AU2747302A - DNA mutagenesis by random fragmentation and reassembly - Google Patents

Description

S&F Ref: 347925D2

AUSTRALIA

PATENTS ACT 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT

ORIGINAL

Name and Address of Applicant: Actual Inventor(s): Address for Service: Invention Title: Affymax Technologies N.V.

De Ruyderkade 62 Curacao Netherlands Antilles Willem P. C. Stemmer, Andreas Crameri Spruson Ferguson St Martins Tower,Level 31 Market Street Sydney NSW 2000 (CCN 3710000177) DNA Mutagenesis by Random Fragmentation and Reassembly The following statement is a full description of this invention, including the best method of performing it known to me/us:- 5845c iealicX1 to DNA mutagenesis by random fragmentation and reassembly.

BACKGROUND OF THE INVENTION Field of the Invention The present invention relates to a method for the production of polynucleotides conferring a desired phenotype and/or encoding a protein having an advantageous predetermined property which is selectable. In an aspect, the method is used for generating and selecting nucleic acid fragments encoding mutant proteins.

Description of the Related Art The complexity of an active sequence of a biological macromolecule, e.g. proteins, DNA etc., has been called its information content The information content of a protein has been defined as the resistance of the active protein to amino acid sequence variation, calculated from the minimum 20 number of invariable amino acids (bits) required to describe a family of related sequences with the same function Proteins that are sensitive to random mutagenesis have a high information content. In 1974, when this definition was coined, 9 .9 protein diversity existed only as taxonomic diversity.

Molecular biology developments such as molecular libraries have allowed the identification of a much larger number of variable bases, and even to select functional sequences from random libraries. Most residues can be varied, although typically not all at the same time, depending on compensating changes in the context. Thus a 100 amino acid protein can contain only 2,000 different mutations, but 20100 possible combinations of mutations.

Information density is the Information Content/unit length of a sequence. Active sites of enzymes tend to have a high information density. By contrast, flexible linkers in enzymes have a low information density Current methods in widespread use for creating mutant proteins in a library format are error-prone polymerase chain reaction (11, 12, 19) and cassette mutagenesis 20, 21, 22, 40, 41, 42), in which the specific region to be c-timized is 2 replaced with a synthetically mutagenized oligonucleotide. In both cases, a 'mutant cloud' is generated around certain sites in the original sequence.

Error-prone PCR uses low-fidelity polymerization conditions to introduce a low level of point mutations randomly over a long sequence. Error prone PCR can be used to mutagenize a mixture of fragments of unknown sequence. However, computer simulations have suggested that point mutagenesis alone may often be too gradual to allow the block changes that are required for continued sequence evolution. The published error-prone

PCR

protocols do not allow amplification of DNA fragments greater than 0.5 to 1.0 kb, limiting their practical application.

Further, repeated cycles of error-prone PCR lead to an accumulation of neutral mutations, which, for example, may make i. 15 a protein immunogenic.

In oligonucleotide-directed mutagenesis, a short sequence is replaced with a synthetically mutagenized oligonucleotide. This approach does not generate combinations of distant mutations and is thus not combinatorial. The limited library size relative to 20 the vast sequence length means that many rounds of selection are j unavoidable for protein optimization. Mutagenesis with synthetic oligonucleotides requires sequencing of individual clones after each selection round followed by grouping into families, arbitrarily choosing a single family, and reducing it to a consensus motif, which is resynthesized and reinserted into a single gene followed by additional selection. This process constitutes a statistical bottleneck, it is labor intensive and not practical for many rounds of mutagenesis.

Error-prone PCR and oligonucleotide-directed mutagenesis are thus useful for single cycles of sequence fine tuning but rapidly become limiting when applied for multiple cycles.

Error-prone PCR can be used to mutagenize a mixture of fragments of unknown sequence (11, 12). However, the published error-prone PCR protocols (11, 12) suffer from a low processivity of the polymerase. Therefore, the protocol is unable to result in the random mutagenesis of an average-sized gene. This inability limits the practical application of error-prone

PCR.

Another serious limitation of error-prone PCR is that the rate of down-mutations grows with the information content of the sequence. At a certain information content, library size, and mutagenesis rate, the balance of down-mutations to up-mutations will statistically prevent the selection of further improvements (statistical ceiling).

Finally, repeated cycles of error-prone PCR will also lead to the accumulation of neutral mutations, which can affect, for example- immunogenicity but not binding affinity.

Thus error-prone PCR was found to be too gradual to allow the block changes that are required for continued sequence evolution 2).

In cassette mutagenesis, a sequence block of a single template is typically replaced by a (partially) randomized sequence. Therefore, the maximum information content that can be 15 obtained is statistically limited by the number of random sequences library size). This constitutes a statistical bottleneck, eliminating other sequence families which are not currently best, but which may have greater long term potential.

Further, mutagenesis with synthetic oligonucleotides requires sequencing of individual clones after each selection round Therefore, this approach is tedious and is not practical for many rounds of mutagenesis.

Error-prone PCR and cassette mutagenesis are thus best 0 suited and have been widely used for fine-tuning areas of S 25 comparatively low information content. One apparent exception is the selection of an RNA ligase ribozyme from a random library using many rounds of amplification by error-prone PCR and selection (13) It is becoming increasingly clear that the tools for the design of recombinant linear biological sequences such as protein, RNA and DNA are not as powerful as the tools nature has developed. Finding better and better mutants depends on searching more and more sequences within larger and larger libraries, and increasing numbers of cycles of mutagenic amplification and selection are necessary. However as discussed above, the existing mutagenesis methods that are in widespread use have distinct limitations when used for repeated cycles.

Evolution of most organisms occurs by natural selection and 4 sexual reproduction. Sexual reproduction ensures mixing and combining of the genes of the offspring of the selected individuals. During meiosis, homologous chromosomes from the parents line up with one another and cross-over part way along their length, thus swapping genetic material. Such swapping or shuffling of the DNA allows organisms to evolve more rapidly (1, In sexual recombination, because the inserted sequences were of proven utility in a homologous environment, the inserted sequences are likely to still have substantial information content once they are inserted into the new sequence.

Marton et al.,(27) describes the use of PCR in vitro to monitor recombination in a plasmid having directly repeated sequences. Marton et al. discloses that recombination will occur during PCR as a result of breaking or nicking of the DNA. This 15 will give rise to recombinant molecules. Meyerhans et al. (23) S"also disclose the existence of DNA recombination during in vitro

PCR.

:..The term Applied Molecular Evolution means the application of an evolutionary design algorithm to a specific, 20 useful goal. While many different library formats for AME have been reported for polynucleotides 11-14), peptides and Sproteins (phage (15-17), lacI (18) and polysomes, in none of

S.

o• these formats has recombination by random cross-overs been usedo• •to deliberately create a combinatorial library.

Sof a5 Theoretically there are 2,000 different single mutants of a 100 amino acid protein. A protein of 100 amino acids has 20100 possible combinations of mutations, a number which is too large to exhaustively explore by conventional methods. It would be advantageous to develop a system which would allow the generation and screening of all of these possible combination mutations.

Winter and coworkers (43,44) have utilized an in vivo site specific recombination system to combine light chain antibody genes with heavy chain antibody genes for expression in a phage system. However, their system relies on specific sites of recombination and thus is limited. Hayashi et al. (48) report simultaneous mutagenesis of antibody CDR regions in single chain antibodies (scFv) by overlap extension and PCR.

Caren et al. (45) describe a method for generating a large population of multiple mutants using random in vivo recombination. However, their method requires the recombination of two different libraries of plasmids, each library having a different selectable marker. Thus the method is limited to a finite number of recombinations equal to the number of selectable markers existing, and produces a concomitant linear increase in the number of marker genes linked to the selected sequence(s).

