Tài liệu Báo cáo Y học: Study of substrate specificity of human aromatase by site directed mutagenesis pdf

Study of substrate specificity of human aromatase by site directed
mutagenesis
P. Auvray
1,
*
,†
, C. Nativelle
1,†
, R. Bureau
2
, P. Dallemagne
2
, G E. Se
´
ralini
1
and P. Sourdaine
1
1
IBBA, Laboratoire de Biochimie et Biologie Mole
´
culaire, Universite
´
de Caen, Esplanade de la Paix, Caen, France;
2
CERMN, Laboratoire de Pharmacochimie, Caen cedex, France
Human aromatase is responsible for estrogen biosynthesis
and is implicated, in particular, in reproduction and estro-
gen-dependent tumor proliferation. The molecular s tructure
model is l argely derived from t he X-ray structure of b acterial
cytochromes sharing only 15–20% identities with

hP-450arom. In the present study, site directed mutagenesis
experiments were performed to examine the role of K119,
C124, I125, K130, E302, F320, D309, H475, D476, S470,
I471 and I474 of aromatase in catalysis and for substrate
binding. The catalytic properties of mutants, transfected in
293 cells, were evaluated using androstenedione, testoster-
one or nor-testosterone as subs trates. In a ddition, inhibition
profiles for these mutants with indane or indolizinone
derivatives were obtained. Ou r results, together w ith com-
puter modeling, show that catalytic properties of mutants
vary in accordance with the substrate used, suggesting
possible differences in substrates positioning within the act-
ive site. In this respect, importance of residues H475, D476
and K130 was discussed. These results allow us to hypo-
thesize that E302 could be involved in the aromatization
mechanism with nor-androgens, whereas D309 remains
involved in androgen aromatization. This study highlights
the flexibility of the substrate–enzyme complex conforma-
tion, and thus sheds new light on residues that may be
responsible for substrate specificity between species or aro-
matase isoforms.
Keywords: aromatase; site-directed mutagenesis; molecular
modeling; androgens; inhibitors.
Estrogens are known to be implicated in reproduction and
estrogen-dependent tumor proliferation [1]. Moreover, an
abnormal expression of aromatase, the enzyme involved in
the conversion of androgens t o e strogens, has been detected
in breast tumors and in surrounding adipose stromal cells
[2,3]. The a romatase enzyme comprises a specific cyto-
chrome P-450 aromatase and the ubiquitous cytochrome

P-450 NADPH reductase. A common treatment for
estrogen-dependent cancers is the use of antiestrogens
and/or aromatase inhibitors [4]. The usual way to develop
new a romatase inhibitors is to screen in vitro chemical
compounds [5–10] from the knowledge of the substrate
structure, or by comparisons with other aromatases [ 11–13].
The design o f more spe cific and efficient a romatase
inhibitors could be improved by a better knowledge of the
enzyme’s active site. A precise modeling of this part of the
molecule is therefore necessary. D espite success in obtaining
aromatase purified to homogeneity [14], crystallization of
this microsomal membrane-anchored protein has not been
reported. Knowledge of the aromatase structure is largely
derived from c omparisons with soluble bacterial cyto-
chromes Pseudomons putida camphor P-450 (P-450cam),
Bacillus megaterium P-450 (P-450BM-3), Pseudomonas
putida a-terpineol P-450 (P-450terp) and Saccharopoly spora
erythreae erythromycin F P-450 (P-450eryF), t hese proteins
being well characterized and crystallized [15–19]. However,
human aroma tase shares only 15–20% homology with these
bacterial cytochromes. A theoretical molecular model of
P-450arom has been proposed [20], but the model revealed a
poor energy profile in the regions between residues 150–250
[21]; the problems seemed to be attributed to the length of
helices F and G, and a model based on cytochrome
P-450camisbetterdefinedinthisregion[21].Itisvery
difficult to produce a more reliable mod el, i rrespective of t he
bacterial cytochrome P-450 used for alignment. In fact,
these later cytochromes P-450, with resolved three-dimen-
sional structures, have very weak sequence homologies with

P-450arom. Moreover,
FASTA
3
AND BLAST
analyses of the
protein databank sequences do not provide more reliable
models of amino-acid sequences. Studies using site-directed
mutagenesis provides a way of v alidating partial or
complete models. Such structure–function studies make it
possible t o identify important regions directly or indirectly
implicated in the aromatizatio n mechanism. These domains
are the substrate access channel (constituted by the
b 1-1 sheet), the FG loop and the aromatase specific region
[20]. G raham-Lorence et al.[20]suggestedthattheB¢C l oop
has an i mportant function in substrate orientation, and t hat
Correspondence to P. Sourdaine, IBBA, Universite
´
de Caen, Es planade
de la Paix, 14032 Caen cedex, F rance. Fax: + 33 2 31 56 53 20,
Tel.: + 33 2 31 56 53 70, E-mail:
Abbreviations: hP-450arom, human P-4 50 a romatase; e P-450arom,
equine P- 450 aromatase; P-4 50cam, P450 camphor f rom Pse u domons
putida; P-450BM-3, P450 f rom Bacillus megaterium; P-450terp,
P-450 a-terpineol from pseudomonas putida ; P -450eryF, P- 450
erythromycin F from Sacch aropolyspora e rythreae;
4-OHA, 4-hydroxyandrostenedione.
Enzyme: cytochrome P450 aromatase (EC 1.14.14.1).
*Present address: Oncodesign S.A., Parc Tech nologique d e la Toison
d’Or, 21000 Dijon, France.
Note: these authors contributed equally to this work.

(Received 17 December 2001, accepted 11 January 2002)
Eur. J. Biochem. 269, 1393–1405 (2002) Ó FEBS 2002
apartoftheb4 sheet (K473–D476) is implicated in the
substrate pocket, particularly in the e xtrahydrophobic
pocket [21]. P revious s ite-directed mutagenesis studies
suggest a role for specific residues such as E302, D309,
T310 [20–24] and K473, H475 [20,25]. Finally, the regions
involved in redox-partner have been defined as being in B,
C, J, J¢, K and L helices.
Recently, the report of three isoforms of porcine aroma-
tase, encoded by distinct genes [26] and the study of their
catalytic differences [27,28] also highlight the need to
understand which residues of the enzyme are involved in
the substrate specificity. In addition to human and porcine
aromatases, e quine aromatase is also well characterized
biochemicaly [29–34]. For example, nor-testosterone is
more rapidly aromatized b y the equine aromatase, despite
a weaker affinity, than the human enzyme [34]; some
inhibition differences have also been described [11,13,31].
Taking into account these r esults and t he thre e-dimensional
model of both human and equine P-450arom w e h ave
suggested that H475 and D476 h ave a role in t he i nteraction
of the indane derivative MR 20814 within t he active sit e [11].
Therefore, the aim of our present study was to explore the
role of H475 and D476 and of o ther human aromatase
residues which might be involved in the orientation and
binding of substrates. Human residues were mutated to the
corresponding aligned eq uine residues (Fig. 1), apart from
D309 and E302, which have been extensively studied in the
literature. H475, which is Asn in horse and other species,

and D476 w hich is absolutely c onserved, were more
extensively studied because of their location within an
extrahydrophobic pocket, or a hydrophobic surface [21].
This could determine differences in inhibition between
human and equine P-450arom by MR 20814 [11].
EXPERIMENTAL PROCEDURES
Chemicals
All chemical products were obtained from Sigma (St
Quentin Fallavier, France) (polye thylenimine, 50 kDa,
was prepared in ddH
2
Oat10m
M
, pH 7.0) or GibcoBRL
(Cergy Pontoise, France). [1b,2b-
3
H]Androstenedione was
from Dupont NEN ( Les Ulis, F rance), testosterone and 19-
nor-testosterone from Sigma ( St Quentin Fallavier, F rance),
solvents from Carlo Erba (Val de Reuil, France) and sds
(Peypin, France), 293 cells (ECACC number: 85120602)
with stable expression of cytochrome P450 reductase
(Kindly provided by V. Luu-The, CHUL, Que
´
bec), Quick-
Change
TM
Site-Directed Mutagenesis kit from Stratagene
(Montigny le Bretonneux, France), alkaline phosphatase
substrate kit from Bio-Rad (Ivry sur Seine, France), culture

media f rom BioWhittaker (Gagny, France), Thermo
Sequenase Kit from Amersham (Les Ulis, France), pCMV
plasmid from Invitrogen (NV Leek, the Netherlands) and
Qiagen Plasmid Maxi Kit from Qiagen ( Courtaboeuf,
France). Human aromatase cDNA was kindly provided by
E. R. Simpson (Monash University, Melbourne, Australia).
The oligonucleotide primers were f rom Pharmacia (Orsay,
France) or EUROBIO (Les Ulis, France). Indane and
indolizinone derivatives were produced by the CERMN
(Caen, France, Fig. 2).
PCMV-human aromatase cDNA construction
The p lasmid used in this study has b een previously described
[13]. Briefly, human aromatase cDNA (2920 bp) [35] was
cloned into pUC18 (2.7 kb) with two fragments of kgt10
HindIII–EcoRI at the 5¢ end (240 bp), and EcoRI–BglII at
the 3 ¢ end (900 bp). This c onstruction w as par tially digested
with EcoRI and the EcoRI–EcoRI fragment (2920 bp) was
cloned into pCMV (EcoRI site at position 753). The
construction orientation was checked by sequencing and
by specific PCR amplification with the primers sense A
(555–575, 5¢-CCATTGACGTCAATGGGAGTT-3¢)and
antisense B (1920–1899, 5¢-TAAGGCTTTGCGCATGAC
CAAG)3¢), which are specific to pCMV and the cDNA,
respectively (the amplicon length was 1368 bp). The
pCMV-cDNA was purified from JM109 bacterial strain
amplification by Qiagen Plasmid Maxi Kit. The length,
concentration and purity of the plasmid-cDNA construc-
tion were checked by 1% agarose electrophoresis and
ethidium bromide staining.
Fig. 1.

