THE JOURNAL OF BIOLOGICAL CHEMISTRY
44021–44027, 2002
© 2002 by The American Society for Biochemistry and Molecular Biology, Inc.
Crystal Structure of an in Vitro Affinity- and Specificity-matured
Anti-testosterone Fab in Complex with Testosterone
IMPROVED AFFINITY RESULTS FROM SMALL STRUCTURAL CHANGES WITHIN THE VARIABLE DOMAINS*
Published, JBC Papers in Press, August 23, 2002, DOI 10.1074/jbc.M208392200
Jarkko Valjakka‡§, Ari Hemminki¶, Seija Niemi储, Hans Söderlund**, Kristiina Takkinen**,
and Juha Rouvinen‡
From the ‡Department of Chemistry, University of Joensuu, P. O. Box 111, 80101 Joensuu, Finland, ¶Finnish Red Cross,
Kivihaantie 7, 00310 Helsinki, Finland, 储Orion Diagnostica, 90460 Oulunsalo, Finland, and **VTT Biotechnology,
P. O. Box 1500, 02044 VTT, Finland
A highly selective, high affinity recombinant anti-tes-
tosterone Fab fragment has been generated by stepwise
optimization of the complementarity-determining re-
gions (CDRs) by random mutagenesis and phage display
selection of a monoclonal antibody (3-C4F5). The best
mutant (77 Fab) was obtained by evaluating the additiv-
ity effects of different independently selected CDR mu-
tations. The 77 Fab contains 20 mutations and has about
40-fold increased affinity (Kd ⴝ 3 ⴛ 10ⴚ10 M) when com-
pared with the wild-type (3-C4F5) Fab. To obtain struc-
tural insight into factors, which are needed to improve
binding properties, we have determined the crystal
structures of the mutant 77 Fab fragment with (2.15 Å)
and without testosterone (2.10 Å) and compared these
with previously determined wild-type structures. The
overall testosterone binding of the 77 Fab is similar to
that of the wild-type. The improved affinity and speci-
ficity of the 77 Fab fragment are due to more compre-
hensive packing of the testosterone with the protein,
which is the result of small structural changes within
the variable domains. Only one important binding site
residue Glu-95 of the heavy chain CDR3 is mutated to
alanine in the 77 Fab fragment. This mutation, origi-
nally selected from the phage library based on improved
specificity, provides more free space for the testoster-
one D-ring. The light chain CDR1 of 77 Fab containing
eight mutations has the most significant effect on the
improved affinity, although it has no direct contact with
the testosterone. The mutations of CDR-L1 cause a rear-
rangement in its conformation, leading to an overall fine
reshaping of the binding site.
Steroid hormones have remained a great challenge for im-
munodiagnostics. They are small, rigid, hydrophobic molecules
with only a few functional groups capable of specific interac-
tions with antibodies. The number of different closely related
steroids in human serum is high, their in vivo concentrations
are low (down to a picomolar level), and their relative concen-
* This work was supported by Wallac, Orion Diagnostica, and the
National Technology Agency (Tekes). 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 1L7S and 1L7T)
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. E-mail: jarkko.
valjakka@joensuu.fi.
This paper is available on line at http://www.jbc.org
Vol. 277, No. 46, Issue of November 15, pp.
Printed in U.S.A.
Received for publication, August 16, 2002
Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007
trations can vary greatly, even between normal healthy indi-
viduals. Furthermore, steroids are poorly immunogenic in mice
and rats; two species from which monoclonal antibodies are
usually generated, making it extremely difficult to produce
high affinity and specificity anti-steroid monoclonal antibodies.
In most diagnostic immunoassays of steroid hormones, rabbit
polyclonal antibodies are used, despite the many drawbacks
associated with the use of antisera. The supply of good poly-
clonal antibody reagents of uniform quality is a severe problem
for the immunodiagnostic industry and requires continuous
immunization of many laboratory animals.
Antibody engineering provides excellent tools to tailor the
properties of antibodies with respect to affinity, specificity, and
performance for different applications. An in vitro process of
antigen-driven selection, based on the display of antibody frag-
ments on the surface of a filamentous bacteriophage, has been
shown to be a powerful method for the selection of specific or
improved binders from a heterogeneous mixture of antibody
fragments (1–3). Many recent publications describe the affinity
and/or specificity improvement of anti-steroid antibodies by
random mutagenesis and phage display selections (4 – 6). We
have previously reported specificity and affinity improvement
of the testosterone-binding Fab fragment initially derived from
a hybridoma cell line (3-C4F5) (7, 8). In the absence of struc-
tural data of the 3-C4F5 antibody, the optimization has been
done by random mutagenesis and phage display selection of
individual CDR1 mutant libraries. The mutant Fab fragment
(A60/HCDR1/LCDR2), which has both a ⬎10-fold improvement
in the relative affinity and a significantly lower cross-reactivity
to dehydroepiandrosterone sulfate (DHEAS), worked in a com-
petitive one-step immunoassay within a wide testosterone con-
centration range of patient serum specimens (8). However, low
testosterone concentrations found in female samples (⬍2 nmol/
liter) deviated from values obtained by the reference immuno-
assay or gas chromatography-mass spectrometry analysis,
most probably due to too high a cross-reaction with DHEAS.
Here we have further refined the binding site of the mutant
Fab fragment (A60/HCDR1/LCDR2) by retargeting the CDR3
regions of the heavy and light chain to mutagenesis. The
individually cloned CDR3 phage libraries were selected in re-
lation to specificity and affinity. In the final step of the binding
site optimization, the additivity of different mutations was
evaluated.
1
The abbreviations used are: CDR, complementarity determining
region; DHEAS, dehydroepiandrosterone sulfate; wt, wild-type; BSA,
bovine serum albumin; CMO, carboxymethyloxime; 5␣-DHT,
5␣-dihydrotestosterone.
44021
44022 Crystal Structure of Mutant 77 Fab Fragment
TABLE I
Data collection and refinement statistics of the mutant 77 Fab fragments
Free Testosterone
Space group C2221 P62
Cell dimensions (Å) a ⫽ 71.71 a ⫽ 82.25
b ⫽ 88.13 b ⫽ 82.25
c ⫽ 143.94 c ⫽ 120.06
Mosaicity (°) 0.755 0.354
Measured reflections 120393 88701
Unique reflections 25386 24335
Rmerge (%) 7.0 5.7
32.6 shell 2.10–2.18 Å 33.0 shell 2.15–2.23 Å
Completeness (%) 93.6 97.0
87.9 shell 2.10–2.18 Å 93.3 shell 2.15–2.23 Å
Overall I/␴(I) 17.8 23.2
3.1 shell 2.10–2.18 Å 3.9 shell 2.15–2.23 Å
Resolution range (Å) 2.10–500 2.15–500
Protein atoms 3353 3357
Hapten atoms 21
Water molecules 291 265

