Báo cáo khoa học: The role of interface framework residues in determining antibody VH ⁄ VL interaction strength and antigen-binding affinity pptx

The role of interface framework residues in determining
antibody V
H
⁄ V
L
interaction strength and antigen-binding
affinity
Kenji Masuda
1
, Kenzo Sakamoto
3
, Miki Kojima
1,2,3
, Takahide Aburatani
3
, Takuya Ueda
1,2
and Hiroshi Ueda
1,2,3,4
1 Department of Integrated Biosciences, Graduate School of Frontier Sciences, University of Tokyo, Chiba, Japan
2 Department of Medical Genome Sciences, Graduate School of Frontier Sciences, University of Tokyo, Chiba, Japan
3 Department of Chemistry and Biotechnology, School of Engineering, University of Tokyo, Tokyo, Japan
4 PRESTO, JST, Kawaguchi, Saitama, Japan
Antibody plays a pivotal role in the humoral immune
response, primarily through the binding of its variable
region to its specific antigen with high affinity. In par-
ticular, the antibody variable region (Fv) and its deriv-
atives are receiving increasing attention in many areas,
including diagnostics and therapy, primarily because of
their relative ease of production by many systems,
including microbial culture. However, because of their

heterodimeric domain structure and weak heavy chain
variable region fragment (V
H
) ⁄ light chain variable
region fragment (V
L
) interaction, Fv and ⁄ or single-
chain Fv (scFv) often show problematic physicochemical
Keywords
antibody variable region; antigen–antibody
interaction; combinatorial mutagenesis;
immunoassay; phage display
Correspondence
H. Ueda, Department of Chemistry and
Biotechnology, School of Engineering,
University of Tokyo, Tokyo 113–8656, Japan
Fax: +81 3 5841 7362
Tel: +81 3 5841 7362
E-mail:
(Received 17 January 2006, revised 14
March 2006, accepted 16 March 2006)
doi:10.1111/j.1742-4658.2006.05232.x
While many antibodies with strong antigen-binding affinity have stable
variable regions with a strong antibody heavy chain variable region frag-
ment (V
H
) ⁄ antibody light chain variable region fragment (V
L
) interaction,
the anti-lysozyme IgG HyHEL-10 has a fairly strong affinity, yet a very

weak V
H
⁄ V
L
interaction strength, in the absence of antigen. To investigate
the possible relationship between antigen-binding affinity and V
H
⁄ V
L
inter-
action strength, a novel phage display system that can switch two display
modes was employed. We focused on the two framework region 2 regions
of the HyHEL-10 V
H
and V
L
, facing each other at the domain interface,
and a combinatorial library was made in which each framework region 2
residue was mixed with that of D1.3, which has a far stronger V
H
⁄ V
L
inter-
action. The phagemid library, encoding V
H
gene 7 and V
L
amber codon
gene 9, was used to transform TG-1 (sup
+

), and the phages displaying
functional variable regions were selected. The selected phages were then
used to infect a nonsuppressing strain, and the culture supernatant contain-
ing V
H
-displaying phages and soluble V
L
fragment was used to evaluate
the V
H
⁄ V
L
interaction strength. The results clearly showed the existence of
a key framework region 2 residue (H39) that strongly affects V
H
⁄ V
L
inter-
action strength, and a marked positive correlation between the antigen-
binding affinity and the V
H
⁄ V
L
interaction, especially in the presence of a
set of particular V
L
residues. The effect of the H39 mutation on the wild-
type variable region was also confirmed by a SPR biosensor as a several-
fold increase in antigen-binding affinity owing to an increased association
rate, while a slight decrease was observed for the single-chain variable

region.
Abbreviations
FR2, framework region 2; Fv, antibody variable region; HEL, hen egg lysozyme; OS, open sandwich; scFv, single-chain Fv; spFv, split
antibody variable region; V
H
, antibody heavy chain variable region fragment; V
L
, antibody light chain variable region fragment.
2184 FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS
behavior, even if parental antibody retains superb sta-
bility and affinity. For example, many Fv dissociate
into the two domains at low protein concentrations
and are too unstable for many applications at physio-
logical temperature [1]. Also, scFv are prone to spon-
taneous dimerization and aggregation, as a result of
their weak V
H
⁄ V
L
interaction, as well as their exposed
interconstant domain surface [2–4]. Moreover, some
Fv lose their affinity by tethering with the interdomain
linker, probably because of local or global conforma-
tional change [5].
On the other hand, previously we found that the Fv
domain of anti-hen egg lysozyme (HEL) IgG (HyHEL-
10) has fairly strong antigen-binding affinity (K
a
¼
2.5 · 10

8
Æm
)1
) [6] yet very weak V
H
⁄ V
L
interaction
strength (K
a
<10
5
Æm
)1
) in the absence of antigen [7]. The
stability of the Fv–HEL complex was later shown to be
maintained by many water-mediated hydrogen bonds at
the imperfect V
H
and V
L
interface [8]. In contrast, in the
case of another anti-HEL IgG – D1.3 – the V
H
⁄ V
L
interaction is very strong (K
a
¼ 10
10