CLUSTALW
1.81 multiple sequence alignment of human and
equine aromatases. Human and equine sequences were, respectively,
from Corbin et al.[35]andTomilinet al. [33].
Fig. 2. Str ucture of inhibitors. 4-OHA is from Brodie et al.[64],MR
20814, MR 20492 and MR 20494 from Auvray et al. [11,13].
1394 P. Auvray et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Aromatase cytochrome P-450 molecular modeling
Initial alignments of cytochrome P-450 BM3 and aroma-
tase cytochrome P-450 were taken from the alignments of
Graham-Laurence et al. [20]. Main chain coordinates for
the core regions were taken directly from the cytochrome
P-450BM3 structure. Using coordinates for the loops
obtained from the loop data base search. The replace
residue command was used t o r eplace a residue. I n t his c ase,
the replacement residue is first aligned to the backbone of
the original residue. After the backbone has been aligned,
the dihedral angles in common with the residue being
replaced are also aligned. New charges and potential
functions types were taken from the r esidue library.
Refinement of the structures involved energy minimization
using
AMBER
[36] and
ESFF FORCE FIELD
(Program
DISCOVER
version 95, />forcefield/esff.html) [37,38]. Difficulties modelling the heme
led to the choice of ESFF force field for th is region. As far as
possible, the atomic parameters were directly determined

from experimental or calculated rather than fit. For the
valence energy in this force field, as with
AMBER
, only
diagonal terms were included. Th e partial charges were
determined by minimizing the electrostatic energy with
respect to the charges, with the co nstraint that the sum of
the c harges is equal to the net charge o n the mo lecule. In this
case, e lectronegativity and hardness were determined. ESFF
was based on electronegativities and hardnesses calculated
using the density functional t heory. For the van der Waals
interactions, ESFF used t he 6-9 potential. T he van d er
Waals parameters were derived using rules consistent with
the charges. The minimization algorithms used were Steep-
est Descents until a gradient of 10 kcalÆmol
)1
ÆA
˚
)1
,then
conjugate gradients until a gradient of 1 kcalÆmol
)1
ÆA
˚
)1
.
The final structure was analyzed by
PROCHECK
[39] and
PROSA

[40]. No specific routine were used for docking of
substrate and inhibitors. Docking of the androstenedione
into the active site was carried out considering the orien-
tation of C(1), C(2) and C(19) above the heme [41] and the
position of the ligand towards D309 and T310. The three-
dimensional structure of androstedione was obtained from
crystallographic data [42]. This docking was followed by a
minimization to refine the complex. This method produced
difficulties in maintaining the overall three-dimensional
structure of the steroid. Indeed, during the minimization,
steric interactions between V370 and the A r ing of th e
androstenedione led to a modification of the conformation
of this ring. W e optimized the position of t he ligand a nd the
orientation o f the h ydrophobic group of V370 (modification
of dihedral angles) t o decrease this in teraction. V370 is
highly conserved suggesting an important role of this
residue in the active site. For the inhibitors, we have
considered an orientation o f t he pyridine group towards the
heme and of the amine group towards t he extrahydrophobic
surface. A discussion on this proposal of docking and on t he
conformation of the inhibitor w as carried out as described
previously [11].
Noncovalent interactions between different residues or
between residues and substrates were determined using the
ISOSTAR
software [43]. Theoretical noncovalent interactions
from the
ISOSTAR
software were calculated from the sum of
the following terms: th e electrostatic energy (attractive and

repulsive Coulombic interaction); the exchange–repulsion
term (sum of an energy lowering due to exchange of
electrons of parallel spin between the molecules and the
repulsive term arising from the Pauli exclusion principle);
the polarizat ion energy (energy gain c aused b y the change of
intramolecular wave function of one molecule due to the
presence of the undistorted c harge distribution o f the second
molecule); the charge-transfer energy (attractive energy
from actual charge transfer between molecules); the disper-
sion energy (calculated a t the second order double e xcitation
level).
The definition of the lipophilicity potential (
MOLCAD
SURFACE
; program
SYBYL
6.0; Tripos Association: St Louis,
MO, USA) is calculated on the basis of t he atomic partial
lipophilicity values [44] and a distance-dependent function
[45].
Site-directed mutagenesis
This step was performed with the QuickChange
TM
Site-
Directed Mutagenesis method from Stratagene. B riefly, this
was b ased on a PCR with two complementary oligonucle-
otide primers containing the mutation. The PCR was
performed with the Pfu DNA polymerase during 16 cycles
(30 s at 9 5 °C, 30 s at 5 5 °Cand13 minat68 °C). The PCR
products were then digested with DpnI which only digests

the parental methylated c DNA. Nicked vector DNA with
the d esired mutations was then transformed into Escherichia
coli XL1-Blue supercompetent cells. Transformed bacteria
were analyzed directly on colonies by PCR with p rime rs 5H-
1 (1361–1379, 5¢-GTC GTGTCATGCTGGACAC-3¢)and
3H-54 (2384–2367, 5¢-GAGGATGACACTATTGGC-3¢)
after 30 cycles (1 min at 95 °C, 1 min at 52 °C, 2 min at
72 °C; 1 cycle: 10 min at 72 °C). T he expected amplicon
length was 1026 bp. Mutations were then checked by
sequencing 10 lL bacterial DNA miniprep with the Ther-
mo Sequenase, as previously described [13]. Plasmid DNA
was extracted as follows: the bacterial pellet was lysed with
8% sucrose, 0.5% Triton X-100, 0.05
M
EDTA, 0.01
M
Tris/HCl pH 8.0 and 10 mgÆmL
)1
lysosyme/0.01
M
Tris/
HCl pH 8.0, boiled and the DNA was then precipitated by
3M NaOAc pH 7.0 and isopropanol. A fter sequencing,
pCMV-cDNA was purified from XL1-Blue supercompe-
tent bacterial strain amplification by means of the Qiagen
Plasmid Maxi Kit, as previously described.
Culture and transfection of 293 cells
Cells were grown in red phenol-free EMEM medium and
supplemented with 2 m
M

glutamine, 10% new-born calf
serum (supreme serum), 1% nonessential amino acids at
37 °C in an atmosphere of 5% CO
2
and 95% air. Cells
(50 000) were grown to 50% confluence on 24-well
cell culture plates 18 h before transfection, washed with
serum-free cell culture medium, supplemented with 500 lL
serum-free medium and t ransiently transfected with 2 lg
pCMV-human aromatase cDNA, using a modification of
the m ethod of Boussif et al. [46]. Briefly, 2 lg p CMV-
cDNA (6 nmol of phosphate) and 54 nmol of polyethylen-
imine were separately diluted with 50 lL 150 m
M
NaCl,
incubated for 10 min at room temperature in a laminar
fume hood, mixed together, incubated for another 1 0 min at
room temperature, and then added to each well. Cells were
incubated for 3–4 h at 37 °C and then supplemented with
Ó FEBS 2002 Structure–function relationships of human aromatase (Eur. J. Biochem. 269) 1395
500 lL medium containing 10% supreme serum. After a
further 18-h incubation, cells were washed with serum-free
medium and the aromatase activity was measured in whole
cells. Evaluation in the whole cell rather than the micro-
somal aromatase activity was used because it may allow
better approximation in the in vivo situation [21]. ELISA
quantification of the aromatase expressed was used as an
indicator of the transfection efficiency.
Whole cell aromatase activity and inhibition
Aromatase activity was assessed in whole cells using the

method described by Zhou et al.[47]bymeasuringthe
3
H
2
Oreleasedfrom[1b,2b-
3
H]-androstenedione. Cells were
washed with serum-free culture medium. Dessica ted radio-
active substra te (200 n
M
and 50– 800 n
M
for I C
50
and k inetic
experiments, respectively) supplemented with 1 l
M
prog-
esterone(usedtoblock5a-reductase activity) and 0–10 l
M
inhibitor (for IC
50
experiments) were mixed w ith s erum-free
culture medium a nd added to each w ell. Cells were
incubated at 37 °C under 5% CO
2
for 45 m in After
incubating cells for 5 min on ice, the culture medium
(1 mL) was sampled and extracted by CHCl
3