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

Average B-factor (Å2) 35.8 37.9
Protein 35.6 37.7
Solvent 37.9 39.7
Hapten 52.7
Rfactor/Rfree (兩F兩 ⬎ 1␴) (%) 21.3/27.5 21.1/27.4
Rmsda bond length (Å) 0.0057 0.0067
Rmsd bond angles (°) 1.34 1.40
a
Rmsd, root mean square deviation.

TABLE II
Affinities and cross-reactivities of the mutant anti-testosterone Fab fragments

To understand the molecular basis for structural changes EXPERIMENTAL PROCEDURES

needed in a high affinity and selectivity binding of a steroid
hormone, we have determined the crystal structures of the
mutant 77 Fab with and without testosterone and compared
these with the recently determined wt (3-C4F5) Fab fragment
structures (9). The crystal structure of 77 Fab in complex with
testosterone reveals that during the stepwise refinement, mu-
tations on the CDR regions that are not in direct contact with
the hapten can have an important influence on the improved
binding. In the 77 Fab, the increased adaptability of the L1 loop
and the binding site mutation E95A of the heavy chain CDR3
facilitate more compact packing of the testosterone to the
protein.
Construction and Selection of CDR3 Mutant Libraries—The VL and
VH CDR3 loops (10) of the mutant Fab fragment (A60/HCDR1/LCDR2)
(8) were retargeted for mutagenesis by using spiked PCR primers (11,
12). The randomness of PCR primers was adjusted by nucleotide doping
during the oligonucleotide synthesis for the CDR loops (62.5% parental-
type (A60/HCDR1/LCDR2) and 12.5% each of three other nucleotides).
Both mutant libraries were constructed into the pComb3 vector (13)
containing the mutant Fab fragment (A60/HCDR1/LCDR2) by utilizing
unique restriction enzyme cleavage sites created before cloning by site-
directed mutagenesis near the CDR3s.
The specificity selection of the CDR-H3 and CDR-L3 mutant libraries
(both ⬃106 clones) was done by first incubating the libraries with a high
concentration of soluble dehydroepiandrosterone-BSA conjugate (dehy-
 Crystal Structure of Mutant 77 Fab Fragment
FIG. 1. The amino acid sequences of the heavy and light chain
variable regions of 77 Fab (numbering is according to Kabat et
al. (10)). The amino acid differences between the 77 Fab and the
wild-type (3-C4F5) are indicated within the CDR residues (underlined).
Mutations of the 77 Fab are shown in bold, and the corresponding wt
residues are shown below.
droepiandrosterone-3-hemisuccinate-BSA, 17–20 mol dehydroepi-
androsterone/mol BSA, 0.4 mg/ml) for 30 min at 37 °C. During the first
two rounds, the phage pools were transferred to microtiter plate wells
coated with 1 ␮g of testosterone-3-carboxymethyloxime (CMO)-BSA,
and during the next three rounds, the phage pools were transferred to
streptavidin-coated wells preincubated for 1 h with testosterone-3-bio-
tin. After a 10-min incubation at 37 °C, the wells were washed (Tris-
buffered saline-1% BSA-0.05% Tween 20) 15 times during a 1-h period
before elution with 50 mM NaOH, pH 12.6, for 15 min. Eluates were
immediately neutralized with 1 M Tris, pH 7.5. A total of five panning
cycles were performed, and during the last three cycles, the concentra-
tion of testosterone-3-biotin used for preincubation of streptavidin-
coated wells was decreased stepwise from 22 to 0.2 nM. The mutant
libraries were also affinity-selected without any competing steroid by
using limited, gradually decreasing concentrations of testosterone-3-
biotin to catch the high affinity binders (14). The libraries were first
incubated with testosterone-3-biotin for 5 min, and then the phage
pools were transferred to streptavidin-coated wells and incubated for 25
min at 37 °C before washing and elution as described above. A total of
five rounds of panning were performed, and during the cycles, the
concentration of testosterone-3-biotin was gradually decreased from 1
nM to 10 pM.
Characterization of the Mutants—After the last panning step, 50
individual clones were first analyzed as soluble Fab fragments on a
competitive enzyme-linked immunosorbent assay, using testosterone-
3-CMO-BSA-coated wells (0.1 ␮g) to catch the testosterone binders and
soluble testosterone and DHEAS to achieve inhibition of binding (7).
The most promising clones, which had decreased DHEAS cross-reactiv-
ity and/or improved relative testosterone binding affinity when com-
pared with the clone A60/HCDR1/LCDR2, were sequenced, cloned into
the pKKtac expression vector (15), and transformed into the Esche-
richia coli strain RV308 for small-scale production. Clones were char-
acterized using goat anti-mouse IgG-coated microtiter wells and testos-
terone-3-CMO-polylysine labeled with fluorescent Europium-chelate
(Wallac) as the label. A one incubation-step assay protocol was used; i.e.
