Báo cáo khoa học: Isolation and characterization of an IgNAR variable domain specific for the human mitochondrial translocase receptor Tom70 potx

Isolation and characterization of an IgNAR variable domain specific
for the human mitochondrial translocase receptor Tom70
Stewart D. Nuttall
1,2
, Usha V. Krishnan
1,2
, Larissa Doughty
1
, Kylie Pearson
2,3
, Michael T. Ryan
2,3
,
Nicholas J. Hoogenraad
2,3
, Meghan Hattarki
1
, Jennifer A. Carmichael
1,2
, Robert A. Irving
1,2
and Peter J. Hudson
1,2
1
CSIRO Health Sciences and Nutrition, and
2
CRC for Diagnostics, Parkville, Victoria, Australia;
3
Department of Biochemistry,
La Trobe University, Bundoora, Victoria, Australia
The new antigen receptor (IgNAR) from sharks is a disul-

phide bonded dimer of two protein chains, each containing
one variable and five constant domains, and functions as an
antibody. In order to assess the antigen-binding capabilities
of isolated IgNAR variable domains (V
NAR
), we have con-
structed an in vitro library incorporating synthetic CDR3
regions of 15–18 residues in length. Screening of this library
against the 60 kDa cytosolic domain of the 70 kDa outer
membrane translocase receptor from human mitochondria
(Tom70) resulted in one dominant antigen-specific clone
(V
NAR
12F-11) after four rounds of in vitro selection. V
NAR
12F-11 was expressed into the Escherichia coli periplasm and
purified by anti-FLAG affinity chromatography at yields of
3mgÆL
)1
. Purified protein eluted from gel filtration columns
as a single monomeric protein and CD spectrum analysis
indicated correct folding into the expected b-sheet confor-
mation. Specific binding to Tom70 was demonstrated by
ELISA and BIAcore (K
d
¼ 2.2 ± 0.31 · 10
)9
M
)1
) indi-

cating that these V
NAR
domains can be efficiently displayed
as bacteriophage libraries, and selected against target anti-
gens with an affinity and stability equivalent to that obtained
for other single domain antibodies. As an initial step in
producing ÔintrabodyÕ variants of 12F-11, the impact of
modifying or removing the conserved immunoglobulin
intradomain disulphide bond was assessed. High affinity
binding was only retained in the wild-type protein, which
combined with our inability to affinity mature 12F-11, sug-
geststhatthisparticularV
NAR
is critically dependent upon
precise CDR loop conformations for its binding affinity.
Keywords: new antigen receptor; variable domain; peptide
display; Tom70; mitochondrial import.
Conventional antibodies recognize antigens through the
combination of six complementarity determining region
(CDR) loops displayed three each upon variable heavy (V
H
)
and variable light (V
L
) chain immunoglobulin domains [1].
These CDR loops vary in size and composition allowing
formation of a large number of conformational antigen-
binding surfaces including planar and ridged topologies [2].
The orientation of the loops is maintained by a combination
of their internal architecture, the underlying immunoglo-

bulin scaffold, and the hydrophobic interaction between the
antibody V
H
and V
L
domains [3]. In contrast, families of
antibody-like molecules characterized recently in camelids
and sharks rely on a single immunoglobulin V
H
-like domain
framework, which presents two or three CDR loops to form
the antigen-binding interface [4–7]. For camelids, these V
H
H
single domain antibodies can bind an extensive range of
antigens, including large proteins, enzymes (either within or
outside the active site clefts), haptens and dyes. Biochemical
and structural data now shows that V
H
H binding affinity
resides in a variety of possible CDR conformations that can
include all three CDRs, or an elongated CDR3 loop alone,
or a combination of CDR and framework side-chain and
main-chain residues [8].
For sharks, the new antigen receptor (IgNAR) from
Ginglymostoma cirratum (nurse sharks) and Orectolobus
maculatus (wobbegong sharks) also utilizes a single V
H
-like
domain, which we herein term V

NAR
[9,10]. Structurally, the
entire intact IgNAR antibody molecule is a disulphide-
bonded dimer of two protein chains, each containing the
single variable and five constant domains. There is no
associated light chain and immunoelectron microscopy
confirms that the V
NAR
domains do not associate together
across a V
H
/V
L
-like interface and thereby provide bivalent
affinity to two separate antigen molecules [11]. There is a
striking evolutionary convergence at the molecular structure
level between the shark V
NAR
and camelid V
H
Hantigen-
binding domains. Similarities include, but are not limited
to (a) the presence of charged rather than hydrophobic
residues in the conventional V
L
interface of the immuno-
globulin framework, which imparts a hydrophilic character
to the solvent-exposed areas; (b) larger CDR3 loops
compared to those found in human and murine antibodies
(murine,


XX ¼ 10; human,

XX ¼ 13; llama V
H
H,

XX ¼ 15;
camel V
H
H,

XX ¼ 17.5; V
NAR

XX ¼ 16) [10,12,13]; and
Correspondence to Stewart Nuttall, CSIRO Health Sciences and
Nutrition, 343 Royal Parade, Parkville, Victoria 3052, Australia.
Fax: + 61 3 9662 7314, Tel.: + 61 3 9662 7100,
E-mail:
Abbreviations: IgNAR, new antigen receptor antibody from sharks;
VNAR, single variable domain of the IgNAR antibody; Tom70,
human 70 kDa component of the translocase of the outer mito-
chondrial membrane; CDR, complementarity determining region.
(Received 14 May 2003, revised 20 June 2003, accepted 1 July 2003)
Eur. J. Biochem. 270, 3543–3554 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03737.x
(c) the frequent presence of disulphide bridges, indicated
by paired cysteine residues, either within the CDR3 loop
(Type1 V
NAR

) or between the CDR1 and CDR3 loops
(Type2 V
NAR
) [5,11]. Camelid V
H
Hs typically possess
disulphide linkages either between the CDR1 and )3, or
CDR2 and )3 loops [13,14]. Despite these similarities, the
camel and shark variable domains are clearly different, both
by sequence alignment (only  20% identity) and the
unusual focus of shark V
NAR
variability into only the
CDR1 and )3 regions (Type2 V
NAR
), or the CDR2 and )3
regions (Type1 V
NAR
)[7].
While the shark IgNARs have yet to be formally
demonstrated as in vivo molecules responsible for immuno-
surveillance, there is strong evidence for their functional
role in antigen binding. First, an analysis of mutational
patterns of membrane-bound and secreted forms of nurse
sharkIgNARsindicatedthattheyaremutatedinthelatter
form, suggesting affinity maturation by somatic hyper-
mutation [15,16]. Second, in a previous study, we showed
that the individual wobbegong V
NAR
s could be expressed

as soluble single molecules in the E. coli periplasm. An
in vitro Type2 V
NAR
library was then designed with
randomized CDR3 loops, and displayed successfully on
the surface of fd bacteriophages and panned for specific
binding molecules [10]. Third, and most recently we
isolated two naturally occurring V
NAR
domains targeting
the kgp protease from Porphyromonas gingivalis with
affinities within the nanomolar range [17]. Finally, a third
type of V
NAR
(Type3) has very recently been identified in
neonatal shark primary lymphoid tissue that probably
functions as a protective low-specificity antibody early in
development, prior to maturation of the Type 1/2 IgNAR
antigen-driven response [7]. The Type3 V
NAR
topology is
characterized by a constant sized CDR3 loop of limited
diversity, probably stabilized by a conserved tryptophan
residue within the CDR1 loop [7].
Here we describe the design and construction of an
expanded in vitro V
NAR
library with more extensive
synthetic CDR3 loop variations, and hypothesize that this
library should contain binding reagents to a wide range of