Calogero et al. (46) and Galizzi et al. (47) report that in vivo .recombination between two homologous but truncated insecttoxin genes on a plasmid can produce a hybrid gene. Radman et al. (49) report in vivo recombination of substantially mismatched SDNA sequences in a host cell having defective mismatch repair Senzymes, resulting in hybrid molecule formation.

It would be advantageous to develop a method for the production of mutant proteins which method allowed for the development of large libraries of mutant nucleic acid sequences S. which were easily searched. The invention described herein is directed to the use of repeated cycles of point mutagenesis, 20 nucleic acid shuffling and selection which allow for the directed molecular evolution in vitro of highly complex linear sequences, such as proteins through random recombination.

Accordingly, it would be advantageous to develop a method which allows for the production of large libraries of mutant DNA, RNA or proteins and the selection of particular mutants for a desired goal. The invention described herein is directed to the use of repeated cycles of mutagenesis, in vivo recombination and selection which allow for the directed molecular evolution in vivo of highly complex linear sequences, such as DNA, RNA or proteins through recombination.

Further advantages of the present invention will become apparent from= the following description of the invention with reference to the attached drawings.

r Summary of the Invention According to a first embodiment of the invention, there is provided a population of polynucleotides resulting from shuffling sequences from substrate polynucleotides encoding antibody chains, whereby the population of polynucleotides comprises recombinant polynucleotides encoding antibody chains having different combinations of heavy chain CDRs and/or light chain CDRs than are encoded by the substrate polynucleotides.

According to a second embodiment of the invention, there is provided an antibody obtainable by expression of polynucleotides in a population according to the first embodiment of the present invention, said antibody being capable of specifically binding to a predetermined antigen and comprising a novel combination of heavy CDR regions S* and/or a novel combination of light chain CDRs.

According to a third embodiment of the invention, there is provided an antibody S. comprising a heavy chain comprising three nonnaturally occurring CDR regions, and a light chain comprising three nonnaturally occurring CDR regions, wherein the antibody specifically binds to a predetermined antigen.

According to a fourth embodiment of the invention, there is provided a method of producing a polynucleotide encoding an antibody chain that specifically binds to a predetermined antigen, the method comprising: denaturing a population of polynucleotide segments encoding CDR regions; S* annealing the denatured population of CDR encoding polynucleotides; incubating the annealed CDR encoding polynucleotides in the presence of a polymerase, thereby forming a population of polynucleotides comprising recombinant polynucleotides encoding antibodies with novel combinations of CDR regions, which novel combinations of CDR regions are not present in an antibody encoded by the population of polynucleotide segments provided in step repeating steps through and screening or selecting the population of polynucleotides to identify at least one polynucleotide that encodes an antibody chain that specifically binds to a predetermined antigen.

There is herein disclosed a method for generating a selected polynucleotide sequence or population of selected polynucleotide sequences, typically in the form of amplified [I:\DayLib\LIBFF]73935spec.doc:gcc 6 and/or cloned polynucleotides, whereby the selected polynucleotide sequence(s) possess a desired phenotypic characteristic encode a polypeptide, promote transcription of linked polynucleotides, bind a protein, and the like) which can be selected for. One method of identifying polypeptides tha: possess a desired structure or functional property, such as binding to a predetermined biological macromolecule receptor), involves the screening of a large library uo polypeptides for individual library members which possess the desired structure or functional property conferre by the amino acid sequence of the polypeotide.

There is disclosed herein a method for generating libraries of displayed polypeptides or displayed antibodies suitable for affinity interaction screening or phenotvoic 1 5 screening. The method comprises obtaining a first plurality of selected library members comprising a displayed polypeptide or displayed antibody and an associated polynucleotide encoding said Sdisplayed polypeptide or displayed antibody, and obtaining said associated polynucleotides or copies thereof wherein said associated polynucieozides comprise a region of substantiall identical sequence, optionally introducing mutations into said polynucleotides or copies, and pooling and fracmenting, typically randomly, said associated polynucleotides or conies form frarments thereof under conditions suitable for P amplification, performing ?CR amplification and opoionall-.

mutagenesis, and thereby honoiogously recombining said fragmento form a shuffled pooi of recombined poiynucleotides, wherebv a substantial fraction greater than 10 percent) .f recombined polynucleotides of said shuffled pool are not present in the first piurality of selected library members, said shuffle.

pool composing a library of displayed polypeptides or disolaved antibodies suitable for affinity interaction screenn.

Optionally, th. method comprises the additional step of screeninthe library members of the snuffled pool to identify individual shuffled library members having the ability to bind or otherw-is interact such as catalytic antibodies) with a predetermined macromolecule, such as for example a oroteinaceous receptor, peptide, oligosaccharide, virion, or other 7 predetermined compound or structure. The displayed polypeptides, antibodies, peptidomimetic antibodies, and variable region sequences that are identified from such libraries can be used for therapeutic, diagnostic, research, and related purposes catalysts, solutes for increasing osmolarity of an acueous solution, and the like), and/or can be subjected to one or nore additional cycles of shuffling and/-r affinity selection. The method can be modified such tha the step of...selfecting is for a phenotypic characteristic other than binding affinity for a predetermined molecule for catalytic activity, stability, oxidation resistance, drug resistance, or detectable phenotype conferred on a host cell).

In one embodiment, the first plurality of selected library members is fragmented and homologously recombined by PCR in 15 vitro.

In one embodiment, the first plurality of selected library members is fragmented in vitro, the resultant fragments transferred into a host cell or organism and homologously recombined to form shuffled library members in vivo.

In one embodiment, the first plurality of selected library members is cloned or amplified on episomally replicable vectors, a multiplicity of said vectors is transferred into a cell and homologously recombined to form shuffled library members in vivo.

In one embodiment, the first plurality of selected library members is not fragmented, but is -cloned or amplified on an episomally replicable vector as a direct repeat, which each repeat comprising a distinct species of selected library member sequence, said vector is transferred into a cell and homologously recombined by intra-vecto recombination to form shuffled library members in vivo.

In an embodiment, zombinations of in vitro and in ivo shuffling are provided to enhance combinatorial diversity.

There is ftrther disclosed herein a method for generating libraries of displayed- antibodies suitable for affinity interaction screening. The method comprises obtaining a first plurality of selected. library members comprising a displayed antibody and an associated polynucleotide encoding said displayed antibody, and obtaining said associated polynucleotides V 8 or copies thereof, wherein said associated polynucleotides comprise a region of substantially identical variable region framework sequence, and pooling and fragmenting said associated polynucleotides or copies to form fragments thereof under conditions suitable for PCR amplification and thereby homologously recombining said fragments to form a shuffled pool of recombincd .polynucleotides comprising novel combinations of CDRs, whlieby a substantial fraction greater than percent) of-the recombined polynucleotides of said shuffled pool comprise CDR combinations are not present in the first plurality of selected library members, said shuffled pool composing a library of displayed antibodies comprising CDR permutations and suitable for affinity interaction screening. Optionally, the shuffled pool is subjected to affinity screening to select shuffled library members which bind to a predetermined epitope (antigen) and thereby selecting a plurality of selected shuffled library members. Optionally, the plurality of selected shuffled library members can be shuffled and screened iteratively, from 1 to about 1000 cycles or as desired until library members having a desired binding affinity are obtained.

Accordingly, there is herein disclosed a method for introducing one or more mutations into a template double-stranded polynucleotide, wherein the template doublestranded polynucleotide has been cleaved into random fraoments of S. a desired size, by adding to the resultant population of doublestranded fragments one or more single or double-stranded oligonucleotides, wherein said cligonucleotides comorise an area of identity and an area of heterology to the temolate polynucleotide; denaturing the resultant mixture cf doubiestranded random fragments and oligonucleotides into sinalestranded fragments; incubating the resultant population of single-stranded fragments with a polymerase under conditions which result in the annealing cf said single-stranded fracments at regions of identity between 'he single-stranded fragments and formation of a mutageni.zed double-stranded polynucleotide; and repeating the above steps as desired.