the label, antibody sample, and competing steroid were all added at the
same time onto the wells (8). Dilution series of testosterone, DHEAS,
5␣-dihydrotestosterone (5␣-DHT), and androstenedione (all in steroid-
free human serum) were used to analyze the relative affinities (ED50
concentrations) for the corresponding steroids. Cross-reactivities were
calculated from these values.
New CDR combinations were created by evaluating the additivity of
the CDR-L3 mutant clones A4 and A46 with other mutations isolated
from the previous selections of randomized CDR libraries (7, 8). The
kinetics of binding of the purified mutant Fab fragments to testoster-
one-3-CMO-BSA immobilized on a dextran-coated sensor chip were
determined from the dependence of the surface plasmon resonance
response on the concentration of the purified Fab fragments injected
into the biosensor (BIAcoreTM; Biacore AB) (16, 17), as described
previously (7).
Crystal Structure Determination—The mutant 77 Fab fragment was
produced on a large scale and purified and crystallized as described
44023
FIG. 2. Overlay plot of the subtractive sensograms obtained
after sequential injection of the wt 3-C4F5 and 77 Fab fragments
Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007
over a CM5 sensor chip surface containing immobilized testo-
serone-3-CMO-BSA. The determined kon (M⫺1s⫺1/105) and koff (s⫺1/
10⫺4) rates and the calculated dissociation constants (M/10⫺9) for the
Fab fragments were as follows: wt Fab, kon ⫽ 3.27, koff ⫽ 43.45, and
Kd ⫽ 13.29; and 77 Fab, kon ⫽ 8.73, koff ⫽ 3.44, and Kd ⫽ 0.39.
previously for the 3-C4F5 wt Fab fragment (9). The diffraction data were
collected on a Rigaku RU-200HB rotating anode equipped with MSC
Confocal Blue Optics and a RAXIS-IIC imaging plate system. Crystals
were not resistant in the x-ray beam at room temperature. Therefore, it
was necessary to collect data by using cryotechniques. After soaking in
the cryoprotectant solution (18), the crystals were quickly placed in a
nitrogen gas coldstream maintained at 120 K using an Oxford Cryosys-
tems Cryostream cooler. The crystal to image plate distance was 120
mm, and the oscillation range was 0.5° in the bound crystal and 1° in
the free crystal. The data were processed with the DENZO program
(19), and the XPREP program (SHELXTL software package) was used
to determine the space groups. The structures were solved with the
coordinates of the wt 3-C4F5 Fab fragment (Protein Data Bank code
1I9J) by using molecular replacement with the AmoRe program (20).
The model building was done with the O program (21), and refinement
was done with the CNS program (22). Structures have been analyzed by
the PROCHECK program (23), and they display good stereochemistry.
Coordinates have been deposited in the Protein Data Bank (24) with the
codes 1L7S and 1L7T. Data collection and refinement statistics are
summarized in Table I.
RESULTS
Isolation and Characterization of the Mutants—A combined
selection procedure that allowed simultaneous selection with
respect to specificity and affinity was set up to retain the
testosterone affinity while decreasing the affinity to DHEAS (7,
8). A high concentration of cross-reactive steroid as a protein
conjugate was incubated with the phage solutions before the
binding of phages to limited concentrations of biotinylated tes-
tosterone immobilized on streptavidin-coated wells. The mu-
tant libraries were also affinity-selected without any competing
steroid with the use of limited, gradually decreasing concen-
trations of biotinylated testosterone to catch the high affinity
binders. The concentration of testosterone used in this ap-
proach was clearly lower (1 nM to 10 pM) compared with the
approach in which the cross-reactive steroid was used.
After the panning steps, about 50 individual clones from both
libraries were analyzed on a competitive enzyme-linked immu-
nosorbent assay test (7), which showed that mutant Fab frag-
ments with improved affinity and/or specificity were selected
only from the light chain CDR3 library (data not shown). The
relative testosterone affinities and cross-reactivities to
DHEAS, 5␣-DHT, and androstenedione were characterized by
a competitive one-step time-resolved fluoroimmunoassay using
steroid standards prepared in human serum mimicking clinical
samples. In this assay system, the relative affinity of the pa-
rental mutant Fab clone A60/HCDR1/HCDR2 was ⬎12-fold
lower and the cross-reactivity with 5␣-DHT was ⬎3-fold higher
44024 Crystal Structure of Mutant 77 Fab Fragment
TABLE III
C␣-atom root mean square difference (Å) of different regions between the 77 Fab and the wt Fab
Mutated 77 Fabbound 77 Fabfree wt Fabbound 77 Fabbound
Residues residues wt Fabbound wt Fabfree wt Fabfree 77 Fabfree