(1 mL).
Steroids were then removed by incubation with 1 mL
charcoal dextran suspension (7%/1.5%), and the
radioactivity o f the aqueous phase was measured as
previously described. The r esults were the mean o f at least
triplicate experiments ± SD and were expressed as
pmolÆmin
)1
Æmg aromatase
)1
. Results from control incuba-
tions, produced by transfecting under the same co nditions
the pCMV plasmid alone inste ad of t he pCMV-cDNA
plasmid, were used to determine the limit of detection.
K
m.app
and V
m.app
determinations were carried out using
linear regression a nalysis of both L ineweaver–Burk and
Hanes–Wolf plots.
Steroid radioimmunoassays
The 17 b-estradiol was assayed in 293 cells supernatant a fter
incubation times of 45 min (nor-testosterone) or 90 min
(testosterone) with 50–1600 n
M
substrate in t he same
conditions as those described above, and after extraction
with 10 volumes of d iethyl ether a s previously described [48].
The supernatant was chosen after demonstrating that the

total part of steroids w as in this compartment (data not
shown). Estradiol rabbit antibodies [(66033)
3
H-estradiol
RIA k it, bioMe
´
rieux, Ch arbonnie
`
res l es Bains, France] were
diluted twofold according to the manufacturer’s instruc-
tions. The extraction efficiency was 80 ± 5% and the
sensitivity of this radioimmunoassay was 10 pgÆmL
)1
.
Results, calculated according to Garnier et al.[49],were
the mean of at least triplicate experiments ± SD and are
expressed as pmolÆmi n
)1
Æmg aromatase
)1
.
Enzyme-linked immuno-sorbent assays
Cells were scraped from c ulture wells (pools of three culture
wells), resuspended in 500 lLofwaterandsonicatedonice
twice at 40 Hz for 20 s. Aromatase in transfected cells was
evaluated by a direct sandwich ELISA method adapted to
our model: a 200-lL cell homogenate or 2 00 lLofNaCl/P
i
containing 2–8 ng of purified equine aromatase (standard
curve) were m ixed with 800 lL o f polyclonal antibody

(1 : 10 000), raised against intact eP-450arom [29], incuba-
ted for 2 h and then added (100 lL per well) to plates
(Nunc, high protein adsorption quality). Plates were p revi-
ously coated overnight at 4 °C with 50 ng per well of
purified equine aromatase, saturated 1 h at 3 7 °Cwith
200 lLNaCl/P
i
/Tween 20 (0.1%)/gelatin (0.5%) and
washed with 150 lLNaCl/P
i
/Tween 20 (0.1%). The
fixation of the anti-(eP-450arom) Ig was then evaluated by
incubating for 1 h at 3 7 °C w ith 100 lLofanti-(rabbitIgG)
Ig coupled to alkaline phosphatase (1 : 6000), washing and
incubating for 1.5 h at 37 °C with 100 lL of the substrate
p-nitrophenylph osphate as described by t he manufacturer.
The absorbance was finally read on a B io-Tek EL800
apparatus (Packard) at 405 nm. Results were the mean of
triplicate experiments ± SD and are expressed in ng
aromatase per culture well. Sensitivity of the assay was
0.2 ng per well of ELISA (Fig. 3) corresponding to 1.6 ng
per well. The antiequine aromatase polyclonal antibodies
were prepared in our laboratory [29]. The to tal protein
quantity was evaluated according to Bradford [50].
Statistical study
Data were compared using the Mann–Whitney test
(
ANOVA
).
RESULTS

The catalytic properties of P450arom mutants are summar-
ized in Tables 1 and 2. Results from an i nvestigation of the
interaction of different aromatase inhibitors w ith mutants,
to test the accuracy of our computer model as well as to
understand the inhibition characteristics of nonsteroidal
inhibitors (MR 20814, MR 20492 and MR 20494) are
shown in Fig. 2. IC
50
are presented in Table 3 and are
analyzed in respect of catalytic properties of mutants with
the substrates tested.
Western blot analysis [29] demonstrated that the poly-
clonal a ntibodies specifically detected h uman aromatase in
microsomes. Aromatase was also evident in E293 cells
Fig. 3. Standard curve of ELISA. (A) The standard curve was
obtained by mixing 200 lLNaCl/P
i
containing 0–8 ng of purified
equine aroma tase to 800 lL of polyclonal a nti-(ep.450arom) Ig;
1 : 10 000. The fixation of the primary antibody was then evaluated
with anti-(rabbit IgG) Ig coupled to alkaline phosphatase and incu-
bation wit h p-nitrophenylphosphate. Absorbance was r ead at 405 nm
on a Bio-tek EL 800 apparatus. (B) Westernblot analysis for
P450arom. Westernblotting of P450arom in mock E293 cells and E293
cells transfecte d with 2 lg p CMV-human aromatase cDNA. The size
and position of the expected 55 kDa aromatase protein is shown.
1396 P. Auvray et al. (Eur. J. Biochem. 269) Ó FEBS 2002
transfected with the p CMV-human aromatase cDNA
construct (Fig. 3).
All mutants could be detected in the 293 transfected cells

using ELISA with an average content per culture well of
3.08 ± 1 .58 ng (mean ± SEM, n ¼ 305), 2.33 ± 0.9 ng
for the wild type (n ¼ 68) and with a range of
2.11 ± 0 .6 ng for D476A (n ¼ 5) to 4.48 ± 2.63 n g for
K130N (n ¼ 33). Aromatase activity was expressed per mg
of P450-arom in order to take i nto account the turnover and
the expression rate of the protein [51].
H475 and D476 residues
H475 or D476 were mutated ( Table 1) to determine the
importance of the residue nature at these positions. With
androstenedione as substrate, six mutations strongly de-
creased a romatase activity: H475N, H475R, H475E,
D476A, D476K and D476L (activities below 1 or 5%
relative to wild-type with 200 n
M
of androstenedione,
Table 1). In contrast, the K
m.app
value for H475A and the
V
m.app
value for D476N decreased and values for D476E
were not different from those of wild-type enzyme. D476N
andD476Ewerealsotestedwithtestosteroneandnor-
testosterone as substrates (Table 2). The decrease of the
binding affinity of testosterone for mutant D476N was
accompanied by a decrease in the V
m.app
value although
catalytic properties of D476N were unchanged with nor-

testosterone. A lower aromatase activity with these sub-
strates was observed for D476E (13% and 8% of wild-type
for testosterone and nor-testosterone, respectively), con-
trasting with the results obtained with a ndrostenedione.
The relative potency of the three inhibitors tested, a ccord-
ing to their IC
50
values (Table 3), was increased for
inhibition of H475A, which had a lower K
m.app
value for
androstenedione than that of the wild-type aromatase. IC
50
values showed D476N to be more sensitive to MR 20814
and MR 20492 than the wild-type. D476E showed
differences in aromatase activity a ccording to the substrate
used, and responded differently to MR 20814 and MR
20492.
Domains of the active site and substrate specificity
D309 was predicted to be directly involved in decarboxy-
lation and a romatization mechanisms [20], a nd its mutation
to Ala induced an activity loss whether androstenedione or
testosterone was substrate. However, D309A had activity
with nor-testosterone (Table 2) with K
m.app
and V
m.app
values similar to those of the wild-type enzyme. Further-
more, E302A was inactive with all substrates tested.
Human residues implicated in the active site of the

aromatase were mutated to corresponding aligned equine
Table 1. Kinetic parameters of wild-type and mutant forms of P450 aromatase using androstenedione as a substrate. The human residues were
mutated i n different d omains of the protein, s ometimes by their c orresponding alig ned equine r esidues. A: Mutations o f residues conserved in b oth
species. B: Non conservative c hanges. C: Conse rvative changes. C ells were transfec ted by h uman P450arom cD NA and t he aromatase a ctivity was
evaluated by the tritiated water assay using [1b,2b-
3
H]-androstenedione as a substrate, as described in Materials and methods. The aromatase
quantity was evaluated by ELISA in ord er to correct the aromatase activity for t ransfection efficiency. Results a re the mean o f at least three
experiments in triplicate. ND, activity not detectable. NC, activity too low to calculate kinetic parameters (NC1, NC2 and NC3: activity below 1%,
5% and 15%, respectively, relative to wild-type with 200 n
M
of androstenedione corresponding to an activity of 1497 ± 300 pmolÆmin
)1
Æmg
aromatase
)1
).
Protein K
m.app
(n
M
) Wild-type K
m
(%) V
m.app
(pmolÆmin
)1
Æmg
)1
) Wild-type V

m
(%)
Wild-type 164 ± 38 100 ± 23 3041 ± 824 100 ± 27
A E302A NC1 NC1
D309A NC1 NC1
D476A NC1 NC1
D476K NC1 NC1
D476L NC2 NC2
D476N 186 ± 71 113 ± 43 1727 ± 605
a
56 ± 19
D476E 125 ± 50 76 ± 30 2671 ± 657 87 ± 21
B K119T NC1 NC1
K119Y 472 ± 51
a
287 ± 30 3240 ± 1461 106 ± 48
K119V 179 ± 84 109 ± 51 4864 ± 301
b
159 ± 9
K119E 57 ± 7
a
34 ± 4 4064 ± 623 133 ± 20
C124Y 151 ± 47 91 ± 28 1496 ± 302
a
49 ± 9
K130N 147 ± 4 89 ± 2 2473 ± 1465 81 ± 48
F320C 111 ± 20
a
67 ± 12 2127 ± 782 69 ± 25
H475N NC2 NC2

