THE JOURNAL OF BIOLOGICAL CHEMISTRY
© 2000 by The American Society for Biochemistry and Molecular Biology, Inc.
Vol. 275, No. 18, Issue of May 5, pp. 13668 –13676, 2000
Printed in U.S.A.
The Rabbit Antibody Repertoire as a Novel Source for the
Generation of Therapeutic Human Antibodies*
Received for publication, December 17, 1999, and in revised form, January 29, 2000
Christoph Rader‡§, Gerd Ritter¶, Sheila Nathan‡储, Marikka Elia‡, Ivan Gout**,
Achim A. Jungbluth¶, Leonard S. Cohen¶, Sydney Welt¶, Lloyd J. Old¶, and
Carlos F. Barbas III‡ ‡‡
From ‡The Skaggs Institute for Chemical Biology and the Department of Molecular Biology, The Scripps Research
Institute, La Jolla, California 92037, the ¶Ludwig Institute for Cancer Research, New York Branch, Memorial
Sloan-Kettering Cancer Center, New York, New York 10021, and the **Ludwig Institute for Cancer Research,
London Branch at University College, London School of Medicine, W1P 8BT London, United Kingdom
The rabbit antibody repertoire, which in the form of
polyclonal antibodies has been used in diagnostic applications for decades, would be an attractive source for
the generation of therapeutic human antibodies. The
humanization of rabbit antibodies, however, has not
been reported. Here we use phage display technology to
select and humanize antibodies from rabbits that were
immunized with human A33 antigen which is a target
antigen for the immunotherapy of colon cancer. We first
selected rabbit antibodies that bind to a cell surface
epitope of human A33 antigen with an affinity in the 1
nM range. For rabbit antibody humanization, we then
used a selection strategy that combines grafting of the
complementarity determining regions with framework
fine tuning. The resulting humanized antibodies were
found to retain both high specificity and affinity for
human A33 antigen.
The growing significance of antibody-based strategies for the
treatment of a variety of diseases demands efficient and reliable routes to human or humanized antibodies with high specificity and affinity. Nonhuman antibodies are highly immunogenic in humans thereby limiting their potential use for
therapeutic applications, especially when repeated administration is necessary. In order to reduce their immunogenicity,
nonhuman antibodies have been humanized using strategies
that are based on rational design, in vitro evolution, or a
combination of both. An alternative route is the direct generation of human antibodies from transgenic mice containing human immunoglobulin (Ig)1 loci or by selection from naive or
synthetic human antibody libraries displayed on phage (1, 2).
* This work was supported in part by The Skaggs Institute for Chemical Biology. The costs of publication of this article were defrayed in part
by the payment of page charges. This article must therefore be hereby
marked “advertisement” in accordance with 18 U.S.C. Section 1734
solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted
to the GenBankTM/EBI Data Bank with accession number(s) AF
245498 –AF 245503.
§ Supported by postdoctoral fellowships from the Swiss National
Science Foundation and the Krebsliga des Kantons Zürich.
储 Current address: CGAT, Faculty of Science and Technology, Universiti Kebangsaan Malaysia, 43600 UKM Bangi, Selangor D.E.,
Malaysia.
‡‡ To whom correspondence should be addressed: 10550 N. Torrey
Pines Rd., La Jolla, CA 92037. Tel.: 858-784-9098; Fax: 858-784-2583;
E-mail: carlos@scripps.edu.
1
The abbreviations used are: Ig, immunoglobulin; CDR, complementarity-determining region; V, variable domain; C, constant domain;
Fab, antibody fragment; Fd, heavy chain fragment; scFv, single-chain
variable domain fragment; PBS, phosphate-buffered saline; ELISA,
Here we exploit a new route to humanized antibodies. We use
phage display technology to select and humanize antibodies
from rabbits that were immunized with a human antigen.
Monoclonal antibodies from multiple species are accessible by
screening combinatorial antibody libraries displayed on phage
(3). The accessibility of antibody repertoires of multiple species
might become substantial in the search for new therapeutic
antibodies. In particular the rabbit antibody repertoire, which
in the form of polyclonal antibodies has been used in diagnostic
applications for decades, would be an attractive source for
therapeutic antibodies. The humanization of rabbit antibodies,
however, has not been reported.
To allow selective antibody targeting, both high specificity of
the antibody and restrictive tissue expression of the antigen
are required. For cancer therapy, only a few antibody-antigen
systems that meet these requirements have been identified (4).
The A33 system which has been established in clinical trials is
one of them (5–7). Mouse monoclonal antibody A33 (5) detects
an antigen that is expressed in normal human colon epithelial
cells and by ⬎95% of human colon cancers. Human A33 antigen
(8, 9) is a transmembrane glycoprotein of the Ig superfamily. It
consists of 2 extracellular Ig folds, a single transmembrane
domain, and a highly polar intracellular tail containing a Spalmitoylation site. The function of the human A33 antigen in
normal and malignant colon tissue is not yet known, however,
several properties of the A33 antigen suggested that it is a
promising target for immunotherapy of colon cancer (10). These
properties include (i) the highly restricted expression pattern of
the A33 antigen, (ii) the expression of large amounts of the A33
antigen on colon cancer cells, (iii) the absence of secreted or
shed A33 antigen, (iv) the fact that upon binding of antibody
A33 to the A33 antigen, antibody A33 is internalized and
sequestered in vesicles, and (v) the good targeting of antibody
A33 to A33 antigen expressing colon cancer in preliminary
clinical studies. Here we chose the human A33 antigen as a
therapeutically relevant target to demonstrate both the applicability and rationale of generating humanized antibodies from
immune rabbits.
EXPERIMENTAL PROCEDURES
Cell Lines—Human colon cancer cell lines LIM1215, SW1222, HT29,
and SW620 were obtained from the cell bank of the Ludwig Institute for
Cancer Research at Memorial Sloan-Kettering Cancer Center.
Recombinant Human A33 Antigen—A 1.6-kilobase XhoI/PstI cDNA
fragment, containing the full-length coding sequence of human A33
antigen (8), was subcloned into pBlueBac4 transfer vector (Invitrogen,
San Diego, CA). Transfection of Sf9 cells, isolation of recombinant
enzyme-linked immunosorbent assay; TBS, Tris-buffered saline; PAGE,
polyacrylamide gel electophoresis.
13668
This paper is available on line at http://www.jbc.org
Generation and Humanization of Rabbit Antibodies
viruses, and subsequent infection of Sf9 cells was performed according
to the manufacturer’s recommendations. Recombinant protein was purified from cell extracts as described (11).
Rabbit Immunization—Over a period of 4 to 5 months, 2 rabbits from
the New Zealand White strain were treated with 3 subcutaneous injections of 106 LIM1215 cells followed by 3 subcutaneous injections of 1 ␮g
of recombinant human A33 antigen in a 1-ml emulsion of Ribi adjuvant
in PBS (Ribi Immunochem Research, Hamilton, MT). Antisera from the
immune animals was analyzed for binding to recombinant human A33
antigen by ELISA using alkaline phosphatase-conjugated goat antirabbit Fc polyclonal antibodies (Cappel, West Chester, PA) as secondary antibodies. Five days after the final boost, spleen and bone marrow
from 1 leg were harvested and used for total RNA preparation with
TRI-REAGENT from Molecular Research Center (Cincinnati, OH).