protein targets. We have chosen as one target antigen the
cytosolic domain of the 70 kDa outer membrane trans-
locase (Tom70) from human mitochondria that is impli-
cated in mitochondrial import processes [18,19]. Studies in
Saccharomyces cerevisiae and the fungus, Neurospora crassa
have characterized Tom70 as a receptor peripheral to the
mitochondrial general insertion pore that preferentially
interacts with a subset of preproteins that typically contain
internal targeting signals and/or require the action of
cytosolic chaperones for their delivery to the mitochondrial
surface [20,21]. The human homologue of Tom70 has only
recently been identified [22], and attempts to generate high
specificity polyclonal or monoclonal antibodies have so far
been unsuccessful, perhaps due to extensive sequence
homology across species. Here, we report the isolation
and characterization of a V
NAR
that binds with high affinity
to human Tom70 as an important demonstration that
V
NAR
libraries can provide novel binding reagents against
refractory and immunosilent targets. Further, to demon-
strate the effect of removing the internal stabilizing disul-
phide bond in the manner of antibody V-domain
ÔintrabodiesÕ [23], residues Cys22 and Cys82 were modified
by alkylation, or replaced with alanine and valine, respect-
ively, and the resultant V
NAR
s evaluated for retention of

binding affinity.
Material and methods
Equipment and reagents
Restriction enzymes and Vent DNA polymerase were
purchased from New England Biolabs; T4 DNA ligase
was from Biotech (Australia). DNA fragment recovery and
purification was by QIAquick Gel Extraction Kit, Qiagen
and small-scale preparations of DNA from E. coli were
performed using the QIAprep Spin Miniprep Kit, Qiagen.
Monoclonal anti-(FLAG) Ig affinity resin was produced as
previously described [17]. BenchMark
TM
Prestained Protein
Ladder Cat. # 10748–010 was from Gibco BRL Life
Technologies. Standard molecular biological techniques
were performed as described [24]. Goat anti-(mouse) IgG
(Fc)-HRP was from Pierce.
E. coli
strains
The cell line used for library propagation and selection and
protein expression was E. coli TG1 (K12 supE D(lac-
proAB) thi hsdD5F¢{traD36 proAB
+
lacI
q
lacZDM15].
E. coli transformants were maintained and grown in
2 · YT broth supplemented with 100 lgÆmL
)1
(w/v) ampi-

cillin and/or 2% (w/v) glucose. Solid media contained 2%
(w/v) Bacto-agar. Transformation of E. coli was by stand-
ard procedures [24] performed using electrocompetent cells.
Isolation of total RNA from Wobbegong sharks
Spotted Wobbegong sharks (Orectolobus maculatus)were
housed and maintained at the Underwater World Aquar-
ium, Mooloolaba, Queensland, Australia. For isolation of
peripheral blood lymphocytes, a blood sample (3 mL) was
taken from the caudal vein of a young male (6.82 kg).
Experiments were performed in accordance with CSIRO
(Health Sciences and Nutrition) animal ethics requirements.
Total RNA was extracted using the AquaPure RNA
Isolation Kit (BIO-RAD, Australia), stored at )80 °C,
and used in reverse transcription-polymerase chain reac-
tions using the Titan one tube RT-PCR system (Roche,
Germany) as described [10].
Library construction and panning
DNA library cassettes encoding the Wobbegong V
NAR
with
randomization of the CDR3 loop were constructed as
described [10], using CDR3 randomization oligonucleotide
primers to generate synthetic CDR3s of 15 residues (#6981);
16 residues (#7211); 17 residues (#6980) and 18 residues
(#7210) in length (Table 1). In addition, natural Wobbe-
gong V
NAR
sequences were amplified direct from cDNA
using only the variable domain terminal primers (Table 1)
[10,17]. Cassette fragments were cut with the restriction

endonucleases NotIandSfiI and ligated into similarly cut
phagemid display vector, pFAB5c.His [25]. Library liga-
tions were purified, pooled and transformed into E. coli
TG1, yielding a total library size of approximately 4.0 · 10
8
3544 S. D. Nuttall et al. (Eur. J. Biochem. 270) Ó FEBS 2003
independent clones, which included a subset ( 7 · 10
6
)of
clones derived from natural CDR3 sequences. Phagemid
particles carrying the V
NAR
–gene3protein fusion were
propagated and isolated by standard procedures [26]. For
biopanning of the phagemid library, recombinant human
Tom70 (5 lgÆmL
)1
in NaCl/P
i
) was coated onto Maxisorb
Immunotubes and incubated at 4 °C overnight. Immuno-
tubes were rinsed (NaCl/P
i
), blocked with NaCl/P
i
/Blotto
(2%, w/v; Diploma skim milk powder, Bonlac Foods Ltd.,
Melbourne, Australia) for 1 h, and incubated with freshly
prepared phagemid particles in NaCl/P
i

/Blotto (2%, w/v)
for 30 min at room temperature with gentle agitation
followed by 90 min without agiation. After incubation,
immunotubes were washed [NaCl/P
i
/Tween20 (0.1%, v/v);
7, 8, 10 and 10 washes for panning rounds 1–4], followed by
an identical set of washes with NaCl/P
i
. Phagemid particles
were eluted using 0.1
M
HCl, pH 2.2/1 mgÆmL
)1
BSA,
neutralized by the addition of 2
M
Tris base [26], and either
immediately reinfected into E. coli TG1 or stored at 4 °C.
Nucleic acid isolation and cloning
Following final selection, phagemid particles were infected
into E. coli TG1 and propagated as plasmids, followed by
DNA extraction. The V
NAR
cassette was extracted as a
NotI/SfiI fragment and subcloned into the similarly restric-
ted cloning/expression vector pGC [27]. DNA clones were
sequenced on both strands using a BigDye terminator cycle
sequencing kit (Applied Biosystems) and a Perkin Elmer
Sequenator. The nucleotide sequence of clone 12F-11 is

deposited in the GenBank database under accession number
AY069988.
Recombinant Tom70 protein
DNA encoding the receptor domain of Tom70 (residues
111–608) was amplified by PCR, cloned into the vector
pET3a as described by Young et al. [21]. For expression,
cells were grown to D
600
¼ 0.6, induced by the addition of
isopropyl thio-b-
D
-galactoside (IPTG, 1 m
M
final), and
recombinant protein purified using Ni-NTA chromatogra-
phy (Qiagen) according to the manufacturer’s instructions
[21].
Soluble expression of V
NAR
constructs from expression
vector pGC
Recombinant proteins were expressed in the bacterial
periplasm as described [10]. Briefly, E. coli TG1 starter
cultures were grown overnight in 2YT medium containing
100 lgÆmL
)1
ampicillin and 2.0% glucose (w/v), diluted
1/100 into fresh 2YT containing 100 lgÆmL
)1
ampicillin and

0.1% glucose (w/v) and then grown at 37 °C and shaken at
200 r.p.m. until D
550
¼ 0.2–0.4. Cultures were then induced
with IPTG (1 m
M
final concentration), grown for a further
16 h at 28 °C and harvested by centrifugation (Beckman
JA-14/6K, 5500 g 10 min, 4 °C). Periplasmic fractions were
isolated by the method of Minsky [28] and either used as
crude fractions or recombinant protein purified by affinity
chromatography using an anti-(FLAG) Ig/Sepharose col-
umn (10 · 1 cm). The affinity column was equilibrated in
NaCl/P
i
, pH 7.4 and bound protein eluted with Immuno-
Pure
TM
gentle elution buffer (Pierce). Eluted proteins were
dialysed against two changes of NaCl/P
i
containing 0.02%
sodium azide, concentrated by ultrafiltration over a 3-kDa
cutoff membrane (YM3, Diaflo), and analysed by FPLC on
a precalibrated Superdex 200 column (Pharmacia) equili-
brated in NaCl/P
i
buffer pH 7.4. Recombinant proteins were
analysed, by SDS/PAGE through 15% Tris/glycine gels.
Enzyme linked immunosorbent assays

