column temperature was 40 °C.
The mobile phase consisted of 65% 25 mM phosphate buffer, pH
3 and 35% methanol and the flow rate was 1 ml/min. The retention
time for estriol-3-glucuronide was 3.8 min. The HPLC analyses for
scopoletin, 1-naphthol, 4-nitrophenol, entacapone and 4-methyl-
umbelliferone were performed as previously described [10,15].
The EED glucuronide was detected as described [15], but using
the Shimadzu HPLC.
Glucuronides of scopoletin and EED were quantified using [14C]-
UDPGA [8], whereas pNP, 1-naphthol, 4-MU, entacapone, estradiol-
3-glucuronide and estriol-3-glucuronide were quantified using
authentic standards [15]. The detection and quantification limits,
respectively, were 0.8 and 2.5
for estriol, 0.3 and 0.9 M for EED, 1.1 and 3.6
and 7.0 M for 4-MU, 0.02 and 0.06 M for scopoletin, 0.03 and
0.1 M for 1-naphthol and 0.6 and 1.9 M for entacapone.
lM for estradiol, 0.3 and 1.0
lM
l
lM for pNP, 2.1
l
l
l
l
2.3. Kinetic analyses
3.1. Activity screens
Enzyme kinetics experiments were performed in triplicates. The
assay conditions and reaction terminations were similar to the
screening assays, but the protein concentrations and incubation
times for these assays were selected according to the preliminary
experiments in order to ensure that product formation was within
the linear range with respect to both protein concentration and
time, and that the substrate consumptions was less than 10%.
Kinetic constants were estimated by fitting the experimental
data using GraphPad Prism for Windows (GraphPad Software,
San Diego, CA, USA). In this study the data was fitted to either
the Michaelis–Menten or the Michaelis–Menten with substrate
inhibition models. The goodness of the fit, as calculated by the
GraphPad Prism program, is indicated alongside the results in
Table 1.
The mutant enzymes were initially studied for glucuronidation
activity toward a panel of substrates. In these screens a single and
relatively high substrate concentration was used and protein con-
centration for the screening experiments of the different UGT1A10
variants was not optimized for each assay. Nevertheless, in no case
more than 10% of the substrate was consumed during the reaction.
The protein concentrations in these assays were mainly similar for
all mutants and the glucuronidation rates were, subsequently, cor-
rected (normalized) according to the relative expression level of
the individual enzyme (see Section 2). It may be added here that
the differences in relative expression levels in the baculovirus-in-
fected insect cells system that was used to produce the recombi-
nant UGTs should not be taken as evidence for the effect of the
mutation on the gene transcription, translation nor folding in the
human intestine, would such mutations occur.
All the mutant enzymes exhibited activity at least towards
scopoletin, suggesting that in each of them the protein is (largely)
folded and that no elements essential for catalysis, such as UDPGA
binding, were lost by the mutations. In addition, the results indi-
cate that some degree of activity from barely detectable to high
rates, was exhibited towards the small substrates by all enzyme
variants, with the exception of 4-MU by F90Y. On the other hand,
several or even the majority of mutants were unable to glucuron-
idate the larger substrates (Fig. 2).
2.4. Modeling
The amino acid sequence of the mature human UGT1A10
(P22309 from the SwissProt database) (http://www.uniprot.org)
was aligned to plant and bacterial glycosyltransferases of known
structure, as described earlier [13]. The sequence-to-structure
match was modified from the published alignment so that the
short helices observed in 2D0Q and 2IYA were used as templates
for the first predicted helix in loop 3. A homology model was con-
structed for the mature human UGT1A10 from the structural align-
ment with the program Modeller 9v6, similarly to the work with
UGT1A1 [13]. The model includes UDPGA.
3.2. pNP
The substrate structures were built and optimized in MO-
PAC2009 (http://openmopac.net) with PM3 methodology [26],
and docked into the active site guided by the interactions with
the reactive site histidine (His37) and the UDPGA. The same
nomenclature for the secondary structure elements was used as
before: the core strands of the globular domains are numbered
sequentially, and named by the domain, e.g. Nb1. The following he-
Glucuronidation screen with pNP as the aglycone substrate re-
vealed sharp differences between mutations at F90 to the corre-
sponding mutations in F93. All the
6 mutations we have
generated at the F90 position severely reduced the pNP glucuron-
idation rate, to 1–15% of wild type activity, whereas replacements
of F93 with the same set of residues did not affect activity that
much in most cases, yielding 10–85% of the wild type glucuronida-
tion rate (Table 1 and Fig. 2). Interestingly, the leucine mutants at
both positions, F90L and F93L, were least affected. Mutant F93G
also exhibited considerable activity, about 60% of control.