Æm
)1
) even in the
absence of antigen [9]. Clearly, there should be struc-
tural differences between these two Fvs that make their
heterodimeric interaction strength quite different. How-
ever, there has been no attempt to clarify experimentally
the determinant of V
H
⁄ V
L
interaction strength or its
relationship with antigen-binding affinity. Previously,
the effect of mutations of the V
H
⁄ V
L
interface residues
on antigen binding has been studied using Fab frag-
ments [10,11]. However, while the primary effect of such
mutations should be altered V
H
⁄ V
L
interaction strength,
in these studies the effect was unclear because of the
covalently and noncovalently interacting C-terminal
constant domains. While Fv and its derivatives are
widely used, to date no attempt has been made to clarify
the effect of the interface mutations on the antigen-

binding affinity. To investigate systematically this
unsought relationship between the antigen-binding
affinity of Fv and the V
H
⁄ V
L
interaction strength, in the
present study we employed a novel phage display system
that can switch two display modes suitable for evaluat-
ing either Fv-antigen or V
H
⁄ V
L
interactions [12].
As a target for the analysis, we focused on the two
framework region 2 (FR2) regions of V
H
and V
L
, each
facing the domain interface of the anti-lysozyme IgG,
HyHEL-10, a weak V
H
⁄ V
L
binder. FR2 regions have
been implicated to play an important role in V
H
⁄ V
L

interaction [13]. In particular, dromedary antibodies
without light chains show characteristic sequence alter-
ation in V
H
FR2 to increase hydrophilicity, and are
stable without pairing with V
L
[14]. Here, we made a
combinatorial library in which each FR2 residue enco-
ded was mixed with that of D1.3, a strong V
H
⁄ V
L
binder, to describe the relationship, and also to iden-
tify key residues in determining the interdomain inter-
action strength.
Results
Construction of an FR2 combinatorial library
To investigate the possible relationship between anti-
gen-binding affinity and V
H
⁄ V
L
interaction strength, a
novel phage display system that can switch two display
modes, named the split Fv (spFv) system, was
employed. The spFv system can either display V
H
and
V

L
fragments on the N termini of M13 phage coat
proteins p9 and p7, respectively, when an amber sup-
pressor strain is used for the phage production, or dis-
play V
H
fragment on the phage p9, and simultaneously
secrete a his-myc tagged V
L
fragment to the culture
supernatant, when a nonsuppressing strain is used as a
host (Fig. 1) [12]. The ability to switch the two display
modes enables side-by-side evaluation of antigen-bind-
ing ability and V
H
⁄ V
L
interaction strength of the tar-
get Fv, and also rapid screening of Fv that is suitable
for open sandwich (OS) immunoassay, an immuno-
assay that utilizes antigen-dependent Fv stabilization
[7]. The potential advantage of the spFv system over
the conventional display system of monodomains on
phage p3 was that it allowed simultaneous mutagenesis
of V
H
and V
L
domains, and also rapid evaluation of
antigen-binding affinity of phage-displayed Fv frag-

ment.
For the efficient library selection with the spFv sys-
tem, first, the original phagemid pKS1 was modified to
stabilize the inserted sequence by inserting two HP ter-
minators [15] both upstream and downstream of the
spFv coding region, to yield a new vector, pKST2.
Based on this vector, a combinatorial library of
HyHEL-10 spFv, where each FR2 residue encoded
was mixed with a D1.3-type residue, was constructed.
HyHEL-10 Fv was chosen as a model, not only
because it has very low V
H
–V
L
affinity, yet retains high
antigen-binding affinity, but also it was efficiently dis-
played as spFv or as sole V
H
on phage, together with
soluble V
L
. Out of 16 heavy and 14 light chain FR2
residues, according to the Kabat database [16], 7 and 6
positions, respectively, are different between HyHEL-10
and D1.3. To make a combinatorial library, two oligo-
nucleotides encoding degenerate codons coding for
both types of amino acids for these positions were
used to amplify the 5¢ half of V
H
and the 3¢ half of V

L
fragments with degenerated FR2 residues (Fig. 2).
Using a linker DNA fragment connecting these two,
an overlap-extension PCR was carried out to yield
K. Masuda et al. V
H
⁄ V
L
interaction and affinity
FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS 2185
the insert DNA, with a total theoretical diversity of
2.6 · 10
5
. The restriction enzyme-digested insert was
ligated with pKST2 digested with the same restriction
enzymes, which was used to transform electrocompe-
tent TG-1(sup
+
) cells. This resulted in a library of
colonies with an estimated size of 7 · 10
7
, which was
considered large enough to cover the theoretical diver-
sity. The Fv-displaying phages were thus prepared
from the harvested cells.
Selection of specific antigen binders
To enrich phages displaying functional Fv, a round
of biopanning was performed on immobilized antigen
HEL. As many combinations of mutations on the
conserved FR2 sequence may destabilize the Fv struc-

ture, and it might be difficult to remove the resultant
nonspecific binders by repeated pannings, we adopted
a combination of a round of biopanning and subse-
quent screening of functional Fv by ELISA. A mini-
mum round of panning would allow selection of
binders with a variety of affinity to antigen, which
might otherwise be lost. After the panning, 1536 colon-
ies were picked, and the corresponding monoclonal
phages were prepared on 96-well plates and screened
by phage ELISA on HEL-immobilized (specific) and
nonimmobilized (blank) wells. Among them, 72 clones
showing sufficient affinity and specificity with an
absorbance of > 0.3, and more than fivefold the blank
absorbance, were re-examined for their antigen specif-
icity. The display efficiencies of V
H
and V
L
fragments
Fig. 2. The framework region 2 (FR2) library. The sequence for FR2
residues, different between HyHEL-10 and D1.3, were mixed to
encode both types of amino acids. As a result of degenerated
codon usage, five out of 13 positions encoded two other codons.
Fig. 1. Split antibody variable region (spFv)
system. (A) Structure of the spFv phagemid
coding region. The vector uses pIX and pVII
of M13 phage to display the antibody heavy
chain variable region fragment (V
H
) and anti-

body light chain variable region fragment
(V
L
) on the phage, respectively. An amber
codon for the switch of display ⁄ secretion of
the V
L
fragment is marked by an arrowhead.
(B) Display of antibody variable region (Fv)
with sup
+
strain (TG1) as a host. Antigen-
binding affinity of the Fv can be evaluated.
(C) Display of the V
H
and secretion of
tagged V
L
with sup
–
strain (HB2151) as a
host. Measurement of the V
H
⁄ V
L
interaction
on the plate immobilized with anti-tag immu-
noglobulin is possible.
V
H