Protein antigens (0.5 lg per well) in NaCl/P
i
were coated
onto Maxisorb Immuno-plates (Nunc) and incubated at
4 °C overnight. Plates were rinsed, blocked with 5% (w/v)
Blotto in NaCl/P
i
for 1 h, and incubated with periplasmic
fractions or recombinant protein for 1 h at room tempera-
ture. Plates were rinsed with NaCl/P
i
, washed three times
with 0.05% Tween20 in NaCl/P
i
, and anti-(FLAG) Ig
(diluted 1/1000 in 5% Blotto in NaCl/P
i
) added. Plates were
incubated and washed as above, and the horseradish
peroxidase conjugated secondary anti-(mouse Fc) Ig added
Table 1. Oligonucleotide primers used in generation of V
NAR
libraries. Sense primer (fi); antisense primer (‹).
Oligonucleotide Number Sequence (5¢ –3¢) Features
5¢ Amplification 8406 (fi)
GTCTCGCGGCCCAGCCGGCCATGGCCACAAGGGTAGACCAAACACC N-terminus ¼ TRVDQTP…
5¢ Amplification 8407 (fi)
GTCTCGCGGCCCAGCCGGCCATGGCCGCAAGGGTGGACCAAACACC N-terminus ¼ ARVDQTP…
5¢ Amplification 8408 (fi)
GTCTCGCGGCCCAGCCGGCCATGGCCGCATGGGTAGACCAAACACC N-terminus ¼ AWVDQTP…

3¢ Amplification 8404 (‹)
CACGTTATCTGCGGCCGCTTTCACGGTTAATGCGGTGCC C-terminus ¼ …GTALTVK
3¢ Amplification 8405 (‹) CACGTTATCTGCGGCCGCTTTCACGGTTAATACGGTGCCAGCTCC C-terminus ¼ …GTVLTVK
CDR3 Library
construction
6981 (‹) GGTTAATACGGTGCCAGCTCCCYYMNNMNNMNNMNNMNNRYHRYH
RYHRYHMNNMNNMNNMNNMNNMNNTGCTCCACACTTATACGTGCCACTG
15 residue randomised loop
CDR3 Library
construction
7211 (‹)
TTTCACGGTTAATACGGTGCCAGCTCCTTTCTCMNNMNNMNNMNNRYHR
YHRYHRYHRYHMNNMNNMNNMNNMNNGNATGCTCCACACTTATACGT
GCC
16 residue randomised loop
CDR3 Library
construction
6980 (‹)
GGTTAATACGGTGCCAGCTCCCYYMNNMNNMNNMNNMNNMNNMNNRYHRY
HRYHRYHMNNMNNMNNMNNMNNMNNTGCTTGACACTTATACGTGCC
ACTG
17 residue randomised loop
CDR3 Library
construction
7210 (‹)
TTTCACGGTTAATACGGTGCCAGCTCCTTTCTCMNNMNNMNNMNNMNN
MNNRYHRYHRYHRYHRYHMNNMNNMNNMNNMNNGNATGCTTGA
CACTTATACGTGCC
18 residue randomised loop
Ó FEBS 2003 An IgNAR variable domain specific for human Tom70 (Eur. J. Biochem. 270) 3545

[1/1000 in NaCl/P
i
/5% Blotto]. Plates were washed again
and developed using 2,2-azino di-(ethyl) benzthiazoline
sulphonic acid (Boehringer Mannheim) and read at A
405
.
Biosensor binding analysis
A BIAcore
TM
1000 biosensor (BIAcore AB, Uppsala,
Sweden) was used to measure the interaction between
V
NAR
protein 12F-11 and Tom70. FPLC-purified Tom70
at a concentration of 20 lgÆmL
)1
in 10 m
M
sodium acetate
buffer, pH 4.5 was immobilized onto a CM5 sensor chip via
amine groups using the amine coupling kit (BIAcore AB)
[29]. The immobilization was performed at 25 °Cand
5 lLÆmin
)1
flow rate. Injection of 28 lL of Tom70 coupled
990 RU to the surface. Binding experiments were performed
in Hepes buffered saline (HBS; 10 m
M
Hepes, 0.15

M
NaCl,
3.4 m
M
EDTA, 0.005% surfactant P20, pH 7.4) at 25 °C
and a constant flow rate of 5 lLÆmin
)1
with a series of
12F-11 concentrations (2.2–17.8 n
M
). Binding experiments
were performed immediately, as prolonged washing with
the HBS buffer resulted in a decrease in activity of the
immobilized Tom70. Regeneration of the Tom70 surface
was achieved by running the dissociation reaction to
completion before the next injection of analyte.
For binding experiments in the reverse orientation,
recombinant protein 12F-11 was immobilized by the
standard amine coupling method. 12F-11 protein at a
concentration of 20 lgÆmL
)1
in 10 m
M
sodium acetate
pH 6.0 was injected for 8 min (40 lL) over an activated
surface to couple 650 RU of protein onto the sensor surface.
The 12F-11 surface was regenerated with a 10-lL aliquot of
50 m
M
HCl with negligible loss of binding activity. Binding

experiments were performed in HBS buffer at 25 °Canda
flow rate of 5 lLÆmin
)1
with a range of Tom70 concentra-
tions (37.0–296 n
M
). The binding data was evaluated with
BIAEVALUATION
3.0.2 [30].
Structural modeling
V
NAR
domains 12F-11 (this study) and 7R-1 [10] were
modelled using the program
MODELLER
[31]. The PDB
database was searched with the 12F-11 V
NAR
sequence and
17 structures with the best Z-scores selected as templates.
The PDB accession numbers for the template molecules
were as follows with the chain Ids indicated: 1B88:B,
1D9K:E, 1H5B:B, 1KB5:A, 1KJ2:A (T-cell receptor alpha
domains); 1BJM:A, 2RHE (Bence–Jones proteins);
1AY1:H, 1BGX:H, 1IGF:J, 1KEN:T, 2IGF:H (V
H
domains); 1IAI:M, 1F6L:L, iIFF:L, 1JNL:L and 35C8:L
(V
L
domains). The resulting model structures were refined

and energy minimized using molecular dynamics restrained
to the template structure, except where gaps occurred in the
alignments and for CDRs 1 and 3. In these cases extensive
loop modelling was undertaken and the final model
selection based on the modeller objective score.
Construction of the 12F-11DCys mutant and 12F-11
reduction/alkylation
For reduction and alkylation of 12F-11, recombinant
protein (1.3–1.5 mg) was denatured/reduced using 6
M
guanidine HCl and 50 m
M
dithiothreitol (pH 8.0) for 1 h
at 45 °C under nitrogen. Cysteine residues were then
alkylated by the addition of 100 m
M
(final) iodo-acetamide
(pH 8.0)/1 h (room temperature) followed by quenching
with additional dithiothreitol. Samples were dialysed against
four changes of NaCl/P
i
, concentrated, and analysed by
FPLC, SDS/PAGE and ELISA, as above. The disulphide
minus variant of 12F-11 incorporating mutations Cys22Ala
and Cys82Val was constructed by overlapping PCR using
oligonucleotide primers N8517 (Forward: 5¢-ACAAGGG
TAGACCAAACACCAAGAACAGCAACAAAAGAG
ACGGGCGAATCACTGACCATCAACgccGTCCTGA
GAGAT-3¢) and N8518 (Reverse: 5¢-TTTCACGGTTAA
TGCGGTGCCAGCTCCCCAACTGTAATAAATACC

AGACAAATTATATGCTCCaacCCTATACGTGCCA
CTG-3¢); followed by secondary PCR using V
NAR
terminal
oligonucleotide primers [17] to complete the framework and
incorporate NotIandSfiI restriction endonuclease sites for
cloning. Bacterial expression was as described above.
Affinity maturation by error-prone PCR
The 12F-11 V
NAR
cassette was mutagenized by error prone
PCR using Taq DNA polymerase [32]. Pools of mutated
V
NAR
cassettes were isolated, cut with SfiI/NotI, cloned into
the phagemid vector pFAB.5c, and transformed into E. coli
TG1 as above. The resulting library ( 1 · 10
6
independent
clones) showed on average 1–2 residue changes/100 amino
acids. Two rounds of panning under high stringency
conditions were performed on immobilized Tom70 as
above, except that the final five washes for each selection
round incorporated a further 2 min incubation to promote
dissociation. The selected V
NAR
cassettes were then rescued,
subcloned and analysed.
Results
Construction of an expanded Wobbegong V