H475R NC2 NC2
H475A 68 ± 16
a
41 ± 10 2388 ± 1518 78 ± 49
H475E NC2 NC2
C I125M ND ND
S470N 290 ± 104
a
176 ± 63 2443 ± 587 80 ± 19
I471M NC3 NC3
I474T 635 ± 251
c
386 ± 152 4107 ± 1583 135 ± 52
a
P < 0.05.
b
P < 0.005.
c
P < 0.001 (
ANOVA
).
Ó FEBS 2002 Structure–function relationships of human aromatase (Eur. J. Biochem. 269) 1397
residues. The mutant K130N had similar catalytic proper-
ties for androstenedione when compared to the wild-type
enzyme (Table 1). K119T also had greatly reduced aroma-
tase activity as was also observed for C124Y (but to a lesser
extent), which had a lower V
m.app
value. F320C increased
the affinity for the substrate. Studies of the nature of the

residue at position K119 showed that K119Y greatly
decreased the binding affinity for androstenedione, whereas
K119E increased it a nd K119V increased the V
m.app
value
for androstenedione. Interestingly, the binding affinity of
K119E for testosterone was unchanged when compared to
the one of wild-type P450arom but the V
m.app
decreased.
Catalytic properties of the mutant K130N for testosterone
and nor-testosterone were differ ent from those observed f or
androstenedione. When compared to t he wild-type enzyme,
this mutant was found to have a h igher b inding affi nity, b ut
alowerV
m.app
value, for testosterone whilst its binding
affinity and the V
m.app
value were lower for nor-testoster-
one. The mutant F320C, weakly active with testosterone,
was found to have a lower V
m.app
value for nor-testosterone
than that of the wild-type P-450arom. K130N and K119E
produced a greater inhibition with MR 20814 or MR 20492,
respectively, whereas F320C and K119V decreased the
inhibition potency of MR 20494 (Table 3).
Furthermore, Table 1 indicates that I125 and I471 could
be d irectly or indirectly implicated in the active s ite s tructure

since their mutations produced weak or inactive proteins.
The K
m.app
values for I474T a nd S470N increased compared
to the wild-type enzyme. Although S470N had a lower
binding affinity for androstenedione, it showed only 2% of
wild-type aromatase activity for testosterone and a slightly
lower V
m.app
value for nor-testosterone. The inhibition
study with these mutants (Table 3) revealed that I474T,
had decreasing affinity for androstenedione and was less
inhibited by the three molecules tested.
DISCUSSION
Based on t he ar omatization c haracteristics and inhibition of
our human mutants, the role o f each residue s tudied may be
discussed taking into account previous knowledge of the
biochemistry of equine aromatase [30–32] and the docu-
mented model of Graham-Lorence et al. [20] together with
results from our molecular model.
The K
m.app
of the recombinant wild-type aromatase for
androstenedione (164 ± 38 n
M
) is in the range of K
m
values
reported in the literature ( 9–150 n
M

) [14,21,24,52–54]. In
our study, the K
m.app
for testosterone is slightly higher than
for androstenedione, as was observed by other groups
Table 2. Kinetic parameters of wild-type and mutant forms of P450 aromatase using testosterone or nor-testosterone as substrates. Cells were
transfected by human P450arom cDNA and the ar omatase activity w as evaluated by radioimmunoassay o f 17b-estradiol, using testosterone or 19
nor-testosterone as substrates, as described in Materials and methods. The aromatase quantity was evaluated by ELISA in order to correct the
aromatase activity for transfection efficiency. Results are the mean of at least three experiments in triplicate, except for D476N and M85V with
testosterone (n ¼ 2). ND: activi ty not detectable; NC: activity too l o w to caculate kinetic parameters (NC1, NC2 and NC3: activity below 1%, 5%
and 15%, respectively, relative to wild-type with 200 n
M
of testosterone corresponding to an activity of 1731 ± 180 pmolÆmin
)1
Æmg arom
)1
.or
with 200 n
M
of nor-testosterone corresponding to an activity o f 2287 ± 477 pmolÆmin
)1
Æmg arom
)1
).
Protein
Testosterone Nor-testosterone
K
m.app
(n
M

) V
m.app
(pmolÆmin
)1
Æmg
)1
) K
m.app
(n
M
) V
m.app
(pmolÆmin
)1
Æmg
)1
)
Wild-type 238 ± 19 4593 ± 613 358 ± 46 4531 ± 887
K119E 256 ± 29 1508 ± 297
a
––
K130N 76 ± 21
a
1610 ± 88
a
791 ± 140
a
835 ± 179
a
E302A ND ND ND ND

D309A NC1 NC1 301 ± 57 4662 ± 2173
F320C NC1 NC1 386 ± 137 2165 ± 698
a
S470N NC2 NC2 338 ± 44 3494 ± 269
a
D476N 480 ± 91 2661 ± 258 414 ± 41 3850 ± 512
D476E NC3 NC3 NC3 NC3
a
P < 0.05 (
ANOVA
).
Table 3. IC
50
values with steroidal and nonsteroidal aromatase inhibi-
tors used with human aromatase mutants. Aromatase ac tivity was
evaluated by measuring the amount of
3
H
2
O released from 200 n
M
[1b,2b-
3
H]-androstenedione incubated in culture medium at 37 °C-5%
CO
2
atmosphere for 45 min in presence of inhibitors. Aromatase ac-
tivities were expressed as percentage of a standard control which was
incubated without inhibitor in t he same con ditions. IC
50

values in l
M
,
were the mean ± SD of three (wild-type) or two (mutants) experiments
in triplicate. IC
50
value with 4OHA, used as control, was
0.45 ± 0.35 l
M
with the wild-type protein.
Protein
IC
50
(l
M
)
MR20814 MR20492 MR20494
Wild-type 10.83 ± 0.80 3.93 ± 0.93 0.23 ± 0.02
K119Y 3.76 ± 1.68
a
5.72 ± 1.10 0.17 ± 0.03
K119V > 10 6.21 ± 2.60 1.55 ± 0.5
a
K119E 9.80 ± 0.42 0.74 ± 0.17
a
0.53 ± 0.45
C124Y 4.40 ± 1.97
a
0.33 ± 0.25
a

0.33 ± 0.18
K130N 0.23 ± 0.16
a
4.27 ± 2.58 0.20 ± 0.13
F320C > 10 5.35 ± 1.12 0.94 ± 0.50
a
I474T > 10 > 10 2.80 ± 0.57
a
H475A 0.92 ± 0.12
a
0.45 ± 0.007
a
0.10 ± 0.00
a
D476N 2.02 ± 1.44
a
0.92 ± 0.12
a
0.15 ± 0.07
D476E 4.68 ± 0.12
a
5.60 ± 0.28
a
0.15 ± 0.07
a
P < 0.05 (
ANOVA
).
1398 P. Auvray et al. (Eur. J. Biochem. 269) Ó FEBS 2002
[14,55,56], whereas other studies have reported similar

values for both substrates [14,35,50]. Furthermore, we
found a 1.5-fold increase in K
m.app
value for nor-testoster-
one when compared to the value obtained for testosterone.
These results are i n accordance with those obtained by
Kellis & Vickery [55] for nor-androstenedione and andros-
tenedione.
Despite the fact that the antibodies used in this study are
polyclonal a nd also recognize t he human aromatase ( Fig. 3)
[29], we cannot exclude that the antibodies bind the human
and the equine enzymes with different affinities. Therefore,
the c omparison of the absolute values of V
m
obtained i n this
study with those reported in the literature is more difficult.
By usin g ELISA, we found aromatase in human placental
microsomes at 210 pmolÆmg proteins
)1
,whichis 2.5-fold
higher than the results reported by Kadohama et al. [51].
Furthermore, the turnover rate observed in our study for
the recombinant wild-type a romatase is of 0.2 min
)1
,which
is 10-fold l ower than reported by Chen et al.[23]inCHO
transfected cells but is in the lower range observed for
human purified aromatase (0.6–35 min
)1
) [14,57,58]. Only