⁄ V
L
interaction and affinity K. Masuda et al.
2186 FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS
were confirmed by ELISA with immobilized anti-flag
and anti-myc IgG, respectively. From these analyses,
64 clones were confirmed to show more than eightfold
specific absorbance than the blank, and for sufficient
display of the two fragments. The phagemids for these
clones were extracted from the stock strain and their
nucleotide sequences were determined. Because the
clones containing amber codons were not suitable for
subsequent analysis with the nonsuppressing strain, 36
clones without any amber codons in the FR2 region
were chosen and used for further analyses.
Evaluation of relative antigen-binding affinity
The relative antigen-binding affinity of 36 clones was
evaluated by phage ELISA after setting the titer of
each clone to 2.5 · 10
8
,1· 10
9
, and 4 · 10
9
colony-
forming units (CFU)ÆmL
)1
. As the widest range of dis-
tribution in absorbance was observed at 1 · 10
9

CFUÆmL
)1
, we decided to compare the antigen-bind-
ing affinity at this titer, and to use the ratio of specific
absorbance minus background absorbance at this titer
against that of wild-type as an index of the relative
affinity to antigen. The ELISA results of representative
three clones are shown in Fig. 3A. Both clones with
higher and lower signals than the wild-type were
observed at similar frequencies. The ELISA signals
and FR2 sequences of all the mutants and the wild-
type, sorted by this index, are summarized in Table 1.
It is worth noting that while two clones with low affin-
ity (1D5 and 4F1) had W at H47, the other 34 clones
had Y at this position, with generally higher affinity. It
is possible that weak binders with W at this position
were counterselected by the biopanning. In addition,
among the 13 highest antigen binders, eight shared a
common four V
L
residues (L41G, L45R, L48V and
L49K), and the combination was not observed for
lower-affinity clones. In addition, 10 out of 13 clones
shared three common residues (L41G, L45R and
L49K), which were not observed in weaker binders.
Evaluation of the V
H
⁄ V
L
interaction strength

To evaluate the V
H
⁄ V
L
interaction strength of these 36
clones, phages were used to infect a nonsuppressing
strain, HB2151, to produce culture supernatant con-
taining V
H
-displaying phage and myc-tagged soluble
V
L
fragment. The culture supernatant was then applied
to either microplate wells immobilized with anti-myc
IgG (specific) or nonimmobilized wells (blank), washed,
and probed with horseradish peroxidase (HRP)-labeled
anti-phage IgG. The specific absorbance minus the
blank absorbance was taken as an index of V
H
⁄ V
L
interaction strength (Table 1). Also, to evaluate the
antigen dependency of the interaction, HEL at three
concentrations was included in the culture supernatant
before ELISA. The results for the OS ELISA of the
representative clones are shown in Fig. 3B. While some
clones showed a similar, or even superior, antigen-
dependent increase in absorbance, others showed a
decreased or diminished antigen-dependency. To evalu-
ate the OS-fitness of the clone, the ratio of the specific

absorbance in the presence of 10 lgÆmL
)1
HEL to that
in the absence of HEL was taken as an index.
To analyze the effect of the type of each FR2 resi-
due on each index, the Student’s t-test was performed
(Supplementary material Table S1). According to the
test, it was clear that the residue type of H39 domi-
nantly affects both V
H
⁄ V
L
interaction strength and the
OS-fitness. When H39 was Lys (K), as in HyHEL-10,
the V
H
⁄ V
L
interaction in the absence of antigen was
generally weak, while the OS-fitness was high. On the
Fig. 3. Representative clones obtained after panning. (A) Phage
ELISA at 1 · 10
9
colony-forming units (CFU) per mL with ⁄ without
immobilized antigen. (B) Open sandwich (OS) ELISA where the
soluble variable region fragment (V
L
) was immobilized with anti-
myc immunoglobulin. Binding of the heavy chain variable region
fragment (V

H
)-phage in the presence ⁄ absence of hen egg lysozyme
(HEL) was detected with horseradish peroxidase (HRP)-anti-M13.
K. Masuda et al. V
H
⁄ V
L
interaction and affinity
FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS 2187
contrary, when H39 was Gln (Q), as in D1.3, the
V
H
⁄ V
L
interaction was generally strong, while the
OS-fitness was close to 1.
Relationship between antigen binding
and V
H
⁄ V
L
interaction
The relationship between the indexes of antigen-bind-
ing affinity and V
H
⁄ V
L
interaction strength was plotted
(Fig. 4). When the plot was classified by the type of
H39, a clear trend was observed in that the clones with

a stronger V
H
⁄ V
L
interaction had Gln at H39 (H39Q
group), and those with a weaker interaction had Lys
at H39 (H39K group). While there appeared to be no
strong correlation between the antigen-binding affinity
and the V
H
⁄ V
L
interaction strength determined, five
clones in the H39Q group showed both high antigen-
binding affinity and strong V
H
⁄ V
L
interaction. In
Table 1. Result of phage ELISA at 1 · 10
9
colony-forming units (CFU) per mL against hen egg lysozyme (HEL)-immobilized and blank wells,
and antibody heavy chain variable region fragment (V
H
) ⁄ antibody light chain variable region fragment (V
L
) interaction strength without HEL
(amyc-blank) determined by the split variable region (split Fv) system. The results are sorted by the HEL-blank value, and shown with partial
framework region 2 (FR2) sequences. HyHEL-10-type, D1.3-type and other residues are shown in roman, hatched and in italic, respectively.
Residues that have possible relationship with antigen binding affinity are shown in bold. Values for the wild-type HyHEL-10 are underlined.