NAR
library
We previously designed and constructed a Wobbegong
IgNAR variable domain (V
NAR
) library with long synthetic
CDR3 loops of either 15 or 17 residues in length inserted into
a mixed scaffold repertoire of 26 naturally occurring V
NAR
domains. This small library ( 3 · 10
7
independent clones)
was displayed on the surface of fd bacteriophage and
successfully panned against protein antigens [10]. In order to
increase the diversity of possible antigen-binding fragments,
this library was expanded in three ways: (a) the extended
library comprised increased complexity with CDR3 lengths
ranging form 15–18 residues to reflect the predominant
natural diversity in Wobbegong and Nurse shark V
NAR
s
(Fig. 1A). Additionally, a different randomization pattern
was used, biased toward the incorporation of cysteine at
CDR3 loop residue positions 1 and 7–11, to enhance the
possibility of inter- and intra CDR disulphide cross-links.
These strategies are summarized in Fig. 1B and Table 1, and
details of the library construction are given in greater detail in
the Materials and methods. (b) The extended library was
based on CDR3 loops grafted into a large scaffold repertoire
of natural V

NAR
domains generated by direct RT-PCR from
total RNA extracted from Wobbegong shark peripheral
blood lymphocytes. Thus, many differing CDR1 sequences
and minor framework variations were represented in the
3546 S. D. Nuttall et al. (Eur. J. Biochem. 270) Ó FEBS 2003
extended library. (c) A subset of the extended library now
also included naturally occurring CDR3 loops derived
directly from the immune repertoire of several sharks. These
natural sequences form part of the matured shark immune
response, generated in response to exposure to antigen in the
natural environment, and have extensive size heterogeneity
(as indicated in Fig. 1A) and different randomizations from
those used for the synthetic CDR3s [17]. Taken together,
these changes resulted in an expanded V
NAR
library consist-
ing of over 4 · 10
8
independent clones, representing signi-
ficantly enhanced diversity compared to our original V
NAR
repertoire. This library was displayed as a fusion with the
gene3protein of fd bacteriophage in the vector pFAB5c.HIS
(Fig. 1C) allowing for standard phage display and selection.
Biopanning against immobilized mitochondrial
import receptor Tom70
Tom70 is an integral membrane protein consisting of a large
globular receptor domain exposed to the cytosol and a short
N-terminal transmembrane anchor. A truncated Tom70

protein construct consisting of residues 111–608 of human
Tom70 was expressed in E. coli, purified as a soluble
extracellular 60 kDa protein, and immobilized on immuno-
tubes as a target protein [21].
The V
NAR
library was transformed into E. coli TG1 and
phagemid particles rescued and panned against the immo-
bilized Tom70 antigen. Four rounds of biopanning were
performed with an increase in the stringency of washing at
each step, and between selection rounds three and four a
significant ( 1000-fold) increase in the titre of eluted
bacteriophage was observed. Colony PCR on transfected
bacteriophage showed that 100% of colonies were positive
for V
NAR
sequences, and this combined with the increase in
the titre, indicated positive selection. Thus, V
NAR
cassettes
were rescued from phagemids, cloned into the periplasmic
expression vector pGC, and transformed into E. coli TG1.
Periplasmic fractions from recombinant clones were tested
for binding to Tom70 and negative control antigens by
ELISA (not shown). Several clones showed significant
binding above background, and all of these proved to be
identical sequences. One of these, designated clone 12F-11,
was chosen for further analysis.
The primary and deduced amino acid sequences of clone
12F-11 are presented in Fig. 2A, including in-frame dual

octapeptide FLAG epitope tags and two alanine linker
regions. The protein represents a 103 residue V
NAR
domain,
with a predicted molecular mass of 14 054 Da (including
the affinity tags). An alignment of protein 12F-11 with four
other V
NAR
proteins showed a high level of sequence
conservation except in the CDR regions (Fig. 2B). How-
ever, and most surprisingly, the alignment also revealed that
the Thr39 residue, present in the other V
NAR
s, was absent in
the 12F-11 protein (Fig. 2B, arrowed). Equally unusual
was the CDR3 structure that at 10 residues (
N
YN
LSGIYYSW
C
) is significantly shorter than the 15–18
residues encoded within the library design. This truncation
in the CDR3 size probably occurred either during the initial
PCR-based library construction, or under selection pressure
within the early rounds of panning. The presence of such
CDR deletions in large libraries is not uncommon [33], and
indeed one V
NAR
protein we reported previously was an
obvious deletion mutant [10].

Recombinant protein 12F-11 is a monomeric,
correctly folded protein
Loss of a nonCDR residue is potentially destructive to
immunoglobulin domain structures. Thus, to determine
Fig. 1. Design of the V
NAR
library. (A) Cumulative frequency histo-
gram of IgNAR CDR3 loop lengths, from Wobbegong sharks (26
sequences) and Nurse sharks (35 sequences). (B) Schematic diagram
of synthetic CDR3s used in V
NAR
library construction, showing the
randomization patterns used for the varying length CDRs. X repre-
sents use of the nucleotide randomization strategy (NNK) that encodes
any residue or an amber stop codon. Surrounding framework regions
are also shown. (C) V
NAR
cassettes were ligated into phagemid vector
pFAB.5c at the SfiIandNotI restriction endonuclease sites. The
phagemid vector incorporates a lacZ promoter, and in-frame PelB
leader, Ala
3
linker, and DGene3 protein domains prior to a translation
termination codon.
Ó FEBS 2003 An IgNAR variable domain specific for human Tom70 (Eur. J. Biochem. 270) 3547
whether the absence of residue Thr39 had an adverse effect
upon protein 12F-11 expression, folding, and stability, we
undertook a thorough protein chemical analysis. Recom-
binant 12F-11 protein was expressed in E. coli and then
isolated from the bacterial periplasm by affinity chromato-

graphy using an anti-(FLAG) Ig affinity resin. Expression
levels obtained were routinely between 2 and 3 mgÆL
)1
of protein from shake-flask cultures. Analysis of affinity-
isolated protein by FPLC through a precalibrated Superdex
200 gel filtration column showed a single peak eluting from
the column at approximately 36 min, corresponding to a
protein of 14–15 kDa molecular mass and consistent with
the size of a monomeric V
NAR
domain. (Fig. 3A). There
was no evidence of protein aggregation, nor were higher
order multimers such as protein dimers or tetramers
observed. Indeed, protein 12F-11 appeared to have at least
equal stability to other V
NAR
proteins we have analysed as
this FPLC profile was maintained consistently, even after
prolonged storage at 4 °C or multiple freeze-thaw cycles,
with no evidence of protein degradation. To confirm that
the isolated protein was being correctly processed in the
E. coli periplasm, N-terminal amino acid sequence analysis
on material eluted from the FLAG column showed that
only one protein species was present (
1
TRVDQTP-)
corresponding to the predicted N-terminus (Fig. 2A). Far
ultraviolet circular dichroism (CD) spectra of aqueous
solutions of protein 12F-11 showed a profile with a negative
band with k

max
at 217–219 nm (Fig. 3B). This spectrum is
characteristic of a protein with a b-sheet structure with
unstructured loops contributed by the CDRs and FLAG
affinity tags, and is not a disordered structure [34]. Indeed,
the 12F-11 specturm is very similar to CD spectra obtained
for other V
NAR
proteins, for example V
NAR
12A-9 that has
different CDR loops and a slightly different b-sheet
framework, and which is shown for comparison (Fig. 3B)
[17]. Together, these results suggest that despite the absence
of Thr39, protein 12F-11 folds into compact, b-sheet
immunoglobulin in the E. coli periplasm. Indeed, in
preliminary structural studies, protein 12F-11 crystallises
in the monoclinic P2 space group (results not shown),
Fig. 2. Nucleotide and deduced amino acid sequences of the V
NAR
12F-
11 variable domain. (A) Nucleotide and deduced amino acid sequences
of clone 12F-11. The conserved termini dictated by the oligonucleotide
primer sequences used in library construction are underlined, and the
alanine linker and dual octapeptide FLAG tags are italicised. The
positions of the CDR1 and )3 regions are indicated in bold type.
(B) Alignment of protein 12F-11 with four other V
NAR
domain amino
acid sequences (GenBank AY069988; AF336094; AF336087;