First-strand cDNA was synthesized using the Superscript Preamplification System for First Strand cDNA Synthesis kit with oligo(dT)
priming (Life Technologies, Inc.).
Rabbit Antibody Library—The first-strand cDNAs from each rabbit
were subjected to separate 35-cycle polymerase chain reactions using
Perkin-Elmer AmpliTaq DNA polymerase (Roche Molecular Systems,
Brunchburg, NY) and 10 primer combinations for the amplification of
rabbit VL (9 ⫻ V␬ and 1 ⫻ V␭) coding sequences and 4 combinations for
the amplification of rabbit VH coding sequences (Table I). The antisense
primers consist of a hybrid rabbit/human sequence designed for the
fusion of rabbit VL and VH coding sequences to human C␬ and CH1
coding sequences, respectively. Human C␬ and CH1 coding sequences
were amplified from a pComb3H-compatible expression vector (3) that
contained the sequence of a human Fab directed to tetanus toxoid (12)
using the primer combination HKC-F (5-ACTGTGGCTGCACCATCTGTC-3⬘)/lead-B (5⬘GCCATGGCTGGTTGGGCAGC-3⬘) and HIgGCH1-F
(5⬘-GCCTCCACCAAGGGCCCATCGGTC-3⬘)/dpseq (5⬘-AGAAGCGTAGTCCGGAACGTC-3⬘), respectively. The antisense primer lead-B hybridizes to a sequence upstream of the Fd fragment coding sequence
and is used to amplify the C␬ coding sequence together with the sequence intervening LC and Fd fragment coding sequences in phagemid
vector pComb3H (3). Based on this strategy, the chimeric rabbit/human
LC and Fd fragment coding sequences are assembled and fused by 2
sequential overlap extension polymerase chain reaction steps. In the
first step, rabbit VL and human C␬ are fused using the primer combination RSC-F (5⬘-GAGGAGGAGGAGGAGGAGGCGGGGCCCAGGCGGCCGAGCTC-3⬘)/lead-B and rabbit VH and human CH1 are fused using
the primer combination lead-VH (5-GCTGCCCAACCAGCCATGGCC3⬘) and dp-EX (5⬘-GAGGAGGAGGAGGAGGAGAGAAGCGTAGTCCGGAACGTC-3⬘). In the second step, the assembled chimeric LC and Fd
fragment coding sequences are fused using the flanking primers RSC-F
and dp-EX. Only LC and Fd fragment coding sequences derived from
the same animal were combined. The final construct was cloned into
phagemid vector pComb3H using 2 asymmetric SfiI sites (3), yielding a
complexity of 2 ⫻ 107 independent transformants. The phage library
displaying the chimeric rabbit/human Fab was panned against immobilized recombinant human A33 antigen using 200 ng of protein in 25 ␮l
of TBS for coating on 1 well of a Costar number 3690 96-well plate,
0.05% (v/v) Tween 20 in TBS for washing, and 10 mg/ml trypsin (Difco)
in TBS for elution. Trypsinization was for 30 min at 37 °C. The washing
steps were increased from 5 in the first round to 10 in the second round
and 15 in the third and fourth round. A round of panning defines a
sequence of binding of the phage to the antigen, removal of unspecific
binders, and elution of specific binders followed by their reamplification
in Escherichia coli. Indicative of a successful selection, the output
numbers strongly increased after the third and fourth round. The
output phage pool of each round was monitored by phage ELISA using
horseradish peroxidase-conjugated sheep anti-M13 phage polyclonal
antibodies (Amersham Pharmacia Biotech, Buckinghamshire, United
Kingdom) as secondary antibodies. An increasing signal above background from round to round was noted. Fourty clones from the final
output were grown and induced with 1 mM isopropyl-␤-D-thiogalactopyransoside. Supernatants were tested for binding to immobilized recombinant human A33 antigen by ELISA using alkaline phosphataseconjugated goat anti-human F(ab⬘)2 polyclonal antibodies (Pierce,
Rockford, IL) as secondary antibody. All clones gave a strong signal
above background and were further analyzed by DNA fingerprinting.
For this, phagemids were isolated, Fab coding sequences amplified
using the flanking primers ompseq (5⬘-AAGACAGCTATCGCGATTGCAG-3⬘) and gback (5⬘-GCCCCCTTATTAGCGTTTGCCATC-3⬘), and digested with the 4-base pair cutter BstOI. Three different, although
highly similar, fingerprints were obtained. Two of them, rabbit clones 1
and 2, were found in 13 and 26 clones, respectively. The third rabbit
clone (3) was represented only once. The sequence of the variable
13669
domains of heavy and light chain of clones 1, 2, and 3 was determined
by DNA sequencing using the primers newpelseq (5⬘-CTATTGCCTACGGCAGCCGCTG-3⬘) and ompseq, respectively. Rabbit antibody sequences have been submitted to GenBank™ under accession numbers
AF 245498 –245503.
Humanized Antibody Library—Human germ-line VL and VH sequences with the highest degree of homology with the corresponding
rabbit sequences were identified from the VBASE Directory of Human
V Gene Sequences by amino acid sequence alignment. The identified
human sequences were diversified at positions that are potentially
involved in antigen binding and used as frameworks for the grafting of
the 6 rabbit CDRs as defined by Kabat et al. (13). Overlapping oligonucleotides were designed accordingly, synthesized, and assembled to
synthetic VL and VH coding sequences in 2 sequential 10-cycle polymerase chain reaction steps using the Expand High Fidelity PCR System
(Roche Molecular Systems). Using the same procedure as described for
the generation of the rabbit antibody library, the synthetic VL and VH
coding sequences were fused to human C␬ and CH1 coding sequences,
respectively, and the resulting humanized light chain and Fd fragment
coding sequences were assembled. The final construct was SfiI cloned
into a variant of pComb3H in which the ampicillin resistance gene had
been replaced by a chloramphenicol resistance. This prevents potential
contaminations with phage derived from the rabbit antibody libraries.
The resulting library with a theoretical complexity of only 256 consisted
of 1 ⫻ 107 independent transformants and was panned against immobilized recombinant human A33 antigen essentially as described for the
screening of the rabbit antibody library but under more stringent conditions. The amount of antigen was decreased from 100 ng in the first
and second round to 50 ng in the third and fourth round and 25 ng in the
fifth and sixth round. Ten washing steps with 0.5% (v/v) Tween 20 in
TBS were performed for each of the 6 rounds. Round 3 and 4 as well as
5 and 6 were linked without phage amplification. For this, phage from
round 3 and 5 were eluted by adding 50 ␮l of 100 mM HCl glycine (pH
2.2), incubated for 10 min at room temperature, collected, neutralized
with 3 ␮l of 2 M Tris base and 50 ␮l of 1% (w/v) bovine serum albumin
in TBS, and directly subjected to another round of panning. Phage from
rounds 1, 2, 4, and 6 were eluted by trypsinization as described before.