V
H
⁄ V
L
interaction and affinity K. Masuda et al.
2188 FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS
addition to Gln at H39, these clones shared four com-
mon residues (L41G, L45R, L48V and L49K) in V
L
,
similar to other high-affinity clones.
The relationship between the indexes of antigen-bind-
ing affinity and the OS-fitness was also plotted (Fig. 5).
In this plot, a clearer H39-dependency was observed,
possibly because the OS-fitness as an absorbance ratio
contained less experimental error owing to the expres-
sion levels of the V
H
⁄ V
L
fragments. Apparently, all the
H39Q members show limited antigen-dependency in
V
H
⁄ V
L
interaction strength. On the contrary, for the
clones in the H39K group, a clone with higher affinity,
as well as higher OS-fitness, was observed, while many
other types of mutants were also observed.

SPR analysis of antigen-binding affinity
For the quantitative evaluation of the H39 mutation,
kinetic analysis for antigen–Fv interaction of the
wild-type and H39KQ mutant of purified Fv and scFv
proteins was performed using an SPR biosensor. As
shown in Table 2, in the concentration range of
50–100 nm where the wild-type V
H
and V
L
are fully
dissociated [7], H39HQ mutant Fv showed 25-fold and
approximately eightfold higher association and dissoci-
ation rate constants, respectively, than the wild-type
Fv, which resulted in a 3.7-fold higher equilibrium
association constant. On the contrary, the scFv with
the H39KQ mutation showed a similar or reduced
association rate and a similar or higher dissociation
rate than the wild-type scFv, which resulted in a 0.58-
fold equilibrium association constant. Apparently, the
mutation to strengthen the V
H
⁄ V
L
interaction was
almost as effective as tethering by the (G
4
S)
3
linker

used in scFv, but no synergistic effect in antigen-bind-
ing affinity was observed.
Discussion
In the present study, we showed a functional analysis
of FR2 residues for the antigen-binding affinity as well
as V
H
⁄ V
L
interaction based on the selected clones from
a combinatorial library. Through the construction of a
sufficient size of combinatorial library and subsequent
analysis, it became clear that a residue near the bottom
of the FR2 loop determines V
H
⁄ V
L
interaction
strength, as well as its dependency on antigen binding.
The importance of H39 in Fv stability has been des-
cribed for the Fv of M29 antibody [17]. Although the
Fv was designed based on HyHEL-10, it is not clear
whether or not H39 is dominantly tuning the V
H
⁄ V
L
interaction of other Fvs, including HyHEL-10. In
addition, the effect of the H39 mutation on antigen
binding has not yet been analyzed. The reason for gen-
erally weak, and stronger, V

H
⁄ V
L
interaction of H39K
and H39Q group Fvs, respectively, may be ascribed
to their ability to form interchain hydrogen bonds
(Fig. 6). While no interchain hydrogen bonds origin-
ating from H39 lysine are observed in the crystal
Fig. 4. Scattered plot of V
H
and V
L
interaction against relative affin-
ity of the Fv to the antigen. The plot is classified by the type of
H39, as indicated.
Fig. 5. Scattered plot of OS-fitness against relative affinity of the
Fv to the antigen classified as in Fig. 4.
Table 2. Kinetic parameters of the wild-type (WT) and H39KQ
mutant in variable region (Fv) and single chain Fv (scFv) formats.
k
on
(10
4
ms
)1
) k
off
(10
)5
s

)1
) K
a
(10
8
M
)1
)
Fv
WT 0.35 ± 0.08 1.02 ± 0.01 3.83 ± 0.83
H39KQ 8.75 ± 1.39 7.93 ± 2.69 12.4 ± 5.6
H39KQ ⁄ WT 25.3 ± 4.0 7.75 ± 2.63 3.67 ± 1.66
scFv
WT 8.31 ± 0.20 8.61 ± 3.96 11.2 ± 5.1
H39KQ 7.10 ± 0.44 12.0 ± 4.35 6.47 ± 2.53
H39KQ ⁄ WT 0.85 ± 0.05 1.40 ± 0.51 0.58 ± 0.23
K. Masuda et al. V
H
⁄ V
L
interaction and affinity
FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS 2189
structure of HyHEL-10, two interchain hydrogen
bonds are formed between two glutamines of H39 and
L38 in the structure of D1.3. In addition, an additional
interchain hydrogen bond (H39Q–L87Y) is possible in
the latter structure.
Both H39 and the corresponding V
L
residue, L38,

are nearly conserved on the genome, and 93% are glu-
tamine in 5355 expressed V
H
and V
L
sequences [18].
Probably, a major part of natural Fvs are stabilized by
the hydrogen bonds between them. During B-cell
development, clones expressing both chains with suffi-
ciently strong interchain (H–L) interaction are believed
to be selected [19]. However, as a result of covalent
linkage of the variable domains through the C-terminal
constant domains, not much is known about the distri-
bution of the V
H
–V
L
interaction strength in natural
B-cell repertoire. Although our model study suggests
expression of clones with a variety of V
H
–V
L
interac-
tion strengths, further study is needed to analyze the
distribution of natural repertoire. The five strong anti-
gen binders shared a common V
L
FR2 sequence in
addition to H39Q. In this group of Fvs, a weak posit-

ive correlation between antigen-binding affinity and
V
H
⁄ V
L
interaction strength, in other words, an appar-
ent positive correlation of the stability of the Fv–anti-
gen complex and that of Fv in the absence of antigen,
was observed. Because these four V
L
FR2 residues seem
to enhance the antigen-binding affinity, irrespective of
H39 type, these V
L
are optimized for high-affinity anti-
gen binding through the mutation at remote sites.
When both V
H
and V
L
fragments are optimized for
antigen binding, this ‘increasing the affinity by increas-
ing the V
H
⁄ V
L
interaction’ might represent a mechan-
ism of increasing Fv affinity. This is also supported by
the kinetic study of purified H39KQ mutant Fv frag-
ment, where enhancement in V