AF336088; AF336089). Amino acids are designated with the single-
letter code, and identical residues (dark shading) and conservative
replacements (light shading; I/V/L/M, D/E, K/R, A/G, T/S, Q/N, F/Y)
are indicated. The framework residue at position 39, absent in 12F-11, is
indicated by the arrow, and the CDR1 and )3 regions are highlighted.
Fig. 3. Size exclusion chromatography and CD analysis of protein
12F-11. (A) Elution profile of affinity purified 12F-11 protein on a
calibrated Superdex 200 gel-filtration column equilibrated in NaCl/P
i
,
pH 7.4 and run at a flow rate of 0.5 mLÆmin
)1
. Protein 12F-11 elutes at
approximately 36 min consistent with a monomeric domain (12F-11
calculated M
r
of 14 054 Da, including linker and dual FLAG octa-
peptide tags). Approximate elution times for a series of protein
standards are indicated by arrows, and the absorbance at A
214
(unbroken line) and A
280
(dashed line) is given in arbitrary units. The
A
214
absorbance peak at approximately 46 min represents sodium
azide. The inset shows the same sample analysed by SDS/PAGE
through a 15% (w/v) polyacrylamide Tris/glycine gel and stained with
Coomassie Brilliant Blue. (B) Circular dichroic spectrum of affinity
purified V

NAR
12F-11 in 0.05
M
sodium phosphate buffer, pH 7.4
(unbroken line). For comparison, the spectrum for the naturally
occurring V
NAR
domain 12A-9 [17] is also shown (dotted line).
3548 S. D. Nuttall et al. (Eur. J. Biochem. 270) Ó FEBS 2003
providing further evidence for folding into an ordered
domain structure.
To explain more fully why protein 12F-11 folds into a
functional protein while missing a framework residue, we
modelled the variable domain structure and compared it to
a model of a conventional V
NAR
(Fig. 4). Our modelling
studies indicate that Thr39 is located at the end of the
C strand, and that the adjacent C-C¢ loop is therefore
amenable to structural changes imposed by the residue
deletion without disruption to the framework. Otherwise,
there is good agreement between the two V
NAR
models,
with the obvious exception of the CDR loop diversity.
Specificity and binding activity of recombinant
protein 12F-11
The specificity of protein 12F-11 for Tom70 was demon-
strated by ELISA (Fig. 5A). Recombinant protein reacted
specifically with Tom70 but not several other antigens tested

and was concentration dependent to 0.6 pmol of protein
(Fig. 5B). The binding kinetics of the 12F)11/Tom70
interaction were also measured by BIAcore biosensor
analysis with Tom70 protein immobilized via amine coup-
ling to the sensor surface. As immobilized Tom70 on the
sensor surface was found to be unstable, with a 70% loss of
binding activity in 24 h, binding experiments were per-
formed immediately upon immobilization and Fig. 6A
shows the interaction of varying concentrations (17.8 n
M
,
8.9 n
M
, 4.5 n
M
, 2.2 n
M
) of peak-purified 12F-11 monomer
with the immobilized Tom70. Analysis of the binding data
with the 1 : 1 Langmuir binding model showed a good fit
(Fig. 6A) for the monovalent analyte (12F-11) binding to a
Tom70 epitope consistent with a 1 : 1 binding interaction.
There was no binding of the 12F-11 monomer to a blank
surface (activated and then blocked with ethanolamine)
indicating that there is no nonspecific interaction with the
sensor surface (Fig. 6A, inset). Kinetic analysis of 12F-11
binding to immobilized Tom70 revealed a rapid association
rate constant (k
a
¼ 1.68 ± 0.27 · 10

6
M
)1
Æs
)1
)andadis-
sociation rate constant (k
D
)of3.49±0.36
–
3Æs
)1
to yield a
K
D
of 2.2 ± 0.31
)9
M
)1
.
V
NAR
protein 12F-11 monomer was also immobilized via
amine coupling onto the sensor chip to measure the binding
interaction in the reverse orientation. The 12F-11 surface
was stable and could be regenerated with 50 m
M
HCl
Fig. 4. Models of V
NAR

domains. V
NAR
s 12F-11 (purple framework
with magenta CDRs) and 7R-1 (dark green framework with light
green CDRs) were modelled on existing immunoglobulin superfamily
variable domain structures. The position of residue Thr39, missing in
12F-11, is indicated.
Fig. 5. Analysis of protein 12F-11 by ELISA. (A) Protein 12F-11 was
purified from the periplasmic fraction of E. coli TG1byaffinity
chromatography through an anti-FLAG M2 antibody column and
150 pmol tested for binding to Tom70, lysozyme, kgp (lysine specific
gingipain protease from Porphyromonas gingivalis), and a-amylase.
Results represent the average of triplicate wells. (B) As for (A) except
serial twofold dilutions of protein 12F-11 were tested for binding to
Tom70 and lysozyme. Results represent the average of duplicate wells.
Ó FEBS 2003 An IgNAR variable domain specific for human Tom70 (Eur. J. Biochem. 270) 3549
without loss of binding activity. The binding data for the
interaction of Tom70 to immobilized 12F-11 however, did
not fit the theoretical Langmuir model for 1 : 1 binding
(Fig. 6B) but displayed biphasic binding characteristics
indicative of multivalent binding. This result is consistent
with the observation that Tom70 elutes as an equilibrium
mixture between dimer (M
r
 136 kDa) and monomer on
size exclusion chromatography.
Removal of the 12F-11 V
NAR
framework disulphide
linkage

Next, protein 12F-11 was used to test the utility of V
NAR
domains as possible ÔintrabodiesÕ for expression and use in
in vivo targeting applications. Specifically, we asked whether
binding affinity was retained in the absence of the conserved
immunoglobulin intradomain disulphide bond. In an initial
series of experiments, recombinant 12F-11 protein was
denatured and reduced using guanidine HCl and dithio-
threitol, followed by alkylation of the cysteine residues and
refolding. However, in a result seen for many V
H
/V
L
antibodies, the modified protein almost exclusively precipi-
tated in the soluble fraction. Only a small proportion
consistently remained in the soluble fraction, and this
protein showed a similar FPLC profile to the unmodified
protein (Fig. 7A). This soluble protein retained binding
affinity for Tom70 by ELISA (Fig. 1C), and we hypothesize
that this fraction represents protein that was not fully
alkylated and was thus able to refold, with probable
reoxidation of the disulphide bond. In contrast, the
insoluble fraction most likely represents irreversibly aggre-
gated alkylated material.
In order to more systematically test this hypothesis, we
elected to eliminate the possibility of disulphide bond
formation genetically by replacement of residues Cys22 and
Cys82 with alanine and valine, respectively, to give a
cysteine minus mutant (12F-11DCys). Use of alanine and
valine was initially determined in a set of competitive

selection experiments [35], and replacement with this pair of
residues maintains a hydrophobic character suitable for
amino acids buried deep within the protein interior and
closely approximates the relative size of the cysteine side
chains. When 12F-11DCys was expressed into the E. coli
periplasm, expression levels comparable to those obtained
for the wild-type were obtained, and both wild-type and
mutant proteins were indistinguishable by gel filtration
chromatography (Fig. 7B). However, when binding to
Tom70 was assessed by ELISA under stringent washing
conditions, 12F-11DCys did not appear to target
Tom70 compared to 12F-11 and the refolded 12F-11
(reduced/alkylated soluble fraction) proteins (Fig. 7C).
This surprising result was further tested by kinetic ana-
lysis by biosensor, which showed an approximately
200-fold decrease in binding affinity (K
D
¼ 2n
M
to
K
D
¼500 n
M
) (Fig. 7D). Interestingly, this loss in bind-
ing affinity was completely attributed to a slower association
phase, with the dissociation curve indistinguishable from the
parent protein. We suggest that removal of the stabilizing
rigidity provided by the disulphide bond results in a slight
perturbation of the orientation of the immunoglobulin