significant differences of V
m.app
between the wild-type and
the mutants will be discussed.
Enzymatic mechanism
The mutation D309A has been already described and the
probable role of D309 is to bring a proton at C(19)
(decarboxylation) and attract a proton at C(2) (aromatiza-
tion), helped by the basic residues H475 or K473 near C(3)
[20]. Moreover, Ahmed [41,59] proposed a mechanism for
the aromatization by ferroxy radical attack on C(19)
(Fig. 4A). According to Graham-Lorence et al. [20], E302,
which was previously thought to interact with the substrate
[60], i s t oo far from C(2). Therefore, these authors suggested
that D309 was a candidate to attract a proton at the C(2)
position. This hypothesis has been checked by mutating
D309 to Ala or Asn. Moreover, Zhou et al. [52] did not
demonstrate any 19-hydr oxy and 19-oxo intermediates with
D309A a nd D309N. These r esults supported D309 acting as
proton donor to T310 during the first ferroxy radical
formation. However, our results showed that D309A was
only active with 19-nor-testosterone, s uggesting that the
aromatization mechanism of nor-androgens was different.
From the model, hydrophobic interactions were observed
between the C(19) of androgens and a hydrophobic surface
formed by the residues F134 and K130 (alkyl chain). The
absence of C(19) for nor-androgens, reducing steric con-
straints, could allow the C(2) of nor-testosterone to come
closer to E302. The aromatization of nor-androgens
(Fig. 4B) could then be different, with a first step corres-

ponding to the ferroxy radical formation, the second step
corresponding to the a ttack of C (1) by t his radical to
hydroxylate this position, the third step corresponding to
the loss of H
2
O and the final step corresponding to the
aromatization o f cycle A, as previously described by A hmed
[59]. T he loss of activity of E302A with nor-testosterone as a
substrate supported this hypothesis. From this proposed
mechanism, a single hydroxylation at C(1) would be
sufficient to allow the aromatization of norandrogens with
a 1-hydroxylated intermediate compound, as was already
suggested by Ganguly et al.[61].
Fig. 4. Enzymatic mechanism of the human
aromatase with androgens (A) and nor-andro-
gens (B). The enzymatic mechanism was
modified from Graham-Lorence [20] and from
Ahmed [41,59].
Ó FEBS 2002 Structure–function relationships of human aromatase (Eur. J. Biochem. 269) 1399
In t heir r ecent s tudy, K ao et al. [ 62] emphasized the
importance of the interaction between E302 and andros-
tenedione based on a decrease in the V
m
value for the
mutant E302D due to a modification of the active site size.
Similarly, our results show that the mutant E302A is not
active with the three substrates tested. These observations
also emphasize the importance of this residue in the active
site structure. However, we propose a slightly different role
for E302 in the structure–function relationships of aroma-

tase than that put forward by Kao et al.[62].
On the other hand, the K473 position within the active
site and o ur kinetic r esults obtained with H475 mutants (see
below), makes the role o f t hese r esidues i n t he enolization of
3-keto group unclear. Our hypothesis is that the driving
elements for the aromatization are the acidic properties of
C2 hydrogens (in a position of the keto group) an d the
electrostatic attractions of D309 towards these hydrogens.
The substrate-binding pocket
Residues toward the keto groups (a-face) of the steroid –
D476. We previously suggested that position 476 could be
important in the active site [11]. We then mutated D476 in
order to understand the role of this residue. Results
showed that position 476, and more partic ularly an acidic
residue, appeared to be important for the aromatase
activity. Like D309, D476 may interact with the C(16)
hydrogens (Fig. 5) in a position of the keto group [C(2)
hydrogens for D309].
This electrostatic interaction, with an acidic group, seems
to be an important part of the stabilization of the steroid
inside the a ctive site (the oretical noncovalent interactions o f
)67 kJÆmol
)1
[43]), as suggested by the results of the
mutations D476A, D476K and D476L leading to inactive
enzyme vs. D476E mutation that did not modify K
m.app
and
V
m.app

values with androstenedione. However, with testos-
terone and nortestosterone, the D476E mutation produced
an inactive compound. In this last case E476, by its longer
lateral chain than D476 and by a straight interaction with
the hydroxyl group (calculated non covalent interaction of
)77 kJÆmol
)1
with
ISOSTAR
software), could destabilize the
position of these ligands inside the a ctive site.
When compared to the wild-type enzyme, D476N had
lower V
m.app
values for C19-androgens but similar K
m.app
and V
m.app
values f or nor-testosterone. These observations
suggest that this mutation produces a modification of the
interaction between C(2) and D309 during aromatization of
the C19-androgens. D4 76 and D 309 line the same face of the
active site and the repulsive force (electrostatic interactions)
between these two acidic residues appears to b e important
for efficient aromatization of C19-androgens. This is also
supported by results obtained with D476A, D476K and
D476L. Moreover, because E302 is located on the opposite
face of the active s ite, aromatization of nor-testosterone was
unchanged. All these results suggest that D476 protrudes
into the active site and may interact with the substrate.

For MR20814, we have previously shown [11] the
existence of a coordination bond between the pyridine group
and iron (Soret band). Because of t his, we have suggested the
following position of the ligand inside t he active site (Figs 6
and 7) corresponding to an orientation of the amino group
towards t he extrahydrophobic surface composed of residues
I474 and L477 [21], an interaction between F134 and the
pyridine group (T stacking), an orientation of t he keto group
towards K130 and the methoxy group positioned in the
polar area. This polar area is formed by H128, Q218
[hydrogen bond to keto group of androstenedione (C17)],
Q225 (electrostatic interactions with D476), H475 and D476.
It has b een previously shown t hat mutation at position Q 225
modified the K
m.app
of the enzyme [24].
The impact of mutations on the IC
50
value is consistent
with the hypothesis. Indeed the mutations D476N, D476A
and H475A not only d ecrease the polar ch aracteristic of the
region called Ôpolar areaÕ that support the interaction with
the methoxy group of the inhibitor for MR20814, but also
the hydrophobic interaction with the pyrrole group for the
two other inhibitors (MR20492, MR20494). The D476E
mutation could stabilize the position of the ligand by an
electrostatic interaction with t he amino g roup of MR 20814.
The K130N mutation avoided the electrostatic repulsion
between the amino group of MR20814 and K130, allowing
a better position of the ligand inside t he active site

(conformational flexibility).
Fig. 5. Relative position o f residues toward androstenedione i n the active
site. D476 and D309 may sta b ilize androstenedione by interacting with
the C(16) a nd C(2) hydrogens, r espectively. K130 by its long aliphatic
lateral chain will reinforce h ydrophobic interactions over the b face o f
the steroid. The quadropole/quadropole interaction of the amino
group of K130 with the aromatic residue F134 co ntributes to steric
constraints over the C(19) methyl.
1400 P. Auvray et al. (Eur. J. Biochem. 269) Ó FEBS 2002
H475. M utation of H 475 of hP-450arom t o Asn lowered t he
activity by 95%, suggesting that this residue is important at
this position. Graham-Lorence et al. [20] suggested that
H475, or K 473, could f acilitate t he enolization o f t he 3-keto
group of the substrate. In order to define the role of this
residue, we transfected the mutants H475R, H475E and
H475A. According to our molecular model, H475 is closer to
O(3) of the substrate, compared to K473 but H475 seemed to
be an unlikely candidate for proton donation to the substrate
3-keto group since H475R was s lightly active, whereas
H475A remained f ully active. However, a hydrophobic
residue seemed to be crucial at this position as observed with
H475N, H475E and H475R, which were weakly active.
H475, located in a polar area and close to the hydrophobic
cluster which begins and ends with I474 and L477, by its
hydrophobic characteristic would then have a role in the
stabilization of the substrate in the active site. Indeed, only
H475A was active and had a 2.5-fold increased binding
affinity for androstenedione, suggesting a modification of
the hydrophobic properties of the cluster. This reinforces
also the inhibitory potencies of the three molecules tested.