Out of 70 clones from the final output that were analyzed by ELISA and
all found to be positive, 24 clones were further analyzed by DNA
sequencing as described before.
The following oligonucleotides were used for humanization, L denotes primers for the VL assembly, H denotes primers for the VH
assembly (r ⫽ A or G; S ⫽ G or C; W ⫽ A or T): L1, 5⬘- GAGCTCCAGATGACCCAGTCTCCATCCTCCCTGTCTGCATCTGTAGGAGACAGAGTCACCATCACTTGCCTGGCCAGTGAGTTCCTTTTTAATGGTGTATCC-3⬘; L2, 5⬘- AGATGGGACCCCAGATTCTAAATTGGATGCACCATAGATCAGGARCTTAGGARCTTTCCCTGGTTTCTGCTGATACCAGGATACACCATTAAAAAGGAACTC-3⬘; L3, 5⬘- AATTTAGAATCTGGGGTCCCATCTCGGTTCAGTGGCAGTGGATCTGGGACAGATTWCACTCTCACCATCAGCAGCCTGCAGSCTGAAGATGTTGCAACT-3⬘; L4,
5⬘- TTTGATCTCCACCTTGGTCCCTCCGCCGAAAGTCAAACCACTACTACCACTATAACCGCCTAGACAGTAATAAGTTGCAACATCTTCAGSCTGCAG-3⬘; L flank sense, 5⬘-GAGGAGGAGGAGGAGGGCCCAGGCGGCCGAGCTCCAGATGACCCAGTCTCCA-3⬘; L antisense flank,
5⬘- GACAGATGGTGCAGCCACAGTTCGTTTGATCTCCACCTTGGTCCCTCC-3⬘; H1, 5⬘- GAGGTGCAGCTGGTGGAGTCTGGGGGAGGCCTGGTCAAGCCTGGGGGGTCCCTGAGACTCTCCTGTGCAGCCTCTGGA-3⬘; H2A, 5⬘- CCAGCTAATGCCATAGTGACTGAAGGTGAATCCAGAGGCTGCACAGGAGAGTCT-3⬘; H2B, 5⬘- CCAGCTAATGCCATAGTGACTGAAGTCGATTCCAGAGGCTGCACAGGAGAGTCT-3⬘; H3, 5⬘TTCAGTCACTATGGCATTAGCTGGGTCCGCCAGGCTCCAGGGAAGGGGCTGGAGTGGGTCGCCTACATTTATCCTAATTATGGGAGTGTAGACTACGCGAGC-3⬘; H4A, 5⬘- GTTCATTTGCAGATACASTGAGTTCTTGGCGTTGTCTCTGGAGATGGTGAATCGGCCATTCACGCTGCTCGCGTAGTCTACACTCCCATA-3⬘; H4B, 5⬘- GTTCATTTGCAGATACASTGAGTTCTGGGCGTTGTCGAGGGAGATGGTGAATCGGCCATTCACGCTGCTCGCGTAGTCTACACTCCCATA-3⬘; H5, 5⬘- AACTCASTGTATCTGCAAATGAACAGCCTGAGAGCCGAGGACACGGCCGTATATTWCTGTGCGAGAGATCGGGGTTATTATTCTGGTAGT-3⬘; H6,
5⬘- TGAGGAGACGGTGACCAGGGTGCCCTGGCCCCAGAGATCCAACCGAGTCCCCCTACTACCAGAATAATAACCCCGATC-3⬘; H flank
sense, 5⬘-GCTGCCCAACCAGCCATGGCCGAGGTGCAGCTGGTGGAGTCTGGGGGA-3⬘; H flank antisense, 5⬘- GACCGATGGGCCCTTGGTGGAGGCTGAGGAGACGGTGACCAGGGTGCC-3⬘.
Fab Preparation—Soluble Fab from rabbit clones 1 and 2 and humanized clones A to F were produced in E. coli strain Xl1-Blue using
gene III fragment-depleted pComb3H as expression vector (3). Fab were
purified from concentrated supernatants and from sonicated lysates of
13670
Generation and Humanization of Rabbit Antibodies
overnight 1-liter flask cultures induced with 1 mM isopropyl-␤-D-thiogalactopyransoside by affinity chromatography using a 5-ml Protein G
HiTrap column (Amersham Pharmacia Biotech) attached to an fast
protein liquid chromatography system (Amersham Pharmacia Biotech).
PBS was used as equilibration and washing buffer and 0.5 M acetic acid
for elution. Eluted fractions were neutralized immediately using 0.5
volumes of 1 M Tris-HCl (pH 9.0), pooled, concentrated, and brought
into PBS. Quality was analyzed by SDS-PAGE followed by Coomassie
Blue staining, quantity by measuring the optical density at 280 nm
using an Eppendorf BioPhotometer (Hamburg, Germany).
Surface Plasmon Resonance—Surface plasmon resonance for the determination of association (kon) and dissociation (koff) rate constants for
binding of rabbit and humanized Fab to recombinant human A33 antigen was performed on a Biacore instrument (Biacore AB, Uppsala,
Sweden). A CM5 sensor chip (Biacore AB) was activated for immobilization with N-hydroxysuccinimide and N-ethyl-N⬘-(3-dimethylaminopropyl)carbodiimide according to the methods outlined by the supplier.
Recombinant human A33 antigen was coupled at a low density to the
surface by injection of 30 to 40 ␮l of a 1 ng/␮l sample in 10 mM sodium
acetate (pH 3.5). Approximately 500 resonance units were immobilized.
Subsequently, the sensor chip was deactivated with 1 M ethanolamine
hydrochloride (pH 8.5). Binding of Fab to immobilized A33 antigen was
studied by injection of Fab at 5 different concentrations ranging from 75
to 200 nM. PBS was used as the running buffer. The sensor chip was
regenerated with 20 mM HCl and remained active for at least 50
measurements. The kon and koff values were calculated using Biacore
AB evaluation software. The equilibrium dissociation constant Kd was
calculated from koff/kon. Data obtained from different sensor chips revealed a high consistency and were further validated as described (19).
Western Blotting—Triton X-100 (0.3% (v/v) in PBS, pH 7.5) lysates of
human colon cancer cell lines were resolved by SDS-PAGE under reducing (5% (v/v) ␤-mercaptoethanol) and nonreducing conditions. Proteins were blotted to polyvinylidene difluoride membranes (Immobilon-P, Millipore, Bedford, MA) and incubated with 0.5 ␮g/ml Fab
followed by alkaline phosphatase-conjugated goat anti-human F(ab⬘)2
polyclonal antibodies (Pierce). Specific binding was visualized by chemiluminescence (Tropix, Bedford, MA).
Flow Cytometry—Flow cytometry was performed using a FACSscan
instrument from Becton-Dickinson (Franklin Lakes, NJ). For each determination, 1 ⫻ 104 cells were analyzed. Indirect immunofluorescence
staining was achieved with 2 ␮g/ml Fab in 1% (w/v) bovine serum
albumin, 25 mM Hepes, 0.05% (w/v) sodium azide in PBS. A 1:100
dilution of fluorescein isithiocyanate-conjugated donkey anti-human
F(ab⬘)2 polyclonal antibodies (Jackson ImmunoResearch Laboratories,
West Grove, PA) was used for detection. Incubation with primary antibodies was for 1 h, with secondary antibodies for 30 min at room
temperature.