In another aspect the present invention is directed to a method of producing recombinant proteins havina biolocical 9 activity by treating a sample comprising double-stranded template polynucleotides encoding a wild-type protein under conditions which provide for the cleavage of said template polynucleotides into random double-stranded fragments having a desired size; S adding to the resultant population of random fragments one or more single or double-stranded oligonucleotides, wherein said oligonucleotides comprise areas of identity and areas of heterology to the template polynucleotide; denaturina the resultant mixture of double-stranded fragments and oligonucleotides into single-stranded fragments; incubating the resultant population of single-stranded fragments with a polymerase under conditions which result in the annealing of said single-stranded fragments at the areas of identity and formation of a mutagenized double-stranded polynucleotide; repeatina the 15 above steps as desired; and then expressing the recombinant protein from the mutagenized double-stranded polynucleotide.

There is further disclosed herein a method for obtaining a chimeric polynucleotide by treating a sample comprising different double-stranded template 20 polynucleotides wherein said different template polvnucleotides contain areas of identity and areas of heterology under conditions which provide for the cleavage of said template polynucleotides into random double-stranded fragments of a desired size; denaturing the resultant random double-stranded fragments contained in the treated sample into single-stranded fragments; incubating the resultant single-stranded fragments with polymerase under conditions which provide for the annealing of the single-stranded fragments at the areas of identity and the *formation of a chimeric double-stranded polynucleotide sequence comprising template polynucleotide sequences; and repeating the above steps as desired.

Still further, there is disclosed herein a method of replicating a template polynucleotide by combining in vitro single-stranded template polynucleotides with small random single-stranded fragments resulting from the cleavage and denaturation of the template polynucleotide, and incubating said mixture of nucleic acid fragments in the presence of a nucleic acid polymerase under conditions wherein a population of doublestranded template polynucleotides is formed.

There is further disclosed herein the use of polynucleotide shuffling, in vitro and/or in vivo to shuffle polynucleotides comprising transcriptional regulatory sequences.

Still further, the disclosure herein provides the use of polynucleotide shuffling to o0 shuffle a population of viral genes capsid proteins, spike glycoproteins, polymerases, proteases, etc.) or viral genomes paramyxoviridae, orthomyxoviridae, herpesviruses, retroviruses, reoviruses, rhinoviruses, etc.). In an embodiment, there is disclosed herein a method for shuffling sequences encoding all or portions of immunogenic viral proteins to generate novel combinations of epitopes as well as novel 15 epitopes created by recombination; such shuffled viral proteins may comprise epitopes or combinations of epitopes which are likely to arise in the natural environment as a consequence of viral evolution such as recombination of influenza virus strains).

There is further disclosed herein a method suitable for shuffling polynucleotide sequences for generating gene therapy vectors and replication-defective gene therapy constructs, such as may be used for human gene therapy, including but not limited to vaccination vectors for DNA-based vaccination, as well as anti-neoplastic gene therapy :and other gene therapy formats.

Brief Description of the Drawings Figure 1 is a schematic diagram comparing mutagenic shuffling over error-prone PCR; the initial library; pool of selected sequences in first round of affinity selection; in vitro recombination of the selected sequences ('shuffling'); pool of selected sequences in second round of affinity selection after shuffling; error-prone PCR; pool of selected sequences in second round of affinity selection after error-prone

PCR.

Figure 2 illustrates the reassembly of a 1.0 kb LacZ alpha gene fragment from 10-50 bp random fragments. Photograph of a gel of PCR amplified DNA fragment having the LacZ alpha gene. Photograph of a gel of DNA fragments after digestion with [I:\DayLib\LIBFF]64191spec.DOC:gcc DNAseI. Photograph of a gel of DNA fragments of 10-50 bp purified from the digested LacZ alpha gene DNA fragment; (d) Photograph of a gel of the 10-50 bp DNA fragments after the indicated number of cycles of DNA reassembly; Photograph of a gel of the recombination mixture after amplification by PCR with primers.

Figure 3 is a schematic illustration of the LacZ alpha gene stop codon mutants and their DNA sequences. The boxed regions are heterologous areas, serving as markers. The stop codons are located in smaller boxes or underlined. indicates a wildtype gene and indicates a mutated area in the gene.

Figure 4 is a schematic illustration of the introduction or spiking of a synthetic oligonucleotide into the reassembly process of the LacZ alpha gene.

S 15 Figure 5 illustrates the regions of homology between a murine IL1-B gene and a human IL1-B gene with E. coli codon usage. Regions of heterology are boxed. The "I indicate crossovers obtained upon the shuffling of the two genes.

Figure 6 is a schematic diagram of the antibody CDR 20 shuffling model system using the scFv of anti-rabbit IgG antibody (AO1B).

Figure 7 illustrates the observed frequency of occurrence of certain combinations of CDRs in the shuffled DNA of the scFv of anti-rabbit IgG antibody (AlOB).

Figure 8 illustrates the improved avidity of the scFv antirabbit antibody after DNA shuffling and each cycle of selection.

Figure 9 schematically portrays pBR322-Sfi-BL-LA-Sfi and in vivo intraplasmidic recombination via direct repeats, as well as the rate of generation of ampicillin-resistant colonies by intraplasmidic recombination reconstituting a functional betalactamase gene.

Figure 10 schematically portrays pBR322-Sfi-2Bla-Sfi and in vivo intraplasmidic recombination via direct repeats, as well as the rate of generation of ampicillin-resistant colonies by intraplasmidic recombination reconstituting a functional betalactamase gene.

Figure 11 illustrates the method for testing the efficiency of multiple rounds of homologous recombination after the

-I

12 introduction of polynucleotide fragments into cells for the generation of recombinant proteins.

Figure 12 schematically portrays generation of a library of vectors by shuffling cassettes at the following loci: promoter, leader peptide, terminator, selectable drug resistance gene, and origin of replication. The multiple parallel lines at each locus represents the multiplicity of cassettes for that cassette.

Figure 13 schematically shows some examples of cassettes suitable at various loci for constructing prokaryotic vector libraries by shuffling.

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention relates to a method for nucleic acid molecule reassembly after random fragmentation and its 15 application to mutagenesis of DNA sequences. Also described is a method for the production of nucleic acid fragments encoding mutant proteins having enhanced biological activity. In particular, the present invention also relates to a method of S 2 repeated cycles of mutagenesis, nucleic acid shuffling and 2 0 selection which allow for the creation of mutant proteins having enhanced biological activity.

The present invention is directed to a method for generating a very large library of DNA, RNA or protein mutants.

This method has particular advantages in the generation of 2. related DNA fragments from which the desired nucleic acid fragment(s) may be selected. In particular the present invention also relates to a method of repeated cycles of mutagenesis, homologous recombination and selection which allow for the creation of mutant proteins having enhanced biological activity.

However, prior to discussing this invention- in further detail, the following terms will first be defined.

Definitions As used herein, the following terms have the following meanings: The term "DNA reassembly" is used when recombination occurs between identical sequences.

By contrast, the term "DNA shuffling" is used herein to 13 indicate recombination between substantially homologous but nonidentical sequences, in some embodiments DNA shuffling may involve crossover via nonhomologous recombination, such as via cre/lox and/or flp/frt systems and the like.

The term "amplification" means that the number of copies of a nucleic acid fragment is increased.

The term "identical" or "identity" means that two nucleic acid sequences have the same sequence or a complementary sequence. Thus, "areas of identity" means that regions or areas of a nucleic acid fragment or polynucleotide are identical or complementary to another polynucleotide or nucleic acid fragment.

