THE JOURNAL OF BIOLOGICAL CHEMISTRY
4183–4190, 2002
© 2002 by The American Society for Biochemistry and Molecular Biology, Inc.
Structural Insights into Steroid Hormone Binding
THE CRYSTAL STRUCTURE OF A RECOMBINANT ANTI-TESTOSTERONE Fab FRAGMENT IN FREE AND
TESTOSTERONE-BOUND FORMS*
Received for publication, June 18, 2001, and in revised form, November 12, 2001
Published, JBC Papers in Press, November 13, 2001, DOI 10.1074/jbc.M105579200
Jarkko Valjakka‡§, Kristiina Takkinenz¶, Tuija Teerinen¶, Hans Söderlund¶,
and Juha Rouvinen‡
From the ‡Department of Chemistry, University of Joensuu, P. O. Box 111, 80101 Joensuu and ¶VTT Biotechnology,
P. O. Box 1500, 02044 VTT, Finland
The monoclonal anti-testosterone antibody (3-C4F5)
has a relatively high affinity (3 ⴛ 108 Mⴚ1) with an overall
good specificity profile. However, the earlier character-
ized binding properties have shown that both the affin-
ity and specificity of this antibody must be improved if it
is intended for use in clinical immunoassays. In this
paper, the crystal structures of the recombinant anti-
testosterone (3-C4F5) Fab fragment have been deter-
mined in the testosterone-bound and free form at reso-
lutions of 2.60 and 2.72 Å, respectively. The high affinity
binding of the (3-C4F5) Fab is mainly determined by
shape complementarity between the protein and testos-
terone. Only one direct hydrogen bond is formed be-
tween the hydroxyl group of the testosterone D-ring and
the main-chain oxygen of Gly100JH. The testosterone is
deeply bound in a hydrophobic pocket, and the close
shape complementarity is mainly formed by the third
complementarity-determining regions (CDR) of the
heavy and light chain. Comparison of the bound struc-
ture with the free structure indicates conformational
changes in the protein upon testosterone binding. The
conformational changes of the side chains of two resi-
dues Glu95H and Tyr99H in the CDR-H3 are particularly
essential for the binding. Interesting similarities in the
binding of different steroids were also observed upon
comparison of the available structures of anti-steroid
antibodies.
The number of different, closely related steroid hormones in
human serum is high, and their relative concentrations, which
usually are very low, can vary greatly even between normal
healthy individuals. Testosterone, the main male sex hormone,
is one of the structurally similar steroid hormones. Measure-
ment of serum testosterone levels is important in the evalua-
tion of conditions associated with hyperandrogenism in women
(1, 2) and to diagnose hypogonadism, impotence, and problems
in spermatogenesis and pubertical development in men (3, 4).
Development of monoclonal antibodies that would fulfill the
high affinity and selectivity requirements needed for the accu-
rate measurement of testosterone levels in serum samples has
* The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby marked
“advertisement” in accordance with 18 U.S.C. Section 1734 solely to
indicate this fact.
The atomic coordinates and structure factors (code 1I9I and 1I9J)
have been deposited in the Protein Data Bank, Research Collaboratory
for Structural Bioinformatics, Rutgers University, New Brunswick, NJ
(http://www.rcsb.org/).
§ To whom correspondence should be addressed: Tel.: 358 –13-251–
2034; Fax: 358 –13-251–3390; E-mail: jarkko.valjakka@joensuu.fi.
This paper is available on line at http://www.jbc.org
Vol. 277, No. 6, Issue of February 8, pp.
Printed in U.S.A.
not been successful with the use of hybridoma technology.
Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007
Rabbit polyclonal antibodies are used in most current immu-
noassays of steroid hormones. However, the supply of poly-
clonal reagents with consistent quality is a severe problem for
the immunodiagnostic industry and requires continuous immu-
nization of many animals. The inherent batch-to-batch varia-
tion of the polyclonal sera makes it necessary to optimize the
assay parameters for each new serum batch.
The three-dimensional structures of steroid binding Fab
fragments are important for a general understanding of the
molecular basis of specific interactions between antibodies
and hydrophobic molecules such as steroids. The molecular
basis of steroid antibody interactions has been studied in
detail by determining the structures of a progesterone-
binding antibody (DB3) in its free and bound form and with
different high affinity cross-reactive progesterone derivatives
(5–7). The binding site of the DB3 antibody is a hydrophobic
pocket, and the binding affinity and specificity are deter-
mined by van der Waals interactions and a few hydrogen
bonds formed with the progesterone hapten (5). The high
affinity binding mode of a number of progesterone derivatives
is due to two alternate docking orientations for the steroid
skeleton (6). In the cases of the anti-digoxin 26-10 and 40-50
antibodies, the high affinity binding mode is mainly deter-
mined by hydrophobic shape complementarity (8 –10). The
most striking differences in these two anti-digoxin Fab struc-
tures is the presence of hydrogen bonds between the Fab
40-50 and ouabain hapten and their absence in the Fab 26-10
complex structure with digoxin. The binding site of the 40-50
antibody is a groove on the protein surface, whereas in the
26 –10 antibody the binding site is a deep cleft (8 –10). The
basis of the binding specificity of steroid hormone binding
antibodies has also been elucidated by determining the crys-
tal structures of the Fv4155 fragment in complex with two
closely related steroid hormone glucoronides (11). The
Fv4155 binds principal urinary metabolites of estradiol, es-
trone ␤-D-glucuronide (E3G)1 and estriol 3-␤-D-glucuronide
(EI3G) with different affinities. The binding affinity of E3G is
about ten times higher than that of EI3G, although EI3G
differs from E3G only in the addition of a hydroxyl group and
the reduction of an adjacent carbonyl at positions 16 and 17
of the steroid D-ring, respectively (11). Both haptens bind in
similar orientations into the long, deep hydrophobic cleft.
The monoclonal antibody (3-C4F5) has been produced by
hyperimmunizing mice with a testosterone-3-carboxymeth-
yloxime-thyroglobulin conjugate, and genes encoding its Fab
1
The abbreviations used are: E3G, estrone ␤-D-glucuronide; EI3G,
estriol 3-␤-D-glucuronide; PEG, polyethylene glycol; CDR, complemen-
tarity-determining regions; DHEAS, dehydroepiandrosterone sulfate.
4183
4184 Crystal Structures of Recombinant Anti-testosterone Fabs
TABLE I
Data collection and refinement statistics
Anti-testosterone Fab fragment