H
⁄ V
L
interaction signifi-
cantly improved the antigen-binding affinity of Fv,
similarly to the level of scFv (Table 2).
Some single V
H
domains of anti-protein IgG, inclu-
ding HyHEL-10 and D1.3, were known to retain the
specificity and affinity to antigen in itself [7,20]. At
least for these antibodies, the role of the V
L
domain
might be to increase the affinity by supporting the V
H
domain. The phenomenon may also be observed in
the chain shuffling experiments. As antigen-induced
V
H
⁄ V
L
rearrangement is also observed for several Fab
fragments [21–23], further verification of the hypothe-
sis is needed. However, this will not be difficult if the
spFv system is utilized.
In the present study, we demonstrated that library
screening is indeed possible using the spFv system. The
results presented suggest that the V
H

⁄ V
L
interaction
strength can be effectively engineered by substituting
H39 (or L38). This, in turn, suggests that antibodies
previously considered unsuitable to OS-immunoassay
can be converted to suitable antibodies by a point
mutation (H. Ueda, unpublished results). Screening of
the mutants with minimal reduction in antigen-binding
affinity may also be possible by applying this FR2
engineering approach.
Experimental procedures
Materials
The Escherichia coli strains used were XL-10 Gold (Strata-
gene, La Jolla, CA, USA) for general cloning, TG1 [supE,
hsdD5, thi, D (lac-proAB), ⁄ F¢ traD36, proAB
+
, lacI
q
, lac-
ZDM15] (Amersham Bioscience, Tokyo, Japan) and
HB2151 [ara, D (lac-proAB), thi ⁄ F¢ proAB
+
, lacI
q
, lac-
ZDM15] (Amersham Bioscience) for phage display. Restric-
tion and modification enzymes were from Takara-Bio
(Otsu, Japan), or New England Biolabs (Ipswich, MA,
USA). Oligonucleotides were from Espec Oligos (Tsukuba,

Japan).
Construction of the spFv phagemid
The spFv expression vector for anti-hen egg lysozyme
HyHEL-10 Fv, pKS1(HyHEL-10), was constructed as des-
cribed previously [12]. To add an SfiI cloning site upstream
A
B
V
L
Gln37
Phe87
Gln38
Lys39
Tyr94
Arg38
Lys39
Phe40
V
H
V
L
Gln37
Tyr87
Gln38
Lys39
Tyr94
Arg38
Gln39
Pro40
V

H
Fig. 6. 3D structures of HyHEL-10 (A) and D1.3 (B) around H39.
Possible hydrogen bonds, calculated by
SWISSPDB VIEWER [26], are
shown as dotted lines.
V
H
⁄ V
L
interaction and affinity K. Masuda et al.
2190 FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS
of NcoI at the 5¢ end of the V
H
sequence, the NcoI–EcoRI
fragment of pKS1(HyHEL-10) was transferred to pCan-
tab5E (Amersham Bioscience), denoted pKS2(HyHEL-10).
To avoid instability of the spFv fragment, possibly as a
result of basal expression of VH-p9 and VL-p7 fusion pro-
teins before induction, two glutamine permease terminators
(tHP) [15] were incorporated upstream and downstream of
the spFv coding sequence. To insert the terminator into the
SapI site upstream of the lac promoter, four 5¢-phosphoryl-
ated oligonucleotides – tHP1 (5¢-AGCGGTACCCGATA
AAAGCGGCTTCCTGAC-3¢), tHP2 (5¢-AGGAGGCCG
TTTTGTTTTGCAGCCCACCTC-3¢), tHP3 (5¢-GCTG
AGGTGGGCTGCAAAACAAAACGGCCT-3¢) and tHP4
(5¢-CCTGTCAGGAAGCCGCTTTTATCGGGTACC-3¢)–
were annealed and ligated to SapI-digested pKS2(HyHEL-
10). To insert tHP downstream of the ORFs, tHP4 and
tHP7 (5¢-AATTGGTACCCGATAAAAGCGGCTTCCTG

AC-3¢), as well as tHP2 and tHP8 (5¢-AATTGAGG
TGGGCTGCAAAACAAAACGGCCT-3¢), were annealed
and ligated to the EcoRI-digested plasmid, described above,
resulting in pKST2(HyHEL-10).
Construction of the FR2 library
A combinatorial library (in which each FR2 residue enco-
ded was mixed with that of D1.3) was constructed by over-
lap extension PCR as follows. First, a DNA fragment
encoding the N-terminal 55 residues (H1–H55) of HyHEL-
10 V
H
, whose FR2 residues were designed to be either
HyHEL-10 or D1.3 type (V
H
FR2), was amplified with
primers MH2BackSfi (5¢-GTCCTCGCAACTGCGGCCC
AGCCGGCCATGGCCSARGTNMAGCTGSAGSAGTC
WGG-3¢) and H10VHframe2 (5¢-ACCACTGTAGCTT
ACGTACCCCAWSYACTCCAGACSKTTACCTGGARR
TTKACGAAYCCAGCTCCAATAATCACTGGT-3¢) with
pKST2(HyHEL-10) as a template. Also, the fragment
encoding L30-L107 of HyHEL-10 with diversified FR2
residues (V
L
FR2) was similarly amplified with primers
H10VLframe2 (5¢-GGCAACAACCTACACTGGTATCA
ACAAAAAYMGSRCRAATCTCCTCRGCTCCTGRTCW
AKTATGCTTCCCAGTCCATCTCT-3¢) and g7EcoFor
(5¢-AGTGAATTCTCATCTTTGACCCCCAGCGATTAT
ACCAA-3¢). As a linker to connect these two, a DNA