b-sheets relative to each other. However, upon binding, the
interaction with antigen functions to lock the b-sheet and
CDR loop conformations, resulting in the original dissoci-
ation kinetics.
Affinity maturation of 12F-11
In order to affinity mature V
NAR
12F-11, a library of
mutant proteins was generated by error-prone PCR. The
resultant bacteriophage-displayed library ( 10
6
independ-
ent clones) was panned against Tom70 under conditions
designed to select for variants with enhanced off-rate
kinetics. After two rounds of selection a >4000-fold
increase in titre was observed indicating strong selection.
Of the resultant clones, a large proportion (64%) were 12F-
11 wild-type, while those with mutations almost exclusively
showed conservative framework variations well-removed
from the V
NAR
binding site. Only one variant, designated
15Z-2, showed changes mapping to either the CDR or
Fig. 6. Analysis of protein 12F-11 by BIAcore. (A) Binding of mono-
meric V
NAR
protein 12F-11 to immobilized Tom70 protein (990 RU)
was measured at a constant flow rate of 5 lLÆmin
)1
with an injection

volume of 35 lL. Dissociation was continued with HBS buffer until
the response returned to the initial value before injecting the next
sample. The inset shows the binding profile of protein 12F-11
(20 lgÆmL
)1
) to immobilized Tom70 and a blank surface (NHS/EDC
activated and blocked with ethanolamine). The circles show the fit to
the data obtained on analysis with the 1 : 1 Langmuir binding model
for evaluation of the kinetic rate constants. (B) Sensorgram showing
the binding of Tom70 (74 n
M
) to immobilized V
NAR
protein 12F-11
monomer (650 RU) at a constant flow rate of 5 lLÆmin
)1
.Thecircles
show the fit to the data on analysis with the 1 : 1 Langmuir binding
model when only the first 110 s of the dissociation phase is analysed.
3550 S. D. Nuttall et al. (Eur. J. Biochem. 270) Ó FEBS 2003
CDR-framework junctions. The two mutations in this clone
were Lys33 fi Arg, which is in the CDR1 loop (Fig. 1A),
and Thr43 fi Ile, which is a conservative framework
variant. Protein 15Z-2 expressed at levels similar to the
parental type and exhibited similar behaviour upon FPLC
analysis (Fig. 8A). When tested by ELISA, 15Z-2 showed a
slightly higher binding response to Tom70, but not negative
control antigens (Fig. 8B). However, this difference was not
apparent in kinetic measurements by biosensor, where
12F-11 and 15Z-2 showed indistinguishable binding kinet-

ics, including no differences in dissociation rates. Moreover,
upon extended storage, protein 15Z-2 showed a slight
tendency to precipitation, suggesting that its coselection
with wild-type, and slightly higher ELISA responses, may be
attributable to an increased tendency toward aggregation.
Discussion
The aim of this study was to generate an in vitro library
of V
NAR
domains, containing both synthetic and natural
CDR3 loops, and then to isolate specific binding molecules
using as an initial target antigen the mitochondrial outer
membrane receptor Tom70. The resulting protein, 12F-11,
shows a high degree of affinity and specificity for Tom70,
andwithaK
D
of  2n
M
compares very well with affinities
reported for camelid V
H
H domains (that vary in the range
2–300 n
M
[36]) and for scFv and disulphide stabilized Fv
fragments [37]. The high affinity of 12F-11 is directly
attributable to a relatively rapid association rate
(k
a
 1.7 · 10

6
M
)1
Æs
)1
), while the dissociation rate (k
Da
 3.5 · 10
)3
Æs
)1
) is more typical of other V
NAR
and camelid
V
H
H single domain antibodies [36]. For example, previously
we described a naturally occurring V
NAR
targeting the kgp
protease from Porphyromonas gingivalis, and there the K
D
was 1.31 · 10
)7
M
, primarily due to the lower association
rate (k
a
¼ 4.3 · 10
4

M
)1
Æs
)1
) [17].
Structurally, protein 12F-11 is a slightly unusual member
of the V
NAR
family. Firstly, the absence of a framework
residue which is commonly present (Thr39), could be
hypothesized to deform the underlying b-framework.
However, the data clearly demonstrates that this is not the
case, and molecular modelling maps this residue to a region
distant to the antigen-binding site, and on the periphery of
the immunoglobulin-like core scaffold and adjacent to the
C-C¢ loop (see Fig. 4). Further, there may well be even
greater latitude for mutations within this region, as in
unrelated experiments we have also discovered a naturally
occurring V
NAR
lacking five residues in this loop position
(S. Nuttall, unpublished data). Secondly, the mean size of
the V
NAR
CDR3 loop is 16 residues, yet protein 12F-11
achieves low nanomolar affinity binding with a CDR3 only
10 residues in length. While this apparently contradicts the
Fig. 7. Effect of removal of the 12F-11 disulphide bond. (A) Elution profiles of affinity purified 12F-11 protein (dotted line), and the soluble fraction
remaining after reduction and alkylation (unbroken line), on a calibrated Superose 12 gel-filtration column equilibrated in NaCl/P
i

, pH 7.4 and run
at a flow rate of 0.5 mLÆmin
)1
. The inset shows relative amounts of material recovered from the soluble (S) and insoluble (P) fractions after
reduction and alkylation compared to untreated protein (U). (B) Elution profiles of affinity purified 12F-11 (unbroken line) and 12F-11DCys
(dotted line) proteins analysed as in (A). (C) ELISA comparing binding of 12F-11, 12F-11 reduced and alkylated soluble fraction, and 12F-11DCys
proteins to Tom70 and negative control antigens. Results represent the average of duplicate wells. (D) Binding of monomeric proteins 12F-11 and
12F-11DCys (40 lgÆmL
)1
) to immobilized Tom70 protein. The inset shows the binding profile of protein 12F-11DCys to immobilized Tom70 and
to a nonspecific antigen and a blank surface (NHS/EDC activated and blocked with ethanolamine).
Ó FEBS 2003 An IgNAR variable domain specific for human Tom70 (Eur. J. Biochem. 270) 3551
theory that V
NAR
s encompass most of their binding affinity
within a long and diverse CDR3 loop, it is significant that
some naturally occurring V
NAR
s, presumably selected by
the shark antigen-driven immune response, have even
shorter CDR3 lengths [10]. Indeed, not all antibody V-like
domains display extended CDR3 loops, by analogy to a
recent structural analysis of camelid V
H
H domains [38] a
significant proportion of the V
NAR
repertoire may comprise
CDR3 loops overlayed onto and across the scaffold surface.
This conclusion means that, in the context of antigen

binding surfaces, shark V
NAR
and camelid V
H
H libraries
will contain structural homologues similar to antibody V
H
and V
L
libraries, as well as providing discrete and distinctly
different structural repertoires.
In an attempt to further improve the 12F-11 binding
affinity, we generated a library of affinity-matured variants.
However, selection failed to isolate any mutants with
improved binding kinetics and instead there was strong
selection for the wild-type protein. This probably reflects a
precise interaction between the CDR1 and relatively short
CDR3 loop with only minor variations being tolerated.
Thus, within the relatively limited context of a library of a
million independent clones covering the entire V
NAR
cassette, it is unlikely that the rare beneficial mutations will
be present. In contrast, in experiments aimed at affinity
maturing other V
NAR
domains, we have isolated several
variants with approaching order-of-magnitude enhanced
affinities from similar sized error-prone PCR libraries.
However, in these cases the mutations targeted an extended
and flexible CDR3 loop, which provided greater latitude for

the introduction of positive mutations (S. Nuttall, unpub-
lished observation).
The experiments testing the impact of removal of the
intradomain disulphide bond provide further evidence that
V
NAR
12F-11 is a tightly folded protein domain relatively
intolerant to manipulation. Reduction and alkylation most
probably had a fatal impact upon protein stability and
folding resulting in aggregated protein. This is not uncom-
mon for immunoglobulin domains, and only relatively few
antibody V
H
or V
L
domains remain functional after such
modification [39–41]. This is unsurprising given that during
alkylation, cysteine is replaced by S-carboxymethylate
cysteine, resulting in a significant increase in molecular
mass that must be accommodated within the otherwise
tightly folded antibody core. More elegant is genetic
replacement of the two half-cystines with an alanine-valine
pair that is virtually mass neutral and represents an
optimum replacement strategy for antibodies determined
in a competitive selection system [35]. Our finding that such
a substitution, with consequent ablation of the intramole-
cular disulphide bond, decreases the kinetics of association
but not dissociation can be explained as due to a slight
perturbation of the orientation of the immunoglobulin
b-sheets relative to each other thereby reducing the associ-

