WO2002000005A1

WO2002000005A1 - Libraries displaying human antibody fragments with hybrid complementarity determining regions

Info

Publication number: WO2002000005A1
Application number: PCT/US2001/019502
Authority: WO
Inventors: Russell Rother; William L. Fodor; Stephen P. Squinto
Original assignee: Alexion Pharmaceuticals, Inc.
Priority date: 2000-06-21
Filing date: 2001-06-18
Publication date: 2002-01-03
Also published as: WO2002000005A8; AU2001268545A1

Abstract

Libraries displaying human antibodies (i.e., whole antibody-like molecules and antigen binding fragments thereof) having nonhuman complementary determining regions are provided. Methods of obtaining human antibodies that recognize self antigens and using such libraries are described.

Description

LIBRARIES DISPLAYING HUMAN ANTIBODY FRAGMENTS WITH HYBRID COMPLEMENTARITY DETERMINING REGIONS

Technical Field

Libraries displaying human antibodies (i.e., whole antibody-like molecules and antigen binding fragments thereof) having nonhuman complementarity determining regions are provided. Methods of obtaining human antibodies that recognize human antigens and using such libraries are described.

Background

It is well known that antibodies are proteins that bind specifically to an antigen. Although each antibody has a unique structure that allows it to bind to a specific antigen, all antibodies have the same overall structure. Whole antibodies are composed of two heavy chains (each about 50 kilodaltons) linked to each other by disulfide bonds and two light chains (each about 25 kilodaltons) each of which is linked to a heavy chain by disulfide bonds. The arnmoterrninal sequences of both the light and heavy chains vary greatly between antibodies and are termed the variable (hereinafter V) domains. The carboxyterrninal sequences of both the light and heavy chains remain constant for antibodies of the same isotype and are termed the constant (hereinafter C) domains. It is further known that the C domains largely are responsible for, inter alia, the effector function of the antibody, while the V domains are responsible for antigen binding. Kabat et al, by comparing sequences of known antibodies, found that within the V domains, regions of greater and lesser variability exist (Kabat et al.,1983, "Sequences of Proteins of Immunological Interest", U.S. Department of Health and Human Services). Kabat defined different elements of these sequences according to the relative sequence identity displayed at equivalent positions.

Specifically, Kabat defined regions of sequence hypervariability and termed them complementarity

determining regions (hereinafter CDR) of the antibody. The remainder of the V regions, found to

be more conserved than the CDR domains, were termed framework regions (hereinafter FR).

It is known in the art that nonhuman (e.g., mouse) monoclonal antibodies prepared using

hybridoma methods, such as those described in Kohler et al, 1975, Nature. 256:495, are

immunogenic in humans and therefore have limited application in human therapy. Early attempts

to overcome this problem provided chimeric antibodies, i.e., antibodies having a mouse, or other

non-human, V domains fused to human C domains. See for example PCT/GB85/00392.

Although these chimeric antibodies were an improvement, problems with immunogenicity

remained.

Winter and coworkers addressed this problem by disclosing the replacement of the CDRs

of a human antibody with specific murine CDRs from an antigen specific antibody, often referred

to in the art as "humanizing antibodies" or "CDR grafting." See, for example, Jones et al, 1986,

Nature. 321:522-525; Riechmann et al, 1988, Nature. 332:323-327; and Verhoeyen et al, 1988,

Science. 239:1534-1536; EP0239400; the contents of all of which are incorporated herein by

reference.

More recently a technique for producing "fully human" antibody-like molecules has been

developed. V domains of antibodies typically generated from the rearranged human repertoire of

variable heavy (hereinafter VH) and/or variable light (hereinafter VL) chain sequences are fused to genes encoding the coat protein of a bacteriophage. Bacteria are then transformed with these

constructs (generally in the presence of helper bacteriophage), which results in the production of

bacteriophage having the antibody-like molecules expressed, or displayed, on their surface as well as the genetic information that codes for these antibody-like molecules. A collection of these

bacteriophage is called a phage display library. The library is then exposed to a desired antigen,

and phage that do not bind to the antigen are removed. See for example U.S. Patent Nos.,

5,427,908; 5,432,018; 5,580,717; 5,723,286; 5,837,500; 5,223,409; 5,403,484; the contents of all

of which are incorporated herein by reference.

More recently fully human synthetic libraries of bacteriophage displaying antibody-like

molecules and/or antigen binding fragments thereof have been described in the art. Such

synthetically generated libraries are modular in nature in that unique cleavage sites (e.g.,

restriction endonuclease sites) flank each CDR. Thus, all six CDRs, either alone or in

combination, can be diversified by substitution. See for example, Knappik et al, 2000, J. Mol.

Biol. 296:57-86 and WO97/08320, the contents of which are incorporated herein by reference.

However, since the above mentioned libraries are generally derived from antibody

sequences that already have undergone rearrangement, and considering that most CDR sequence

combinations capable of recognizing human antigens are eliminated during this process, libraries

made in this manner contain a restricted source of antibodies directed toward human antigens.

Since many therapeutically relevant antibodies must recognize human antigens, there exists a need

in the art to provide libraries displaying a higher diversity of antibodies (and/or antigen binding

fragments thereof) capable of recognizing the human antigens.