Immunohistochemistry—All immunochemical stainings were done
on snap-frozen tissue samples, embedded in O.C.T. compound (Tissue
Tek, Torrance, CA). 5-␮m cuts (HM503 cryostat, Zeiss, Walldorf, Germany) were mounted on slides for immunohistochemistry (Superfrost
Plus, Fisher Scientific, Pittsburgh, PA). Serial sections were used, so as
to compare staining results of the different antibody preparations. After
cutting, the slides were fixed in cold acetone for 10 min and then air
dried. Reactivity of the humanized Fab was analyzed using the colon
cancer cell line SW1222 xenografted into nude mice. A working concentration of Fab (1 ␮g/ml) was established by titering. The humanized Fab
was detected by biotinylated goat-anti human F(ab)2 polyclonal antibodies (1:200; Vector, Burlingame, CA) and an avidin-biotin complex
system (ABC/Elite kit, Vector). Diaminobenzidine tetrahydrochloride
(Biogenex, San Ramon, CA) was used as a chromogen. Reactivity of the
humanized Fab was also evaluated in human colonic adenocarcinoma
samples. In order to prevent immunoreactivity of endogenous human
immunoglobulin, a special technique for the detection of humanized
Fab was utilized. Prior addition to tissue, the humanized Fab (1 ␮g/ml)
was incubated with biotinylated goat-anti human F(ab)2 polyclonal
antibodies in a test tube. The optimal ratio of humanized Fab to secondary antibody was determined in separate titration assays. Incubation of humanized Fab and secondary antibody was done at room
temperature for 1 h and followed by an addition of human serum in
order to block the activity of unbound secondary antibody. Again, the
optimal ratio of human serum to secondary antibody was determined in
separate titration assays.
RESULTS
Our strategy for the generation and humanization of rabbit
antibodies directed to human A33 antigen is outlined in Fig. 1.
FIG. 1. Flow chart outlining the process of generating and
humanizing rabbit antibodies directed to a target antigen.
Two rabbits were immunized with the human colon cancer cell
line LIM1215 which expresses large amounts of the A33 antigen. A LIM1215 cDNA library had been used to clone the cDNA
for human A33 antigen (8). The initial immunization and 2
boosts elicited a weak immune response against recombinant
human A33 antigen as revealed by the analysis of the rabbit
sera by ELISA (data not shown). Both rabbits were further
boosted with recombinant human A33 antigen derived from a
baculovirus expression system. A strong immune response
against recombinant human A33 antigen in sera from both
rabbits resulted (data not shown). Our sequential cell and
protein immunization strategy aimed at targeting the humoral
immune response to native epitopes of the protein that are
accessible at the cell surface.
A rabbit antibody library displayed on phage was generated
as follows. RNA was isolated from bone marrow and spleen of
the immune rabbits, retro-transcribed, and VL and VH coding
sequences were amplified using a variety of primer combinations designed to amplify most of the known rabbit antibody
sequences (Table I). The primer combinations used here extend
the primer sets reported previously (14 –16). Importantly, the
rabbit antibody library was based on a chimeric Fab format.
Variable domains from rabbit light and heavy chains were
fused to the corresponding human constant domains. The use
of human constant domains suggested several advantages.
First, while antigen binding is confined to the variable domains
and, thus, is not expected to be influenced by constant domain
swapping, the human constant domains confer established and
standardized detection and purification means on Fab derived
from multiple species. Second, the use of human constant regions was found to improve the E. coli expression level of Fab
(17, 18). Lastly, a Fab with human constant domains is already
partially humanized and can be readily channeled into recently
reported strategies for complete humanization (19).
The phage library displaying chimeric rabbit/human Fab
was selected by panning against immobilized recombinant human A33 antigen. Among 40 selected clones that demonstrated
strong reactivity in ELISA on immobilized recombinant human
A33 antigen, only 3 distinct sequences were identified (Fig. 2).
Data bank screening revealed that the sequences corresponding to the variable domains were rabbit sequences. The clones
with distinct sequences were designated rabbit clones 1 to 3.
Rabbit clones 1 and 2 shared identical V␬ coding sequences and
90% identical VH coding sequences. Rabbit clone 3 consisted of a
V␬ coding sequence with 90% identity to rabbit clones 1 and 2 and
Generation and Humanization of Rabbit Antibodies
13671
TABLE I
Primers for the amplification of rabbit V␬, V␭, and VH coding sequences
Note that the 3⬘ antisense primers are designed to allow the fusion of rabbit variable domains to human constant domains (M ⫽ A or C; R ⫽ A
or G; S ⫽ G or C; Y ⫽ C or T).
V␬ 5⬘ sense primers
1, 5⬘-GGGCCCAGGCGGCCGAGCTCGTGMTGACCCAGACTCCA-3⬘
2, 5⬘-GGGCCCAGGCGGCCGAGCTCGATMTGACCCAGACTCCA-3⬘
3, 5⬘-GGGCCCAGGCGGCCGAGCTCGTGATGACCCAGACTGAA-3⬘
V␬ 3⬘ antisense primers
1, 5⬘-ACAGATGGTGCAGCCACAGTTAGGATCTCCAGCTCGGTCCC-3⬘
2, 5⬘-GACAGATGGTGCAGCCACAGTTTTGATTTCCACATTGGTGCC-3⬘
3, 5⬘-GACAGATGGTGCAGCCACAGTTTTGACSACCACCTCGGTCCC-3⬘
V␭ 5⬘ sense primer
5⬘-GGGCCCAGGCGGCCGAGCTCGTGCTGACTCAGTCGCCCTC-3⬘
V␭ 3⬘ antisense primer
5⬘-CGAGGGGGCAGCCTTGGGCTGGCCTGTGACGGTCAGCTGGGTCCC-3⬘
VH 5⬘ sense primers
1, 5⬘-GCTGCCCAACCAGCCATGGCCCAGTCGGTGGAGGAGTCCRGG-3⬘
2, 5⬘-GCTGCCCAACCAGCCATGGCCCAGTCGGTGAAGGAGTCCGAG-3⬘
3, 5⬘-GCTGCCCAACCAGCCATGGCCCAGTCGYTGGAGGAGTCCGGG-3⬘
4, 5⬘-GCTGCCCAACCAGCCATGGCCCAGSAGCAGCTGRTGGAGTCCGG-3⬘
VH 3⬘ antisense primer
5⬘-CGATGGGCCCTTGGTGGAGGCTGARGAGAYGGTGACCAGGGTGCC-3⬘
FIG. 2. Amino acid sequence alignment of selected rabbit clones 1–3 and humanized clones A-F. Shown are frameworks (FRs) and
CDRs of VL and VH sequences. Dashes indicate identical amino acids. The humanized library with diversified human FRs and rabbit CDRs is
shown in the center. Diversified positions are shown in bold type. Underlined positions indicate a coupled diversification that limits the selection
to either all human or all rabbit sequence.
shared an identical VH coding sequence with rabbit clone 3. The
high similarity of the sequences included the hypervariable VDJ
and VJ joint regions HCDR3 and LCDR3. Thus, it is likely that
all the selected sequences originated from 1 B cell clone in 1 of the
2 rabbits that had undergone somatic diversification.