encoding H49 to L39 with intervening sequences (link-
erFR2) was also amplified with primers H10linkRV
(5¢-GGGTACGTAAGCTACAGTG-3¢) and H10linkFR
(5¢-GATACCAGTGTAGGTTG-3¢). The three fragments
(V
H
FR2, V
L
FR2, and linker FR2) were assembled by
splice overlap extension (SOE)-PCR as follows: a thermal
cycling without primer (94 °C for 5 min, followed by seven
cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for
1 min), followed by a normal cycling (30 cycles of 94 °C
for 30 s, 55 °C for 30 s, and 72 °C for 1 min) with primers
MH2BackSfi and g7EcoFR, and Ex Taq DNA polymerase
(Takara-Bio).
The amplified 0.9 kb fragment encoding both V
H
and V
L
(split Fv fragment) was recovered from a 1.5% agarose gel,
digested with NcoI and NotI, repurified on the gel and
ligated with pKST2 digested with the same enzymes. The
ligation mix was electroporated to E. coli TG-1, and plated
on a YTAG agar plate (8 gÆL
)1
tryptone, 5 gÆL
)1
yeast
extract, 5 gÆL

)1
NaCl, 100 lgÆmL
)1
ampicillin, 1% glucose,
15 gÆL
)1
agar). Several colonies were selected for extraction
of the phagemid to check the quality of the library. Nucleo-
tide sequencing was performed using a 3100 Genetic
Analyzer (Applied Biosystems, Tokyo, Japan) and BigDye
Terminator Cycle Sequencing Kit (Applied Biosystems)
with primers M13RV (Takara-Bio) and OmpARV
(5¢-ACAGCTATCGCGATTGCAGTG-3¢).
Preparation of phage library pKST2 vector digested with
NcoI and NotI (2.13 pmol), and the split Fv fragment diges-
ted with the same (21.3 pmol), was ligated by adding 0.5 vol
of Ligation High (Toyobo, Osaka, Japan) and incubated
at 16 °C for 2 h. After ethanol precipitation, the pellet was
dissolved in 20 lL of Milli-Q water. The solution was divi-
ded into two, each was mixed with 100 lL of electrocompe-
tent TG-1 cells, and electroporated with Easyject (EquiBio,
Ashford, UK) with 2 mm cuvettes. Then, 900 lL of 2YT
medium was added, and the solution was transferred to a
microtube and incubated at 37 °C for 30 min. One microli-
tre was taken for colony counting after serial dilution, and
the rest was plated onto a YTAG agar large square plate
(Sumitomo Bakelite, Tokyo, Japan), and incubated at 30 °C
for 16 h. The TG-1 colonies on the plate were harvested
with 5 mL of 2YT (16 gÆL
)1

tryptone, 10 gÆL
)1
yeast
extract, 5 gÆL
)1
NaCl, pH 7.6), mixed and stored at )80 °C
after adding 0.5 vol of 50% glycerol, except for 50 lL,
which was used to inoculate 100 mL of 2YTAG, and
shaken at 37 °C until the attenuance (D) at 600 nm reached
0.5. Then, 10 mL of culture was added with helper phage,
M13KO7, at a multiplicity of infection (m.o.i.) of 20, incu-
bated without shaking at 37 °C for 30 min, and centrifuged
in a desktop centrifuge at 760 g for 15 min at 4 °C. After
removal of the supernatant, the pellet was resuspended in
50 mL of 2YT containing 100 lgÆmL
)1
ampicillin and
50 lgÆmL
)1
kanamycin (2YTAK) medium in a 500 mL baf-
fled flask, and vigorously shaken at 30 °C, 250 r.p.m. for
20 h. After incubation, the culture was centrifuged at 6500 g
for 10 min, 0.2 vol of 20% polyethylene glycol 6000 ⁄ 2.5 m
NaCl (PEG ⁄ NaCl) was added to the supernatant and incu-
bated on ice for 1 h. After centrifugation (6500 g,4°C,
30 min), the pellet was resuspended in 1 mL of 10 mm
Tris ⁄ HCl, 1 mm EDTA, pH 8.0 (TE), centrifuged at
15 000 g,4°C for 20 min, and the supernatant was used as
the phage library.
Biopanning and phage ELISA

Thirty wells of Falcon 3912 microplate (Becton Dickinson,
Oxnard, CA, USA) were coated overnight with 100 lL per
K. Masuda et al. V
H
⁄ V
L
interaction and affinity
FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS 2191
well of 10 lg ÆmL
)1
HEL in 10 mm NaCl ⁄ P
i
, blocked at
room temperature for 2 h with NaCl ⁄ P
i
containing 2%
skim milk (2% MPBS), washed three times with NaCl ⁄ P
i
containing 0.1% Tween-20 (PBST), the phage library (10
11
CFU) in 1% MPBS was added to the well and incubated
at 30 °C for 90 min. After discarding the phage solution,
200 lL of PBST was added and incubated for 1 min. This
was repeated once, and twice with 200 lL NaCl ⁄ P
i
. Then,
100 lL of 0.2 m glycine-HCl (pH 2.2) containing
1mgÆmL
)1
BSA was added and incubated for 10 min, to

elute well-bound phages. The eluate was recovered and
neutralized with 1 vol of 2 m Tris base.
In the case of phage ELISA, the plates, after incuba-
tion with phage, were washed three times with PBST and
incubated at room temperature for 1 h with 100 lL per
well of 5000-fold diluted HRP-conjugated mouse anti-M13
(Amersham Bioscience) in MPBS. The plate was washed
three times with PBST and developed with 100 lL
per well of substrate solution (100 lgÆmL
)1
3,3¢,5,5¢-
tetramethylbenzidine; Sigma, Tokyo, Japan; 0.04 l LÆmL
)1
H
2
O
2
, in 100 mm NaOAc, pH 6.0). After incubation
for 5–30 min, the reaction was stopped with 50 lL
per well of 1 m sulfuric acid and the absorbance read
at 450 nm, with the absorbance at 650 nm used as a
control.
Measurement of V
H
⁄ V
L
interaction strength
HB2151 cells carrying each spFv-encoding phagemid were
used to prepare culture supernatant containing V
H