U.S. Patent 5,885,793 to Griffiths states that it is directed to methods for the production

of antiself antibodies using phage display libraries. It is stated that the nucleic acid sequences of

such libraries may be derived "from . . . rearranged V genes of an unimmunized mammal . . ." or

"synthetic recombination of V gene segments, which may be germline V gene sequences." See column 6, line 15 et seq. Artificially recombining germline V gene segments theoretically increases the chance of generating antiself antibodies, since the resultant antibodies are not subject

to clonal deletion events that would otherwise occur during natural rearrangements. See Example 5, beginning at column 30 and column 31 line 45 to column 32, line 51. However, the probability of finding antiself antibodies in a library created in this manner is still limited. Since human V

gene segments have evolved in the presence of human antigens, it is likely that many gene family members capable of recognizing self have been eliminated. Therefore, use of purely human gene segments in the generation of the synthetic human library may still contain some biasing against antihuman antibodies. Furthermore, considering that synthetic gene rearrangement can not truly duplicate the diversity generated during natural gene rearrangement, this strategy may have further theoretical limitations. This patent does not teach the generation of hybrid libraries comprised of sequences from more than one species.

U.S. Patent 5,565,332 to Hoogenboom is directed to methods of making human antibody polypeptide dimers specific for an antigen of interest from murine or other nonhuman species through chain shuffling and CDR imprinting. Chain shuffling involves the use of VL or VH sequences from a non-human antibody that recognize a specific human antigen to find a human VH or VL sequence, respectively, that when paired with a nonhuman V region recognizes the original antigen. This specific human VH or VL sequence is then used to find a corresponding human VL or VH sequence, respectively, that recognizes the human antigen, resulting in a totally human antibody. It is suggested that this approach could be extended to include not only nonhuman VL or VH sequences to a particular antigen, but also component parts thereof, e.g., CDRs. In this case the component parts of an antigen binding site of a nonhuman antibody known to bind a particular antigen are combined with the repertoire of component parts of an antigen binding site of human antibodies, in which case the procedure would be "CDR imprinted selection." Indeed, it is stated at column 12, line 22 of the patent:

"[I]n examples 2 and 3, the CDR3 of the original mouse heavy chain was

retained, and combined with a repertoire of human heavy chain variable domains by PCR amplification of a repertoire of human heavy chain variable domains with a primer incorporating the mouse CDR3. The retention of mouse VH-CDR3 may be particularly advantageous in that CDR3 of the heavy chain is often most important for antigen binding.

This principle could be extended to a mouse CDR3 repertoire by

amplifying the rearranged mouse VH genes with human 5' primers (forward

primers) and human VH framework 3' primers (backward primers). These primers

would have to be designed with homology to both mouse and human V-genes. The amplified DNA repertoire of mouse CDR3s could then be assembled by PCR with a repertoire of human heavy chain genes. " It will be appreciated that this chain shuffling/CDR imprinting technique necessarily is restricted to the use of one nonhuman V region or component(s) of one nonhuman V region, since by definition, the nonhuman V region or component(s) thereof must be imprinted on a corresponding human V region or component(s) thereof.

Further, this method of deriving murine CDRs, i.e., using the polymerase chain reaction (PCR) with primers homologous to both the nonhuman and human sequences, limits the repertoire obtainable from the nonhuman source and therefore, does not maximize the generation of libraries displaying a high diversity of antibodies (and/or antigen binding fragments thereof) capable of recognizing human antigens. Accordingly, a need exists for improved methods of generating libraries displaying a high diversity of antibodies (and/or antigen binding fragments

thereof) capable of recognizing human antigens while ma taining nonimmunogenic

characteristics.

Summary

Libraries displaying a high diversity of antibodies (and or antigen binding fragments

thereof) capable of recognizing human antigens, but which maintain an immunoprivileged status,

are advantageously provided. Such diverse display libraries allow for the identification and

isolation of nucleotide sequences which encode specific antigen binding pairs (ABP) to human

antigens. The ABP preferably includes both human and nonhuman sequences in each of the VH

and VL regions.

In an exemplary embodiment, the libraries utilize a genetic display unit (GDU) for

combining (a) nucleic acid sequence(s) which encode a genetically diverse repertoire, or representation thereof (e.g., consensus sequences), of human antibody VH regions, including the

CDR and FR sequences contained therein, and (b) nucleic acid sequence(s) which encode a

genentically diverse repertoire, or representation thereof, of human antibody VL regions,

including the CDR and FR sequences contained therein; wherein at least one of the human CDR

sequences from the human heavy chain and at least one of the human CDR sequences from the

human light chain is replaced by a repertoire of corresponding nonhuman CDR sequences, or

representation thereof, to form a library of nucleic acid sequences encoding ABPs. The ABPs

each consist of a human heavy chain polypeptide, including both human and nonhuman sequences,

and a human light chain polypeptide, including both human and nonhuman sequences, which in combination form an antigen binding site of an ABP specific for an antigen of interest, which is preferably a human antigen.

By way of example but not limitation, the display libraries are selected from the group consisting of a ribosomal display library, yeast display library, bacterial display library, and viral display library. See for example, U.S. Patent Nos. 5,643,756; 5,723,287; and 5,952,474, the contents of which are incorporated herein by reference. Presently, preferred libraries include bacteriophage display libraries such as filamentous bacteriophage display libraries. Suitable bacteriophage display libraries include class I and class II bacteriophage such as fd, Ml 3, Ifl, Ike, ZJ/Z, Ff, XF, Pfl, PB, etc. By way of example but not limitation, libraries may display whole antibodies and/or antigen binding fragments thereof. Suitable antigen binding fragments include Fv, disulfide-linked Fv, scFv, Fab, F(ab')₂ and other fragments capable of binding antigen.