The term "corresponds to" is used herein to mean that a polynucleotide sequence is homologous is identical, not strictly evolutionarily related) to all or a portion of a 5 reference polynucleotide sequence, or that a polypeptide sequence s identical to a reference polypeptide sequence. In contradistinction, the term "complementary to" is used herein to mean that the complementary sequence is homologous to all or a portion of a reference polynucleotide sequence. For illustration, 20 the nucleotide sequence "TATAC" corresponds to a reference sequence "TATAC" and is complementary to a reference sequence

S"GTATA".

~The following terms are used to describe the sequence relationships between two or more polynucleotides: "reference sequence", "comparison window", "sequence identity", "percentage of sequence identity", and "substantial identity". A "reference sequence" is a defined sequence used as a basis 'for a sequence comparison; a reference sequence may be a subset of a larger sequence, for example, as a segment of a full-length cDNA or gene sequence given in a sequence listing, such as a polynucleotide sequence of Fig. 1 or Fig. or may comprise a complete cDNA or gene sequence. Generally, a reference sequence is at least nucleotides in length, frequently at least 25 nucleotides in length, and often at least 50 nucleotides in length. Since two polynucleotides may each comprise a sequence a portion of the complete polynucleotide sequence) that is similar between the two polynucleotides, and may further comprise a sequence that is divergent between the two polynucleotides, sequence comparisons between two (or more) polynucleotides are typically performed by comparing sequences of the two polynucleotides over a "comparison window" to identify and compare local regions of sequence similarity.

A "comparison window", as used herein, refers to a conceptual segment of at least 20 contiguous nucleotide positions wherein a polynucleotide sequence may be compared to a reference sequence of at least 20 contiguous nucleotides and wherein the portion of the polynucleotide sequence in the comparison window may comprise additions or deletions gaps) of 20 percent or less as compared to the reference sequence (which does not comprise additions or deletions) for optimal alignment of the two sequences. Optimal alignment of sequences for aligning a comparison window may be conducted by the local homology 15 algorithm of Smith and Waterman (1981) Adv. Appl. Math. 2: 482, by the homology alignment algorithm of Needleman and Wunsch (1970) J. Mol. Biol. 48: 443, by the search for similarity method of Pearson and Lipman (1988) Proc. Natl. Acad. Sci. 2444, by computerized implementations of these algorithms (GAP, BESTFIT, FASTA, and TFASTA in the Wisconsin Genetics Software Package Release 7.0, Genetics Computer Group, 575 Science Dr., Madison, WI), or by inspection, and the best alignment resulting in the highest percentage of homology over the comparison window) generated by the various methods is selected.

The term "sequence identity" means that two polynucleotide sequences are identical on a nucleotide-bynucleotide basis) over the window of comparison. The term "percentage of sequence identity" is calculated by comparing two optimally aligned sequences over the window of comparison, determining the number of positions at which the identical nucleic acid base A, T, C, G, U, or I) occurs in both sequences to yield the number of matched positions, dividing the number of matched positions by the total number of positions in the window of comparison the window size), and multiplying the result by 100 to yield the percentage of sequence identity.

The terms "substantial identity" as used herein denotes a characteristic of a polynucleotide sequence, wherein the polynucleotide comprises a sequence that has at least 80 percent sequence identity, preferably at least 85 percent identity and often 90 to 95 percent sequence identity, more usually at least 99 percent sequence identity as compared to a reference sequence over a comparison window of at least 20 nucleotide positions, frequently over a window of at least 25-50 nucleotides, wherein the percentage of sequence identity is calculated by comparing the reference sequence to the polynucleotide sequence which may include deletions or additions which total 20 percent or less of the reference sequence over the window of comparison.

Conservative amino acid substitutions refer to the interchangeability of residues having similar side chains. For example, a group of amino acids having aliphatic side chains is glycine, alanine, valine, leucine, and isoleucine; a group of amino acids having aliphatic-hydroxyl side chains is serine and 15 threonine; a group of amino acids having amide-containing side chains is asparagine and glutamine; a group of amino acids having aromatic side chains is phenylalanine, tyrosine, and tryptophan; a group of amino acids having basic side chains is lysine, arginine, and histidine; and a group of amino acids having 20 sulfur-containing side chains is cysteine and methionine.

Preferred conservative amino acids substitution groups are: :valine-leucine-isoleucine, phenylalanine-tyrosine, lysine-_ arginine, alanine-valine, and asparagine-glutamine.

The term "homologous" or "homeologous" means that one single-stranded nucleic acid sequence may hybridize to a complementary single-stranded nucleic acid sequence. The degree of hybridization may depend on a number of factors including the amount of identity between the sequences and the hybridization conditions such as temperature and salt concentration as discussed later. Preferably the region of identity is greater than about 5 bp, more preferably the region of identity is greater than 10 bp.

The term "heterologous" means that one single-stranded nucleic acid sequence is unable to hybridize to another singlestranded nucleic acid sequence or its complement. Thus areas of heterology means that nucleic acid fragments or polynucleotides have areas or regions in the sequence which are unable to hybridize to another nucleic acid or polynucleotide. Such regions or areas are, for example, areas of mutations.

The term "cognate" as used herein refers to a gene sequence that is evolutionarily and functionally related between species. For example but not limitation, in the human genome, the human CD4 gene is the cognate gene to the mouse CD4 gene, since the sequences and structures of these two genes indicate that they are highly homologous and both genes encode a protein which functions in signaling T cell activation through.MHC class II-restricted antigen recognition.

The term "wild-type" means that the nucleic acid fragment does not comprise any mutations. A "wild-type" protein means that the protein will be active at a level of activity found in nature and will comprise the amino acid sequence found in nature.

The term "related polynucleotides" means that regions or 15 areas of the polynucleotides are identical and regions or areas of the polynucleotides are heterologous.

The term "chimeric polynucleotide" means that the polynucleotide comprises regions which are wild-type and regions 2 which are mutated. It may also mean that the polynucleotide 20 comprises wild-type regions from one polynucleotide and wild-type regions from another related polynucleotide.

The term "cleaving" means digesting the polynucleotide with enzymes or breaking the polynucleotide.

The term "population" as used herein means a collection of components such as polynucleotides, nucleic acid fragments or proteins. A "mixed population" means a collection of components which belong to the same family of nucleic acids or proteins are related) but which differ in their sequence are not identical) and hence in their biological activity.

The term "specific nucleic acid fragment" means -a nucleic acid fragment having certain end points and having a certain nucleic acid sequence. Two nucleic acid fragments wherein one nucleic acid fragment has the identical sequence as a portion of the second nucleic acid fragment but different ends comprise two different specific nucleic acid fragments.

The term "mutations" means changes in the sequence of a wild-type nucleic acid sequence or changes in the sequence of a peptide. Such mutations may be point mutations such as 17 transitions or transversions. The mutations may be deletions, insertions or duplications.

In the polypeptide notation used herein, the lefthand direction is the amino terminal direction and the righthand direction is the carboxy-terminal direction, in accordance with standard usage and convention. Similarly, unless specified otherwise, the lefthand end of single-stranded polynucleotide sequences is the 5' end; the lefthand direction of doublestranded polynucleotide sequences is referred to as the direction. The direction of 5' to 3' addition of nascent

RNA

transcripts is referred to as the transcription direction; sequence regions on the DNA strand having the same sequence as the RNA and which are 5' to the 5' end of the RNA transcript are referred to as "upstream sequences"; sequence regions on the DNA 15 strand having the same sequence as the RNA and which are 3' to the 3' end of the coding RNA transcript are referred to as "downstream sequences".

The term "naturally-occurring" as used herein as applied to an object refers to the fact that an object can be 20 found in nature. For example, a polypeptide or polynucleotide sequence that is present in an organism (including viruses) that can be isolated from a source in nature and which has not been intentionally modified by man in the laboratory is naturallyoccurring. Generally, the term naturally-occurring refers to an i object as present in a non-pathological (undiseased) individual, such as would be typical for the species.