-display-
ing phage and soluble V
L
. The overnight culture was centri-
fuged at 6500 g for 30 min and the supernatant was
recovered and stored at 4 °C.
To perform OS-ELISA, 100 mL per well of 1 lgÆmL
)1
anti-myc (9E10) IgG was coated overnight in NaCl ⁄ P
i
.
After blocking at room temperature for 2 h with MPBS,
the plate was washed three times with PBST and incubated
at room temperature for 1 h with 100 lL per well of
culture supernatant mixed with HEL, if necessary, which
was mixed for 1 h before and preincubated. The plate was
washed six times with PBST and phages were detected as
described above. Data are presented as an average of three
measurements.
Preparation of Fv ⁄ single chain Fv and kinetic
analysis with Biacore
To prepare soluble scFv fragment, primers M13RV and
ScFvH10VHF (5¢-CCAGAGCCACCTCCGCCTGAACCG
CCTCCACCGCTCGAGACGGTGACCG-3¢), ScFvH10
VLbk (5¢-CAGGCGGAGGTGGCTCTGGCGGTGGCGG
ATCGACGGACATTGAGCTCAC-3¢) and SplitVLSeqFor
(5¢-CCTCTTCTGAGATGAGTTTTTGTTCT-3¢) were
used to amplify the fragments encoding linker-tagged V
H
and V

L
, respectively. The gel-purified fragments were
assembled by SOE PCR, digested with NcoI and NotI, and
ligated with pET20b digested with the same. To express Fv,
a fragment encoding a stop codon and Shine–Dalgarno
sequence were inserted between BstEII and Sse8387I sites
of the above plasmid, which was amplified with primers
VHBstBack (5¢-GGGACCACGGTCACCGTCTCGAGCT
GAGCTGCTGACTAC-3¢), H10VkNotFor (5¢-AGCCGC
GGCCGCGCTTATTTCCAGCGCGGTCCCCCCTCC-3¢)
and pKST2(HyHEL-10) as a template, and digested by the
same enzymes. The H39KQ mutation was introduced by
the QuikChange mutagenesis kit (Stratagene) using primers
H39KQU (5¢-TTGGAGCTGGATACGGCAATTCCCAG
GGA-3¢) and H39KQD (5¢-TCCCTGGGAATTGCCGT
ATCCAGCTCCAA-3¢). After confirmation of the
sequences, BL21(DE3, pLysS) was transformed with the
plasmid, cultured stepwise into 100 mL of Luria–Bertani
medium (LB; containing 100 lgÆmL
)1
ampicillin and
34 lgÆ mL
)1
chloramphenicol) at 27 °C until the D
600
reached 0.5, when the culture was induced with 0.1 mm iso-
propyl thio-b-d-galactoside. After shaking for 16 h at
27 °C, the culture supernatant was precipitated with 65%
saturated ammonium sulfate, and the scFv protein was
purified from the precipitate using Talon affinity resin (BD

Bioscience, Tokyo, Japan) according to the manufacturer’s
protocol. Alternatively, the protein was purified using a
HEL affinity column made from HiTrap NHS-activated
HP column (Amersham Biosciences) essentially as described
previously [24].
The purified protein was quantified based on the absorb-
ance at 280 nm [25], and subjected to SPR kinetic analysis
with Biacore 2000 at 25 °C with 300 resonance units of
immobilized HEL on a CM5 sensorchip, and HBS-EP
(Biacore, Tokyo, Japan) as a buffer at a flow rate of
20 lLÆmin
)1
. Three measurements were performed for each
analyte concentration of 50 and 100 nm, and kinetic and
equilibrium constants were calculated using biaevaluation
4.1 program (Biacore).
Acknowledgements
We are grateful to Drs Shinya Tsukiji and Teruyuki
Nagamune for allowing the use of Biacore. This work
was supported by a Grant-in-Aid for Scientific
Research (B14350430, B17360394) from JSPS, and a
Grant-in-Aid for Exploratory Research (14655303)
from MEXT, Japan.
References
1 Glockshuber R, Schmidt T & Pluckthun A (1992)
The disulfide bonds in antibody variable domains:
effects on stability, folding in vitro, and functional
expression in Escherichia coli. Biochemistry 31,
1270–1279.
V

H
⁄ V
L
interaction and affinity K. Masuda et al.
2192 FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS
2 Reiter Y, Brinkmann U, Kreitman RJ, Jung SH, Lee B
& Pastan I (1994) Stabilization of the Fv fragments in
recombinant immunotoxins by disulfide bonds engi-
neered into conserved framework regions. Biochemistry
33, 5451–5459.
3 Raag R & Whitlow M (1995) Single-chain Fvs. FASEB
J 9, 73–80.
4 Nieba L, Honeggaer A, Krebber C & Pluckthun A
(1997) Disrupting the hydrophobic patches at the
antibody variable ⁄ constant domain interface:
improved in vivo folding and physical characterization
of an engineered scFv fragment. Protein Eng 10,
435–444.
5 Asano R, Takemura S, Tsumoto K, Sakurai N, Tera-
mae A, Ebara S, Katayose Y, Shinoda M, Suzuki M,
Imai K et al. (2000) Functional construction of the anti-
mucin core protein (MUC1) antibody MUSE11 variable
regions in a bacterial expression system. J Biochem
(Tokyo) 127, 673–679.
6 Tsumoto K, Nishimiya Y, Kasai N, Ueda H, Naga-
mune T, Ogasahara K, Yutani K, Tokuhisa K, Matsu-
shima M & Kumagai I (1997) Novel selection method
for engineered antibodies using the mechanism of Fv
fragment stabilization in the presence of antigen. Protein
Eng 10, 1311–1318.