The repertoires of the human heavy and light chain FR and/or CDR sequences may be naturally or synthetically generated from sequences that have undergone gene rearrangement. Alternatively, the repertoires of human heavy and light chain FR and/or CDR sequences can be generated from nonrearranged germline sequences which may be artificially rearranged. The repertoires of nonhuman CDR sequences may also be naturally or synthetically generated from sequences that have undergone gene rearrangement. Alternatively, the repertoires of nonhuman CDR sequences can be generated from nonrearranged germline sequences which may be artificially rearranged. The repertoires of nonhuman CDR sequences are generated from a nonhuman species, preferable a mammalian species (e.g., a rodent or nonhuman primate). The repertoires of nonhuman CDR sequences may be generated from an unimmunized naϊve animal or an animal previously immunized with a preselected antigen (e.g., a human antigen). Mixtures of

the repertoires described herein are also contemplated.

The repertoires of human heavy and light chain FR and/or CDR sequences represented in

the display library can be generated to be compatible with the insertion of repertoires of

nonhuman CDR sequences. Similarly, the repertoires nonhuman CDR sequences can be

generated to be compatible with their insertion into the repertoires of human heavy and light chain

FR sequences of the display libraries. For example display libraries are provided wherein the

human heavy and light chain FR sequences include unique cleavage sites, which preferably flank

each of the CDR sequences. The corresponding nonhuman CDR sequences include terminal

cleavage sites that are compatible with the aforementioned cleavage sites.

In another aspect, there are provided methods of producing an ABP, which include

providing a library displaying a repertoire of human heavy and light chain FRs and human CDRs

contained therein, wherein one or more of the human CDRs is replaced by a repertoire of corresponding nonhuman CDRs, and selecting an ABP by exposing the library to a preselected

human antigen and recovering therefrom an ABP that binds to the preselected human antigen. It

will be appreciated that the display libraries can also be screened for other desired properties such

as affinity, expression yields, stability and solubility. It shall also be appreciated that the

preselected human antigen includes purified antigens, multimeric antigens, crude antigens and

antigens associated with other proteins (e.g., on the surface of a cell). The display libraries may

also be screened to remove undesired ABPs as in library subtraction.

In another aspect, there is provided a method wherein the selected ABP is modified to

produce a derivative thereof (e.g., addition of constant region sequences to VH and VL sequences

to generate whole antibodies). Detailed Description

Technical and scientific terms used herein have the meanings commonly understood by one of ordinary skill in the art to which the present teachings pertain, unless otherwise defined herein (e.g., see Terminology below). Reference is made herein to various methodologies known to those of skill in the art. Publications and other materials setting forth such known

methodologies to which reference is made are incorporated herein by reference in their entireties as though set forth in full. Practice of the methods described herein will employ, unless otherwise indicated, conventional techniques of chemistry, molecular biology, microbiology, recombinant DNA, and immunology, which are within the skill of the art. Such conventional techniques are explained fully in the literature. See, e.g., Sambrook, Fritsch, and Maniatis, Molecular Cloning; Laboratory Manual 2nd ed. (1989); DNA Cloning, Volumes I and JJ (D.N Glover ed. 1985); Oligonucleotide Synthesis (M.J. Gait ed, 1984); Nucleic Acid Hybridization (B.D. Hames & S.J. Higgins eds. 1984); the series, Methods in Enzymology (Academic Press, Inc.), particularly Vol. 154 and Vol. 155 (Wu and Grossman, eds.); PCR-K Practical Approach (McPherson, Quirke, and Taylor, eds., 1991); Immunology, 2d Edition, 1989, Roitt et al, C.V. Mosby Company, and New York; Advanced Immunology, 2d Edition, 1991, Male et al, Grower Medical Pubhshing, New York.; DNA Cloning: A Practical Approach, Volumes I and 13, 1985 (D.N. Glover ed.); Oligonucleotide Synthesis, 1984, M.L. Gait ed); Transcription and Translation, 1984 (Hames and Higgins eds.); Animal Cell Culture, 1986 (R.I. Freshney ed.); Immobilized Cells and Enzymes, 1986 (IRL Press); Perbal, 1984, A Practical Guide to Molecular Cloning; and Gene Transfer Vectors for Mammalian Cells, 1987 (J. H. Miller and M. P. Calos eds., Cold Spring Harbor Laboratory); WO97/08320; US. Patent Nos. 5,427,908; 5,885,793; 5,969,108; 5,565,332; 5,837,500; 5,223,409; 5,403,484; 5,643,756; 5,723,287; 5,952,474; Knappik et al, 2000, J. Mol.

Biol. 296:57-86; Barbas et al, 1991, Proc. Natl. Acad. Sci. USA 88:7978-7982; Schaffitzel et al

1999, J. Immunol. Meth. 10:119-135; Kitamura, 1998, Int. J. Hematol.. 67:351-359; Georgiou et

al, 1997, Nat. Biotechnol. 15:29-34; Little, et al, 1995, J. Biotech. 41:187-195; Chauthaiwale et

al, 1992, Microbiol. Rev.. 56:577-591; Aruffb, 1991, Curr. Opin. Biotechnol. 2:735-741;

McCafferty (Editor) et al, 1996, Antibody Engineering: A Practical Approach.

Any suitable materials and/or methods known to those of skill can be utilized in carrying

out the methods described herein; however, preferred materials and/or methods are described.

Materials, reagents and the like to which reference is made in the following description and

examples are obtainable from commercial sources, unless otherwise noted.