Hapten - Testosterone
Space group I222 P21212
Cell dimensions (Å) a ⫽ 90.46 a ⫽ 89.87
b ⫽ 90.85 b ⫽ 95.56
c ⫽ 137.10 c ⫽ 67.22
Mosaicity (°) 0.954 0.879
Resolution limit (Å) 2.72 2.60
Measured reflections 103739 52788
Unique reflections 15557 16913
Rmerge (%) 6.9 8.5
30.4 shell 2.72–2.82 Å 29.9 shell 2.60–2.69 Å
Completeness (%) 99.9 91.7
99.9 shell 2.72–2.82 Å 95.7 shell 2.60–2.69 Å
Overall I/␴(I) 11.7 12.3
4.4 shell 2.72–2.82 Å 4.2 shell 2.60–2.69 Å
Refinement

Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007

Resolution range (Å) 2.72–500 2.60–500
Protein atoms 3345 3345
Hapten atoms - 21
Water molecules 88 337
Average B-factor (Å2) 49.7 25.8
Protein 50.0 25.2
Solvent 36.6 31.7
Hapten - 21.0
Rfactor/Rfree (兩F兩 ⬎ 1␴) (%) 19.5/25.2 18.8/25.5
r.m.s.d. bond length (Å)a 0.0066 0.0062
r.m.s.d. bond angles (°) 1.40 1.34
a
r.m.s.d., root mean square deviation.

FIG. 1. Amino acid sequences of (a)

the heavy and (b) the light chain vari-
able regions of the 3-C4F5 antibody.
The CDR regions are underlined. The
numbering is according to Kabat et al.
(29).

fragment have been previously cloned. The (3-C4F5) antibody
has a reasonably high affinity for testosterone (Ka ⬃ 3 ⫻ 108
⫺1
M ) with an overall good specificity profile, e.g. its cross-
reactivity with 5␣a-dihydrotestosterone and androstenedione
is low (12). The main problem with the 3-C4F5 antibody is its
1% cross-reaction with dehydroepiandrosteronesulphate
(DHEAS), due to the high DHEAS concentration in sera, this
cross-reaction prevents the diagnostic use of the 3-C4F5 an-
tibody. The binding site of the recombinant (3-C4F5) Fab
fragment has recently been targeted to random mutagenesis,
and mutants with significantly improved affinity and speci-
ficity have been isolated by phage display selections (12, 13).
In this report we describe the structures of the free and
testosterone-bound form of the recombinant (3-C4F5) Fab at
2.72 and 2.60 Å, respectively. On the basis of the structures, we
are able to describe the details of the testosterone binding and
also interpret previous results obtained in the fine-tuning of
the anti-testosterone binding site (12, 13). When the results
 Crystal Structures of Recombinant Anti-testosterone Fabs 4185
TABLE II
CDRs of the anti-testosterone Fab fragment
Comparison of the root mean square deviations (Å) between the free and testosterone-bound CDR regions and analysis of the shortest distances
between CDR residues and testosterone. The CDR-loops, defined by Kabat (29), have been superimposed and the shortest distances measured
using the XtalView program (version 4.0, The Scripps Research Institute, San Diego, CA) (42).
Region r.m.s.d.a
Canonical C␣ All Residue Distance Testosterone
CDR-loop structure atoms atoms