I474. This position was extensively studied [21,53] in order
to evaluate the extra hydrophobic pocket of the enzyme.
According to Kao et al. [21], I474 would be on an
extrahydrophobic s urface r ather than in a hydrophobic
pocket, as described by Graham-Lorence et al .[20],based
on the f act t hat mutation o f I 474 t o bulky and hydrophobic
aromatic residues (I474Y, I474W) maintained the potency
of the 7a-APTA inhibitor. Similar results were obtained by
Zhou et al. [53] with mutant I474F. I n our st udy, the
mutant I474T presented a weak binding affinity for
androstenedione, but a V
m.app
value similar to that of the
wild-type aromatase, and it was less inhibited by the
nonsteroidal inhibitors known previously to interact with
the hydrophobic s urface [11–13]. Presence of a polar re sidue
on this hydrophobic surface of the active site could reduce
desolvation effects leading to a decreasing binding of the
substrates and inhibitors (MR 20814, MR 20492 and MR
20494). T his hypothesis is supported by the observation t hat
a previously described mutant I474N showed decreasing
binding affinity for androstenedione whereas mutations
with hydrophobic residues (I474F, I474W, Y474Y, I474M)
did not [21,53]. Furthermore, r esults obtained for th e I474T
mutant, with those obtained for H475N, suggest that this
face of the active site has a slightly reduced hydrophobic
interaction with t he substrate for the e quine enzyme
compared with the human counterpart. This could be also
the case for the porcine isoforms (K474 and N475).
F320 and S470. These residues are not located in the

substrate-binding site but their mutations could indirectly
influence its structure. With regard to the F320 residue,
positioned in the helix I, the results indicated that its
mutation to Cys had weakly increased binding affinity for
androstenedione and decreased V
m.app
value for n or-
testosterone. The F320 residue is bulky and seems to be
close t o S470 (4 .5 A
˚
). The F320C mutation could lead t o the
creation of a disulfide bond between C320 and C 467 ( 3.5 A
˚
)
as was previously proposed [11]. This new bond may t hen
modify the loop structure S470–S478 leading to a modifi-
cation of the p olar area (H475 and D476) and t he
hydrophobic surface (L477, I474) explaining t he impact
on the affinity of the substrates, particularly testosterone,
and of the inhibitors by disfavoring the interaction of the
pyridyl group with the heme. This could be explained by
the impact of the position of an acidic group towards the
affinity of testosterone (D476E and this result). T hese results
also confirm the different p osition of testosterone and nor-
testosterone in the active site (see Discussion on enzymatic
mechanism). S470N (polar residue) m utation s howed also a
variation in K
m.app
value for a ndrostenedione and a decrease
Fig. 7. View of th e h ydrophobic surface (

MOLCAD SURFACE
). The
definition of the lipophilicity potential made it possible to define
the hydrophobic surface (brown), the neutral surface (green) and the
hydrophilic surface (blue). T he position of MR20814 inside the a ctive
site corresponds to an orientation of its amino group toward the
hydrophobic surface, previously described as the extrahydrophobic
surface (21), compared of residue I474 and L477.
Fig. 6. Hypothesis on the position of MR20814 inside the active site.
The position of MR20814 inside the active site correspond to: an
orientation of the amino group toward the hydrophobic area com-
posed of residues I474 and L477, an interaction between the pyridine
group and F134, an orientation of the keto group toward K130 and
the methoxy group positioned in the polar area. H128, Q218, Q225,
H475 and D476 form this polar area.
Ó FEBS 2002 Structure–function relationships of human aromatase (Eur. J. Biochem. 269) 1401
in V
m.app
value f or no r-testosterone. T hese effects, similar to
those observed for F320C, led to the hypothesis that this
mutation has an impact on the loop structure.
Residues toward the methyl groups (b-face) of the steroid
– K130. According to the theoretical model of Graham-
Loren ce et al. [20], K130 could be involved in a salt-bridge
with E302, this possibly aiding reprotonation of D309 and
T310. Furthermore, if the amino group of K130 pairs with
E302, its long aliphatic lateral chain will reinforce hydro-
phobic i nteractions, c ontrolled a lso b y F134, with the b-face
of the steroid. The role of K130 in reprotonation was not
confirmed, as the mutation of K 130 to Asn had no effect on

the catalytic properties of the enzyme with androstenedione
as a substrate. According to our molecular model, repro-
tonation of D309 could be inferred from the conserved
residue K473, which is closer to D309 than to the 3-keto
group of the substrate. Another possibility is that t he amino
group of K130 strongly interacts with the aromatic residue
F134 (quadrupole–quadrupole interactions) rather than
with E302, and this complex contributes to steric cons traints
over the C(19) meth yl group of androgens. However,
K130N showed 70% decrease in the K
m.app
value using
testosterone as a substrate, but th ere was a t wofold increase
of the K
m.app
value with nor-testosterone as a substrate, and
the V
m.app
was 36% and 18% that of the wild-type for
testosterone and nor-testosterone, r espectively. These e ffects
on substrate binding could be e xplained by taking into
account the roles of several residues: (a) the stabilizing role
of the polar area towards the polar group on C17 (keto or
hydroxyl group); (b) the impact of an acidic group on the
17-hydroxyl of testosterone and nor-testosterone (D476E);
(c) the steric constraints of the K130-F134 complex to the
C(19) methyl of testosterone; and (d) the role of D309 and
of E302 in aromatization of androgens and nor-androgens,
respectively. Thus, N130, located at the opposite face of
D476, could create a second polar area (Fig. 8) formed by

H128 and N130, leading to new electrostatic interactions
with the 17-hydroxyl group of testosterone and of nor-
testosterone. Consequently, the holding of the C19 methyl
near the heme iron could be affected, which would provide
an explanation to the lower K
m.app
and V
m.app
values
observed for testosterone. Contrasting with the results
obtained with testosterone, the K130N mutation largely
decreased the binding affinity for nor-testosterone, suggest-
ing that this residue, along with F134, may have a more
important role on substrate binding of nor-androgens as
suggested previously. The 80% drop of the V
m.app
value for
nor-testosterone may result from the wrong holding of ring
A near the heme iron and near E302.
Residues toward the D-ring of the steroid – K119.
Laughton et al. [25] suggested that this residue would
interact with the h eme c arboxylate group wher eas Graham-
Loren ce et al. [20] indicated that K119 was on t he external
surface o f the protein and orientated outwards. In our
molecular model, K119 was also pointing towards the
solvent and located in a short b sheet between two glycines
involved in loop structures (GS KLG). External ionic
residues, such as glutamic acid, could m aintain this
structure, whereas hydroxylated residues, such as threonine
and tyrosine, will modify it. The other possibility is that

K119 is not outside the protein surface but rather inside, as
our mutants modified the binding affinity for androsten-
edione and the inhibitory potential of inhibitors. In this
case, by introducing a hydroxyl group, accessibility of the
substrate to the active site is reduced, as observed with the
mutants K119T and K119Y. By keeping an ionic residue,
the binding affinity for androstenedione and the inhibitory
potential of MR 20492 are increased as observed for the
K119E mutant. This residue appears to be significant and
may be involved indirectly in the structure of the L122–
H128 domain interacting with the
D
-ring of the steroid as
suggested by the catalytic differences of K119E observed
with androstenedione or testosterone as substrates.
I125. According to our molecular model, this residue is in
the L122–H128 domain of the active site, which has been
implicated in an extrahydrophobic pocket surrounding the
D
-ring of the steroid and the polar area (H128). Interest-
ingly, I125M, which was not susceptible to modify the
hydrophobic surface, had no activity. These results may be
explained by the possibility of a strong interaction of the
methionine with the aromatic residue F
221
(3 A
˚
). This
interaction could increase the steric constraints over the
D

-ring of the steroid and could be involved in the lower
efficiency of the 16a-hydrotestosterone aromatization by the
equine enzyme than by the human one [63]. Differences
between the two enzymes was also apparent in the C124Y
mutant suggesting that an increase o f t he hydrophobic
character of this part of the human active site modified the
aromatization efficiency of androstenedione but increased
the inhibition potency of MR 20814 and MR 20492.
CONCLUSIONS
In this study, we explored the role of residues within the
active site of the human more specifically those involved in
substrate i nteractions and c atalysis. O ur results indicate that
some mutants show variable activity with androstenedione,
testosterone or nor-testosterone. These results suggested
that the residues involved in substrate stabilization may
vary, or had differing i mportance d epending on the
Fig. 8. Representation of the pos ition of this new polar area toward s the
last po lar area. N13 0 located at the opposite f ace of D476 could c reate
a second polar area, with H128, leading to new electrostatic interac-
tions with the C(17) group of the ste roid.
1402 P. Auvray et al. (Eur. J. Biochem. 269) Ó FEBS 2002
substrate. For instance, we highlight the importance of the
L122–H128 domain (results obtained with the K119, I124,
I125 mutants) of the loop structure corresponding to S470–
S478 (results obtained with the S470, I471, I474, H475,
D476 mutants) in the stabilization of the steroid (D-ring)
and particularly the role of D476. Positioning of the steroid
inside the active site will differ according to its C(17) group
and m odification of the polarity of these two domains
(L122–H128 and S470–S478) in different species. Further-

more, residues K130 and F134 appear to form a hydro-
phobic surface close t o t he heme and control the position of
the substrate inside the active site (differences for nor-
androgens and androgens). These results are the first report
using nor-testosterone a s substrate for aromatase mutants.
They suggest that E302 interacts with C (2) of nor-androgens
and is implicated in their aromatization, whereas decar-
boxylation and aromatization of androgens involve D309.
Most of the substitutions of human residues by their
corresponding aligned equine residues, conservatives or not,
produced differences in the catalytic properties of th e
enzyme suggesting that incompletely conserved residues
may b e i nvolved in biochemical specificities between species,
including differences in inhibition. This is also supported by
the recent work of Corbin et al. [27] and Kao et al. [28],
which demonstrate catalytic differences between porcine
isoforms of aromatase. The main differences, which could
account for differences in biochemical specificities between
equine and human enzymes, are residues N475, N130 and
M125. Such comparative studies of aromatase across
species has shed new light on the aromatization mechanism
and the structure of the active site.
ACKNOWLEDGEMENTS
This work was i nitiated and developed t hrough a grant from t he Ligue
Nationale Contre Le Cancer (Comite
´
de la Manche et du Calvados) for
the biological s tudies, and a studentship to P. A., a nd throu gh a grant
from the Region de Basse-Normandie and, FEDER (Comite
´