The term "agent" is used herein to denote a chemical S. compound, a mixture of chemical compounds, an array of spatially localized compounds a VLSIPS peptide array, polynucleotide array, and/or combinatorial small molecule array), a biological macromolecule, a bacteriophage peptide display library, a bacteriophage antibody scFv) display library, a polysome peptide display library, or an extract made from biological materials such as bacteria, plants, fungi, or animal (particularly mammalian) cells or tissues. Agents are evaluated for potential activity as antineoplastics, anti-inflammatories, or apoptosis modulators by -inclusion in screening assays described hereinbelow. Agents are evaluated for potential activity as specific protein interaction inhibitors an agent which selectively inhibits a binding interaction between two predetermined polypeptides but which does not substantially interfere with cell viability) by inclusion in screening assays described hereinbelow.

As used herein, "substantially pure" means an object species is the predominant species present on a molar basis it is more abundant than any other individual macromolecular species in the composition), and preferably a substantially purified fraction is a composition wherein the object species comprises at least about 50 percent (on a molar basis) of all macromolecular species present. Generally, a substantially pure composition will comprise more than about to 90 percent of all macromolecular species present in the 15 composition. Most preferably, the object species is purified to essential homogeneity (contaminant species cannot be detected in the composition by conventional detection methods) wherein the composition consists essentially of a single macromolecular species. Solvent species, small molecules (<500 Daltons), and elemental ion species are not considered macromolecular species.

As used herein the term "physiological conditions" refers to temperature, pH, ionic strength, viscosity, and like biochemical parameters which are compatible with a viable organism, and/or which typically exist intracellularly in a 5 viable cultured yeast cell or mammalian cell. For example, the intracellular conditions in a yeast cell grown under typical laboratory culture conditions are physiological conditions.

Suitable in vitro reaction conditions for in vitro transcription cocktails are generally physiological conditions. In general, in vitro physiological conditions comprise 50-200 mM NaCl-or KC1, pH 6.5-8.5, 20-45 0 C and 0.001-10 mM divalent cation Mg Ca+) preferably about 150 mM NaCl or KC1, pH 7.2-7.6, 5 mM divalent cation, and often include 0.01-1.0 percent nonspecific protein BSA). A non-ionic detergent (Tween, NP-40, Triton X-100) can often be present, usually at about 0.001 to 2%, typically 0.05-0.2% Particular aqueous conditions may be selected by the practitioner according to conventional methods.

For general guidance, the following buffered aqueous conditions 19 may be applicable: 10-250 mM NaC1, 5-50 mM Tris HC1, pH 5-8, with optional addition of divalent cation(s) and/or metal chelators and/or nonionic detergents and/or membrane fractions and/or antifoam agents and/or scintillants.

Specific hybridization is defined herein as the formation of hybrids between a first polynucleotide and a second polynucleotide a polynucleotide having a distinct but substantially identical sequence to the first polynucleotide), wherein the first polynucleotide preferentially hybridizes to the second polynucleotide under stringent hybridization conditions wherein substantially unrelated polynucleotide sequences do not form hybrids in the mixture.

As used herein, the term "single-chain antibody" refers to a polypeptide comprising a VH domain and a VL domain in 15 polypeptide linkage, generally linked via a spacer peptide [Gly-Gly-Gly-Gly-Ser]x), and which may comprise additional amino acid sequences at the amino- and/or carboxy- termini. For example, a single-chain antibody may comprise a tether segment :for linking to the encoding polynucleotide. As an example, a 20 scFv is a single-chain antibody. Single-chain antibodies are generally proteins consisting of one or more polypeptide segments of at least 10 contiguous amino acids substantially encoded by genes of the immunoglobulin superfamily see The Immunoqlobulin Gene Superfamily, A.F. Williams and A.N. Barclay, in Immunoqlobulin Genes, T. Honjo, F.W. Alt, and T.H. Rabbitts, eds., (1989) Academic Press: San Diego, CA, pp.361-387, which is incorporated herein by reference), most frequently encoded by a rodent, non-human primate, avian, porcine, bovine, ovine, goat, or human heavy chain or light chain gene sequence. A functional single-chain antibody generally contains a sufficient portion of an immunoglobulin superfamily gene product so as to retain the property of binding to a specific target molecule, typically a receptor or antigen (epitope).

As used herein, the term "complementarity-determining region" and "CDR" refer to the art-recognized term as exemplified by the Kabat and Chothia CDR definitions also generally known as hypervariable regions or hypervariable loops (Chothia and Lesk (1987) J. Mol. Biol. 196: 901; Chothia et al. (1989) Nature 342: 877; E.A. Kabat et al., Sequences of Proteins of Immunological Interest (National Institutes of Health, Bethesda, MD) (1987); and Tramontano et al. (1990) J. Mol. Biol. 215: 175). Variable region domains typically comprise the amino-terminal approximately 105-115 amino acids of a naturally-occurring immunoglobulin chain amino acids 1-110), although variable domains somewhat shorter or longer are also suitable for forming single-chain antibodies.

An immunoglobulin light or heavy chain variable region consists of a "framework" region interrupted by three hypervariable regions, also called CDR's. The extent of the framework region and CDR's have been precisely defined (see, "Sequences of Proteins of Immunological Interest," E. Kabat et al., 4th Ed., U.S. Department of Health and Human Services, Bethesda, MD (1987)). The sequences of the framework regions of different light or heavy chains are relatively conserved within a species. As used herein, a "human framework region" is a framework region that is substantially identical (about 85% or more, usually 90-95% or more) to the framework region of a 20 naturally occurring human immunoglobulin. The framework region of an antibody, that is the combined framework regions of the constituent light and heavy chains, serves to position and align the CDR's. The CDR's are primarily responsible for binding to an epitope of an antigen.

S2._ As used herein, the term "variable segment" refers to a portion of a nascent peptide which comprises a random, pseudorandom, or defined kernal sequence. A variable segment can comprise both variant and invariant residue positions, and the

S.

degree of residue variation at a variant residue position may be limited; both options are selected at the discretion of the practitioner. Typically, variable segments are about 5 to amino acid residues in length 8 to 10), although variable segments may be longer and may comprise antibody portions or receptor proteins, such as an antibody fragment, a nucleic acid binding protein, a receptor protein, and the like.

As used herein, "random peptide sequence" refers to an amino acid sequence composed of two or more amino acid monomers and constructed by a stochastic or random process. A random peptide can include framework or scaffolding motifs, which may comprise invariant sequences.

As used herein "random peptide library" refers to a set of polynucleotide sequences that encodes a set of random peptides, and to the set of random peptides encoded by those polynucleotide sequences, as well as the fusion proteins containing those random peptides.

As used herein, the term "pseudorandom" refers to a set of sequences that have limited variability, so that for example the degree of residue variability at one position is different than the degree of residue variability at another position, but any pseudorandom position is allowed some degree of residue variation, however circumscribed.

As used herein, the term "defined sequence framework" 15 refers to a set of defined sequences that are selected on a nonrandom basis, generally on the basis of experimental data or structural data; for example, a defined sequence framework may comprise a set of amino acid sequences that are predicted to form a 3-sheet structure or may comprise a leucine zipper heptad 20 repeat motif, a zinc-finger domain, among other variations.

A

"defined sequence kernal" is a set of sequences which encompass a limited scope of variability. Whereas a completely random 10-mer sequence of the 20 conventional amino acids can be any of (20)10 sequences, and a pseudorandom 10-mer sequence of the 5 20 conventional amino acids can be any of (20) 1 0 sequences but will exhibit a bias for certain residues at certain positions and/or overall, a defined sequence kernal is a subset of sequences which is less that the maximum number of potential sequences if each residue position was allowed to be any of the allowable 20 conventional amino acids (and/or- allowable unconventional amino/imino acids). A defined sequence kernal generally comprises variant and invariant residue positions and/or comprises variant residue positions which can comprise a residue selected from a defined subset of amino acid residues), and the like, either segmentally or over the entire length of the individual selected library member sequence. Defined sequence kernals can refer to either amino acid sequences or polynucleotide sequences. For illustration and not limitation, the sequences (NNK) 1 0 and (NNM) 1 0 where N represents A, T, G, or C; K represents G or T; and M represents A or.C, are defined sequence kernals.