It may be noted here that the screening assays were done prior to
kinetic analyses with the selected mutants. As became clear to us la-
ter, and will become clear to the reader below, the calculated Vmax
values for some of the mutants were, eventually, higher than for
UGT1A10, even if these mutants exhibited lower glucuronidation
rates in the screening assays (Table 1). This was mainly due to a large
lix bears the same name, e.g. N
when the inter-strand segment contains more than one helix, e.g.
3-1. Amino acid numbers refer to the nascent full length
a1, followed by additional numbers
N
a
UGT1A10.
3. Results
We have previously identified F90 and F93 of the human
UGT1A10 as residues that form covalent bonds with 4-azido-2-

1468
C. Höglund et al. / Steroids 76 (2011) 1465–1473
Fig. 2. Glucuronidation rates of the eight substrates by UGT1A10 and the 12 mutants. These screening assays were carried out at a single substrate concentration, as follow:
scopoletin, pNP, entacapone and estriol, 500 M; 4-MU, 200 M; estradiol, EED and 1-naphthol, 100 M. The rates were corrected (normalized) according to the relative
l
l
l
expression levels of the UGT in each preparation (see Section 2). The units for all the Y axes are nmol/min/mg (see arrow on the left side) and the mutants are indicated on the
X-axes (see arrows at the bottom).
increase in the apparent Km value of these mutants, an important
and interesting effect of the respective mutations that was only
appreciated during the subsequent kinetic analyses (see below).
idation at about twice the rate of the control UGT1A10. Changing F93
to either G or Y did not affect the 1-naphthol glucuronidation rate,
while the activity of F93I was about half the control rate. The F–V
replacements in both positionsyielded about one third of the control
rate and the least active mutant was F90G (Fig. 2).
3.3. 1-Naphthol
The glucuronidation of 1-naphthol was least affected by the
mutations, particularly those in F93. Similarly to the results with
pNP, the leucine mutations at both positions were the most active
in comparison to the corresponding mutations, respectively: F90L
exhibited wild type rate, while F93L catalyzed 1-naphthol glucuron-
3.4. 4-MU
The glucuronidation of 4-MU turned out to be highly sensitive
to the mutations we have prepared. None of the 12 mutants exhib-
ited similar, let alone higher 4-MU glucuronidation rate than the

Table 1
Kinetic parameters for 4-nitrophenol (pNP), estradiol (at the 3-OH) and ethinylestradiol (EED) glucuronidation by UGT1A10 and the mutants. Empty boxes indicate that kinetic analyses were not done due to very low, if any, detectable
activity. The Vmax values were corrected for relative expression level (see Section 2).