Agrobio).
We would also like to thank Prof. W.R. Miller (Edinburgh, UK) and
C. Joubel for their c areful perusal of the manuscript and Dr S. Moslemi
for helpful discussions on the equine aromatase .
REFERENCES
1. Simpson, E.R., Zhao, Y., Agarwal, V.R., Michael, M.D., Bulun,
S.E., Hinshelwood, M.M., Graham-Loren ce, S., Sun, T., Fisher,
C.R.,Qin,K.&Mendelson,C.R.(1997)Aromataseexpressionin
health and disease. Recent Prog. Horm. Res. 52, 185–213.
2. Reed, M.J. ( 1994) The role of a romatase in breast tumors. Breast
Cancer Res. Treat. 30, 7–17.
3. Miller, W.R. (1997) Aromatase inhibitors and breast cancer.
Cancer Treat. Rev. 23, 171–187.
4. Griffiths, C.T., Hall, T.C., Saba, Z., Barlow, J.J . & N evinny, H.B.
(1973) Preliminary trial of aminoglutethimide in breast cancer.
Cancer 32, 31–37.
5. Le Borgne, M., Marchand, P., Duflos, M., Delevoye-Seiller, B.,
Piessard-Robert, S., Le Baut, G., Hartmann, R.W. & Palzer, M.
(1997) Synthesis and in v i tr o evaluation of 3-(1-azo lylmethyl)-
1H-indoles and 3-(1-azo lyl-1-phenylmethyl)-1H-indoles as inhi-
bitors of P450 arom. Arch. Pharm. 330, 141–145.
6. Numazawa, M., Oshibe, M. & Yamaguchi, S. (1997) 6-Alkylan-
drosta-4,6-diene-3,17-diones and their 1,4,6-triene analogs as
aromatase i nhibitors. Structure–activity r elationships. Ste roids 62,
595–602.
7. Okada, M., Yoden, T., K awaminami, E., Shimada, Y., Kudoh,
M., Isomura, Y., Shikama, H. & Fujikura, T. (1996) Studies on
aromatase inhibitors. I. Synthesis and biological evaluation of
4-amino-4H-1,2,4-triazole derivatives. Chem. Pharm. Bull. 44,
1871–1879.

8. Okada, M., Yoden, T., K awaminami, E., Shimada, Y., Kudoh,
M. & Isomura, Y. (1997) Studies on aromatase inhibitors. IV.
Synthesis and biological evaluation of N,N-disubstituted-5-ami-
nopyrimidine derivatives. Chem.Pharm.Bull.45, 1293–1299.
9. Okada, M., Yoden, T., K awaminami, E., Shimada, Y., Kudoh,
M. & Isomura, Y. (1997) Studies on aromatase inhibitors. III.
Synthesis and biological evaluation of [(4-bromobenzyl)
(4-cyanophenyl) amino] azoles and their azine analogs. Chem.
Pharm. Bull. 45, 482–486.
10. Okada, M., Yoden, T., Kawaminami, E., Shimada, Y., Kudoh,
M. & Isomura, Y. (1997) Studies on aromatase inhibitors. II.
Synthesis and biological evaluation of 1-amino-1H-1,2,4-triazole
derivatives. Chem.Pharm.Bull.45, 333–337.
11. Auvray, P., Moslemi, S ., So urdaine, P., Se
´
ralini, G E., Engue-
hard, C., Dallemagne, P., Sonnet, P., Bureau, R. & Rault, S.
(1998) Evidence for a new non-steroidal human aromatase
inhibitors and comparison with equine aromatase inhibition for an
understanding of mammalian active site. Eur. J. Med. Chem. 33,
451–462.
12. Sonnet, P., Auvray, P., Guillon, J., Dallemagne, P., Bureau, R.,
Rault, S., M oslemi, S., Sourdaine, P., G alopin, S. & Se
´
ralini, G.E.
(1998) New aromatase inhibitors. Synthesis and bi ological activity
of aryl-substituted pyrrolizine a nd indolizine derivatives. Bioorg.
Med. Chem. Lett. 8, 1041–1044.
13. Auvray, P., Sourdaine, P., Moslemi, S., Se
´

ralini, G. E., Sonnet, P.,
Dallemagne, P., Bureau, R. & Rault, S. (1999) MR 20492 and MR
20494: two indolizinone derivatives that strongly inhibit human
aromatase. J. Steroid Biochem. Mol. Biol. 70, 59–71.
14. Yoshida, N. & Osawa, Y. (1991) Purification o f human plac ental
aromatase cytochrome P-450 with monoclonal antibody and its
characterization. Biochemistry 30, 3003–3010.
15. Poulos,T.L.,Finzel,B.C.&Howard,A.J.(1987)High-resolution
crystal structure of cytochrome P450cam. J. Mol. Biol. 195, 687–
700.
16. Ravichandran, K.G., Boddupalli, S.S ., Hasemann, C.A., Peter-
son, J.A. & Deisenhofer , J. (1993) Crystal structure of hemopro-
tein domain of P450BM-3, a prototype for microsomal P450s.
Science 261, 731–736.
17. Hasemann, C.A., Ravichandran, K.G., Peterson, J.A. &
Deisenhofer, J. (1994) Crystal structure and refinement of cyto-
chrome P450terp at 2.3 A
˚
resolution. J. Mol. Biol. 236, 1169–1185.
18. Cupp-Vickery, J.R. & Poulos, T.L. (1995) Structure of cyto-
chrome P450eryF involved in erythromycin biosynthesis. Nat.
Struct. Biol. 2, 144–153.
19. Cupp-Vickery, J.R. & Poulos, T.L. (1997) Structure of cyto-
chrome P450eryF: substrate, inhibitors, and model compounds
boundintheactivesite.Steroids 62, 112–116.
20. Graham-Lorenc e, S., Amarne h, B., White, R.E., P eterson, J.A. &
Simpson, E.R. (1995) A three-dimensional model of aromatase
cytochrome P450. Protein Sci. 4, 1065–1080.
21. Kao, Y.C., Cam, L.L., Laughton, C.A., Zhou, D. & Chen, S.
(1996) Binding ch aracte ristics o f seve n in hibitors o f hu man

aromatase: a site-directed mutagenesis s tudy. Cancer Res. 56,
3451–3460.
22. Chen, S. & Zhou, D. (1992) Functional domains of aromatase
cytochrome P450 inferred from comparative analyses of amino
acid sequences and substantiated by site-directed mutagenesis
experiments. J. Biol. Chem. 267, 22587–22594.
23. Chen, S., Zhou, D., Swiderek, K.M., Kadohama, N., Osawa, Y. &
Hall, P.F. (1993) Effect of androstenedione on growth of
untransfected and aromatase-transfected MCF-7 cells in culture.
J. Steroid Biochem. Mol. Biol. 44, 347–356.
Ó FEBS 2002 Structure–function relationships of human aromatase (Eur. J. Biochem. 269) 1403
24. Amarneh, B., Corbin , C.J., Peterson, J.A., Simpson, E.R. &
Graham-Lorence, S. (1993) Functional domains of human
aromatase cytochrome P450 characterized by linear alignment and
site-directed mutagenesis. Mol. Endocrinol. 7, 1617–1624.
25. Laughton, C.A., Zvelebil, M.J. & Neidle, S. (1993) A detailed
molecular model f or human a romatase. J. Steroid Biochem. Mol.
Biol. 44, 399–407.
26. Graddy, L.G., Kowalski, A.A., Simmen, F.A., Davis, S.L.F.,
Baumgartner, W.W. & Simmen, R.C.M. (2000 ) Multiple isoforms
of porcine aromatase are encoded by three distinct genes.
J. Steroid Biochem. Mol. Biol. 73, 49–57.
27. Corbin, C.J., Trant, J.M., Walters, K.W. & Conley, A.J. (1999)
Changes i n testosterone metabolism associated with the evolution
of placental and gonadal isozymes of porcine aromatase cyto-
chrome P450. Endocrinology 140, 5202–5210.
28. Kao, Y.C., Higashiyama, T., S un, X., Okudo, T., Yab orough, C.,
Choi, I., Osawa, Y., Simmen, F.A. & Chen, S. (2000) Catalytic
differences between porcine blastocyst and placental aromatase
isozy me s. Eur. J. Biochem. 267, 6134–6139.