ation rate. In contrast, antigen binding presumably locks the
b-sheet and CDR loop conformation, resulting in the
original dissociation kinetics. This interpretation is consis-
tent with the current dogma that antigen binding is critically
dependant upon the precise interaction of CDR1, CDR3,
and underlying framework residues. Specifically, we noted
that the three CDR3 tyrosine residues in 12F-11 (Tyr
residues are over-represented in CDR loop structures [42]),
may combine with Tyr35 which lies just C-terminal to the
CDR1 loop, to form the predominant antigen binding site.
Additionally, the noncanonical disulphide bridge often
found in V
NAR
domains, that links the CDR1 and )3
loops together providing additional conformational stabi-
lity, is absent in this case. However, any more detailed
analysis of the 12F-11 paratope and assessment of the
varying contributions of CDR and framework regions
clearly requires definitive structural data; this analysis is
currently in progress.
The isolation of 12F-11 from the extended V
NAR
library
demonstrates that synthetic CDR3 libraries selected in vitro
can generate proteins with antigen-binding affinities at least
equal to those of natural systems (i.e. immunization of
animals followed by isolation of the variable gene reper-
toire). This is especially important where conventional (i.e.
murine) antibodies are difficult to generate. The human
Tom70 receptor is an important example of such refractory

targets, and the generation of a high-affinity reagent specific
for the cytosolic domain is a potentially valuable tool in the
study of preprotein targeting. However, in a series of in vitro
import inhibition experiments, designed to test the ability of
12F-11 protein to block the import of precursor proteins
Fig. 8. Affinity maturation of 12F-11. (A) Elution profiles of affinity
purified 12F-11 (dotted line) and 15Z-2 (unbroken line) proteins on a
calibrated Superose 12 gel-filtration column equilibrated in NaCl/P
i
,
pH 7.4 and run at a flow rate of 0.5 mLÆmin
)1
. (B) ELISA comparing
binding of 12F-11 and 15Z-2 protein to Tom70 and negative control
antigens. Results represent the average of duplicate wells.
3552 S. D. Nuttall et al. (Eur. J. Biochem. 270) Ó FEBS 2003
into rat liver mitochondria, no inhibition was observed,
suggesting that 12F-11 binds an epitope of the Tom70
soluble domain that is inaccessible in the endogenous form,
for example through oligomerization [43], or by it facing the
membrane. This, combined with the lowered affinity of the
12F-11DCys variant, raises doubts as to the utility of these
particular V
NAR
s as intracellular targeting reagents.
The V
NAR
library described here is of sufficient size and
diversity to be an important resource, particularly for
screening against large proteinaceous antigens. Future

advances likely to expand these applications include
improvements in initial panning and screening, followed
by the application of more sophisticated affinity maturation
techniques beyond error-prone PCR, for example DNA
shuffling or in vivo mutagenesis and selection [44,45]. We
predict that such strategies, especially those targeting the
CDR1 region, will result in an improvement in affinity,
particularly by decreasing the dissociation rate. Alternat-
ively, increased V
NAR
functional affinity (avidity) may also
be achieved by domain multimerization. Such strategies are
well established for scFv fragments (e.g., adjustment of
linker lengths resulting in various geometric conformations
of V
H
/V
L
interactions [46], or production of helix-stabilized
Fv fragments [47]) and for camelid V
H
Hs (production of a
bivalent camelid antibody with dual affinities by coupling
two V
H
H domains through a semirigid linker [48]). Mul-
timerization of V
NAR
domains, either through addition of
C-terminal dimerization domains, or by using single chain-

like linkers, will result in relatively small sized proteins
( 30 kDa for the dimer, equivalent to one scFv) making
them attractive reagents for future diagnostic and possibly
therapeutic purposes [49].
Acknowledgements
We thank Drs Alex Kortt, Lindsay Sparrow, Jacqui Gulbis and Mike
Gorman, and Ms Chaille Webb for advice and discussions. We are
indebted to Mr Andreas Fischer for his assistance with collection of
blood samples from Wobbegong sharks. We thank Dr Terry Mulhern
for performing the CD spectral analysis and Mr N. Bartone for
oligonucleotide synthesis, and Mr P. Strike for assistance with
biosensor measurements.
References
1. Chothia, C., Lesk, A.M., Tramontano, A., Levitt, 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. & Poljak, R.J. (1989)
Conformations of immunoglobulin hypervariable regions. Nature
28, 877–883.
2. MacCallum, R.M., Martin, A.C. & Thornton, J.M. (1996) Anti-
body–antigen interactions: contact analysis and binding site
topography. J. Mol. Biol. 262, 732–745.
3. Padlan, E.A. (1994) Anatomy of the antibody molecule. Mol.
Immunol. 31, 169–217.
4. Hamers-Casterman, C., Atarhouch, T., Muyldermans, S., Rob-
inson, G., Hamers, C., Songa, E.B., Bendahman, N. & Hamers, R.
(1993) Naturally occurring antibodies devoid of light chains.
Nature 363, 446–448.
5. Greenberg, A.S., Avila, D., Hughes, M., Hughes, A., McKinney,
E.C. & Flajnik, M.F. (1995) A new antigen receptor gene family
that undergoes rearrangement and extensive somatic diversifica-

tion in sharks. Nature 374, 168–173.
6. Nuttall, S.D., Irving, R.A. & Hudson, P.J. (2000) Immuno-
globulin VH domains and beyond: design and selection of single-
domain binding and targeting reagents. Current Pharmaceut.
Biotechnol. 1, 253–263.
7. Diaz, M., Stanfield, R.L., Greenberg, A.S. & Flajnik, M.F. (2002)
Structural analysis, selection, and ontogeny of the shark new
antigen receptor (IgNAR): identification of a new locus pre-
ferentially expressed in early development. Immunogenetics 54,
501–512.
8. Muyldermans, S. (2001) Single domain camel antibodies: current
status. J. Biotechnol. 74, 277–302.
9. Flajnik, M.F. (1996) The immune system of ectothermic verte-
brates. Vet. Immunol. Immunopathol. 54, 145–150.
10. Nuttall, S.D., Krishnan, U.V., Hattarki, M., De Gori, R., Irving,
R.A. & Hudson, P.J. (2001) Isolation of the new antigen receptor
from Wobbegong sharks, and use as a scaffold for the display of
protein loop libraries. Mol. Immunol. 38, 313–326.
11. Roux, K.H., Greenberg, A.S., Greene, L., Strelets, L., Avila, D.,
McKinney, E.C. & Flajnik, M.F. (1998) Structural analysis of the
nurse shark (new) antigen receptor (NAR): molecular convergence
of NAR and unusual mammalian immunoglobulins. Proc. Natl
Acad. Sci. USA 95, 11804–11809.
12. Wu, T.T., Johnson, G. & Kabat. E.A. (1993) Length distribution
of CDRH3 in antibodies. Proteins 16, 1–7.
13. Vu, K.B., Ghahroudi, M.A., Wyns, L. & Muyldermans, S. (1997)
Comparison of llama VH sequences from conventional and heavy
chain antibodies. Mol. Immunol. 34, 1121–1131.
14. Muyldermans, S., Atarhouch, T., Saldanha, J., Barbosa, J.A. &
Hamers, R. (1994) Sequence and structure of VH domain from