Rabbit clones 1 and 2 were produced as soluble Fab in E. coli
and purified by protein G affinity chromatography (Fig. 3).
Analysis by flow cytometry revealed that both Fab specifically
bound to human colon cancer cells expressing native human
A33 antigen (Fig. 4). Binding of the Fab to human A33 antigen
was found to be very strong with affinities in the 1 nM range as
determined by surface plasmon resonance using a Biacore instrument. Rabbit clones 1 and 2 gave Kd values of 390 pM and
1.6 nM, respectively (Fig. 5, Table III). While both a higher
association and a slower dissociation rate constant contributed
to the higher affinity of rabbit clone 1, rabbit clone 2 gave
consistently higher expression yields in E. coli. Considering the
fact that the majority of selected clones contained the rabbit
clone 2 sequence, it can be assumed that the higher expression
level of rabbit clone 2 led to its preferential selection despite the
higher affinity of rabbit clone 1.
As a further step in the generation of therapeutic antibodies
directed to human A33 antigen, the selected rabbit variable
domains were humanized. We used a humanization strategy
that combines CDR grafting with framework fine tuning by
phage display. This type of strategy has been used for the
13672
Generation and Humanization of Rabbit Antibodies
humanization of mouse antibodies (20, 21). The selected rabbit
sequences were first aligned with human V and J genes. The
selected rabbit VH sequence was found to be most homologous
to the human V gene DP-77/3-21 from the VH3 family and
human J gene JH1. The selected rabbit V␬ sequence was found
to be most homologous with the human V gene DPK-4/A20 from
the V␬1 family and human J gene J␬4. These human sequences
not only gave the best alignment with the selected rabbit se-
FIG. 3. Analysis of purified rabbit and humanized Fab by SDSPAGE and Coomassie Blue staining. Fab were purified from E. coli
cultures by Protein G affinity chromatography. Numbers on the right
indicate molecular masses of standard proteins in kDa.
FIG. 4. Flow cytometry histograms demonstrating that the selected rabbit clones 1 and 2, as well as the selected humanized
clones A-F, bind specifically to native human A33 antigen expressed on the cell surface. For indirect immunofluorescence staining, cells were incubated with Fab (except for the control) followed by
fluorescein isithiocyanate-conjugated secondary antibodies. Human colon cancer cell lines LIM1215 (bold line) and SW1222 (fine line) express
human A33 antigen, whereas HT29 (dotted line) is not. The y axis gives
the number of events in linear scale, the x axis the fluorescence intensity in logarithmic scale.
FIG. 5. Representative Biacore sensorgrams obtained for the binding of
rabbit Fab 1 to immobilized human
A33 antigen. For association, Fab were
injected at 5 different concentrations
(200, 150, 125, 100, and 75 nM; top to
bottom) between t ⫽ 125 s and t ⫽ 370 s
using a flow rate of 5 ␮l/min. For dissociation, the flow rate was increased to 50
␮l/min. RU, resonance units.
quences, but are frequently found in the human antibody repertoire in vivo (22) and are highly related to human V genes
DP-47/3-23 and DPK-9/O2 whose frameworks have been used
for a number of mouse antibody humanizations and have given
high expression yields in E. coli (23). In fact, pairs of VH3
family heavy chains and V␬1 family light chains are the most
frequent combinations found in native human antibodies (24).
Thus, this framework combination is expected to be non-immunogenic in humans. The CDR sequences of rabbit clone 2 were
chosen because of the higher expression level of this clone. The
6 rabbit CDR sequences, 3 from each variable domain as defined by Kabat et al. (13), were grafted into the appropriate
human framework sequences. For further humanization, tryptophan at position 62 in the rabbit HCDR2 was converted to
serine. For framework fine tuning, residues at 6 positions in
the human VH framework sequences and residues at 4 positions in the human V␬ framework sequences were diversified as
shown in Table II. These residues were chosen out of a set of
key framework residues that are known to be involved in antigen binding, either indirectly, by supporting the confirmation
of the CDR loops, or directly, by contacting the antigen (25).
Key framework residues of rodent monoclonal antibodies have
been identified structurally by crystallization and molecular
modeling as well as empirically by antibody humanization. No
such data are available for rabbit monoclonal antibodies. However, the similarity of the primary structure of rabbit antibodies with rodent antibodies (13) makes an analogous set of key
framework residues likely. From this set, 10 positions that
differed between the rabbit and the human frameworks were
identified. These positions were then diversified to allow the
selection of basically either the rabbit or the human sequence
(Table II). Position 43 in framework 2 of VL was diversified to
allow the selection of either human residue alanine or valine,
reflecting the natural diversity of this position in light chains of
the V␬1 family (13).
The phage library displaying Fab with a diversified human
framework sequence and rabbit CDR sequences was selected
by panning against immobilized recombinant human A33 antigen using highly stringent conditions (see “Experimental Procedures”). All selected clones demonstrated strong reactivity in
ELISA on immobilized recombinant human A33 antigen. The
sequences of 6 selected clones, designated as humanized clones
A-F, are given in Fig. 2. These clones reflect the diversity found
in 24 clones that were sequenced. A consensus sequence was
found for the diversified framework of VH. Positions 27 and 28
in framework 1 and positions 71 and 75 in framework 3 were
found to contain the original rabbit residues isoleucine, aspartic acid, leucine, and glutamine, respectively, in 16 out of 24
clones. While 3 clones contained the human residues phenylalanine and threonine at positions 27 and 28, respectively,
none of the clones contained the human residues arginine and
Generation and Humanization of Rabbit Antibodies
TABLE II
Key framework residues targeted for diversification
Linked positions (VH 27–28 and VH 71–75, respectively) indicate a
coupled diversification that limits the selection to either all-human or
all-rabbit sequence.
Position
VL
43
46
71
80
VH
27
28
71
75
78
91
Human
Rabbit
A
L
F
P
P
F
Y
A
F
T
R
K
L
Y
I
D
L
Q
V
F
Diversification
A,
L,
F,
P,
V
F
Y
A
 FI 
TD
RL
KQ
L, V
Y, F
lysine at positions 71 and 75, respectively. Interestingly, 2 of
the diversified positions were found to contain mutations, i.e.
residues that were not included in the diversification. Both
mutations were due to a single point mutation that was probably generated by misincorporation during oligonucleotide synthesis or assembly. A glycine at position 28 was found in 3
clones including clones B and F (Fig. 2). A phenylalanine at
position 71 was found in 2 different clones, A and E (Fig. 2).