The repertoires of the human heavy and light chain FR and/or CDR sequences represented

in the display libraries disclosed herein may be naturally or synthetically generated from sequences that have undergone gene rearrangement. Alternatively, the repertoires of human heavy and light

chain FR and/or CDR sequences can be generated from nonrearranged geπnline sequences which

may be artificially rearranged. The repertoires of human heavy and light chain FR and/or CDR

sequences represented in the display library will be generated to be compatible with the insertion

of repertoires of nonhuman CDR sequences. Similarly, the repertoires of nonhuman CDR sequences represented in the display libraries disclosed herein may be naturally or synthetically

generated from sequences that have undergone gene rearrangement. Alternatively, the repertoires

of nonhuman CDR sequences can be generated from nonrearranged germline sequences which

may be artificially rearranged. The repertoires of nonhuman CDR sequences will be generated to

be compatible with their insertion into the repertoires of human heavy and light chain FR

sequences of the display libraries. The repertoires of nonhuman CDR sequences are generated from a nonhuman species, preferable a mammalian species (e.g., a rodent or nonhuman primate).

The repertoires of nonhuman CDR sequences may be generated from an uriirnmunized naϊve

animal or an animal previously immunized with a preselected antigen (e.g., a human antigen).

Mixtures of the repertoires described herein are also contemplated.

In one embodiment, the repertoires of human heavy and light chain FR and/or

CDR sequences represented in the display libraries disclosed herein are generated from in vivo

rearranged variable genes (naturally generated). This genetically diverse repertoire is typically

generated by PCR amplification of naturally rearranged V gene repertoirs using primers sets

specific for the V region of each chain (see for example Barbas et al., 1991, Proc. Natl. Acad.

Sci., 88:7978; Marks et al., 1991, J. Mol. Biol., 222:581, the contents of each of which are

incorporated herein by reference). PCR primers generally consist of a 5' set that is upstream of

and may include FR1 sequences and a 3' set that is downstream of and may include FR4

sequences, which amplifies the repertoire of variable regions for both the heavy and light chains

including FRs 1-4 and CDRs 1-3. Alternatively, the repertoire of human VH and VL chain sequences could be generated

where the 3' primer set is specific for FR3 sequences resulting in a repertoire of human VH and

VL chain sequences that include FRs 1-3 and CDRs 1 and 2. A CDR3 staffer sequence could be

included in the 3' primer set, as a component of the library vector (e.g., phagemid) or it could be

omitted all together. Human or nonhuman FR4 sequence(s) could also be included in the 3'

primer set or as a component of library vector and could consist of a unique or repertoire of

sequences. Most importantly, a unique cleavage site may be included that either flanks each side

of the staffer CDR3 sequence, or in the absence of CDR3, a unique cleavage site will be included

between FR3 and FR4 sequences. By including unique cleavage sites between FR3 and FR4 sequences, a repertoire of nonhuman CDR3 sequences with compatible cleavage sites could be

added to the human display library. This strategy could also be employed to generate a library

from nonrearranged germline sequences considering that FRs 1-3 and CDRs 1 and 2 are part of

the VH or VL chain sequences and not a product of gene rearrangement.

Similarly, a human display library representing a repertoire of human VH and VL chain

sequences that include FRs 2-4 and CDRs 2 and 3 may be generated. Human or nonhuman FRl

sequence(s) and a CDR1 staffer sequence is included in the library vector. Alternatively the 5'

primer set or the CDR1 staffer sequence optimally may be omitted. The unique cleavage site

would either flank each side of the staffer CDR1 sequence, or in its absence, the unique cleavage

site would be included between FRl and FR2 sequences. The inclusion of unique cleavage sites

between FRl and FR2 sequences, allows for the addition to a human display library of a

repertoire of nonhuman CDR1 sequences with compatible cleavage sites. Components of a library generated using this strategy may also be derived from nonrearranged germline sequences.

In another embodiment, a human display library representing a repertoire of human VH

and VL chain sequences that include FRs 1-4 and CDRs 1 and 3 may be generated. In this case,

the generation of the human display library occurs in two steps. Initially, a repertoire of human

VH or VL chain FRl, CDR1 and FR2 sequences (library fragment 1) is generated by PCR, where

the 5' primer set is upstream and may include FRl sequences and the 3' primer set is specific for

FR2 sequences. Subsequently, a repertoire of human VH and VL chain FR3, CDR3 and FR4

sequences (library fragment 2) is generated by PCR, where the 5' primer set is specific for FR3

sequences and the 3' primer set is downstream and may include FR4 sequences. Then library fragments 1 and 2 are combined in the library vector with fragment 1 always occuring upstream of

fragment 2. Included between fragments 1 and 2 is a unique cleavage site that enables a repertoire of nonhuman CDR2 sequences with a compatible cleavage site to be added to the

human display library. This method of generating a display library further increases the repertoire

of ABPs by generating unique combinations of CDR sequences. Components of a library

generated using this strategy may also be derived from nonrearranged germline sequences.

In another embodiment, the human heavy and light chain FR and/or CDR sequences

represented in the display libraries disclosed herein are synthetically generated from in vivo

rearranged variable gene sequences or germline sequences. For example, WO97/08320 and

Knappik et al, 1999, supra disclose the design, construction and analysis of a human antibody

library concept designated HuCAL (Human Combinatorial Antibody Library). Each of the human

VH and VL subfamilies that is frequently used during an immune response is represented by one

consensus framework, resulting in seven master framework genes for the heavy chain repertoire

and seven for the light chain repertoire. Pairing of the different heavy and light chain master

framework regions yields a total of 49 possible combinations. The 14 HuCAL master framework

sequences are synthetically generated with the following considerations: antibody structure,

amino acid sequence diversity and germline usage. The inclusion of useful cleavage sites flanking

each CDR results in the generation of modular framework genes that contain readily accessible

CDRs.