UGT
4-Nitrophenol (pNP)
Estradiol
Km M)
Ethinylestradiol (EED)
Km M) Vmax (nmol/min/ Kcat (Vmax
a
Km
(lM)
Vmax (nmol/min/ Kcat (Vmax
/
Goodness of fit
(
l
Vmax (nmol/min/ Kcat (Vmax
mg) Km
/
Goodness of fit
(l
/
Goodness of fit (R2)
mg)
Km
)
(R2)
)
(R2)
mg)
Km)
1A10 386.9 73.5
F90A
2.46 0.14
6.36 ꢀ 10ꢁ6 0.93
3.11 ꢀ 10ꢁ6 0.99
11.06 1.37 6.10 0.17
5.5 ꢀ 10ꢁ4
0.97
122.5 18.0
0.27 0.02
2.20 ꢀ 10ꢁ6 0.98
F90L
F90G
F90V
F90I
12331 1916 38.3 4.5
12.1 1.7
0.78 0.03
6.45 ꢀ 10ꢁ5 0.96
F90Y
F93A 16210 4780 16.6 4.1
1.02 ꢀ 10ꢁ6 0.99
54.0 9.5
67.6 7.3
1.99 0.12
8.63 0.35
3.69 ꢀ 10ꢁ5 0.92
1.28 ꢀ 10ꢁ4 0.99
65.7 16.4 Ki
0.21 0.03
3.66 0.2
3.20 ꢀ 10ꢁ6 0.99 (substrate
(l
M)
inhibition)
454 231
F93L
7418 2344
10.2 2.3
5.23 0.6
1.38 ꢀ 10ꢁ6 0.97
1.01 ꢀ 10ꢁ6 0.99
F93G 5203 870
30.9 3.6 Ki
1.18 ꢀ 10ꢁ4 0.99 (substrate
(l
M)
inhibition)
401 86
F93V 4064 408
1.65 0.10
1.74 0.07
3.03 0.08
0.41 ꢀ 10ꢁ6 0.99
0.35 ꢀ 10ꢁ6 0.99
1.27 ꢀ 10ꢁ6 0.99
F93I
5021 308
F93Y 2383 121
89.0 4.9
3.49 0.08
3.92 ꢀ 10ꢁ5 0.99
a
The Km values for the mutants, as well as the Ki values in the cases of observed substrate inhibition kinetics, were determined by extrapolation since they were outside the (broad) tested concentration range.

1470
C. Höglund et al. / Steroids 76 (2011) 1465–1473
3.5. Scopoletin
3.8. Estriol
The activity of the tested enzymes toward scopoletin was radi-
cally different from that toward 4-MU, regardless the structural
similarity of the two compounds. The scopoletin glucuronidation
rates of the majority of the mutants were higher than control, up
to 6-fold the (normalized) rate of UGT1A10 (Fig. 2). The two leu-
cine mutants, F90L and F93L, were the most active ones; whereas
the corresponding F–Y mutants were the only ones exhibiting
scopoletin glucuronidation rates that were lower than control
UGT1A10 (Fig. 2).
All the F90 and 4 out of the 6 F93 mutants were practically devoid
of estriol glucuronidation activity. The only two mutants that
exhibited considerable activity toward this estrogen were F93A
and, notably, F93G. The activity of F93A in the estriol glucuronida-
tion screen was about 60% of the control activity, whereas F93G,
the mutant with the smallest side chain, catalyzed estriol glucuron-
idation at about three times higher rate than control UGT1A10
(Fig. 2).
3.9. Ethinylestradiol
3.6. Entacapone
Ethinylestradiol is generally regarded as a specific substrate for
UGT1A1, at least when human liver microsomes are assayed [24].
Nevertheless, we have noticed that UGT1A10, too, can catalyze
EED glucuronidation at low rates, and included this aglycone sub-
strate in the current study. The results of the EED glucuronidation
screen revealed low but detectable activity by UG1A10, somewhat
higher in the F93A mutant and, strikingly, very high activity to-
ward this synthetic steroid by F93G, at least in comparison to
UGT1A10 (Fig. 2).
The drug entacapone is a UGT1A9-specific substrate when hu-
man liver microsomes are analyzed, but it is also glucuronidated
by the (largely) extra-hepatic UGTs 1A7, 1A8 and 1A10 [15]. As a
molecule, entacapone differs significantly from the other sub-
strates that were selected for this study since it is rather large, very
flexible and contains two strongly polar groups, a cyano and a nitro
groups (Fig. 1). The screening results for entacapone glucuronida-
tion rates show high variability: in 6 mutants the rates were very
low, or below detection limit, while in F93L the rate was 2.5-fold
higher than in control UGT1A10 (Fig. 2).