Most notably, these 5 clones containing the mutations demonstrated the strongest reactivity in ELISA. The remaining 2
positions that were diversified in the framework of VH, i.e.
positions 78 and 91 in framework 3, did not give a significant
consensus sequence but reflected a rather random selection of
either the human or the rabbit residue. A similar result was
obtained in the framework of VL at positions 43 and 46 in
framework 2 and position 71 in framework 3. The human
proline residue at diversified position 80 of framework 3 of VL
was found in 18 out of 24 clones including the 5 mutated clones
that demonstrated the strongest reactivity in ELISA. The consensus sequence selected at 5 out of 10 diversified positions
with a preference of the rabbit sequence at 4 positions demonstrates the potency of the framework fine tuning approach. The
fact that further mutations were found in diversified but not in
constant positions supporting our set of key framework residues. Considering the potential immunogenicity of rabbit residues in a human framework, we analyzed the natural diversity of the human framework using the Kabat data base. This
analysis revealed that all 10 diversified positions are naturally
diverse. In fact, only 2 positions in our framework diversification yielded residues that have not been found at these positions in human frameworks. These are aspartic acid and glycine at position 28 in framework 1 of VH and leucine and
phenylalanine at position 71 in framework 3 of VH. In a procedure termed resurfacing mouse antibodies have been humanized by just replacing a small set of surface accessible residues
(26, 27). This was based on the suggestion that the humanization of surface accessible residues of a mouse antibody would
decrease its immunogenicity (28). The probability of a human
framework residue to be exposed at the antibody surface has
been determined by comparing three-dimensional structures of
human antibodies (28, 29). Among our 10 diversified framework residues, only 3 positions are likely to be exposed. These
are position 80 of VL and positions 28 and 75 of VH (Table II).
From these only position 28 of VH yielded potentially immunogenic residues as detailed above.
Humanized clones A-F include 4 of the 5 mutated clones (A,
B, E, and F) as well as 2 additionally selected clones (C and D)
that demonstrated weaker reactivity in ELISA and were included for further characterization for comparison (Fig. 2). All
13673
TABLE III
Binding parameters of rabbit and humanized Fab directed
to human A33 antigen
Association (kon) and dissociation (koff) rate constants were determined using surface plasmon resonance. Human A33 antigen was immobilized on the sensor chip. The dissociation constant (Kd) was calculated from koff/kon.
Clone
kon/104
⫺1
M
Rabbit
1
2
Humanized
A
B
C
D
E
F
s⫺1
koff/10⫺4
Kd
s⫺1
nM
30.7 ⫾ 0.9
17.4 ⫾ 1.0
1.2 ⫾ 0.1
2.8 ⫾ 0.2
10.5 ⫾ 0.7
35.2 ⫾ 1.5
10.9 ⫾ 0.3
6.5 ⫾ 0.3
6.5 ⫾ 0.2
19.2 ⫾ 1.2
5.9 ⫾ 0.1
6.1 ⫾ 0.1
19.7 ⫾ 0.3
19.0 ⫾ 0.6
5.0 ⫾ 0.1
6.8 ⫾ 0.2
0.39
1.6
5.6
1.7
18.1
29.2
7.7
3.5
6 clones were produced as soluble Fab by E. coli expression and
purified by protein G affinity chromatography (Fig. 3). The
recovered yields of Fab ranged from 0.5 to 2 mg per 1-liter
shake flask culture. The affinity of each of the humanized Fab
was measured by surface plasmon resonance using a Biacore
instrument (Table III). Humanized clone B (Kd ⫽ 1.7 nM) demonstrated the highest affinity, matching that of the parental
rabbit clone 2 (Kd ⫽ 1.6 nM). The other mutated humanized
clones A, E, and F had slightly lower affinities, while the
unmutated humanized clones C and D gave the lowest affinities as anticipated based on their ELISA reactivity. The mutations were found to affect both association and dissociation rate
constants. The dissociation rate constants of all 4 mutated
humanized clones A, B, E, and F were found to be very similar,
that is about 2-fold higher compared with parental rabbit clone
2 and about 3-fold lower compared with unmutated humanized
clones C and D. Interestingly, humanized clones B and F revealed the highest association rate constants indicating that
the substitution of the surface accessible and negatively
charged residue aspartic acid by glycine at position 28 of VH
improves the association rate.
Analysis by flow cytometry revealed that the 6 selected humanized Fab bound specifically to human colon cancer cells
expressing native human A33 antigen (Fig. 4). This result was
confirmed with a rosetting assay in which surface bound humanized Fab were detected by rabbit anti-human F(ab⬘)2 polyclonal antibodies followed by adherence of protein A-coated
human red blood cells to human colon cancer cells expressing
native human A33 antigen (data not shown). Cell lines that do
not express human A33 antigen were negative in both assays
(Fig. 4; data not shown). The 4 mutated clones A, B, E, and F
demonstrated the strongest reactivity in flow cytometry and
rosetting assays. The differences between the humanized
clones were found to correlate with their differences in affinity
to immobilized recombinant human A33 antigen. The humanized Fab were further analyzed for reactivity with human A33
antigen extracted from colon cancer cell lines by Western blotting. As shown for humanized clone B (Fig. 6), the humanized
Fab strongly reacted with a band of about 43 kDa under nonreducing conditions. No reactivity was observed using reducing
conditions (data not shown), suggesting the recognition of a
conformational epitope on human A33 antigen (30). Taken
together, these results demonstrate that the selected humanized antibodies bind to a native epitope on human A33 antigen
that is fully accessible on the cell surface.
In order to evaluate the selectivity of the humanized Fab in
an independent system of higher complexity, their reactivity
with tumor tissue sections was analyzed by immunohistochem-
13674
Generation and Humanization of Rabbit Antibodies
istry. As shown for humanized clone B, the humanized Fab
reacted strongly with xenografts of the human colon cancer cell
line SW1222 grown in nude mice (Fig. 7, A and B). The humanized Fab revealed immunoreactivity similar to the mouse
monoclonal antibody A33 with an intense staining of dysplastic
glandular structures in tissue sections of human colon adenocarcinoma after blocking endogenous human immunoglobulins
(Fig. 7, C and D). A comparison of corresponding tissue sections
stained with the full blocking step and stained without blocking of the endogenous human immunoglobulins illustrates the
FIG. 6. Western blot reactivity of humanized Fab B with Triton
X-100 extracts of human A33 antigen-expressing (LIM1215,
SW1222) and -nonexpressing (SW620) human colon cancer cell
lines. Specific binding was detected by alkaline phosphatase-conjugated goat anti-human F(ab⬘)2 polyclonal antibodies and visualized
using chemiluminescence. Numbers on the right indicate molecular
masses of standard proteins in kDa.
FIG. 7. Immunohistochemical reactivity of humanized Fab B in human
colon cancer tissue sections. A and B,
xenograft of human colon cancer cell line
SW1222 in nude mice; C-F, serial sections
of moderately differentiated human colon
adenocarcinoma. Scale bar ⫽ 300 ␮m.
Specific binding was detected by biotinylated goat anti-human F(ab⬘)2 polyclonal
antibodies and visualized using an avidin-biotin complex system and diaminobenzidine tetrahydrochloride as a chromogen. A, humanized Fab B showing
intense staining in SW1222 xenograft. B,
buffer only without application of humanized Fab (negative control) showing no
staining in SW 1222 xenograft. C, mouse
monoclonal antibody A33 showing intense
staining of dysplastic glandular structures in human colon adenocarcinoma. D,
humanized Fab B revealing similar staining in corresponding carcinoma areas after blocking of endogenous human immunoglobulins. No staining of additional
tissue components due to endogenous human immunoglobulins is detectable. E,
buffer only without application of humanized Fab (negative control) but with blocking of endogenous human immunoglobulins showing no staining. F, buffer only
without application of humanized Fab
(negative control) and omitting the blocking of endogenous human immunoglobulins showing intense staining of endogenous human immunoglobulins.
amount of internal reactivity (Fig. 7F) and its complete blocking (Fig. 7E).