In another embodiment, the repertoires of nonhuman CDR sequences can be generated by

techniques known in the art, such as PCR amplification of CDR repertoires from naturally

rearranged V genes using primer sets specific for the framework and/or constant regions flanking

each particular CDR. In a preferred embodiment, a repertoire of nonhuman CDR3 sequences

may be generated by PCR amplification using a 5' primer set specific for sequences upstream of

CDR3 (e.g., FR3 sequences) and a 3' primer set specific for sequences downstream of CDR3 (e.g., FR4 and/or C region sequences). Repertoires of nonhuman CDR3 sequences can also be

generated by artificial rearrangement and/or randomization of nonrearranged germline sequences.

The artificially rearranged CDR3 sequences may be synthetically generated, PCR generated or

generated using a combination of the two techniques. In a preferred embodiment, sequences of

both primer sets will include cleavage sites that are compatible with cleavage sites that have been

incorporated into the human library sequences. Alternatively, the primer sets specific for

nonhuman CDR sequences can be designed with homology to compatible sequences in the human

library to enable insertion of the repertoires of nonhuman CDR sequences into the human libraries

using conventional PCR technologies (e.g., overlapping PCR).

Similarly, repertoires of nonhuman CDR1 and CDR2 sequences can be generated via PCR

amplification of naturally rearranged V genes. In the case of CDR1 sequences, the 5' primer set is

specific for sequences upstream of CDR1 (e.g., FRl and/or leader sequences) and the 3' primer

set would be specific for sequences downstream of CDR1 (e.g., FR2 sequences). A repertoire of

nonhuman CDR2 sequences would be generated using a 5' primer set specific for sequences

upstream of CDR2 (e.g., FR2 sequences) and a 3' primer set specific for sequences downstream

of CDR2 (e.g., FR3 sequences). PCR amplification of repertoires of nonhuman CDR1 and CDR2

sequences may also be accomplished using nonrearranged germline sequences. The CDR1 and

CDR2 sequences may be synthetically generated, PCR generated, or generated using a

combination of the two techniques. In a preferred embodiment, sequences of all primer sets will

include cleavage sites that are compatible with cleavage sites that have been incorporated into the

human library sequences. Alternatively, the primer sets specific for nonhuman CDR sequences

can be designed with homology to compatible sequences in the human library to enable insertion of the repertoires of nonhuman CDR sequences into the human libraries using conventional PCR

technologies (e.g., overlapping PCR).

In another embodiment, the repertoire of nonhuman CDR sequences may be synthetically

generated from mixed trinucleotides while biasing towards natural antibody sequences (e.g., using

the HuCAL method described above; Virnekas et al, 1994, Nucl. Acids Res.. 22:5600-5607;

which is incorporated herein by reference). The CDR sequences are designed such that the

naturally occurring diversity is covered, both in terms of length and amino acid composition. In a

preferred embodiment, a repertoire of nonhuman CDR3 sequences may be synthetically

generated. Nonhuman CDR sequences can be collected from publically available sources (e.g.,

Kabat et al, 1991, Sequences of Proteins of Immunological Interest, 5^th edition, Publication

number 91-3242; Benson et al, 1997, Nucl. Acids Res., 25:1-6) or may be determined by

perforrning sequencing reactions on a representation of antibody variable regions from the

predetermined nonhuman species. Nonhuman CDR sequences may be derived from gerrnline

antibody sequences, or preferably, from a collection of rearranged and mutated antibody

sequences. Alignment of nonhuman CDR sequences (using, a minimum of 3 different antibody

sequences) allows the identification and positions of commonly occurring amino acid residues.

For example, the VH chain CDR3 sequences from at least three different antibody sequences may

be aligned using standard commercially available alignment software well known to those of skill

in the art (e.g., DNA Star). This alignment information allows nucleotide biasing during synthetic

trinucleotide mutagenesis resulting in the generation of a diverse but biased population of

nonhuman VH chain CDR3 sequences. The repertoires of nonhuman CDR sequences are

designed to contain unique terminal cleavage sites that are compatible with cleavage sites that

have been incorporated into the human library sequences. In a preferred embodiment, repertoires of nonhuman CDR sequences, whether generated, through PCR or synthetic means, will contain flanking unique cleavage sites that will allow their

insertion into a human display library that has been designed to include compatible unique

cleavage sites flanking the respective FR sequences. For example, a human display library may be

generated to include a unique cleavage site(s) flanking the VH chain CDR3 sequences. The

repertoire of nonhuman VH chain CDR3 sequences also may be designed to contain flanking

sequence cleavage sites compatible with the cleavage sites contained in the human display library.

By specific enzymatic digestions of the human library sequences and the nonhuman VH chain

CDR3 sequences, the repertoire of nonhuman CDR3 sequences may be inserted into the human

display library at the appropriate site using an enzymatic ligation reaction.

Alternatively, repertoires of nonhuman CDR sequences may be inserted into human

display libraries that have been generated without the inclusion of unique cleavage sites. Primer

sets utilized to PCR amplify repertoires of nonhuman CDR sequences can be generated with homology to corresponding regions of the human display library allowing the nonhuman CDR

sequences to be inserted in the human library through conventional PCR technologies. For

example, a repertoire of nonhuman CDR3 sequences may be generated by PCR amplification

using a 5' primer set specific for FR3 sequences and a 3' primer set specific for FR4 sequences

where the 5' and 3' primer sets contain sequences homologous to human FR3 and FR4 sequences

respectively. The repertoire of nonhuman CDR3 sequences then may be inserted into the human

display library using overlapping PCR techniques.