3.10. Kinetic analyses
3.7. Estradiol
As useful as the screening results are for gaining an overall view
of the effects of different mutations, much more detailed data
about the interplay between enzyme and substrate can be gained
from kinetics studies. To keep this study within manageable size,
full kinetic assays were carried out with three of the substrates,
pNP, estradiol and EED. The mutant enzymes that exhibited signif-
icant and easily detectable activity toward these compounds were
included in these assays and the results are presented in Fig. 3. The
reactions generally followed either Michaelis–Menten or Michae-
lis–Menten with substrate inhibition kinetics. The derived kinetic
constants, Km, Ki and Vmax (expression-normalized), are listed in Ta-
ble 1. As can be seen from the results, in the cases of pNP and estra-
diol, the Km values of the tested mutants were mostly higher than
the corresponding values in UGT1A10. Interestingly, in the two
This estrogen has two hydroxyl groups, 3-OH and 17-OH, both
of which can be glucuronidated. However, since UGT1A10 gener-
ates estradiol-3-glucuronide at nearly 100-fold higher rate than
estradiol-17-glucuronide [5], we have concentrated in this study
on the formation of estradiol-3-glucuronide. The results show that
only 4 of the 12 mutants exhibited considerable estradiol glucu-
ronidation rate. Two of them, F93G and F93Y, catalyzed estradiol
glucuronidation at similar rates to control 1A10, while F90L and
F93A exhibited about a quarter of the control rates (Fig. 2). The lat-
ter findings are in agreement with the results in a previous study
that examined the estrogens glucuronidation activity of the F90L
and F93A mutants of UGT1A10, along with several other UGTs [25].
Fig. 4. Sequence to structure alignment between structurally-solved glycosyltransferases (6 upper sequences) and selected human UGTs. The secondary structure elements
of the crystal structures of the glycosyltransferases are shown above the sequences; those present in all proteins are colored black, and those found in some but not all crystals
are gray. The boxed areas mark segments that are fully super-imposable among all structures. The predicted helices and strands of the UGT sequences are indicated with gray
background. UGT1A10 sequence is in boldface and its F90 and F93 are marked below the sequences. Numbers indicate residues before and after the shown fragments
(including the signal sequences, in case of the UGTs), or the number of residues omitted from five protein structures before N
structures, and the structure codes, are: UGT72B1/2VCE, the hydroquinone glucosyltransferase from Arabidopsis thaliana; GtfB/1iir, the TDP-epi-vancosaminyltransferase GtfB
from Amycolatopsis orientalis; GtfA/1PN3, GtfA from the latter organism; UrdGT2/2P6P, the C–C bond-forming DTDP- -olivose-transferase from Streptomyces fradiae; calG3/
3d0q, the enediyne glycosyltransferase from Micromonos poraechinospora; OleiI/2iya, the oleandomycin glycosyltransferase from Streptomyces antibioticus.
a3-1. The enzymes with known crystal
D

C. Höglund et al. / Steroids 76 (2011) 1465–1473
1471
mutants that catalyzed considerable rates of EED glucuronidation,
the ratio of kinetic constant values between mutants and control
were different. Mutant F93G exhibited lower Km and much higher
Vmax values than UGT1A10, while in F93A both Km and Vmax values
were lower than in the control (Table 1 and Fig. 3). The kinetic
analyses of the EED glucuronidation reactions also indicated minor
substrate inhibition in both F93G and F93A.
be helical (Na3-1) by both the PredictProtein and Jpred programs
[23,2], respectively). This early part of loop 3 shows major struc-
tural variability and mobility among the resolved structures of gly-
cosyltransferases from plant and bacteria, while the surrounding
areas of the proteins overlap well (Fig. 4). Interestingly, this is a
rather variable segment among the human UGTs, too. The two
phenylalanines are conserved in the highly homologous UGTs
1A7–1A10 (Fig. 4), but not in many other UGTs. At position 90, F,
Y, I, L, V and T are observed in other UGTs, while at position 93
the other UGTs carry either F, Y, H, I, L, M or Q. Notably, neither
A nor G, and not any charged residues, are found in any of the se-
quences at the homologous positions (Fig. 4).
3.11. Homology model
In order to interpret the large amount of results we have gener-
ated in this study, it is important to locate F90 and F93 with re-
spect to the catalytic center of the enzyme, His37 and the bound
UDPGA. To this end, we have modeled the protein similarly to
our recent work with UGT1A1 [13]. The model indicates that phen-
ylalanines 90 and 93 are located 15 and 18 residues, respectively,
after Nb1 (C72–K75), in a segment that is strongly predicted to
The predicted helix Na3-1 is amphipathic. The derived model of
the substrate binding site is shown with pNP and estradiol docked
to the reactive center (Fig. 5). Both substrates are aromatic, rigid
and planar, but differ largely in size. The much smaller pNP does
not reach up to the studied phenylalanines, while the twice larger
estradiol does. In the vicinity of the reactive center, however, the
substrates occupy the same space. It may also be noted here that
both phenylalanines are free to adopt several rotamers.