Variable 19 0.73 1.13 0.67 1.09

Heavy 1–111 4 0.43 0.45 0.31 0.38
Light 1–107 15 0.91 1.47 0.62 1.45
H1 Thr-31-Ser-35 1 0.18 0.16 0.19 0.18
H2 Ser-50-Gly-65 0 0.37 0.28 0.22 0.34
H3 Glu-95-Tyr-102 3 0.67 0.71 0.45 0.31
L1 Arg-24-Glu-34 8 1.80 2.75 0.77 2.36
L2 Lys-50-Ser-56 4 0.23 0.29 0.17 0.16
L3 Phe-89-Thr-97 3 0.52 0.62 0.40 0.27

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

FIG. 3. Structure of the variable domain of 77 Fab in complex
with testosterone (blue). The mutations of the 77 Fab are shown in
red. The figure has been drawn by the SETOR program (39).

than those determined previously by a two-step immunoassay

protocol within buffer-based standards (8). In the case of the
best light chain CDR3 mutant, A4, isolated from the combined
affinity/specificity panning approach, the DHEAS and 5␣-DHT
cross-reactivities were clearly decreased (3-fold and 2-fold, re-
spectively), whereas the relative testosterone affinity was pre-
served at the same level as in the parental mutant A60/
HCDR1/HCDR2 (Table II). The relative testosterone affinity of
the light chain CDR3 mutant A46, selected from the affinity
panning, was improved 3-fold, but all cross-reactivities were
increased. The deduced amino acid sequences of these reopti-
mized light chain CDR3s revealed two mutations in both mu-
tants, Gln-90 3 Asp and Thr-97 3 Lys in A4 and Gln-90 3 Glu
and His-93 3 Gln in A46.
New CDR combinations were created by evaluating the ad-
ditivity of the CDR-L3 mutants A4 and A46 with other muta-
tions isolated from the previous selections of randomized CDR
libraries (7, 8). Interestingly, additivity effects were observed
only with A4 mutant combinations. The mutant A46, which
was isolated from the affinity selection, was not compatible
with other CDR mutations. Two previously characterized mu-
tated CDR regions were found to have higher additivity than FIG. 4. Superposition of the L1 loop structures of 77 Fab (yel-
low) and wt Fab (gray). The side chains of the mutated residues are
those present in the mutant A60/HCDR1/HCDR2 when com- shown. a, conformational difference of the 77 Fab and wt Fab L1 loop.
bined with the mutant A4. A CDR-L2 sequence (KAYKRFPT) Testosterone is shown in blue. b, influence of the cooperation of the
had been isolated from the combined affinity and specificity CDR-L1 and CDR-L3 mutations of 77 Fab on L1 loop conformation.
panning (8). When compared with the CDR-L2 of the A60/ Superposition of the C␣ atoms was made by the O program (21), and the
figure has been drawn by the SETOR program (39).
HCDR1/HCDR2 mutant, there are two alternative mutations,
namely, Arg-52 3 Tyr and Tyr-53 3 Lys. This alternative
CDR-L2 sequence had a slightly decreased testosterone affinity ditional mutation Tyr-102 3 His, in comparison with the
and was therefore not used in the A60/HCDR1/HCDR2 Fab mutant 44 (7).
fragment. The other new alternative CDR found within the The evaluation of the best CDR combinations resulted in the
additivity evaluation is the CDR-H3 isolated during the first mutant 77 with a high affinity and specificity binding profile
CDR3 specificity optimization. This CDR-H3 contains the ad- (Table II). In comparison with the mutant A60/HCDR1/
 Crystal Structure of Mutant 77 Fab Fragment 44025
HCDR2, the testosterone affinity of 77 Fab is slightly higher,
combined with almost 3-fold decreased DHEAS and ⬃2-fold
decreased 5␣-DHT cross-reactivities. The 77 Fab contains 19
mutations within five CDR regions as compared with the wt
3-C4F5 (Fig. 1). The heavy chain CDR2 is the only unmutated
region. The heavy chain mutation Ala-113 3 Ser originates
from an introduced XhoI restriction site, which was used for
the cloning of the randomized CDR-H3 library.
The binding kinetics of the purified Fab fragments (77 and
wt 3-C4F5) to testosterone-3-CMO-BSA were determined by
BIAcoreTM (Fig. 2). The most significant change in the rate
constants is the ⬃15-fold slower dissociation rate of the 77 Fab
fragment. In comparison with the original monoclonal 3-C4F5,
the mutant 77 Fab fragment has about a 40-fold increase in
affinity (Kd ⫽ 3 ⫻ 10⫺10 M).
Comparison of the 77 Fab and the wt 3-C4F5 Fab Struc-