The repertoires of nonhuman CDR sequences may be derived from any nonhuman species,

preferably a mammalian species. Suitable nonhuman mammalian species include but are not limited to nonhuman primates, camel, cattle, sheep, goats, pigs, rabbits, rats, guinea pigs and mice.

The repertoires of nonhuman CDR sequences may be derived from a nonhuman species that has not been immunized with the preselected antigen (naϊve animal). In a preferred

embodiment, the nonhuman CDR sequences can be derived from a nonhuman species previously immunized with a preselected human antigen. For example, immunization of the nonhuman species can be accomplished by injecting the animal with purified human antigen, components of the antigen (e.g., specific peptide or carbohydrate epitopes), crude preparations of the antigen, structures displaying the antigen (e.g., membrane preparations, whole cells, etc.) or antigen in the context of a carrier (e.g., hapten). Multiple injections with the antigen source are typically performed at specific time intervals (e.g., two weeks) and the immunizations can be given in the presence and/or absence of various adjuvants (e.g., complete or incomplete Freund's adjuvant).

Repertoires of nonhuman CDR sequences generated by PCR amplification of naturally rearranged antibody sequences can be derived from various tissue sources. Sources of tissue include spleen, thymus, peripheral blood lymphocytes, B cells or any other tissue source that contains rearranged antibody sequences. Tissues from the nonhuman species may be derived from a naϊve uriimmunized animal or an animal previously immunized with a preselected human antigen. The PCR template from each tissue can be derived from isolated RNA that has been reverse transcribed by methods that are known in the art (see McPherson et al., supra). In one embodiment, ABPs specific for a human antigen may be produced as follows:

1) constructing a library in a library vector (e.g., phagemid) in accordance with any of the methods disclosed here and above, of human VH and VL chain sequences including CDR and FR sequences contained therein, wherein at least one of the human CDR sequences from the human heavy chain and at least one of the human CDR sequences from the

human light chain is replaced by a repertoire of corresponding nonhuman CDR sequences

to form a library of nucleic acid sequences;

2) introducing the library into a recombinant host cell in order to package the library into a

GDU; wherein each specific GDU contains the nucleic acid sequence encoding a specific

ABP and displays the ABP on the GDU surface as a component of the GDU;

3) selecting the ABP of interest by exposing the library of GDUs displaying a repertoire of

ABPs to a preselected human antigen or derivative thereof, such as those disclosed herein

above, and recovering an ABP that binds to the preselected human antigen;

4) amplifying the selected GDUs displaying the specific ABP before subsequent further

selection and screening and therefore enriching GDUs that encode a desired ABP;

5) determining the nucleic acid sequence encoding the selected ABP by performing

sequencing reactions (well known in the art, see for example Sambrook et al., supra) on

isolated DNA of the GDU displaying the specific ABP.

Useful libraries may include ribosomal display libraries, yeast display libraries, bacterial

libraries and viral display libraries (see for example, U.S. Patent Nos. 5,643,756; 5,723,287;

5,952,474; Schaffitzel et al. 1999, supra; Kitamura, 1998, supra; Georgjou et al, supra; Little, et

al, 1995, supra; Chauthaiwale et al. 1992, supra; Aruffo, 1991, supra). In a preferred

embodiment, the display libraries will be filamentous bacteriophage libraries.

Each ABP of a library of ABPs may be expressed as a single polypeptide chain (e.g., an

scFv fragment) or as two polypeptide chains (e.g., an Fv or an Fab fragment). Alternatively, the

two polypeptide chains may be expressed as fusion partners with nonantibody domains which will interact, either covalently or non-covalently, to hold the variable domains in a conformation which allows the antibody polypeptide chains to form an ABP.

In one embodiment, the ABPs are expressed as a single polypeptide chain where the VH and VL chain sequences are linked into a continuous sequence that preferably contains a sequence separating the two chains encoding a peptide linker. The single polypeptide chain is expressed in

a manner that is suitable to enable the display of the ABPs in association with the GDU. In a preferred embodiment, nucleic acid sequences encoding the single polypeptide chain are cloned

into the 5' region of a gene encoding a phagemid coat protein (e.g., gene HI, gHI) where the

vector is derived from a bacteriophage, such as fd, Ml 3, Ifl, Ike, ZJ/Z, Ff, XF, Pfl, Pf3, etc. In another embodiment, the ABPs are expressed as two polypeptide chains where the VH and VL chain sequences are not linked, but are preferably both encoded by the same GDU. The two polypeptide chains are expressed in a manner that is suitable to enable the display of the ABPs in association with the GDU. In a preferred embodiment, the nucleic acid sequences

encoding one chain of the ABPs are cloned into the 5' region of a gene encoding a phagemid coat

protein (e.g., gene m) where the vector is derived from a bacteriophage, such as fd, M13, Ifl, Ike, ZJ/Z, Ff, XF, Pfl, Pf3, etc. The second chain of ABP is expressed (preferably from the same vector) as a second polypeptide chain which is transported to the periplasm of the host cell. The heavy and light chains then come together into an ABP during phage assembly. Since the ABP is fused to the N terminus of the bacteriophage coat protein, the ABP is displayed on the outer surface of the bacteriophage particles as they extrude from the host cell.

Selected ABPs may be expressed as soluble molecules that may be isolated in free form.

In a preferred embodiment, phagemid vectors that encode an ABP of interest can be manipulated to produce a soluble form of the ABP that is otherwise expressed on the surface of a bacteriophage. For example, an amber mutation can be used to allow expression of the free form under certain conditions.