4. Discussion
Little is currently known about the aglycone substrate binding
site, or sites, of individual UGTs and about the effects of different
amino acids on the complex substrate selectivity of these enzymes.
To address these issues we have explored the role(s) of F90 and F93
of UGT1A10, two residues that were previously reported to react
with photoactive phenolic probes [28]. We have replaced these
phenylalanine residues with either the aromatic tyrosine (least dif-
ferent), or with hydrophobic residues of increasing size, glycine,
alanine, valine, leucine and isoleucine. In addition, we have se-
lected several different substrates for this study, large and small
(Fig. 1), in order to obtain a better view of the substrate selectivity
of the different mutants. The substrates are all formally neutral,
although they included compounds with the strongly polar nitro
and cyano groups.
The high variability in the effects of different mutations, partic-
ularly at position 93 (Fig. 2), and the apparent dependence of these
effects on the tested substrates, appears at first sight to be nearly
impossible to understand. However, closer looks at the size and
flexibility of the different substrates (Fig. 1), as well as the new
model for the active site of UGT1A10 (Fig. 5), provide interesting
insights and likely explanations for part of the findings.
As seen in the structures of homologous proteins, the active
sites of the UGTs, and other glycosyltransferases related to them,
lie in the cleft between the N and C domains [20]. The binding
pocket is closed at one side by Na1, and the substrates in the re-
solved structures have only limited positional freedom. The crystal
structures show fully buried substrate analogs bound to the active
site [1], suggesting that for a substrate molecule to enter and prod-
uct to leave, the surface loops must rearrange. The surface loops of
the N domain are very likely critical for the specificity differences
between the human UGTs, as all the UGT1A enzymes are identical
from E286 onward. A static model like ours cannot represent the
whole sequence of events, but as no details on the activation mech-
anism of the UGTs are known, we consider the model an acceptable
starting point.
The substrates used in this study can be divided, according to
their size, into two groups. The small substrates group includes
pNP, 1-naphthol, 4-MU, and scopoletin, whereas the large ones
are entacapone and the three estrogens (Fig. 1). The effect of the
size difference on the results can be addressed by placing the reac-
tive hydroxyl groups of each substrate between the active site his-
tidine, H37, and the bound UDPGA, while the rest of the substrate
Fig. 5. A structural model of UGT1A10. (A) An overall view of the lumenal part of
the enzyme. The N-terminal domain is shown in white cartoon representation, the
C-terminal domain is in dark gray cartoon, and the envelop helices are shown as a
black Ca-trace (see Laakkonen and Finel [13] for a details description of the domains
in a UGT enzyme). The spheres represent the locations of F90, F93 and the ‘‘catalytic
His’’, H37. The location of the reactive site is marked with a space-filling mesh. The
endoplasmic reticulum membrane is just underneath the shown model. (B) A close-
up view of the active site of the enzyme. The N-terminal domain (in white) is up,
and the C-terminal domain (in dark gray) is down. Residues discussed in this study,
as well as the structural elements they interact with, are shown and named.
Estradiol and 4-nitrophenol (pNP) were docked to putative reactive distance and
orientation with respect to His 37 and the bound UDPGA. Estradiol is shown in
white ball-and-stick representation, and pNP in dark gray.

1472
C. Höglund et al. / Steroids 76 (2011) 1465–1473
molecule is directed toward phenylalanines 90 and 93. In such an
orientation, only the large substrates, represented in Fig. 5 by
estradiol, but not the small substrates like pNP, reach to the seg-
ment containing these two Phe residues (Fig. 5).
suggestion that these residues are involved in substrate binding
by UGT1A10 (not merely phenolic substrates, though) [28].
Nevertheless, it is most likely that these two phenylalanines are
not the only major player in estrogens glucuronidation by
UGT1A10. If they were, the highly homologous enzyme, UGT1A9,
that has both Phe residues, should have also exhibited high glucu-
ronidation rates toward the 3-OH of estradiol, but it is practically
inactive with estradiol [5]. Still, the latter study suggested that
UGT1A9 binds estradiol since it inhibited the scopoletin glucuron-
idation activity of the enzyme [5], leaving open the possibility that
either or both these phenylalanines play a role in estrogens binding
to UGT1A9, too.