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

tures—The free and testosterone-bound structures of the 77
Fab fragment are of high quality, and clear electron density
was observed almost throughout the protein. Comparison of
the 77 Fab and wt Fab structures reveals that the overall
structures are quite similar (Table III). However, there are
clear and important local differences. Generally, mutations
give more adaptability to the entire variable fragment. Clear
conformational changes can be observed when free and bound
forms of the CDR loops are compared. The root mean square
deviation of the superimposed C␣ atoms is 1.09 Å for the 77 Fab
and 0.67 Å for the wt Fab. The bound forms of the wt Fab
fragment and the 77 Fab fragment are closer to each other
(0.73 Å) than the free forms (1.13 Å). The CDR loops H1, H2,
and L2 of the 77 Fab have very similar conformations in com-
parison with the wt Fab (0.2– 0.4 Å). The H3 and L3 loops,
which make the important contacts with testosterone, have
larger differences (0.5– 0.7 Å). The L1 loop of 77 Fab, located on
the protein surface, has the largest difference in conformation
(1.8 Å) in comparison with the wt Fab. This correlates clearly
with the number of mutations (Fig. 3). The hairpin loop of L1
fluctuates most, the weak electron density from Val-27c to
Gly-29 of L1 prevents the exact positioning of the loop tip in the
free 77 Fab structure.
FIG. 5. The packing of testosterone (in purple) in the binding
The L1 Loop—The conformational change of the L1 loop is
sites of 77 Fab (A) and wt Fab (B) shown by van der Waals fields.
the significant structural difference between 77 Fab and wt The figure has been made by the XtalView program (40).
Fab. The CDR-L1 contains eight mutations, and the His-27d 3
Thr, Ser-27e 3 Arg, and Tyr-32 3 Pro mutations are especially distance criterion of 4.1 Å, a total of 15 residues in the 77 Fab
structurally significant (Fig. 4a). The Tyr-32 3 Pro mutation make van der Waals contacts with the testosterone. Residues
makes room, with the cooperation of the mutation His-27d 3 Ala-95, Tyr-99, Val-100, Gly-100j, and Leu-100k of the H3 loop
Thr, for the Ser-27e 3 Arg mutation, providing space for the and Phe-89, Gly-91, Val-94, and Pro-96 of the L3 loop mainly
side chain of the arginine. The mutation of the CDR-L3 clone form the binding site. Residues Ala-33 and Ser-35 of the H1
A4, Gln-90 3 Asp, turns the side chain of Ile-27b into a new loop, Ser-50 and Tyr-58 of the H2 loop, and both Arg-27e and
position (Fig. 4b). This rotation of Ile-27b in the Fab 77 causes Glu-34 of the L1 loop also take part in the binding site forma-
“flipping of side chain” from Val-27c to Gly-29. The other tion. The Trp-47H is the only framework residue that is in
CDR-L3 mutation, His-93 3 Thr, provides space for the new contact (3.9 Å) with testosterone. Residues (Trp-47H, Tyr-99H,
conformation of the L1 loop. Due to these mutations, the tip of and Gly-100jH) that play a critical role in the testosterone
the L1 loop is able to bend toward the binding site. In this interaction are located at the same positions in the binding site
position, the side chain of Arg-27e forms two hydrogen bonds of 77 Fab as in the wt Fab (9) (Fig. 6). A similar rotation of the
with the Gly-91 of the L3 loop, leading to tighter packing by side chain of the Tyr-99H, which opens the binding site upon
filling the cavity near testosterone when compared with the wt testosterone binding, is observed in the 77 Fab as in the wt Fab.
Fab (Fig. 5). Other CDR-L1 mutations, Gln-27 3 Glu, Ser-27a The heavy chain CDR3 mutation Glu-95 3 Ala makes the
3 Val, Asn-30 3 Tyr, Ser-31 3 Thr, and Leu-33 3 Ile, alter testosterone binding more favorable because the alanine mu-
the number of hydrogen bonds. The L1 loop of the 77 Fab has tation eliminates the need for the side chain rotation of Glu-95
only one internal hydrogen bond (the total number of hydrogen observed in the wt Fab (9). Only one direct hydrogen bond,
bonds of the L1 structure is 19), whereas the wt Fab has 12 which is the same as that in the wt Fab structure, is formed
internal hydrogen bonds (the total number of hydrogen bonds between the Gly-100jH and the hydroxyl group of the testos-
is 29), and therefore, the adaptability of the L1 loop has been terone D-ring. As in the wt Fab structure, one water molecule
increased. makes indirect hydrogen bonds between the D-ring hydroxyl
The Structure of the Binding Site—The binding site is formed group and the carbonyl oxygen of the Ala-33H and the side
mainly by the hydrophobic H3 and L3 loops. Using a maximum chain of the Ser-35H.
44026 Crystal Structure of Mutant 77 Fab Fragment