Å
L1: Arg24–Glu34 4 0.77 1.58 Y32CG 6.3 B-ring, C7
Q34OE1 4.5 D-ring, C16
50 56
L2: Lys –Ser 1 0.18 0.52 K50N 9.5 D-ring, C16
L3: Phe89–Thr97 1 0.41 0.97 F89CE2 3.6 D-ring, C15
V94CG2 3.7 A-ring, C4
V94CG2 4.6 A-ring, O3
P96CG 3.8 B-ring, C6
H1: Thr31–Ser35 1 0.18 0.30 A33CB 4.0 C-ring, C12
S35OG 3.4 C18

Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007

50 65
H2: Ser –Gly 1 0.22 0.76 S50OG 3.7 C-ring, C11
V52CG1 3.7 A-ring, C1
Y58ring 4.1 A-ring, C2
H3: Glu95–Tyr102 - 0.45 1.73 Y99ring 3.8 A-ring, C1
Y99O 3.8 D-ring, C15
G100'N 3.6 D-ring, C16
G100'O 2.7 D-ring, OH20
a
r.m.s.d., root mean square deviation.

were compared with other available structures of anti-steroid

antibodies, interesting, similar features in the binding mecha-
nism among these antibodies were observed.
EXPERIMENTAL PROCEDURES
Production and Purification of the Recombinant Anti-testosterone
Fab Fragment—For the large scale production of the (3-C4F5) Fab
fragment, the pKKtac expression vector (12) was transformed into the
Escherichia coli strain RV308 (su⫺, ⌬DlacX74, galISII::OP308, strA).
Cells were grown in a bioreactor in a defined medium using the fed-
batch technique (14, 15). The production level of the anti-testosterone
Fab fragment was about 50 mg/liter. The Fab fragment was purified
from the culture supernatant first by a cation exchange column (SP-
Sepharose, Amersham Biosciences, Inc.) followed by a protein G column
(Hi Trap, Amersham Biosciences, Inc.). Coomassie Blue-stained protein
gels showed that the purified Fab fragments were essentially homoge-
neous and correctly assembled. Amino acid sequences of the heavy and
the light chain variable regions are presented below in Fig. 1.
Crystallization—Crystals of the free and testosterone-complexed
3-C4F5 Fab fragment were grown as described previously (16). Crystals
without testosterone were grown by the hanging-drop method, from 9
mg/ml protein solutions buffered with 0.1 M BTP (Bis-Tris propane) (pH FIG. 2. Superposition of the variable regions of the anti-testos-
5.6) and using 15% PEG3350 as the precipitant. Crystals with testos- terone Fab fragments. Comparison of the C␣ atom traces of testos-
terone were grown by the same method, but the crystallization solution terone bound (black) and free (gray) structures. The testosterone mol-
contained 0.1 M MgCl2 as its additive and 20% PEG3350 as its precip- ecule is shown in the binding pocket. Superposition of the C␣ atoms was
itant. In the case of the testosterone complex, the 0.1 M BTP (pH 5.6) made by the O program (22), and the figure has been drawn by the
buffer was added into the well and not to the drops. For diffraction data SETOR program (40).
collection, all crystals were harvested into a cryoprotectant solution,
and flash-frozen in a cold stream at 120 K (using an Oxford Cryosystem Molecular Replacement and Structure Refinement—The molecular
cryostream cooler). The PEG3350 as a 35% solution worked as the replacement was performed with the AmoRe program (19). The immu-
cryoprotectant for the free crystals. In the case of testosterone complex noglobulin 4-4-20 Fab (20), having high sequence homology with the
crystals, the cryoprotectant solution was 20% PEG3350 with 20% anti-testosterone Fab, was used as the search model for rotation and
ethanol. translation calculations. The homologies of the light chain and the
Diffraction Data Collection—All diffraction data were collected on a heavy chain were 72 and 90%, respectively. Initial searching for rota-
Rigaku RU-200HB rotating-anode generator operating at 50 kV and tion solution with the entire model was successful, but it failed with the
100 mA, equipped with a new MSC Confocal Blue Optics (17) and an translation solution. For this reason, it was necessary to perform a
RAXIS-II imaging plate system. In all cases, the crystal to image plate search for the constant and variable domains separately. Each model
distance was 120 mm and the oscillation range was 1°. All data were was used in separate rotation searches using data in the range 8- to 4-Å
indexed and processed with the DENZO program (Yale University, New resolution and a sphere radius of 18 Å. The correlation coefficient for
Haven, CT) (18), and space groups were defined by the XPREP program the variable domain rotation was 21.0 (next peak 19.5); for the constant
(SHELXTL software package). Table I (see below) summarizes the data domain the rotation correlation coefficient was 17.3 (next peak 16.7).
collection statistics for both the free and testosterone-bound data sets. Once the rotation orientation was determined, the AmoRe program
The crystals were not resistant in the x-ray beam at room temperature, was used to search for a translation solution. The first 20 peaks from
and the crystal decay was significant after a few hours’ exposure. the rotation function were used in the translation search at 8- to 4-Å
However, it was possible to collect data from single crystals by using resolution. Initially, each model was used separately in the translation
cryotechniques. After soaking, crystals were rapidly placed in a nitro- searches. Numbers 10 and 11 solutions from the rotation function of the
gen gas coldstream maintained at 120 K. Data from individual image constant domain produced peaks 20.9 and 21.1 ␴ above the mean,
plates were scaled, and reflection intensities were merged using the respectively. Fixing solution number 11 of the constant model with a
SCALEPACK program (Yale University). subsequent search for a variable model provided the best solution, as
4186 Crystal Structures of Recombinant Anti-testosterone Fabs
judged by the highest correlation coefficient 33.0. Rigid body refinement
was carried out using data from 8 to 4 Å, and rigid body refinement gave
an R factor of 44.3%. Data refinement, including individual tempera-
ture factor refinement, was done by the CNS program (21), and after
that an electron density map was calculated. Manual rebuilding of the
anti-testosterone Fab structure was performed with the O program (22)
by mutating the 4-4-20 Fab sequence in those positions where it differed
from the anti-testosterone Fab fragment. Water molecules were added
manually, as they became apparent in the map. Water molecules, which
had lower temperature factor than 60 Å2, were kept in the model.
Testosterone atoms in the bound structure were not included in the
refinement until refinement of the protein had reached convergence
and the electron density was clear enough. Omit maps were calculated
by eliminating atoms of interest from the model followed by a round of
simulated annealing refinement. The refined structure of the testoster-
one-bound structure was used as the search model for the molecular
replacement into the free data.
A summary of both refinement results is given in Table I. The final
free structure has a conventional crystallographic R factor of 19.5%
free structure was also unclear. Residues (129H-133H) in the
constant domains were not visible, and they were, therefore,
presumed to be disordered in both structures. These residues
have generally been observed to be disordered in other Fab
structures (5, 8, 25–28).
Complementarity-determining Regions—To do pairwise com-
parisons, the CDR regions (according to the Kabat definition
(29)) of the free and testosterone-bound structures were super-
imposed (Table II)I. O. The root mean square deviations of all
atoms show that the free and testosterone-bound Fabs differ
significantly (⬎1.0 Å) in the CDR-L1 and CDR-H3. The elec-
tron density for the CDR-L1 is weak, and, therefore, the real
structure rearrangement caused by testosterone binding or
normal thermal motion is uncertain. Differences between all
atoms in residues of the CDR-H3 have also been observed to be
large in previous studies (30, 31). The CDR-H1, CDR-H2, and

Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007

with an Rfree of 25.2% (21) for all data between 8- and 2.72-Å resolution. the CDR-L2 loops do not undergo a large change upon testos-
In comparison, the testosterone-bound structure has a crystallographic terone binding even though some residues form close contacts
R factor of 18.8% with an Rfree of 25.5% for all data between 8- and 2.6-Å with testosterone.
resolution. As reported by the PROCHECK program (23), all residues of Five of the CDR-loops, L1, L2, L3, H1, and H2, of the anti-
the protein lie in or close to allowed regions of the Ramachandran plot.
testosterone Fab, have their size and residues at specific sites
According to this analysis, 79.5% of the residues in the free structure
and 87.2% of the residues in the testosterone-bound structure lie in the that indicate previously reported canonical structures (32–34).
most favored areas with further residues of 18.4 and 12.2%, respec- Currently, there is no available specific classification for the
tively, occurring in additional allowed areas. Co-ordinates for the struc- CDR-H3 loop, but many authors have analyzed and classified
tures have been deposited with the Protein Data Bank (24) (codes 1I9I CDR-H3 loops in several forms. In the case of the anti-testos-
and 1I9J). terone Fab, which contains Arg94H but lacks Asp101H, the
RESULTS
Description and Quality of the Structures—The determined
structures show the typical Fab immunoglobulin fold, and the
backbone atoms of the protein frameworks can be superim-
posed with other general Fab structures. The free and testos-
terone-bound structures display different elbow angles making
the overall superposition difficult. The superposition of the C␣
atoms of the testosterone-bound and free structures of the
variable domains is shown in Fig. 2. The variable domains
show an overall root mean square deviation for all atoms and
for C␣ atoms of 0.99 and 0.55 Å, respectively. When the vari-
able domains of the heavy and light chain were superimposed
individually, the deviation of C␣ atoms was 0.31 and 0.62 Å,
respectively. We have also compared the previously build mo-
lecular model of the testosterone-3-C4F5 antibody complex (12)
and the testosterone-bound structure. The root mean square
values of C␣ atoms and all atoms in variable domains were 1.17
and 1.73 Å, respectively.
The electron density map allowed to trace all six CDR loops FIG. 3. 2Fo ⴚ Fc electron density omit map of the testosterone
of the Fab-testosterone structure. Weak electron density was molecule. Testosterone atoms were eliminated from the map calcula-
tion. Electron densities of Trp47H, Tyr58H, and Tyr99H residues are
observed for the side chains of Gln27L to Val27cL and Asn30L also displayed. The direct hydrogen bond between testosterone and
residues in the CDR-L1 of the free structure. The electron Gly100JH and the indirect hydrogen bond between testosterone and
density of the residues His27dL to Gly29L in the CDR-L1 of the Ser35H through a water molecule are shown.