DISCUSSION
Antibody engineering technology has facilitated the generation of antibodies to virtually any antigen of interest as well as
their improvement in terms of affinity, specificity, and immunogenicity (2, 31). The generation of human or humanized
antibodies to human antigens is of particular interest for the
treatment of a variety of diseases. Here we report a new route
to therapeutically relevant humanized antibodies. Using phage
display, we selected and humanized antibodies from rabbits
immunized with a human antigen.
The rabbit Ig gene repertoire is well characterized (32). This
has allowed the generation of combinatorial rabbit antibody
libraries displayed on phage whose screening resulted in the
selection of rabbit monoclonal antibodies (14 –16). While the
generation of rabbit monoclonal antibodies by hybridoma technology has also been reported (33), the phage display approach
with its inherent linkage of phenotype and genotype provides
ready access to antibody sequences and facilitates further in vitro
optimizations such as humanization or affinity maturation.
Compared with the other existing sources of human or humanized antibodies, immune rabbits are an attractive alternative for several reasons. First, humanized antibodies from immune rabbits extend the accessible epitope repertoire of a given
antigen. Epitopes that are not immunogenic in mice, a species
from which the vast majority of monoclonal antibodies to human antigens has been generated, might be immunogenic in
rabbits. This is of particular interest for the development of
therapeutic human antibodies that are evaluated in mouse
models and are required to recognize both the human antigen
Generation and Humanization of Rabbit Antibodies
and its mouse homologue. In contrast, humanized antibodies
that are derived from immune mice, either indirectly through
humanization (e.g. Ref. 19) or directly through transgenic mice
containing human Ig loci (34), are negatively selected against
epitopes displayed by the mouse homologue. Second, in contrast to human antibodies derived from large naive combinatorial antibody libraries that are selected in vitro (35), humanized antibodies derived from immune animals have been
subjected to in vivo selection and, thus, are more likely to
recognize a given antigen selectively. Lastly, as we demonstrate here for the first time, rabbit antibodies can be converted
to humanized antibodies that retain both high specificity and
affinity to the antigen.
Using Fab display we selected rabbit antibodies with dissociation constants as low as 390 pM. In contrast, rabbit antibodies selected by scFv display revealed 1 to 2 orders of magnitude
weaker affinity (14, 16). While different antigens as well as
different immunization schedules are likely to contribute to
these differences, Fab display in general may result in the
selection of antibodies with higher affinity. This is based on the
fact that Fab are displayed monovalently at the phage surface
and are selected based on affinity and expression, whereas the
selection of scFv is also influenced by avidity due to their
tendency to dimerize or form higher order aggregates (2, 3). As
a result, antibodies selected as scFv do not generally bind their
antigen well when they are converted in the Fab format. In
contrast, we found that the opposite conversion from Fab to
scFv format yielded scFv with high apparent affinity, for instance, in the case of humanized clone B which is described
here (data not shown). Fab are also readily convertible to whole
antibody formats by simple fusion with Fc coding sequences
which does not affect the antigen-binding site. In summary,
selections based on the Fab format generate antibodies that are
well suited for therapeutical evaluation in multiple formats.
We chose human A33 antigen as target for the development
of humanized antibodies from immune rabbits. Human A33
antigen has been established as a target for radioimmunotherapy of colon cancer in phase I and II clinical trials with 131Iand 125I-labeled mouse monoclonal antibody A33 (5–7). Antitumor effects were demonstrated in these studies despite the
fact that only a single dose could be administered due to development of a human anti-mouse antibody response. A33 was
retained for prolonged periods (up to 6 weeks) in tumors but
was cleared rapidly from normal colon (5– 6 days). In fact, bone
marrow suppression was the dose-limiting toxicity in the radioimmunotherapy trials and no normal colon toxicity was
observed. To make repeated administrations feasible, A33 was
humanized by CDR grafting (36) and evaluated in phase I and
II clinical trials. While these studies confirmed antitumor effects, a number of patients developed a human anti-human
antibody response after several administrations, forcing the
termination of the treatment.2 Analysis of sera from the patients revealed evidence for an anti-idiotypic immune response,
i.e. the sera contained antibodies that recognized the antigenbinding site of humanized antibody A33.3 Administration of an
equally potent humanized antibody that is unreactive to the
anti-idiotypic immune response generated by humanized antibody A33 would allow therapy to continue.
The humanized antibodies described here compare favorably
to the previously described humanized antibody A33. They
were found to recognize a cell surface epitope of the A33 antigen with a specificity as high as mouse and humanized antibody A33. The monovalent affinity of humanized clone B is
2
3
S. Welt, G. Ritter, and L. J. Old, manuscript in preparation.
G. Ritter, S. Welt, and L. J. Old, manuscript in preparation.
13675
5–10-fold higher than the monovalent affinity reported for
mouse and humanized antibody A33 (36, 37). Using surface
plasmon resonance (data not shown), we further found that the
humanized antibodies described here (i) bind to an epitope on
the A33 antigen distinct from the epitope recognized by mouse
and humanized antibody A33, and (ii) are not recognized by
sera from patients who had developed an anti-idiotypic immune response after repeated treatment with humanized antibody A33. Thus, regarding therapeutic application, our humanized antibodies derived from the rabbit antibody repertoire
complement humanized antibody A33 derived from the mouse
antibody repertoire. Based on these findings, immune rabbits
may be an important source for the generation of therapeutic
human antibodies.
Acknowledgments—We thank John A. Neves and Clarence Williams,
Jr. for excellent technical assistance. Special thanks go to Dr. Peter
Steinberger for his expertise and support.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
Barbas, C. F., III (1995) Nat. Med. 1, 837– 839
Hudson, P. J. (1999) Curr. Opin. Immunol. 11, 548 –557
Rader, C., and Barbas, C. F., III (1997) Curr. Opin. Biotechnol. 8, 503–508
Scott, A. M., and Welt, S. (1997) Curr. Opin. Immunol. 9, 717–722
Welt, S., Divgi, C. R., Real, F. X., Yeh, S. D., Garin-Chesa, P., Finstad, C. L.,
Sakamoto, J., Cohen, A., Sigurdson, E. R., Kemeny, N., Carswell, E. A.,
Oettgen, H. F., and Old, L. J. (1990) J. Clin. Oncol. 8, 1894 –1906
Welt, S., Divgi, C. R., Kemeny, N., Finn, R. D., Scott, A. M., Graham, M.,
Germain, J. S., Richards, E. C., Larson, S. M., Oettgen, H. F., and Old, L. J.
(1994) J. Clin. Oncol. 12, 1561–1571
Welt, S., Scott, A. M., Divgi, C. R., Kemeny, N. E., Finn, R. D., Daghighian, F.,
Germain, J. S., Richards, E. C., Larson, S. M., and Old, L. J. (1996) J. Clin.