FIG. 6. Superposition of binding

sites of the 77 Fab (top panel) and wt
Fab (bottom panel) in the free and
testosterone-bound form. The free
structures are shown in pink, and testos-
terone complex is shown in black. Stere-
opairs were prepared with the Molscript
program (29).

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

DISCUSSION
Development of an Anti-testosterone Fab with High Affinity
and Specificity—The production of anti-hapten antibodies with
both improved affinity and specificity has proven to be chal-
lenging even with the new antibody engineering techniques.
The sequential CDR optimization strategy has been success-
fully used to develop high affinity antibodies binding protein
antigens (25–27). We have employed the sequential CDR opti-
mization strategy to simultaneously improve the affinity and
specificity of an anti-testosterone antibody. The best mutant
variant, 77 Fab, characterized in this study, contains 19 mu-
tations when compared with the wt Fab. In the final refine-
ment step, the additivity of different combinations of CDR
mutations to further improve affinity or specificity was evalu-
ated. When compared with the wt Fab, the 77 Fab has about a
40-fold increase in affinity (Kd ⫽ 3 ⫻ 10⫺10 M) and has signif-
icantly lower cross-reactivities to DHEAS, 5␣-DHT, and andro- FIG. 7. The binding site in the 77 Fab (A) and in the wt Fab (B)
stenedione. The overall specificity refinement was an interest- structures. The solvent-accessible surface was produced using the O
program (21).
ing finding because only DHEAS was used as the competitive
steroid in the specificity selections throughout the optimization
procedure. the structure refinement of 77 Fab are interesting. The most
We targeted the diversification to CDRs, but randomization significant structural change in the 77 Fab when compared
of the whole variable regions by error-prone PCR combined with the wt (3-C4F5) is the increased adaptability of the L1
with phage, bacterial, or yeast display selection has also been loop. This L1 loop, which was originally isolated based on a
shown to be a very efficient strategy to improve binding prop- 10-fold affinity increase (8), has adapted the loop structure by
erties of anti-hapten antibodies (5, 14, 28, 30, 31). These stud- mutations His-27d 3 Thr and Tyr-32 3 Pro, making space for
ies have demonstrated that framework mutations can also the Ser-27e 3 Arg mutation (Fig. 4a). The other CDR-L1 mu-
make significant contributions to improved binding properties. tations minimize the internal hydrogen bond network between
Both these diversification strategies have the drawback that loop residues. The CDR-L3 mutation Gln-90 3 Asp cooperates
only one or few mutant lines can proceed for further mutagen- with the L1 loop by turning the side chain of Ile-27b into a new
esis rounds because library construction is a laborious step, position (Fig. 4b). This causes “flipping of the side chains” from
and the transformation step limits the size of the library. The Val-27c to Gly-29, inducing the tip of the L1 loop to bend toward
recently demonstrated ribosome display technique, allowing a the binding site. In this new L1 loop position Arg-27e is able to
“built-in” in vitro evolution, combined with propagation of sev- make two hydrogen bonds with the Gly-91 of the L3 loop,
eral mutant lineages totally in in vitro conditions provides mediating more convenient interaction of the binding site by
excellent, new tools to improve antibody properties (32, 33). the filling the cavity near testosterone (Fig. 5).
Jermutus et al. (34) showed the efficiency of the ribosome The cascade of the important mutations leads to a new,
display technique evolving the off-rate and thermodynamic improved conformational fit between testosterone and antibody
stability of single-chain antibodies. (Fig. 7). Surface complementarity between testosterone and 77
Implications of Mutations for the Highly Selective Testoster- Fab is clearly better when compared with the wt Fab. The