The ABPs selected by the methods described above may be used directly, or may be further

fused with additional sequences (e.g., human constant regions) in order to equip antibody

molecules with specific properties (e.g., effector function, longer half-life etc.). The variable

domains may also be fused to sequences encoding, enzymes or toxins (e.g., ricin, Shigella toxin,

diptheria toxin) to permit antibody-directed targeting of these molecules to particular cell types

(e.g., tumor cells).

ABPs, isolated by, and according to, the teachings herein include antibodies, or fragments

thereof, directed against human antigens. Human antigens include, but are not limited to: cell

surface molecules, such as cell adhesion molecules (e.g., ICAM-1, ICAM-2, VCAM-1, CTLA-4,

the B7s, VLA-4, LFA-1, CD 11 a/CD 18, LPAM-1), selectins, (L-, E- and P-selectin); cluster of

differentiation (CD) antigens (e.g., CD2, CD3, CD28, CD45R, CD58); blood group antigens

(e.g., A, B, O), transplantation antigens (e.g., MHC class I and class π molecules); tumor-

associated antigens; immunoglobulins (e.g., IgE, IgA, IgM, IgG, IgD); soluble proteins and

cytokines, such as interleukins (e.g., IL-lα, IL-lβ, DL-2, IL-3, JX4, etc), interferons (e.g., IFNα,

EFNβ, JJFNγ), growth factors (e.g., GMCSF, TGF-β) and tumor necrosis factors (e.g., TNF-α,

TNF-β).

Terminology

By Genetic Display Unit (GDU) is meant a biological structure or organism which can

display an antigen binding pair (ABP). The ABP is encoded by genetic information that is a

component of the structure or organism, therefore linking the display of the ABP and the nucleic acid sequence that is encoding it. The biological structure or organism may be a ribosome, virus (e.g., bacteriophage such as as fd, Ml 3, Ifl, Ike, ZJ/Z, Ff, XF, Pfl, Pf3, etc), bacteria or yeast.

By antigen binding pair (ABP) is meant the pairing of two polypeptide chains of an

antibody, wherein this pairing results in a molecule capable of binding an antigen. For example,

exemplary ABPs include scFv, Fv, Fab, F(ab')₂ and whole antibodies.

By scFv is meant a single chain variable fragment consisting of a VL and a VH chain that

are joined in a continuous sequence that preferably contains a sequence of a predetermined length

separating the two chains which encodes a peptide linker.

By Fv is meant a VL and VH chain that are encoded by separate sequences. The VL and

VH chains may be associated by noncovalent interactions or may be covalently linked through

disulfide bonding. Alternatively, the VL and VH chains may be expressed as fusion partners with

nonantibody domains which will interact, either covalently or noncovalently, to hold the variable domains in a conformation which allows the VL and VH chains to form an ABP.

By Fab is meant VL, VH, CL and CHI chains that are encoded by separate sequences.

The polypetide chains may be associated by noncovalent interactions or may be covalently linked

through disulfide bonding. Alternatively, the polypeptide chains may be expressed as fusion

partners with nonantibody domains which will interact, either covalently or noncovalently, to hold

the variable domains in a conformation which allows the polypeptide chains to form an ABP.

By F(ab')₂ is meant two Fab fragments linked to one another by a disulfide bond found in the hinge region following the CHI chain. Alternatively, the Fab fragments may be expressed as

fusion partners with nonantibody domains which will interact, either covalently or noncovalently,

to hold the Fab fragements in a conformation which allows the two Fab fragments to form an

F(ab')₂. By vector is meant a DNA molecule, which is capable of replication (amplification) in a

host organism. Generally the vector is amenable to gene insertion for amplification and/or

expession of the inserted gene. The vector is also capable of being introduced into the host cell.

By naturally generated is meant V chain sequences that are generated through

amplification (e.g., PCR) of existing sequences. Existing sequences include sequences of

nonrearranged germline genes and/or rearranged genes.

By synthetically generated is meant V chain sequences that are generated by total

synthesis. Synthesis of sequences may include nonrearranged germline genes and/or rearranged

genes. Synthetic sequences can include a representation of a group of sequences as in consensus

sequences.

By artificially rearranged is meant V chain sequences rearranged by means other than

natural gene rearrangement.

The term repertoire is used to indicate a collection of sequences that represent genetic

diversity.

By display library is meant a collection of nucleotide sequences that encode a diverse

population of ABPs that are displayed in association with GDUs.

By human antigen is meant a molecule of human origin that is capable of, either alone

and/or in combination with other agents, eliciting an antibody response.

By primer set is meant one or more oligonucleotide sequences that is capable of

recognizing a majority of members in a family of sequences (e.g., human VH chain FR3 sequences) in order that, in combination with a compatible primer set, the family of sequences can

by amplified by PCR. Oligonucleotides within a primer set may contain varying degrees of degeneracy. By cleavage site is meant a stretch of nucleic acid sequence that is recognized by a specific enzyme wherein the specific sequence can be cleaved by that enzyme (e.g., restriction endonuclease sites).

It will be understood that various modifications may be made to the embodiments disclosed herein. For example, any ABP derived from a library produced according to the

methods herein may be modified to produce a derivative thereof, or used to prepare a therapeutic or prophylactic medicament or a diagnostic product. Also contemplated are kits including any of the display libraries or library components disclosed here and above. Therefore, the above description should not be construed as limiting, but merely as exemplifications of preferred embodiments. Those skilled in the art will envision other modifications within the scope of the claims appended hereto.