Oncol. 14, 1787–1797
Heath, J. K., White, S. J., Johnstone, C. N., Catimel, B., Simpson, R. J., Moritz,
R. L., Tu, G. F., Ji, H., Whitehead, R. H., Groenen, L. C., Scott, A. M., Ritter,
G., Cohen, L., Welt, S., Old, L. J., Nice, E. C., and Burgess, A. W. (1997)
Proc. Natl. Acad. Sci. U. S. A. 94, 469 – 474
Ritter, G., Moritz, R. L., Ji, H., Cohen, L. S., Nice, E. C., Catimel, B., Heath,
J. K., White, S. J., Welt, S., Burgess, A. W., Old, L. J., and Simpson, R. J.
(1997) Biochem. Biophys. Res. Commun. 236, 682– 686
Welt, S., and Ritter, G. (1999) Semin. Oncol. 26, 683– 690
Moritz, R. L., Ritter, G., Catimel, B., Cohen, L. S., Welt, S., Old, L. J., Burgess,
A. W., Nice, E. C., and Simpson, R. J. (1998) J. Chromatogr. Sect. A 798,
91–101
Barbas, C. F., III, Kang, A. S., Lerner, R. A., and Benkovic, S. J. (1991) Proc.
Natl. Acad. Sci. U. S. A. 88, 7978 –7982
Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S., and Foeller, C. (1991)
Sequences of Proteins of Immunological Interest, 5th Ed., Public Health
Service, National Institutes of Health, Bethesda, MD
Ridder, R., Schmitz, R., Legay, F., and Gram, H. (1995) Bio/Technology 13,
255–260
Lang, I. M., Barbas, C. F., III, and Schleef, R. R. (1996) Gene (Amst.) 172,
295–298
Foti, M., Granucci, F., Ricciardi-Castagnoli, P., Spreafico, A., Ackermann, M.,
and Suter, M. (1998) J. Immunol. Methods 213, 201–212
Carter, P., Kelley, R. F., Rodrigues, M. L., Snedecor, B., Covarrubias, M.,
Velligan, M. D., Wong, W. L. T., Rowland, A. M., Kotts, C. E., Carver, M. E.,
Yang, M., Bourell, J. H., Shepard, H. M., and Henner, D. (1992) Bio/
Technology 10, 163–167
Ulrich, H. D., Patten, P. A., Yang, P. L., Romesberg, F. E., and Schultz, P. G.
(1995) Proc. Natl. Acad. Sci. U. S. A. 92, 11907–11911
Rader, C., Cheresh, D. A., and Barbas, C. F., III (1998) Proc. Natl. Acad. Sci.
U. S. A. 95, 8910 – 8915
Rosok, M. J., Yelton, D. E., Harris, L. J., Bajorath, J., Hellstrom, K.-E.,
Hellstrom, I., Cruz, G. A., Kristensson, K., Lin, H., Huse, W. D., and Glaser,
S. M. (1996) J. Biol. Chem. 271, 22611–22618
Baca, M., Presta, L. G., O’Connor, S. J., and Wells, J. A. (1997) J. Biol. Chem.
272, 10678 –10684
de Wildt, R. M., Hoet, R. M. A., van Venrooij, W. J., Tomlinson, I. M., and
Winter, G. (1999) J. Mol. Biol. 285, 895–901
Presta, L. G., Chen, H., O’Connor, S. J., Chisholm, V., Meng, Y. G., Krummen,
L., Winkler, M., and Ferrara, N. (1997) Cancer Res. 57, 4593– 4599
Brezinschek, H. P., Foster, S. J., Dorner, T., Brezinschek, R. I., and Lipsky,
P. E. (1998) J. Immunol. 160, 4762– 4767
Foote, J., and Winter, G. (1992) J. Mol. Biol. 224, 487– 499
Roguska, M. A., Pedersen, J. T., Keddy, C. A., Henry, A. H., Searle, S. J.,
Lambert, J. M., Goldmacher, V. S., Blattler, W. A., Rees, A. R., and Guild,
B. C. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 969 –973
Roguska, M. A., Pedersen, J. T., Henry, A. H., Searle, S. M. J., Roja, C. M.,
Avery, B., Hoffee, M., Cook, S., Lambert, J. M., Blättler, W. A., Rees, A. R.,
and Guild, B. C. (1996) Protein Eng. 9, 895–904
Padlan, E. A. (1994) Mol. Immunol. 31, 169 –217
Pedersen, J. T., Henry, A. H., Searle, S. J., Guild, B. C., Roguska, M., and Rees,
A. R. (1994) J. Mol. Biol. 235, 959 –973
Catimel, B., Ritter, G., Welt, S., Old, L. J., Cohen, L., Nerrie, M. A., White,
S. J., Heath, J. K., Demediuk, B., Domagala, T., Lee, F. T., Scott, A. M., Tu,
G. F., Ji, H., Moritz, R. L., Simpson, R. J., Burgess, A. W., and Nice, E. C.
13676
Generation and Humanization of Rabbit Antibodies
(1996) J. Biol. Chem. 271, 25664 –25670
31. Carter, P., and Merchant, A. M. (1997) Curr. Opin. Biotechnol. 8, 449 – 454
32. Knight, K. L., and Crane, M. A. (1994) Adv. Immunol. 56, 179 –218
33. Spieker-Polet, H., Sethupathi, P., Yam, P. C., and Knight, K. L. (1995) Proc.
Natl. Acad. Sci. U. S. A. 92, 9348 –9352
34. Mendez, M. J., Green, L. L., Corvalan, J. R., Jia, X. C., Maynard-Currie,
C. E.,Yang, X. D., Gallo, M. L., Louie, D. M., Lee, D. V., Erickson, K. L.,
Luna, J., Roy, C. M., Abderrahim, H., Kirschenbaum, F., Noguchi, M.,
Smith, D. H., Fukushima, A., Hales, J. F., Klapholz, S., Finer, M. H., Davis,
C. G., Zsebo, K. M., and Jakobovits, A. (1997) Nat. Genet. 15, 146 –156
35. Vaughan, T. J., Williams, A. J., Pritchard, K., Osbourn, J. K., Pope, A. R.,
Earnshaw, J. C., McCafferty, J., Hodits, R. A., Wilton, J., and Johnson, K. S.
(1996) Nat. Biotechnol. 14, 309 –314
36. King, D. J., Antoniw, P., Owens, R. J., Adair, J. R., Haines, A. M. R.,
Farnsworth, A. P. H., Finney, H., Lawson, A. D. G., Lyons, A., Baker, T. S.,
Baldock, D., Mackintosh, J., Gofton, C., Yarranton, G. T., McWilliams, W.,
Shochat, D., Leichner, P. K., Welt, S., Old, L. J., and Mountain, A. (1995)
Br. J. Cancer 72, 1364 –1372
37. Catimel, B., Nerrie, M., Lee, F. T., Scott, A. M., Ritter, G., Welt, S., Old, L. J.,
Burgess, A. W., and Nice, E. C. (1997) J. Chromatogr. Sect. A 776, 15–30