naturally occurring camel heavy chain immunoglobulins lacking
light chains. Protein Eng. 7, 1129–1135.
15. Diaz, M., Greenberg, A.S. & Flajnik, M.F. (1998) Somatic
hypermutation of the new antigen receptor gene (NAR) in the
nurse shark does not generate the repertoire: possible role in
antigen-driven reactions in the absence of germinal centers. Proc.
NatlAcad.Sci.USA95, 14343–14348.
16. Diaz, M., Velez, J., Singh, M., Cerny, J. & Flajnik, M.F. (1999)
Mutational pattern of the nurse shark antigen receptor gene
(NAR) is similar to that of mammalian Ig genes and to sponta-
neous mutations in evolution: the translesion synthesis model of
somatic hypermutation. Int. Immunol. 11, 825–833.
17. Nuttall, S.D., Krishnan, U.V., Doughty, L., Alley, N., Hudson,
P.J., Pike, R.N., Kortt, A.A. & Irving, R.A. (2002) A naturally
occurring NAR variable domain against the Gingipain K protease
from Porphyromonas gingivalis. FEBS Lett. 516, 80–86.
18. Hoogenraad, N.J., Ward, L.A. & Ryan, M.T. (2002) Import and
assembly of proteins into mitochondria of mammalian cells.
Biochim. Biophys. Acta 1592, 97–105.
19. Suzuki, H., Maeda, M. & Mihara, K. (2002) Characterization of
rat TOM70 as a receptor of the preprotein translocase of the
mitochondrial outer membrane. J. Cell. Sci. 115, 1895–1905.
20. Brix, J., Dietmeier, K. & Pfanner, N. (1997) Differential recogni-
tion of preproteins by the purified cytosolic domains of the
mitochondrial import receptors, Tom20, Tom22, and Tom70.
J. Biol. Chem. 272, 20730–20735.
21. Young, J.C., Hoogenraad, N.J. & Hartl, F.U. (2003) Molecular
chaperones Hsp90 and Hsp70 deliver preproteins to the
mitochondrial import receptor Tom70. Cell 112, 41–50.
22. Alvarez-Dolado, M., Gonzalez-Moreno, M., Valencia, A., Zenke,

M., Bernal, J. & Munoz, A. (1999) Identification of a mammalian
homologue of the fungal Tom70 mitochondrial precursor protein
import receptor as a thyroid hormone-regulated gene in specific
brain regions. J. Neurochem. 73, 2240–2249.
23. Huston, J.S. & George, A.J. (2001) Engineered antibodies take
center stage. Hum. Antibodies 10, 127–142.
Ó FEBS 2003 An IgNAR variable domain specific for human Tom70 (Eur. J. Biochem. 270) 3553
24. Ausubel, F.M., Brent, R., Kingston, R.E., Moore, D.D., Seid-
man, J.G., Smith, J.A. & Struhl, K. (1989) Current Protocols in
Molecular Biology. John Wiley and Sons, New York.
25. Engberg, J., Andersen, P.S., Nielsen, L.F., Dziegiel, M., Johansen,
L.K. & Albrechtsen, B. (1996) Phage-display libraries of murine
and human antibody Fab fragments. Molec. Biotechnol. 6, 287–
310.
26. Galanis, M., Irving, R.A. & Hudson, P.J. (1997) Bacteriophage
library construction and selection of recombinant antibodies. In
Current Protocols in Immunology, pp. 17.1.1–17.1.45. John Wiley
and Sons, New York.
27. Coia, G., Ayres, A., Lilley, G.G., Hudson, P.J. & Irving, R.A.
(1997) Use of mutator cells as a means for increasing production
levels of a recombinant antibody directed against Hepatitis B.
Gene 201, 203–209.
28. Minsky, A., Summers, R.G. & Knowles, J.R. (1986) Secretion of
beta-lactamase into the periplasm of Escherichia coli: evidence for
a distinct release step associated with a conformational change.
Proc.NatlAcad.Sci.USA83, 4180–4184.
29. Gruen, L.C., Kortt, A.A. & Nice, E. (1993) Determination of
relative binding affinity of influenza virus N9 sialidases with the
Fab fragment of monoclonal antibody NC41 using biosensor
technology. Eur. J. Biochem. 217, 319–325.

30. Kortt, A.A., Nice, E. & Gruen, L.C. (1999) Analysis of the binding
of the Fab fragment of monoclonal antibody NC10 to influenza
virus N9 neuraminidase from tern and whale using the BIAcore
biosensor: effect of immobilization level and flow rate on kinetic
analysis. Anal. Biochem. 273, 133–141.
31. S
ˇ
ali, A. & Blundell, T.L. (1993) Comparative protein modelling by
satisfaction of spatial restraints. J. Mol. Biol. 234, 779–815.
32. Leung, D.W., Chen, E. & Goeddel, D.V. (1989) A method for
random mutagenesis of a defined DNA segment using a modified
polymerase chain reaction. Technique 1, 11–15.
33. Hoogenboom, H.R., de Bruine, A.P., Hufton, S.E., Hoet, R.M.,
Arends, J.W. & Roovers, R.C. (1998) Antibody phage display
technology and its applications. Immunotechnology 4, 1–20.
34. Brahms, S. & Brahms, J. (1980) Determination of protein
secondary structure in solution by vacuum ultraviolet circular
dichroism. J. Mol. Biol. 138, 149–178.
35. Proba, K., Worn, A., Honegger, A. & Pluckthun, A. (1998)
Antibody scFv fragments without disulfide bonds made by
molecular evolution. J. Mol. Biol. 275, 245–253.
36. Muyldermans, S. & Lauwereys, M. (1999) Unique single-domain
antigen binding fragments derived from naturally occurring camel
heavy-chain antibodies. J. Mol. Recognit. 12, 131–140.
37. Reiter, Y., Brinkmann, U., Lee, B. & Pastan, I. (1996) Engineering
antibody Fv fragments for cancer detection and therapy: disulfide-
stabilized Fv fragments. Nat. Biotechnol. 14, 1239–1245.
38. Desmyter, A., Spinelli, S., Payan, F., Lauwereys, M., Wyns, L.,
Muyldermans, S. & Cambillau, C.J. (2002) Three camelid VHH
domains in complex with porcine pancreatic alpha-amylase.

Inhibition and versatility of binding topology. J. Biol. Chem. 277,
23645–23650.
39. Worn, A. & Pluckthun, A. (1998) An intrinsically stable antibody
scFv fragment can tolerate the loss of both disulfide bonds and
fold correctly. FEBS Lett. 427, 357–361.
40. Martineau, P. & Betton, J.M. (1999) In vitro folding and thermo-
dynamic stability of an antibody fragment selected in vivo for high
expression levels in Escherichia coli cytoplasm. J. Mol. Biol. 292,
921–929.
41. Tavladoraki, P., Girotti, A., Donini, M., Arias, F.J., Mancini, C.,
Morea, V., Chiaraluce, R., Consalvi, V. & Benvenuto, E. (1999) A
single-chain antibody fragment is functionally expressed in the
cytoplasm of both Escherichia coli and transgenic plants. Eur
J. Biochem. 262, 617–624.
42. Lo Conte, L., Chothia, C. & Janin, J. (1999) The atomic structure
of protein–protein recognition sites. J. Mol. Biol. 285, 2177–2198.
43. Wiedemann, N., Pfanner, N. & Ryan, M.T. (2001) The three
modules of ADP/ATP carrier cooperate in receptor recruitment
and translocation into mitochondria. EMBO J. 20, 951–960.
44. Crameri, A., Cwirla, S. & Stemmer, W.P. (1996) Construction and
evolution of antibody-phage libraries by DNA shuffling. Nat.
Med. 2, 100–102.
45. Coia, G., Hudson, P.J. & Irving, R.A. (2001) Protein affinity
maturation in vivo using E. coli mutator cells. J. Immunol.
Methods 251, 187–193.
46. Kortt, A.A., Dolezal, O., Power, B.E. & Hudson, P.J. (2001)
Dimeric and trimeric antibodies: high avidity scFvs for cancer
targeting. Biomol. Eng. 18, 95–108.
47. Arndt, K.M., Mu
¨

ller, K.M. & Plu
¨
ckthun, A. (2001) Helix-stabi-
lized Fv (hsFv) antibody fragments: substituting the constant
domains of a Fab fragment for a heterodimeric coiled-coil
domain. J. Mol. Biol. 312, 221–228.
48. Els Conrath, K., Lauwereys, M., Wyns, L. & Muyldermans, S.
(2001) Camel single-domain antibodies as modular building units
in bispecific and bivalent antibody constructs. J. Biol. Chem. 276,
7346–7350.
49. Hudson, P.J. & Souriau, C. (2001) Recombinant antibodies
for cancer diagnosis and therapy. Expert Opin. Biol. Ther. 1,
845–855.
3554 S. D. Nuttall et al. (Eur. J. Biochem. 270) Ó FEBS 2003