In summary, the results of this study suggest that F93 of the hu-
man UGT1A10 is involved in substrate binding, whereas the role of
F90 is somewhat different. The new detailed structural model for
this enzyme, as well as the wealth of activity results, are likely to
stimulate new studies that will take us closer to understanding
the complex substrate selectivity of the human UGTs, and the dif-
ferent steps in their catalytic cycle.
The three estrogens differ from each other in ring D substitu-
tions (Fig. 1) and among them the trend in the glucuronidation
rates could be explained in the following way. Estradiol is glucu-
ronidated effectively by UGT1A10, and two of the mutants at
F93, F93G and F93Y, exhibit as high as control activity toward it
(Fig 2). The glucuronidation activity of UGT1A10 and, particularly
F93Y, decreases sharply with additional substituents on the
steroid, either the hydroxyl at C16 in the case of estriol, or the eth-
inyl group at C17 of EED. However, replacing phenylalanine at po-
sition 93 of UGT1A10 with the smallest side-chain amino acid,
glycine, results in higher or even much higher glucuronidation
rates of estriol and EED, respectively, in comparison to control
UGT1A10 (Fig. 2). It is thus suggested that in the enzyme-substrate
complex, the estradiol substituents on ring D of the steroid back-
bone occupy the same space as residue 93 of the protein. The out-
come of this is that increasing the substrate size in this part of the
binding site can be partly compensated by reducing the size of the
side chain at position 93, as seen in estriol glucuronidation by mu-
tant F93A and the glucuronidation of both estriol and EED by the
F93G mutant (Fig. 2).
Financial support
This study was supported, in parts, by the Sigrid Juselius Foun-
dation (M.F.), the Magnus EhrnroothFoundation (N.S. and L.L.), and
by the N.I.H. Grant GM075893 (AR-P).
It may be added here that the F90 mutants, even F90L that
exhibited about 25% of the control rate in estradiol glucuronida-
tion, are nearly inactive toward estriol and EED (Fig. 2).
Acknowledgements
Scopoletin has an extended substitutent on the naphthol back-
bone, the methoxy group at a beta position with respect to the
reactive hydroxyl (Fig. 1). The activity results of the mutants with
scopoletin are a clear exception (Figs. 1 and 2) that we currently
cannot explain. Nevertheless, it may be suggested that a combina-
tion of structural and physico-chemical properties makes the glu-
curonidation of scopoletin less dependent on interactions with
other residues than the catalytic His and the bound UDPGA. In this
respect, it may be noted here that addition of 0.2% of the detergent
Triton X-100 to the full-length membrane-bound UGT1A9, or the
preparation of a water-soluble mutant of UGT1A9 by truncating
the enzyme before the predicted start of the trans-membrane helix
and the cytoplasmic tail affected the scopoletin glucuronidation
activity of the enzyme differently than glucuronidation of other
aglycone substrates [9,10], respectively). It is thus assumed that
the differences in glucuronidation rates of the F90 and F93 mutants
toward scopoletin on one hand, and toward 4-MU on the other
hand (Fig. 2) are linked in some way to the previous observations
with UGT1A9 and these substrates [9,10]. Nevertheless, it remains
to be studied how these findings are connected.
We would like to thank Johanna Mosorin, Kaisa Laajanen and
Johanna Troberg for skilful technical assistance at different stages
of this study, and Orion Pharma Ltd. for providing entacapone.
References
[1] Brazier-Hicks M, Offen WA, Gershater MC, Revett TJ, Lim EK, Bowles DJ, et al.
Characterization and engineering of the bifunctional N- and O-
glucosyltransferase involved in xenobiotic metabolism in plants. Proc Natl
Acad Sci USA 2007;104:20238–43.
[2] Cole C, Barber JD, Barton GJ. The Jpred 3 secondary structure prediction server.
Nucleic Acids Res 2008;36(web server issue):W197–201.
[3] Dellinger RW, Chen G, Blevins-Primeau AS, Krzeminski J, Amin S, Lazarus P.
Glucuronidation of PhIP and N–OH–PhIP by UDP-glucuronosyltransferase
1A10. Carcinogenesis 2007;28:2412–8.