7 Ueda H, Tsumoto K, Kubota K, Suzuki E, Nagamune
T, Nishimura H, Schueler PA, Winter G, Kumagai I &
Mahoney WC (1996) Open sandwich ELISA: a novel
immunoassay based on the interchain interaction
of antibody variable region. Nat Biotechnol 14, 1714–
1718.
8 Kondo H, Shiroishi M, Matsushima M, Tsumoto K
& Kumagai I (1999) Crystal structure analysis of
anti-hen lysozyme antibody (HyHEL10) Fv-antigen
complex: conformational changes of protein antigen
and water mediated interactions of Fv-antigen
and VH–VL interfaces. J Biol Chem 274, 27623–
27631.
9 Berry MJ & Pierce JJ (1993) Stability of immunoadsor-
bents comprising antibody fragments: Comparison of
Fv fragments and single-chain Fv fragments. J Chroma-
togr A 629, 161–168.
10 Chatellier J, Van Regenmortel MH, Vernet T &
Altschuh D (1996) Functional mapping of conserved
residues located at the VL and VH domain interface of
a Fab. J Mol Biol 264, 1–6.
11 Khalifa MB, Weidenhaupt M, Choulier L, Chatellier J,
Rauffer-Bruyere N, Altschuh D & Vernet T (2000)
Effects on interaction kinetics of mutations at the VH–
VL interface of Fabs depend on the structural context.
J Mol Recognit 13, 127–139.
12 Aburatani T, Sakamoto K, Masuda K, Nishi K,
Ohkawa H, Nagamune T & Ueda H (2003) A general
method to select antibody fragments suitable for non-
competitive detection of monovalent antigens. Anal

Chem 75, 4057–4064.
13 Essen L-O & Skerra A (1994) The de novo design of an
antibody combining site: crystallographic analysis of the
V
L
domain confirms the structural model. J Mol Biol
238, 226–244.
14 Hamers-Casterman C, Atarhouch T, Muyldermans S,
Robinson G, Hamers C & Songa EB (1993) Naturally
occurring antibodies devoid of light chains. Nature 363,
446–448.
15 Nohno T, Saito T & Hong JS (1986) Cloning and com-
plete nucleotide sequence of the Escherichia coli gluta-
mine permease operon (glnHPQ). Mol Gen Genet 205,
260–269.
16 Kabat EA, Wu TT, Perry HM, Gottesman KS &
Foeller C (1991) Sequences of protein of immunological
interest. 5th edn. U.S. Government Printing Office,
Bethseda.
17 Essen L-O & Skerra A (1993) Single-step purification of
a bacterially expressed antibody Fv fragment by immo-
bilized metal affinity chromatograph in the presence of
betaine. J Chromatogr A 657, 55–61.
18 Chothia C, Gelfand I & Kister A (1998) Structural
determinants in the sequences of immunoglobulin vari-
able domain. J Mol Biol 278, 457–479.
19 Melchers F, Karasuyama H, Haasner D, Bauer S, Kudo
A, Sakaguchi N, Jameson B & Rolink A (1993) The
surrogate light chain in B-cell development. Immunol
Today 14, 60–68.

20 Ward ES, Gu
¨
ssow D, Griffiths AD, Jones PT & Winter
G (1989) Binding activities of a repertoire of single
immunoglobulin variable domains secreted from Escher-
ichia coli. Nature 341, 544–546.
21 Bhat TN, Bentley GA, Fischmann TO, Boulot G &
Poljak RJ (1990) Small rearrangements in structures of
Fv and Fab fragments of antibody D 1.3 on antigen
binding. Nature 347, 483–485.
22 Herron JN, He XM, Ballard DW, Blier PR, Pace PE,
Bothwell AL, Voss EW Jr & Edmundson AB (1991) An
autoantibody to single-stranded DNA: comparison of the
three-dimensional structures of the unliganded Fab and a
deoxynucleotide-Fab complex. Proteins 11, 159–175.
23 Stanfield R, Takimoto-Kamimura M, Rini J, Profy A
& Wilson I (1993) Major antigen-induced
domain rearrangements in an antibody. Structure 1,
83–93.
24 Tsumoto K, Ueda Y, Maenaka K, Watanabe K,
Ogasawara K, Yutani K & Kumagai I (1994) Contribu-
tion to antibody–antigen interaction of structurally
perturbed antigenic residues upon antibody binding.
J Biol Chem 269, 28777–28782.
25 Gill SC & von Hippel PH (1989) Calculation of protein
extinction coefficients from amino acid sequence data.
Anal Biochem 182, 319–326.
K. Masuda et al. V
H
⁄ V

L
interaction and affinity
FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS 2193
26 Guex N & Peitsch MC (1997) SWISS-MODEL and the
Swiss-PdbViewer: An environment for comparative pro-
tein modeling. Electrophoresis 18, 2714–2723.
Supplementary material
The following supplementary material is available
online:
Table S1. Summary of the Student’s t-test for the
effect of individual framework region 2 (FR2) substitu-
tions.
This material is available as part of the online article
from
V
H
⁄ V
L
interaction and affinity K. Masuda et al.
2194 FEBS Journal 273 (2006) 2184–2194 ª 2006 The Authors Journal compilation ª 2006 FEBS