Claims

What is claimed is:

1. A display library comprising a repertoire of human heavy and light chain FR and human

CDR sequences contained therein, wherein at least one of the human CDR sequences from the

human heavy chain and at least one of the human CDR sequences from the human light chain is

replaced by a repertoire of corresponding nonhuman CDR sequences.

2. The library of claim 1, wherein the library is a selected from the group consisting of a

ribosomal display library, yeast display library, bacterial display library, and viral display library.

3. The library of claim 1, wherein the library is a bacteriophage display library.

4. The library of claim 1, wherein the library is a filamentous bacteriophage display library.

5. The library of claim 3, wherein the bacteriophage is selected from the group consisting

of class I bacteriophages and the class π bacteriophages.

6. The library of claim 1, which displays a member of the group consisting essentially of

whole antibodies and antigen binding fragments thereof.

7. The library of claim 6, wherein the antigen binding fragments comprise fragments

selected from the group consisting of Fv, disulfide-linked Fv, scFv, Fab, F(ab')₂, and other

fragments capable of binding an antigen.

8. The library of claim 1, wherein the repertoire of nonhuman CDR sequences is from an

animal immunized with a preselected antigen.

9. The library of claim 1, wherein the repertoire of nonhuman CDR sequences is from an

uriimmunized naive animal.

10. The library of claim 1, wherein the repertoire of nonhuman CDR sequences is

synthetically generated.

11. The library of claim 1, wherein the repertoire of nonhuman CDR sequences is naturally

generated.

12. The library of claim 1, wherein the repertoire of nonhuman CDR sequences is

generated from nonrearranged germline sequences.

13. The library of claim 12, wherein the germline sequences are artificially rearranged.

14. The library of claim 1, wherein the repertoires of nonhuman CDR sequences comprise

mixtares of the nonhuman CDR sequences of claims 10, 11 and/or 12.

15. The library of claim 1, wherem the repertoire of human heavy and light chain FR and

human CDR sequences is synthetically generated.

16. The library of claim 1, wherein the repertoire of human heavy and light chain FR and

human CDR sequences is naturally generated.

17. The library of claim 1, wherein the repertoire of human heavy and light chain FR and

human CDR sequences is generated from nonrearranged germline sequences.

18. The library of claim 17, wherein the geimline sequences are artificially rearranged.

19. The library of claim 1, wherein the repertoire of human heavy and light chain FR and

human CDR sequences comprises mixtares of the human heavy and light chain FR and human

CDR sequences from claims 15, 16 and/or 17.

20. The library of claim 1, wherein the human heavy and light chain framework regions

comprise unique flanking restriction endonuclease sites.

21. The library of claim 20, wherein the corresponding nonhuman CDR sequences are

flanked by restriction endonuclease sites compatible with the restriction endonuclease sites of

claim 20.

22. A method of producing an antigen binding pair comprising: providing a library displaying a repertoire of human heavy and light chain FR and

human CDR sequences contained therein, wherein at least one of the human CDR sequences from

the human heavy chain and at least one of the human CDR sequences from the human light chain

is replaced by a repertoire of corresponding nonhuman CDRs;

and selecting an antigen binding pair by exposing the library to a preselected human

antigen and recovering therefrom an antigen binding pair that binds to the preselected human antigen.

23. The method of claim 22, wherein the library is a selected from the group consisting of a

ribosomal display library, a yeast display library, a bacterial display Ubrary, and a viral display library.

24. The method of claim 22, wherein the library is a bacteriophage display library.

25. The method of claim 24, wherein the library is a filamentous bacteriophage display library.

26.The method of claim 24, wherein the bacteriophage is selected from the group

consisting of the class I bacteriophages and the class JJ bacteriophages.

27. The method of claim 22, wherein said repertoire which is displayed comprises whole

antibodies and/or antigen binding fragments thereof.

28. The method of claim 27, wherein the antigen binding fragments comprise fragments

selected from the group consisting of Fv, disulfide-linked Fv, scFV, Fab, F(ab')₂ and other

fragments capable of binding an antigen.

29. The method of claim 22, wherein the repertoire of non-human CDR sequences is from

an animal immunized with a preselected antigen.

30. The method of claim 22, wherein the repertoire of non-human CDR sequences is from an unimmunized naive animal.

31. The method of claim 22, wherein the non-human CDR sequences are synthetically generated.

32. The method of claim 22, wherein the repertoire of nonhuman CDR sequences is naturally generated.

33. The method of claim 22, wherein the repertoire of nonhuman CDR sequences is generated from nonrearranged germline sequences.

34. The method of claim 33, wherein the geimline sequences are artificially rearranged.

35. The method of claim 22, wherein the repertoires of nonhuman CDR sequences comprise mixtures of the nonhuman CDR sequences of claims 31, 33, and/or 34.

36. The method of claim 22, wherein the repertoire of human heavy and light chain FR and human CDR sequences is synthetically generated.

37. The method of claim 22, wherein the repertoire of human heavy and light chain FR and human CDR sequences is synthetically generated.

38. The method of claim 22, wherein the repertoire of human heavy and light chain FR and human CDR sequences is generated from nonrearranged germline sequences.

39. The method of claim 38, wherein the germline sequences are artificially rearranged.

40. The method of claim 22, wherein the human heavy and light chain FR and human CDR sequences contained therein comprise mixtures of the repertoires of claims 36, 37, and/or 39.

41. The method of claim 22, wherein the human heavy and light chain FR comprise unique restriction sites.

42. The method of claim 41, wherein the restriction sites flank each of the CDR sequences.

43. The method of claim 22, wherein the selected antigen binding agent is modified to produce a derivative thereof.

44. An antigen-binding agent produced by the method of claim 22.