As used herein "epitope" refers to that portion of an antigen or other macromolecule capable of forming a binding interaction that interacts with the variable region binding pocket of an antibody. Typically, such binding interaction is manifested as an intermolecular contact with one or more amino acid residues of a CDR.

As used herein, "receptor" refers to a molecule that has an affinity for a given ligand. Receptors can be naturally occurring or synthetic molecules. Receptors can be employed in an unaltered state or as aggregates with other species.

Receptors can be attached, covalently or noncovalently, to a 15 binding member, either directly or via a specific binding substance. Examples of receptors include, but are not limited to, antibodies, including monoclonal antibodies and antisera e* reactive with specific antigenic determinants (such as on 0 viruses, cells, or other materials), cell membrane receptors, S" 20 complex carbohydrates and glycoproteins, enzymes, and hormone receptors.

As used herein "ligand" refers to a molecule, such as a random peptide or variable segment sequence, that is recognized by a particular receptor. As one of skill in the art will j recognize, a molecule (or macromolecular complex) can be both a receptor and a ligand. In general, the binding partner having a smaller molecular weight is referred to as the ligand and the binding partner having a greater molecular weight is referred to as a receptor.

As used herein, "linker" or "spacer" refers to a molecule or group of molecules that connects two molecules, such as a DNA binding protein and a random peptide, and serves to place the two molecules in a preferred configuration, so that the random peptide can bind to a receptor with minimal steric hindrance from the DNA binding protein.

As used herein, the term "operably linked" refers to a linkage of polynucleotide elements in a functional relationship.

A nucleic acid is "operably linked" when it is placed into a functional relationship with another nucleic acid sequence. For instance, a promoter or enhancer is operably linked to a coding sequence if it affects the transcription of the coding sequence.

Operably linked means that the DNA sequences being linked are typically contiguous and, where necessary to join two protein coding regions, contiguous and in reading frame.

Methodology Nucleic acid shuffling is a method for in vitro or in vivo homologous recombination of pools of nucleic acid fragments or polynucleotides. Mixtures of related nucleic acid sequences or polynucleotides are randomly fragmented, and reassembled to yield a library or mixed population of recombinant nucleic acid molecules or polynucleotides.

15 In contrast to cassette mutagenesis, only shuffling and error-prone PCR allow one to mutate a pool of sequences blindly (without sequence information other than primers) The advantage of the mutagenic shuffling of this invention over error-prone PCR alone for repeated selection can best be explained with an example from antibody engineering. In Figure 1 is shown a schematic diagram of DNA shuffling as described herein. The initial library can consist of related sequences of diverse origin antibodies from naive mRNA) or can be derived by any type of mutagenesis (including shuffling) of a 2. single antibody gene. A collection of selected complementarity determining regions ("CDRs") is obtained after the first round of f affinity selection (Fig.l). In the diagram the thick CDRs confer S: onto the antibody molecule increased affinity for the antigen.

Shuffling allows the free combinatorial association of all of the CDRls with all of the CDR2s with all of the CDR3s, etc. -(Fig.l).

This method differs from PCR, in that it is an inverse chain reaction. In PCR, the number of polymerase start sites and the number of molecules grows exponentially. However, the sequence of the polymerase start sites and the sequence of the molecules remains essentially the same. In contrast, in nucleic acid reassembly or shuffling of random fragments the number of start sites and the number (but not -size) of the random fragments decreases over time. For fragments derived from whole plasmids 1 24 the theoretical endpoint is a single, large concatemeric molecule.

Since cross-overs occur at regions of homology, recombination will primarily occur between members of the same sequence family. This discourages combinations of CDRs that are grossly incompatible (eg. directed against different epitopes of the same antigen). It is contemplated that multiple families of sequences can be shuffled in the same reaction. Further, shuffling conserves the relative order, such that, for example, CDR1 will not be found in the position of CDR2.

Rare shufflants will contain a large number of the best (eg.

highest affinity) CDRs and these rare shufflants may be selected based on their superior affinity (Fig. 1).

CDRs from a pool of 100 different selected antibody 15 sequences can be permutated in up to 1006 different ways. This large number of permutations cannot be represented in a single library of DNA sequences. Accordingly, it is contemplated that multiple cycles of DNA shuffling and selection may be required depending on the length of the sequence and the sequence 20 diversity desired.

Error-prone PCR, in contrast, keeps all the selected CDRs in the same relative sequence (Fig. generating a much smaller mutant cloud.

The template polynucleotide which may be used in the methods of this invention may be DNA or RNA. It may be of various lengths depending on the size of the gene or DNA fragment to be 0 recombined or reassembled. Preferably the template polynucleotide is from 50 bp to 50 kb. It is contemplated that entire vectors containing the nucleic acid encoding the protein of interest can be used in the methods of this invention, and in fact have been successfully used.

The template polynucleotide may be obtained by amplification using the PCR reaction Patent No. 4,683,202 and 4,683,195) or other amplification or cloning methods. However, the removal of free primers from the PCR product before fragmentation provides a more efficient result. Failure to adequately remove the primers can lead to a low frequency of crossover clones.

The template polynucleotide often should be double-stranded.

A double-stranded nucleic acid molecule is required to ensure that regions of the resulting single-stranded nucleic acid fragments are complementary to each other and thus can hybridize to form a double-stranded molecule.

It is contemplated that single-stranded or double-stranded nucleic acid fragments having regions of identity to the template polynucleotide and regions of heterology to the template polynucleotide may be added to the template polynucleotide at this step. It is also contemplated that two different but related polynucleotide templates can be mixed at this step.

The double-stranded polynucleotide template and any added double-or single-stranded fragments are randomly digested into fragments of from about 5 bp to 5 kb or more. Preferably the size of the random fragments is from about 10 bp to 1000 bp, more S 15 preferably the size of the DNA fragments is from about 20 bp to I' 500 bp.

~Alternatively, it is also contemplated that double-stranded nucleic acid having multiple nicks may be used in the methods of othis invention. A nick is a break in one strand of the double- S stranded nucleic acid. The distance between such nicks is o preferably 5 bp to 5 kb, more preferably between 10 bp to 1000 bp.

The nucleic acid fragment may be digested by a number of different methods. The nucleic acid fragment may be digested 0 0with a nuclease, such as DNAseI or RNAse. The nucleic acid may be randomly sheared by the method of sonication or by passage *o through a tube having a small orifice.

It is also contemplated that the nucleic acid may also be partially digested with one or more restriction enzymes, such that certain points of cross-over may be retained statistically.

The concentration of any one specific nucleic acid fragment will not be greater than 1% by weight of the total nucleic acid, more preferably the concentration of any one specific nucleic acid sequence will not be greater than 0.1% by weight of the total nucleic acid.

The number of different specific nucleic acid fragments in the mixture will be at least about 100, preferably at least about.

500, and more preferably at least about 1000.

26 At this step single-stranded or double-stranded nucleic acid fragments, either synthetic or natural, may be added to the random double-stranded nucleic acid fragments in order to increase the heterogeneity of the mixture of nucleic acid fragments.

It is also contemplated that populations of double-stranded randomly broken nucleic acid fragments may be mixed or combined at this step.

Where insertion of mutations into the template polynucleotide is desired, single-stranded or double-stranded nucleic acid fragments having a region of identity to the template polynucleotide and a region of heterology to the template polynucleotide may be added in a 20 fold excess by weight as compared to the total nucleic acid, more preferably the 15 single-stranded nucleic acid fragments may be added in a 10 fold excess by weight as compared to the total nucleic acid.