29. Almadhidi, J., Se
´
ralini, G.E., Fresnel, J., Silberzahn, P. & Gail-
lard, J.L. (1995) Immunohistochemical localization of cytochrome
P450 aromatase in equine gonads. J. Histochem. Cytochem. 43,
571–577.
30. Almadhidi, J., Moslemi, S., Drosdowsky, M.A. & Se
´
ralini, G.E.
(1996) Equine cytochrome P450 aromatase exhibits an estrogen
2-hydroxylase a ctivity in vitro . J. Steroid B iochem. Molec. B iol. 59,
55–61.
31. Moslemi, S. & Se
´
ralini, G. E. (1997) Inhibition and inactivation o f
equine aromatase by steroidal and non-steroidal co mpounds. A
comparison with human aromatase inhibition. J. Enz. Inhib. 12,
241–254.
32. Moslemi, S., Auvray, P., Sourdaine, P., Drosdowsky, M.A. &
Se
´
ralini, G.E. (1997) Structu re–function relationships for equine
and human a romatases. A comparative study. Ann. NY Acad. Sc i.
839, 576–577.
33. Tomilin, A., A uvray, P., Moslemi, S., So urdaine, P., Drosdowsky,
M.A. & S e
´
ralini, G.E. (1996) Molecular cloning and character-
ization of equine cytochrome P450 aromatase. Abstract no. A22.
IV International Aromatase Conference, Tahoe-City, California.

34. Dintinger, T., Gaillard, J.L., Mos lemi, S., Zwain, I. & Silberzahn,
P. (1989) Androgen and 1 9-n orandrogen aromatization by equine
and human placental microsomes. J. Steroid Biochem. 33,949–
954.
35. Corbin, C.J., Graham-Lorence, S., McPhaul, M., Mason, J.I.,
Mendelson, C.R. & Simpson, E.R. (1988) Isolation of a full-length
cDNA insert encoding human aromatase system cytochrome
P-450 and its expression in nonsteroidogenic cells. Proc. Natl
Acad. Sci. USA 85, 8948–8952.
36. Pearlman, D.A., Case, D.A., Caldwell, J.C., Seobe, G.L., Chandra
Singh, U ., Weiner, P. & Kollman, P.A. (1991) AMBER 4.0 .
University of California, San Francisco, CA, USA.
37. O’Hare, D., Barlow, S . & Rohl, A.L. ( 1996) Molecular m echanics
study of oligomeric models for poly (ferrocenylsilanes) using the
extensible systematic forcefield (ESFF). J. Am. Chem. Soc. 118,
7578–7592.
38. Asensio, J .L., Martin-Pastor, M. & J imenz-Barbero, J. ( 1997) The
use of the MM3 and ESFF f orce fields i n conformational analysis
of carbohydrate molecules in solution. The methyl alpha-lactoside
case. J. Mol. S truc t. 395/396, 245–270.
39. Morris, A.L., MacArthur, M.W., Hutchinson, E .G. & Thornton,
J.M. (1992) Stereochem ical quality o f protein structure
coordinate. Proteins 12, 345–364.
40. Sippl, M.J. (1993) Recognition of errors in three-dimensional
structures of proteins. Proteins 17, 355–362.
41. Ahmed, S. (1998) Review of the molecular modelling studies of the
cytochrome P-450 estrogen synthetase enzyme, aromatase. Drug
Design Disc ov. 15, 239–252.
42. Busetta, B., Com berton, G., Course ille, C. & Hospital, M. (1972)
Cryst. Struct. Commun. 1, 129.

43. Bruno, I.J., Cole, J.C., Lommerse, J.P.M., Rowland, R.S., Taylor,
R. & Verdonk, M.L. (1997) IsoStar: a library of information
about nonbonded interactions. J. Computer-Aided Mol Design 11,
525–537.
44. Ghose, A.K., Viswanadhan, V.N. & Wendoloski, J.J. (1998)
Prediction of hydrophobic (li pophilic) p roperties o f s mall organic
molecules usi ng fr agment al me thods : an analysis of ALO GP a nd
CLOGP methods. J. Phys. Chem. A102, 3762–3772.
45. Audry, E., Dubost, J.P., Colleter, J.C. & Dallet, P. (1986) Une
nouvelle approche des re latio ns structu re–activite
´
:leÔpotentiel de
lipophilie mole
´
culaireÕ. Eur. J. Med. Chem Chim. T her. 21, 71– 72.
46.Boussif,O.,Lezoualc’h,F.,Zanta,M.A.,Mergny,M.D.,
Scherman,D.,Demeneix,B.&Behr,J.P.(1995)Aversatile
vector for gene and oligonucleotide transfer into cells in culture
and in vivo: polyethylenimine. Proc. Natl Acad. Sci. USA 92, 7297–
7301.
47. Zhou, D .J., Pompon, D. & Chen, S.A. ( 1990) Stable e xpression of
human aromatase complementary DNA in mammalian cells: a
useful system for aromatase inhibitor s creening. Cancer Res. 50,
6949–6954.
48. Terqui,M.,Dray,F.&Cotta,J.(1973)Variationsof17beta-
estradiol conc entration in peripheral sheep blo od in the course o f
the estrus cycle. C. R. Acad. Sci. 277, 1795–1798.
49. Garnier,D.H.,Cotta,Y.&Terqui,M.(1978)Androgenradio-
immunoassay in the ra m: result of direct plasma testosterone and
dehydroepiandrosterone measurement and physioogical evalua-

tion. Ann. Biol. Anim. 18, 265–281.
50. Bradford, M.M. (1976) A rapid and sensitive method for the
quantitation of microgram quantities of protein utilizing the
principle of protein-dye binding. Anal. Biochem. 72, 248–254.
51. Kadohama, N., Zhou, D., Chen, S. & Osawa, Y. (1 993) Catalytic
efficiency of expressed aromatase following site-directed muta-
genesis. Biochim. Biophys. Acta. 1163, 195–200.
52. Zhou,D.J.,Korzekwa,K.R.,Poulos,T.&Chen,S.A.(1992)A
site-directed mutagenesis study of human placental aromatase.
J. Biol. Chem. 267, 762–768.
53. Zhou, D., Cam, L. L., L aughton, C.A., Korzekwa, K.R. & Chen,
S. (1994) Mutagenesis study at a postulated hydroph obic region
neartheactivesiteofaromatasecytochromeP450.J. Biol. Chem.
269, 19501–19508.
54. Purohit, A. & Oakey, R.E. (1989) Evidence for separate sites for
aromatisation of androstenedione and 16 alpha-hydroxyandro-
stenedione in human placental microsomes. J. Steroid Biochem.
133, 439–448.
55. . Kellis, J.T. & Vickery, L.E. (1987) The active site of aromatase
cytochrome P-450. Differential e ffects of c yanide provide evidence
for proximity of heme-iron a nd carbon-19 in the enzyme-substrate
complex. J. Biol. Chem. 262, 8840–8844.
56. Watanabe,S.,Tohma,Y.,Chiba,H.,Shimizu,Y.,Saito,H.&
Yanaihara, T. (1994) Biochemical significance of 19-hydro-
xytestosterone in the process of aromatization in human corpus
luteu m. Endocrinol. J. 41, 421–427.
57. Sethumadhavan, K., Bellino, F.L. & Thotakura, N.R. (1991)
Estrogen synthetase (aromatase). The cytochrome P-450 compo-
nent of the human p lacen tal enzyme i s a glycoprotein. Mol. Cell.
Endocrinol. 78, 25–32.

58. Hagerman, D.D. (1987) Human placenta estrogen synthetase
(aromatase) purified by affinity chromatography. J. Biol. Chem.
262, 2398–2400.
59. Ahmed, S. (1997) The m echanism of a P-450 enzyme-aromatase; a
molecular modelling perspective for the removal of the C(19)
methyl and aromatisation o f the s teroid A r ing. J. Enz. Inhib. 12,
59–70.
60. Graham-Lorence, S., Khalil, M.W., Lorence, M.C., Mendelson,
C.R. & Simpson, E.R. (1991) Structure–function relationships
1404 P. Auvray et al. (Eur. J. Biochem. 269) Ó FEBS 2002
of human aromatase cytochrome P-450 using molecular mode-
ling and site-directed mutagenesis. J. Biol. Chem. 266, 11939–
11946.
61. Ganguly, M., Cheo, K.L. & Brodie, H.J. (1976) Estrogen bio-
synthesis and 1beta-hydroxylation using C19 and 19-nor steroid
precursors. Biochim. Biophys. Acta 43, 326–334.
62. Kao, Y.C., Korzekwa, K.R., Laughton & Chen, C.A., S.
(2001) Evaluation of the mechanism of aromatase cytochrome
P450. A site-directed mutagenesis study. Eur. J. Biochem. 268,
243–251.
63. Gaillard, J.L. (1991) Equine test icular aromatase: substrates
specificity and kinetic cheracteristics. Comp. Biochem. Physiol.
100B, 107–115.
64. Brodie, A. (1991) Aromatase and its inhibitors – an overview.
J. Steroid Biochem. Mol. Biol. 40, 255–261.
Ó FEBS 2002 Structure–function relationships of human aromatase (Eur. J. Biochem. 269) 1405