one Binding—The mutations with an important contribution to residues Val-94L and Tyr-99H of the 77 Fab are also closer to
 Crystal Structure of Mutant 77 Fab Fragment
testosterone than in the wt Fab. The distance between the
Tyr-99H and the A-ring is 3.4 –3.7 Å in the 77 Fab and 3.7– 4.5
Å in the wt Fab. The distance between Val-94L and the car-
bonyl oxygen of the A-ring is 3.6 Å in 77 Fab and 4.6 Å in the
wt Fab. It is worth noting that DHEAS has a sulfate substit-
uent in the position of this carbonyl group. New orientation of
Val-94L might explain the decreased DHEAS cross-reactivity
of 77 Fab.
Generally, hydrogen bonds are important for selectivity, as
also shown for the steroid-binding antibodies (35–37). As in wt
Fab, only one hydrogen bond is formed between the hydroxyl
group of the testosterone D-ring and the carbonyl oxygen of
Gly-100jH in the 77 Fab. The higher cross-reactivity of 77 Fab
with 5␣-DHT than with DHEAS and androstenedione might be
partly explained by this hydrogen bond. 5␣-DHT has an iden-
tical substituent in the D-ring, and it is thus capable of making
a similar hydrogen bond. On the other hand, DHEAS and
androstenedione have a carbonyl substituent in the D-ring,
which is not able to form the corresponding hydrogen bond.
The Structural Basis of in Vivo and in Vitro Affinity Matu-
ration—The structural basis of in vivo affinity maturation has
been studied by comparing the high-resolution crystal struc-
tures of the anti-hapten antibody 48G7 and its germ-line pred-
ecessor (38). None of the nine somatic mutations of the 48G7
antibody, leading to a significant increase in the affinity (3 ⫻
103-fold), directly contacts the hapten. The somatic mutations
induced small structural changes in the binding domains of
48G7 that resulted in the affinity improvement. The results
obtained from the in vitro affinity maturation of antibodies by
random diversification methods show that it resembles the in
vivo process. For example, the in vitro maturation of the anti-
digoxin 26-10 antibody and anti-fluorescein 4-4-20 antibody
showed that the majority of the mutations leading to higher
affinity correspond to residues distant from the binding site
(30, 31).
Our results obtained within the fine-tuning of the testoster-
one binding site also indicate that positive binding site muta-
tions are not easily evolved during in vitro affinity and speci-
ficity maturation. In the case of anti-testosterone antibody,
only one important binding site residue, namely, CDR-H3 Glu-
95, is mutated to alanine in the 77 Fab. This mutation has two
important consequences: it eliminates the need for side chain
rotation of Glu-95 observed to take place in the wt Fab, and it
provides more free space for the testosterone D-ring (Fig. 6).
However, very recently, Short et al. (41) showed that in the case
of the anti-digoxin 26-10 antibody, the binding site could be
reordered by mutations of two important contact residues and
still retain the high binding affinity. Smaller changes (fine-
tuning) can be achieved via the cooperation of mutations that
are not directly in contact with the hapten. The rational design
of these kinds of mutations is a very difficult task because the
number of randomized possibilities increases greatly. There-
fore, different random mutagenesis approaches combined with
phage or ribosome display selections and in vitro evolution
strategies are valuable tools for the in vitro maturation of
antibody binding sites.
Acknowledgments—We thank Armi Boman, Maarit Nyman, and
Reetta Kallio-Ratilainen for excellent technical assistance and Lasse
Hautoniemi for providing the steroid conjugates.
44027
REFERENCES
1. Winter, G., Griffiths, A. D., Hawkins, R. E., and Hoogenboom, H. R. (1994)
Annu. Rev. Immunol. 12, 433– 455
2. Hoogenboom, H. R. (1997) Trends Biotechnol. 15, 62–70
3. Rader, C., and Barbas, C. F., III. (1997) Curr. Opin. Biotechnol. 8, 503–508
4. Chames, P., Coulon, S., and Baty, D. (1998) J. Immunol. 161, 5421–5429
5. Saviranta, P., Pajunen, M., Jauria, P., Karp, M., Pettersson, K., Mäntsälä, P.,
and Lövgren, T. (1998) Protein Eng. 11, 143–152