Where a mixture of different but related template polynucleotides is desired, populations of nucleic acid fragments from each of the templates may be combined at a ratio of less than about 1:100, more preferably the ratio is less than about 1:40. For example, a backcross of the wild-type polynucleotide with a population of mutated polynucleotide may be desired to eliminate neutral mutations mutations yielding an :0 00 insubstantial alteration in the phenotypic property being selected for). In such an example, the ratio of randomly digested wild-type polynucleotide fragments which may be added to the randomly digested mutant polynucleotide fragments is approximately 1:1 to about 100:1, and more preferably from 1:1 to 40:1.

The mixed population of random nucleic acid fragments are denatured to form single-stranded nucleic acid fragments and then reannealed. Only those single-stranded nucleic acid fragments having regions of homology with other single-stranded nucleic acid fragments will reanneal.

The random nucleic acid fragments may be denatured by heating. One skilled in the art could determine the conditions necessary to completely denature the double stranded nucleic acid. Preferably the temperature is from 80 °C to 100 more preferably the temperature is from 90 °C to 96 OC. Other methods which may be used to denature the nucleic acid fragments include pressure (36) and pH.

The nucleic acid fragments may be reannealed by cooling.

Preferably the temperature is from 20 °C to 75 oC, more preferably the temperature is from 40 oC to 65 OC. If a high frequency of crossovers is needed based on an average of only 4 consecutive bases of homology, recombination can be forced by using a low annealing temperature, although the process becomes more difficult. The degree of renaturation which occurs will depend on the degree of homology between the population of single-stranded nucleic acid fragments.

Renaturation can be accelerated by the addition of polyethylene glycol or salt. The salt concentration is preferably from 0 mM to 200 mM, more preferably the salt concentration is from 10 mM to 100 mM. The salt may be KC1 or NaC1. The concentration of PEG is preferably from 0% to more preferably from 5% to The annealed nucleic acid fragments are next incubated in the presence of a nucleic acid polymerase and dNTP's dATP, :dCTP, dGTP and dTTP). The nucleic acid polymerase may be the S" Klenow fragment, the Taq polymerase or any other DNA polymerase known in the art.

The approach to be used for the assembly depends on the S* minimum degree of homology that should still yield crossovers. If the areas of identity are large, Taq polymerase can be used with an annealing temperature of between 45-650C. If the areas of identity are small, Klenow polymerase can be used with an annealing temperature of between 20-30 0 C. One skilled in the art could vary the temperature of annealing to increase the number of cross-overs achieved.

The polymerase may be added to the random nucleic acid fragments prior to annealing, simultaneously with annealing or after annealing.

The cycle of denaturation, renaturation and incubation in the presence of polymerase is referred to herein as shuffling or reassembly of the nucleic acid. This cycle is repeated for a desired number of times. Preferably the cycle is repeated from 2 to 50 times, more preferably the sequence is repeated from to 40 times.

The resulting nucleic acid is a larger double-stranded polynucleotide of from about. 50 bp to about 100 kb, preferably the larger polynucleotide is from 500 bp to 50 kb.

This larger polynucleotide fragment may contain a number of copies of a nucleic acid fragment having the same size as the template polynucleotide in tandem. This concatemeric fragment is then digested into single copies of the template polynucleotide.

The result will be a population of nucleic acid fragments of approximately the same size as the template polynucleotide. The population will be a mixed population where single or doublestranded nucleic acid fragments having an area of identity and an area of heterology have been added to the template polynucleotide 15 prior to shuffling.

These fragment are then cloned into the appropriate vector and the ligation mixture used to transform bacteria.

It is contemplated that the single nucleic acid fragments may be obtained from the larger concatemeric nucleic acid fragment by amplification of the single nucleic acid fragments prior to cloning by a variety of methods including PCR (U.S.

Patent No. 4,683,195 and 4,683,202) rather than by digestion of the concatemer.

The vector used for cloning is not critical provided that it will accept a DNA fragment of the desired size. If expression of the DNA fragment is desired, the cloning vehicle should further comprise transcription and translation signals next to the site of insertion of the DNA fragment to allow expression of the DNA fragment in the host cell. Preferred vectors include the pUC series and the pBR series of plasmids.

The resulting bacterial population will include a number of recombinant DNA fragments having random mutations. This mixed population may be tested to identify the desired recombinant nucleic acid fragment. The method of selection will depend on the DNA fragment desired.

For example, if a DNA fragment which encodes for a protein with increased binding efficiency to a ligand is desired, the proteins expressed by each of the DNA fragments in the population or library may be tested for their ability to bind to the ligand by methods known in the art panning, affinity chromatography). If a DNA fragment which encodes for a protein with increased drug resistance is desired, the proteins expressed by each of the DNA fragments in the population or library may be tested for their ability to confer drug resistance to the host organism. One skilled in the art, given knowledge of the desired protein, could readily test the population to identify DNA fragments which confer the desired properties onto the protein.

It is contemplated that one skilled in the art could use a phage display system in which fragments of the protein are expressed as fusion proteins on the phage surface (Pharmacia, Milwaukee WI). The recombinant DNA molecules are cloned into the phage DNA at a site which results in the transcription of a S. 15 fusion protein a portion of which is encoded by the recombinant "oDNA molecule. The phage containing the recombinant nucleic acid molecule undergoes replication and transcription in the cell.

I"...The leader sequence of the fusion protein directs the transport of the fusion protein to the tip of the phage particle. Thus the fusion protein which is partially encoded by the recombinant DNA molecule is displayed on the phage particle for detection and selection by the methods described above.

It is further contemplated that a number of cycles of nucleic acid shuffling may be conducted with nucleic acid *eefragments from a subpopulation of the first population, which subpopulation contains DNA encoding the desired recombinant •ego protein. In this manner, proteins with even higher binding affinities or enzymatic activity could be achieved.

It is also contemplated that a number of cycles of nucleic acid shuffling may be conducted with a mixture of wild-type nucleic acid fragments and a subpopulation of nucleic.acid from the first or subsequent rounds of nucleic acid shuffling in order to remove any silent mutations from the subpopulation.

Any source of nucleic acid, in purified form can be utilized as the starting nucleic acid. Thus the process may employ DNA or RNA including messenger RNA, which DNA or RNA may be single or double stranded. In addition, a DNA-RNA hybrid which contains one strand of each may be utilized. The nucleic acid sequence may be of various lengths depending on the size of the nucleic acid sequence to be mutated. Preferably the specific nucleic acid sequence is from 50 to 50000 base pairs. It is contemplated that entire vectors containing the nucleic acid encoding the protein of interest may be used in the methods of this invention.

The nucleic acid may be obtained from any source, for example, from plasmids such a pBR322, from cloned DNA or RNA or from natural DNA or RNA from any source including bacteria, yeast, viruses and higher organisms such as plants or animals.

DNA or RNA may be extracted from blood or tissue material. The template polynucleotide may be obtained by amplification using the polynucleotide chain reaction (PCR) Patent no.

4,683,202 and 4,683,195). Alternatively, the polynucleotide may 15 be present in a vector present in a cell and sufficient nucleic acid may be obtained by culturing the cell and extracting the nucleic acid from the cell by methods known in the art.

Any specific nucleic acid sequence can be used to produce the population of mutants by the present process. It is only necessary that a small population of mutant sequences of the specific nucleic acid sequence exist or be created prior to the present process.

The initial small population of the specific nucleic acid sequences having mutations may be created by a number of 5 different methods. Mutations may be created by error-prone PCR.

Error-prone PCR uses low-fidelity polymerization conditions to introduce a low level of point mutations randomly over a long sequence. Alternatively, mutations can be introduced into the template polynucleotide by oligonucleotide-directed mutagenesis.