[4] Fujiwara R, Nakajima M, Yamanaka H, Yokoi T. Key amino acid residues
responsible for the differences in substrate specificity of human UDP-
glucuronosyltransferase (UGT)1A9 and UGT1A8. Drug Metab Dispos
2008;37:41–6.
[5] Itäaho K, Mackenzie PI, Ikushiro S, Miners JO, Finel M. The configuration of the
17-hydroxy group variably influences the glucuronidation of beta-estradiol
and epiestradiol by human UDP-glucuronosyltransferases. Drug Metab Dispos
2008;36:2307–15.
The homology model provides an explanation for the effects of
the mutations on the activity of UGT1A10 toward the large
substrates. It does not, however, yield a clear explanation for the
variable results with the smaller substrates, such as pNP, 4-MU
and 1-naphthol. Another observation that is not currently ex-
plained by the model is the differences between F90 and F93,
and the much higher sensitivity of UGT1A10 to the replacement
of F90 by other residues, except leucine (Fig. 2). The main current
limitation in building a detailed model for the active site of the
UGTs concerns the loops at the back of the substrate binding pock-
et. This is because the homologous glycosyltransferases from plant
and bacteria for whom a solved structure is available (as a tem-
plate for the modeling) mostly lack these loops (Fig. 4). In any case,
these loops are predicted to be mobile and flexible to some extent,
and they may be affected by the binding of UDPGA in an ‘‘induced
fit’’ type of movement.
[6] Itäaho K, Court MH, Uutela P, Kostiainen R, Radominska-Pandya A, Finel M.
Dopamine is a low-affinity and high-specificity substrate for the human UDP-
glucuronosyltransferase 1A10. Drug Metab Dispos 2010;37:768–75.
[7] Itäaho K, Laakkonen L, Finel M. How many and which amino acids are
responsible for the large activity differences between the highly homologous
UDP-glucuronosyltransferases (UGT) 1A9 and UGT1A10? Drug Metab Dispos
2010;38:687–96.
[8] Kaivosaari S, Salonen JS, Mortensen J, Taskinen J. High-performance liquid
chromatographic method combining radiochemical and ultraviolet detection
for determination of low activities of uridine 50-diphosphate-
glucuronosyltransferase. Anal Biochem 2001;292:178–87.
[9] Kurkela M, Garcia-Horsman JA, Luukkanen L, Morsky S, Taskinen J, Baumann
M, et al. Expression and characterization of recombinant human UDP-
glucuronosyltransferases (UGTs). UGT1A9 is more resistant to detergent
inhibition than other UGTs and was purified as an active dimeric enzyme. J
Biol Chem 2003;278:3536–44.
[10] Kurkela M, Hirvonen J, Kostiainen R, Finel M. The interactions between the N-
terminal and C-terminal domains of the human UDP-glucuronosyltransferases
are partly isoform-specific, and may involve both monomers. Biochem
Pharmacol 2004;68:2443–50.
[11] Kurkela M, Patana AS, Mackenzie PI, Court MH, Tate CG, Hirvonen J, Goldman
A, Finel M. Interactions with other human UDP-glucuronosyltransferases
attenuate the consequences of the Y485D mutation on the activity and
substrate affinity of UGT1A6. Pharmacogenet Genomics 2007;17:115–26.
The results have shown that different replacements at positions
90 and 93 of UGT1A10 affect activity and kinetics in different ways.
The results and the model are in agreement with the original

C. Höglund et al. / Steroids 76 (2011) 1465–1473
1473
[12] Kuuranne T, Kurkela M, Thevis M, Schanzer W, Finel M, Kostiainen R.
Glucuronidation of anabolic androgenic steroids by recombinant human
UDP-glucuronosyltransferases. Drug Metab Dispos 2003;31:1117–24.
[22] Patana AS, Kurkela M, Finel M, Goldman A. Mutation analysis in UGT1A9
suggests a relationship between substrate and catalytic residues in UDP-
glucuronosyltransferases. Protein Eng Des Sel 2008;21:537–43.
[23] Rost B, Yachdav G, Liu J. The predict protein server. Nucleic Acids Res
2004;32(web server issue):W321–6.