6. Lamminmäki, U., Paupério, S., Westerlund-Karlsson, A., Karvinen, J.,
Virtanen, P. L., Lövgren, T., and Saviranta, P. (1999) J. Mol. Biol. 291,
589 – 602
7. Hemminki, A., Niemi, S., Hoffrén, A.-M., Hakalahti, L., Söderlund, H., and
Takkinen, K. (1998) Protein Eng. 11, 311–319
8. Hemminki, A., Niemi, S., Hautoniemi, L., Söderlund, H., and Takkinen, K.
(1998) Immunotechnology 4, 59 – 69
9. Valjakka, J., Takkinen K., Teerinen, T., Söderlund, H., and Rouvinen J. (2002)
J. Biol. Chem. 277, 4183– 4190
10. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S., and Foeller, C. (1991)
in Sequences of Protein of Immunological Interest, 5th ed., National Insti-
tutes of Health, Bethesda, MD
11. Hermes, J. D., Parekh, S. M., Blacklow, S. C., Köster, H., and Knowles, J. R.
(1989) Gene (Amst.) 84, 143–151
Downloaded from www.jbc.org at JOENSUUN YLIOPISTON KIRJASTO on September 4, 2007
12. Hermes, J. D., Blacklow, S. C., and Knowles, J. R. (1990) Proc. Natl. Acad. Sci.
U. S. A. 87, 696 –700
13. Barbas, C. F., III, Kang, A. S., Lerner, R. A., and Benkovic, S. J. (1991) Proc.
Natl. Acad. Sci. U. S. A. 88, 7978 –7982
14. Hawkins, R. E., Russell, S. J., and Winter, G. (1992) J. Mol. Biol. 226, 889 – 896
15. Takkinen, K., Laukkanen, M.-L., Sizmann, D., Alfthan, K., Immonen, T.,
Vanne, L., Kaartinen, M., Knowles, J. K. C., and Teeri, T. T. (1991) Protein
Eng. 4, 837– 841
16. Jönsson, U., Fägerstam, L., Ivarsson, B., Johnsson, B., Karlsson, R., Lundh,
K., Löfas, S., Persson, B., Roos, H., Rönnberg, I., Sjölander, S., Stenberg, E.,
Ståhlberg, R., Urbaniczky, C., Östlin, H., and Malmqvist, M. (1991) Bio-
Techniques 11, 620 – 627
17. Karlsson, R., Michaelsson, A., and Mattson, A. (1991) J. Immunol. Methods
145, 229 –240
18. Valjakka J., Hemminki A., Teerinen T., Takkinen, K., and Rouvinen, J. (2000)
Acta Crystallogr. Sect. D Biol. Crystallogr. D 56 218 –221
19. Otwinowski, Z., and Minor, W. (1997) Methods Enzymol. 276, 307–326
20. Navaza, J. (1994) Acta Crystallogr. 50, 157–163
21. Jones, T. A., Zou, J.-Y., Cowan, S. W., and Kjeldgaard, M. (1991) Acta Crys-
tallogr. Sect. A 47, 110 –119
22. 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
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., and Bourne, P. E. (2000) Nucleic Acids Res. 28, 235–242
25. Yang, W.-P., Green, K., Pinz-Sweney, S., Briones, A. T., Burton, D. R., and
Barbas, C. F. (1995) J. Mol. Biol. 254, 392– 403
26. Schier, R., McCall, A., Adams, G. P., Marshall, K. W., Merritt, H., Yim, M.,
Crawford, R. S., Weiner, L. M., Marks, C., and Marks, J. D. (1996) J. Mol.
Biol. 263, 551–567
27. Chen, Y., Wiesmann, C., Fuh, G., Li, B., Christinger, H. W., McKay, P., de Vos,
A. M., and Lowman, H. B. (1999) J. Mol. Biol. 293, 865– 881
28. Gram, H., Marconi, L. A., Barbas, C. F., Collet, T. A., Lerner, R. A., and Kang,
A. S. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 3576 –3580
29. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946 –950
30. Daugherty, P. S., Chen, G., Iverson, B. L., and Georgiou, G. (2000) Proc. Natl.
Acad. Sci. U. S. A. 97, 2029 –2034
31. Boder, E. T., Midelfort, K. S., and Wittrup, K. D. (2000) Proc. Natl. Acad. Sci.
U. S. A. 20, 10701–10705
32. Hanes, J., and Plückthun, A. (1997) Proc. Natl. Acad. Sci. U. S. A. 94,
4937– 4942
33. Hanes, J., Jermutus, L., Weber-Bornhauser, S., Bosshard, H. R., and
Plückthun, A. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 14130 –14135
34. Jermutus, L., Honegger, A., Schwesinger, F., Hanes, J., and Plückthun, A.
(2001) Proc. Natl. Acad. Sci. U. S. A. 98, 75– 80
35. Arevalo, J. H., Taussig, M. J., and Wilson, I. A. (1999) Nature 365 859 – 863
36. 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
37. Trinh, C. H., Hemmigton, S. D., and Verhoeyen, M. V. (1997) Structure 5
937–948
38. Yang, P., L., and Schultz, P. (1999) J. Mol. Biol. 294, 1191–1201
39. Evans, S. V. (1993) J. Mol. Graphics 11, 134 –138
40. McRee, D. E. (1999) XtalView 4.0, The Scripps Research Institute, San Diego,
CA
41. Short, M. K., Krykbaev, R. A., Jeffrey, P. D., and Margolies, M. N. (2002) J.
Biol. Chem. 277, 16365–16370