FIG. 4. Superposition of the free

and testosterone-bound structures il-
lustrating the changes in the binding
site residues of the anti-testosterone
Fab upon testosterone binding. The
testosterone-bound structure is shown in
gray. The light chain in the free structure
is shown in red and the heavy chain in
green. Testosterone is shown in yellow.
 Crystal Structures of Recombinant Anti-testosterone Fabs
FIG. 5. The OMIT maps showing the positions of carbonyl oxy-
gens in the (A) free and (B) testosterone-bound Fab fragments.
The Omit electron-density maps were calculated without sigma cut-off
from Gly98H to Val100H after residues had been removed. Map calcu-
lations of Gly98H and Val100H were separated from each other. The
simultaneous “pep-flip” of Gly98H is a requirement for testosterone
binding.
so-called torso region is not related to either the bulged confor-
mation or the non-bulged one (35–37).
The Testosterone Binding Site—The testosterone molecule is
deeply bound between the light and heavy variable domains in
the binding pocket, which is mainly formed by the CDR-H3 and
-L3 loops. The molecular surface complementarity between the
4187
Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007
FIG. 6. The steroid hormone haptens of the antibodies used in
the structure comparisons. The conjugation position of testosterone
and progesterone is marked with an asterisk.
FIG. 7. Molecular surfaces of the binding sites viewed directly
upon the CDR regions. Positive potential is indicated in blue and
negative potential in red. Top panel, anti-testosterone variable Fab
fragment (a) free and (b) testosterone-bound form; middle panel, anti-
progesterone variable Fab fragment (c) free and (d) progesterone-bound
form; bottom panel, (e) digoxin-bound Fab 26-10 and (f) estriol-bound
Fv4155 fragment. This figure was produced with the use of the GRASP
program (39).
Fab fragment and hapten is evident (Fig. 3). The cyclohexene
D-ring with the 20-hydroxyl group is buried deepest in the
pocket. The cyclopentane A-ring with the 3-keto group, which
4188 Crystal Structures of Recombinant Anti-testosterone Fabs
has been used for the conjugation of testosterone to thyroglob-
ulin in the original immunogen, is directed toward the solvent
(12). The testosterone-Fab complex shows that the highly se-
lective binding mode is mainly determined by shape comple-
mentarity between the protein and testosterone. Testosterone
is placed in a sandwich between the ring structures of the
framework residue Trp47H and Tyr99H of the CDR-H3. The
two methyl groups of the testosterone molecule are directed
toward the Trp47H. Tyr58H of the CDR-H2 is also in a close
vicinity to the A-ring of the testosterone molecule.
Using a maximum distance criterion of 4.1 Å, a total of 15
residues make van der Waals contacts with the testosterone.
Residues Glu95H and Tyr99H-Leu100KH of the CDR-H3
largely form the binding pocket with the co-operation of the
residues Phe89L, Gly91L, Val94L, and Pro96L of the CDR-L3
loop. Residues Ala33H and Ser35H from the CDR-H1 loop;

Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007

Ser50H, Val52H, and Tyr58H of the CDR-H2 loop; and Glu34L
of the CDR-L1 loop also take part in the binding site formation.
The CDR-L2 is not directly involved in the binding of the
testosterone. The two negatively charged residues, Glu95H and
Glu34L, at the bottom of the binding site make it less hydro-
phobic. The only direct hydrogen bond is between the D-ring
hydroxyl group at position 17 and the main chain oxygen in
Gly100JH. Two indirect hydrogen bonds are formed between
the D-ring oxygen of testosterone and the main-chain oxygen of
Ala33H and the side chain of Ser35H through a water mole-
cule. Upon testosterone binding, the largest spatial movements
occur in the side chains of residues Glu95H and Tyr99H (Fig.
4). In the free conformation, these residues sterically block
testosterone access, so rearrangements appear to be necessary
for the binding of testosterone. The residues Glu95H and
Tyr99H are packed against each other in the free structure.
These conformational changes from free form to testosterone-
bound form are related by two rotations in Glu95H (C␦d-C␥-
C␤-C␣) from ⫺70° to 173° and (C␥-C␤-C␣-N) from ⫺65° to 70°
and one in Tyr99H (C␥-C␤-C␣-C) from 87° to 177°. At medium
resolution it is not possible to firmly determine main-chain
conformation. However, by inspecting omit electron density
maps in the region Gly98H to Val100H in the CDR-H3 loop, we
found evidence for conformational flexibility in the main chain
(Fig. 5). The conformational change in the Glu95H side-chain
results in the “pep-flip” in the position of Gly98H (the carbonyl
oxygen has been moved to point in the opposite direction). This
allows the formation of a new hydrogen bond between the side
chain of Glu95H and the main-chain N atom of Tyr99H.
DISCUSSION
Anatomy of the Binding Sites of Anti-steroid Antibodies—We
have compared the structure of the anti-testosterone Fab with
those of anti-progesterone (5), anti-digoxin (8), and Fv4155
(11). Haptens of these antibodies are shown in Fig. 6. The
structure of the immunogen originally used in the production of
different anti-steroid monoclonal antibodies as well the affinity
and specificity thresholds of the screening methods used in the
isolation of the hybridoma clones have influence in the anat-
omy of the binding site.
In the case of the anti-testosterone (3-C4F5) and anti-proges-
terone (DB3) antibodies, the aim has been to develop high
binding selectivity toward the free steroid (12, 38), and the
surface representations of the binding sites of these antibodies
show that the binding sites are narrow, deep pockets (Fig. 7).
The structures of the free and steroid-bound structures show
the conformational changes of Tyr99H in the anti-testosterone
Fab (Fig. 7, a and b) and Trp100H in the anti-progesterone Fab
(Fig. 7, c and d) upon binding. The binding site of the anti-
testosterone Fab is partly negative (Fig. 7b), this is due to the
residue Glu34L being located at the bottom of the binding site.
FIG. 8. Comparison of the binding mechanism of the anti-tes-
tosterone Fab (top), anti-progesterone Fab DB3 (5) (middle), and
anti-digoxin Fab (8) (bottom). The free structures are shown in pink
and steroid complexes in black. Stereopairs were prepared with the
Molscript program (41).
The other binding sites studied here do not have a similar
negatively charged region. The molecular surface of the anti-
digoxin Fab fragment (Fig. 7e) shows that the digoxin binding
site is also a pocket, whereas the biding site of the Fv4155
fragment (Fig. 7f) is a groove. In the Fv4155 structure the
carbohydrate group of the hapten makes three hydrogen bonds
to the protein (11). The carbohydrate moiety is oriented to the
solvent also in the digoxin-Fab complex, but no hydrogen bonds
or salt-bridge interactions are formed between the digoxin and
the antibody. Molecular surfaces of the anti-digoxin Fab frag-
ment (Fig. 7e) and the Fv4155 fragment (Fig. 7f) show that
larger steroids are binding in the deep pocket or in the groove
on the surface of the antibody.
Similarities in the Steroid Binding in the Known Antibody
Structures—The molecular interactions that affect antibody
specificity for steroid molecules have been previously discussed
in a few articles (5– 8, 11). Two to four hydrogen bonds are
observed (5) in the anti-progesterone complex structures,
whereas a single hydrogen bond exists between the steroid and
the protein in the Fv4155 structures (11) and in the anti-
testosterone Fab structure. Thus, high affinity and specificity
displayed by steroid binding antibodies primarily arises from
shape complementarity (5– 8, 11). The binding mode is very
similar in the steroid-related Fab structures. The binding sites
of steroid-specific Fab structures are mainly hydrophobic.
There are only two charged residues in the binding site of the
anti-testosterone Fab fragment. One of them, Glu95H, rotates
away and leaves space for testosterone. Another residue,
 Crystal Structures of Recombinant Anti-testosterone Fabs 4189

FIG. 9. Sequence alignments of selected binding site regions of steroid binding antibodies. The residues affecting the binding mode are
boxed, and those residues forming the sandwich structure characteristic for steroid binding are shown in gray.

Glu34L, makes charged surroundings in the location of testos-

terone on the bottom of the binding site.

Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007

A comparison between the anti-testosterone Fab and the
anti-progesterone Fab structures showed clear structural and
functional similarities. The sandwich structure made by
Trp47H and Tyr99H in the anti-testosterone Fab, and Trp50H
and Trp100H in the anti-progesterone Fab structure (5) will
have an important role in the steroid binding (Fig. 8). The
anti-testosterone Fab shows similarly “open” and “closed” con-
formation as the anti-progesterone Fab. In both cases, the free
binding site is occupied by an aromatic residue. In the presence
of the steroid hapten, this aromatic residue will be reconsti-
tuted as a part of a “sandwich structure.” The digoxin molecule
is also sandwiched between the aromatic residues Tyr50H and
Trp100H, but no similar open and closed conformational
changes between complexed and free structures were observed.
In the absence of free structure of the Fv4155, it can only be
speculated whether Tyr100H of the Fv4155 will have a similar
role in the closed conformation (11). The Fv4155 structures
tightly bind with metabolites of estradiol by making a sand-
wich between Leu95H and Tyr100H residues, and there are no
aromatic residues available in the favorable surround.
Three residues, which are located in a similar position in the
amino acid sequence (Fig. 9) and in the three-dimensional
structure, presumably affect steroid hormone binding mode
(Fig. 10). At the bottom of the binding pocket there is a bulky FIG. 10. Schematic diagrams of the steroid binding sites. A, the
hydrophobic residue, 100KH on the CDR-H3. This residue is anti-testosterone Fab; B, the anti-progesterone Fab (5); C, the anti-
leucine, phenylalanine, or methionine. The residues 96L (from digoxin Fab 26-10 (8), and D, the Fv4155 (10) fragments.
the CDR-L3 loop) and 95H (from the CDR-H3 loop) allow ste-
roids to enter or restrict steroids from entering into the interior
of the Fab variable domain. 96L positions are occupied by the
side chains of proline or leucine. In the position 95H, the side A60, contained the sequence RSSEVIVTRNGYTPIE, and the
chain of Glu95H is oriented away from the binding pocket in other, designated A58, contained the sequence RSSQRM-
the anti-testosterone Fab complex structure. In the anti-pro- VQRNGHTPLE. When compared with the wild-type CDR-L1
gesterone Fab, there is a small glycine residue in this position. loop sequence, RSSQSIVHSNGNSYLE, the A60 clone contains
In the cases of the anti-digoxin Fab 26-10 and the Fv4155 eight and the A58 clone seven mutations, respectively. Surpris-
fragments, the binding pocket is not so large; there is leucine or ingly, these mutated CDR-L1 loops, which improve the affinity
serine in position 95H. The shortest distance between 96L and about 10-fold, are not in direct contact with the testosterone.
95H in the side of the binding site can range from a narrow 4 Å The mutated residues primarily locate on the surface of the
in the Fv4155 to a larger 11 Å in the anti-testosterone Fab. protein. Therefore, it is difficult to explain their impact on
Tuning the Properties of the Anti-testosterone Fab Frag- affinity improvement. Probably, the main influence is on the
ment—The recent CDR random mutagenesis approach com- mobility of the loop. The increased flexibility in the CDR-L1
bined with phage display selections led to the isolation of mu- loop may allow more compact binding of testosterone to the Fab
tants of the anti-testosterone Fab with significantly improved
fragment.
binding properties (12, 13). Only one of the important binding
The determined free and testosterone-bound structures now
site residues has been changed in the isolated mutants, namely
allow more rational planning of mutations. However, improve-
the CDR-H3 residue Glu95H, located at the side of the binding
ments of the binding properties in the anti-testosterone anti-
site. Based on the structure, the mutation Glu95HAla, which
was selected from the mutant library based on improved spec- body (3-C4F5) can also be achieved by mutating CDR residues,
ificity, will provide more free space for the testosterone D-ring. which are not in direct contact with the hapten.
In the affinity maturation selection, two different clones Acknowledgments—We thank Reetta Kallio-Ratilainen and Armi
having extensive mutations within the CDR-L1 loop were iso- Boman for skillful technical assistance and Tekes, the National Tech-
lated (13). One clone of the mutant CDR-L1 loop, designated nology Agency, for its financial support.
4190 Crystal Structures of Recombinant Anti-testosterone Fabs
REFERENCES
1. Wheeler, M. J. (1994) The Immunoassay Handbook, pp. 365–378, Macmillan