[24] Soars MG, Ring BJ, Wrighton SA. The effect of incubation conditions on the
enzyme kinetics of UDP-glucuronosyltransferases. Drug Metab Dispos
2003;31:762–7.
[13] Laakkonen L, Finel M.
A
molecular model of the human UDP-
glucuronosyltransferase 1A1, its membrane orientation, and the interactions
between different parts of the enzyme. Mol Pharmacol 2010;77:931–9.
[14] Lautala P, Ethell BT, Taskinen J, Burchell B. The specificity of glucuronidation of
entacapone
and
tolcapone
by
recombinant
human
UDP-
glucuronosyltransferases. Drug Metab Dispos 2000;28:1385–9.
[25] Starlard-Davenport A, Xiong Y, Bratton S, Gallus-Zawada A, Finel M,
Radominska-Pandya A. Phenylalanine(90) and phenylalanine(93) are crucial
amino acids within the estrogen binding site of the human UDP-
glucuronosyltransferase 1A10. Steroids 2007;72:85–94.
[15] Luukkanen L, Taskinen J, Kurkela M, Kostiainen R, Hirvonen J, Finel M. Kinetic
characterization of the 1A subfamily of recombinant human UDP-
glucuronosyltransferases. Drug Metab Dispos 2005;33:1017–26.
[16] Mackenzie PI, Bock KW, Burchell B, Guillemette C, Ikushiro S, Iyanagi T, et al.
Nomenclature update for the mammalian UDP glycosyltransferase (UGT) gene
superfamily. Pharmacogenet Genomics 2005;15:677–85.
[26] Stewart JJ. Optimization of parameters for semiempirical methods V:
modification of NDDO approximations and application to 70 elements. J Mol
Model 2007;13:1173–213.
[17] Miley MJ, Zielinska AK, Keenan JE, Bratton SM, Radominska-Pandya A, Redinbo
MR. Crystal structure of the cofactor-binding domain of the human phase II
[27] Wells PG, Mackenzie PI, Chowdhury JR, Guillemette C, Gregory PA, Ishii Y,
Hansen AJ, Kessler FK, Kim PM, Chowdhury NR, Ritter JK. Glucuronidation and
the UDP–glucuronosyltransferases in health and disease. Drug Metab Dispos
2004;32:281–90.
drug-metabolism enzyme UDP-glucuronosyltransferase 2B7.
2007;369:498–511.
J Mol Biol
[18] Ohno S, Nakajin S. Determination of mRNA expression of human UDP-
glucuronosyltransferases and application for localization in various human
tissues by real-time reverse transcriptase-polymerase chain reaction. Drug
Metab Dispos 2009;37:32–40.
[19] Oleson L, Court MH. Effect of the beta-glucuronidase inhibitor saccharolactone
on glucuronidation by human tissue microsomes and recombinant UDP-
glucuronosyltransferases. J Pharm Pharmacol 2008;60:1175–82.
[20] Osmani SA, Bak S, Møller BL. Substrate specificity of plant UDP-dependent
glycosyltransferases predicted from crystal structures and homology
modeling. Phytochemistry 2009;70:325–47.
[28] Xiong Y, Bernardi D, Bratton S, Ward MD, Battaglia E, Finel M, et al.
Phenylalanine 90 and 93 are localized within the phenol binding site of
human UDP-glucuronosyltransferase 1A10 as determined by photoaffinity
labeling, mass spectrometry, and site-directed mutagenesis. Biochemistry
2006;45:2322–32.
[29] Zhang H, Tolonen A, Rousu T, Hirvonen J, Finel M. Effects of cell differentiation
and assay conditions on the UDP-glucuronosyltransferases activity in Caco-2
cells. Drug Metab Dispos 2011;39:456–64.
[30] Zielinska A, Lichti CF, Bratton S, Mitchell NC, Gallus-Zawada A, Le VH, et al.
Glucuronidation of monohydroxylated warfarin metabolites by human liver
[21] Patana AS, Kurkela M, Goldman A, Finel M. The human UDP-
glucuronosyltransferase: identification of key residues within the
nucleotide-sugar binding site. Mol Pharmacol 2007;72:604–11.
microsomes and human recombinant UDP-glucuronosyltransferases.
Pharmacol Exp Ther 2008;324:139–48.
J