Press, Hants, United Kingdom
2. Wilke, T. J., and Utley, D. J. (1987) Clin. Chem. 33, 1372–1375
3. Ismail, A. A., Astley, P., Burr, W. A., Cawood, M., Short, F., Wakelin, K., and
Wheeler, M. J. (1986) Ann. Clin. Biochem. 23, 113–134
4. Wang, C., and Swerdloff, R. S. (1997) Ann. Med. 29, 365–370
5. Arevalo, J. H., Stura, E. A., Taussig, M. J., and Wilson, I. A. (1993) J. Mol. Biol.
231, 103–118
6. Arevalo, J. H., Taussig, M. J., and Wilson, I. A. (1993) Nature 365, 859 – 863
7. Arevalo, J. H., Hassing, C. A., Stura, E. A., Sims, M. J., Taussig, M. J., and
Wilson, I. A. (1994) J. Mol. Biol. 241, 663– 690
8. Jeffrey, P. D., and Sheriff, S. (1993) Proc. Natl. Acad. Sci. U. S. A. 90,
10310 –10314
9. Jeffrey, P. D., Schildbach, J. F., Chang, C. Y., Kussie, P. H., Margolies, M. N.,
and Sheriff, S. (1995) J. Mol. Biol. 248344 –28360
10. Walliman, P., Marti, T., Fürer, A., and Diederich, F. (1997) Chem. Rev. 97,
1567–1608
11. Trinh, C. H., Hemmigton, S. D., Verhoeyen, M. E., and Phillips, S. E. V. (1997)
Structure 5, 937–948
12. Hemminki, A., Niemi, S., Hoffren, A.-M., Hakalahti, L., Söderlund, H., and
Takkinen. K. (1998) Protein Eng. 11, 311–319
13. Hemminki, A., Niemi, S., Hautoniemi, L., Söderlund, H., and Takkinen. K.
(1998) Immunotechnology 4, 59 – 69
14. Pack, P., Kujau, M., Schroeckh, V., Knupfer, U., Wenderoth, R, Riesenberg, D.,
and Plückthun, A. (1993) Bio/Technology 11, 1271–1277
15. Plückthun, A., Krebber, A., Krebber, C., Horn, U., Knüpfer, U., Wenderoth, R.,
Nieba, L., Proba, K., and Riesenberg, D. (1996) Antibody Engineering: A
Practical Approach, pp. 203–252, IRL, Oxford
16. Valjakka, J., Hemminki, A., Teerinen, T., Takkinen, K and Rouvinen, J. (2000)
Acta Crystallogr. Sect. D Biol. Crystallogr. 56, 218 –221
17. Yang, C., Courville, A., and Ferrara, J. D. (1999) Acta Crystallogr. Sect. D Biol.
Crystallogr. 55, 1681–1689
18. Otwinowski, Z., and Minor, W. (1997) Methods Enzymol. 276, 307–326
19. Navaza, J. (1994) Acta Crystallogr. Sect. A 50, 157–163
20. Herron, J. N., He, X. M., Voss, E. W., Jr., and Edmundson, A. P. (1989) Proteins
5, 271
21. Brünger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-
Kunstleve, W., Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N. S., Read,
R. J., Rice, L. M., Simonson, T., and Warren, G. L. (1998) Acta Crystallogr.
Sect. D Biol. Crystallogr. 54, 905–921
22. Jones, T. A., Zou, J.-Y., Cowan, S. W., and Kjeldgaard, M. (1991) Acta Crys-
tallogr. Sect. A 47, 110 –119
23. Laskowski, R. A., MacArthur, M. W., Moss, D. S., and Thornton, J. M. (1993)
J. Appl. Cryst. 26, 283–291
24. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H.,
Shindyalov, I. N., Bourne, P. E. (2000) Nucleic Acids Res. 28, 235–242
25. Stanfield, R. L., and Wilson, I. A. (1993) Immunomethods 3, 211–221
26. Brady, R. L., Edwards, D. J., Hubbard, R. E., Jiang, J. S., Roberts, L. S. M., and
Todd, R. J. (1992) J. Mol. Biol. 227, 253–264
27. Rini, J. M., Schulze-Gahmen, U., and Wilson, I. A. (1992) Science. 225,
959 –965
28. Tulip, W. R., Varghese, J. N., Laver, W. G., Webster, R. G., and Colman, P. M.
(1992) J. Mol. Biol. 227, 122–148
29. Kabat, E. A., Wu, T. T., Perry, H. M. Gottesman, K. S., and Foeller, C. (1991)
Sequences of Protein of Immunological Interest, 5th Ed., National Institutes
of Health, Bethesda, MD
30. Chothia, C., and Lesk, A. M. (1987) J. Mol. Biol. 196, 901–917
31. Chothia, C., Lesk, A. M., Tramontano, A., Lewitt, M., Smith-Gill, S. J., Air, G.,
Sheriff, S., Padlan, E. A., Davies, D., Tulip, W. R., Colman, P. M., Spinelli,
S., Alzari, P. M., and Poljak, R. J. (1989) Nature 342, 877– 883
Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007
32. Chothia, C., Lesk, A. M., Gherardi, E., Tomlinson, I. M., Walter, G., Marks,
J. D., Llewelyn, M. B., and Winter, G. (1992) J. Mol. Biol. 227, 799 – 817
33. Martin, A. C. R., and Thornton, J. M. (1996) J. Mol. Biol. 263, 800 – 815
34. Oliva, B., Bates, P. A., Querol, E., Avilés, F. X., and Sternberg, J. E. (1998) J.
Mol. Biol. 279, 1193–1210
35. Shirai, H., Kidera, A., and Nakamura, H. (1996) FEBS Lett. 399, 1– 8
36. Morea, V., Tramontano, A., Rustici, M., Chothia, C., and Lesk, A. M. (1998) J.
Mol. Biol. 275, 269 –294
37. Kim, S. T., Shirai, H., Nakajima, N., Higo, J., and Nakamura, H. (1999)
Proteins 37, 683– 696
38. Gani, M., Coley, J., Piron, J., Humphreys, A. S., Arevalo, J., Wilson, I. A., and
Taussig, M. J. (1994) J. Steroid Biochem. Mol. Biol. 48, 277–282
39. Nicholls, A., Sharp, K. A., and Honig, B. (1991) Proteins 11, 281–296
40. Evans, S. V. (1993) J. Mol. Graphics 11, 134 –138
41. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946 –950
42. McRee, D. E. (1999) J. Struct. Biol. 125, 156 –165