In oligonucleotide-directed mutagenesis, a short sequence of the polynucleotide is removed from the polynucleotide using restriction enzyme digestion and is replaced with a synthetic polynucleotide in which various bases have been altered from the original sequence. The polynucleotide sequence can also be altered by chemical mutagenesis. Chemical mutagens include, for example, sodium bisulfite, nitrous acid, hydroxylamine, hydrazine or formic acid. Other agents which are analogues of nucleotide precursors include nitrosoguanidine, 5-bromouracil, 2-

I

31 aminopurine, or acridine. Generally, these agents are added to the PCR reaction in place of the nucleotide precursor thereby mutating the sequence. Intercalating agents such as proflavine, acriflavine, quinacrine and the like can also be used. Random mutagenesis of the polynucleotide sequence can also be achieved by irradiation with X-rays or ultraviolet light. Generally, plasmid DNA or DNA fragments so mutagenized are introduced into E. coli and propagated as a pool or library of mutant plasmids.

Alternatively the small mixed population of specific nucleic acids may be found in nature in that they may consist of different alleles of the same gene or the same gene from different related species cognate genes). Alternatively, they may be related DNA sequences found within one species, for example, the immunoglobulin genes.

15 Once the mixed population of the specific nucleic acid sequences is generated, the polynucleotides can be used directly or inserted into an appropriate cloning vector, using techniques well-known in the art.

The choice of vector depends on the size of the polynucleotide sequence and the host cell to be employed in the methods of this invention. The templates of this invention may be plasmids, phages, cosmids, phagemids, viruses retroviruses, parainfluenzavirus, herpesviruses, reoviruses, paramyxoviruses, and the like), or selected portions thereof 2' coat protein, spike glycoprotein, capsid protein). For example, cosmids and phagemids are preferred where the specific nucleic acid sequence to be mutated is larger because these vectors are able to stably propagate large nucleic acid fragments.

If the mixed population of the specific nucleic acid sequence is cloned into a vector it can be clonally amplified by inserting each vector into a host cell and allowing the host cell to amplify the vector. This is referred to as clonal amplification because while the absolute number of nucleic acid sequences increases, the number of mutants does not increase.

Utility The DNA shuffling method of this invention can be performed blindly on a pool of unknown sequences. By adding to the reassembly mixture oligonucleotides (with ends that are homologous to the sequences being reassembled) any sequence mixture can be incorporated at any specific position into another sequence mixture. Thus, it is contemplated that mixtures of synthetic oligonucleotides, PCR fragments or even whole genes can be mixed into another sequence library at defined positions. The insertion of one sequence (mixture) is independent from the insertion of a sequence in another part of the template. Thus, the degree of recombination, the homology required, and the diversity of the library can be independently and simultaneously varied along the length of the reassembled DNA.

This approach of mixing two genes may be useful for the humanization of antibodies from murine hybridomas. The approach of mixing two genes or inserting mutant sequences into genes may be useful for any therapeutically used protein, for example, interleukin I, antibodies, tPA, growth hormone, etc. The approach may also be useful in any nucleic acid for example, promoters or Sintrons or 3' untranslated region or 5' untranslated regions of genes to increase expression or alter specificity of expression 20 of proteins. The approach may also be used to mutate ribozymes or aptamers.

Shuffling requires the presence of homologous regions separating regions of diversity. Scaffold-like protein structures may be particularly suitable for shuffling. The conserved scaffold determines the overall folding by selfassociation, while displaying relatively unrestricted loops that mediate the specific binding. Examples of such scaffolds are the immunoglobulin beta-barrel, and the four-helix bundle This shuffling can be used to create scaffold-like proteins with various combinations of mutated sequences for binding.

In vitro Shuffling The equivalents of some standard genetic matings may also be performed by shuffling in vitro. For example, a 'molecular backcross' can be performed by repeated mixing of the mutant's nucleic acid with the wild-type nucleic acid while selecting for the mutations of interest. As in traditional breeding, this approach can be used to combine phenotypes from different sources 33 into a background of choice. It is useful, for example, for the removal of neutral mutations that affect unselected characteristics immunogenicity). Thus it can be useful to determine which mutations in a protein are involved in the enhanced biological activity and which are not, an advantage which cannot be achieved by error-prone mutagenesis or cassette mutagenesis methods.

Large, functional genes can be assembled correctly from a mixture of small random fragments. This reaction may be of use for the reassembly of genes from the highly fragmented DNA of fossils In addition random nucleic acid fragments from fossils may be combined with nucleic acid fragments from similar genes from related species.

It is also contemplated that the method of this invention 15 can be used for the in vitro amplification of a whole genome from a single cell as is needed for a variety of research and diagnostic applications. DNA amplification by PCR is in practice limited to a length of about 40 kb. Amplification of a whole genome such as that of E. coli (5,000 kb) by PCR would require about 250 primers yielding 125 forty kb fragments. This approach is not practical due to the unavailability of sufficient sequence data. On the other hand, random digestion of the genome with DNAseI, followed by gel purification of small fragments will provide a multitude of possible primers. Use of this mix of random small fragments as primers in a PCR reaction alone or with the whole genome as the template should result in an inverse chain reaction with the theoretical endpoint of a single concatemer containing many copies of the genome.

100 fold amplification in the copy number and an average fragment size of greater than 50 kb may be obtained -when only random fragments are used (see Example 2) It is thought that the larger concatemer is generated by overlap of many smaller fragments. The quality of specific PCR products obtained using synthetic primers will be indistinguishable from the product obtained from unamplified DNA. It is expected that this approach will be useful for the mapping of genomes.

The polynucleotide to be shuffled can be fragmented randomly or non-randomly, at the discretion of the practitioner.

In Vivo Shuffling In an embodiment of in vivo shuffling, the mixed population of the specific nucleic acid sequence is introduced into bacterial or eukaryotic cells under conditions such that at least two different nucleic acid sequences are present in each host cell. The. fragments can be introduced into the host cells by a variety of different methods. The host cells can be transformed with the fragments using methods known in the art, for example treatment with calcium chloride. If the fragments are inserted into a phage genome, the host cell can be transfected with the recombinant phage genome having the specific nucleic acid sequences. Alternatively, the nucleic acid sequences can be introduced into the host cell using electroporation, transfection, lipofection, biolistics, conjugation, and the like.

15 In general, in this embodiment, the specific nucleic acids sequences will be present in vectors which are capable of stably replicating the sequence in the host cell. In addition, it is contemplated that the vectors will encode a marker gene such that host cells having the vector can be selected. This ensures that the mutated specific nucleic acid sequence can be recovered after introduction into the host cell. However, it is contemplated that the entire mixed population of the specific nucleic acid sequences need not be present on a vector sequence. Rather only 2 a sufficient number of sequences need be cloned into vectors to ensure that after introduction of the fragments into the host cells each host cell contains one vector having at least one specific nucleic acid sequence present therein. It is also S. contemplated that rather than having a subset of the population of the specific nucleic acids sequences cloned into vectors, this subset may be already stably integrated into the host cell.

It has been found that when two fragments which have regions of identity are inserted into the host cells homologous recombination occurs between the two fragments. Such recombination between the two mutated specific nucleic acid sequences will result in the production of double or triple mutants in some situations.

It has also been found that the frequency of recombination is increased if some of the mutated specific nucleic acid sequences are present on linear nucleic acid molecules.

Therefore, in a preferred embodiment, some of the specific nucleic acid sequences are present on linear nucleic acid fragments.

After transformation, the host cell transformants are placed under selection to identify those host cell transformants which contain mutated specific nucleic acid sequences having the qualities desired. For example, if increased resistance to a particular drug is desired then the transformed host cells may be subjected to increased concentrations of the particular drug and those transformants producing mutated proteins able to confer increased drug resistance will be selected. If the enhanced ability of a particular protein to bind to a receptor is desired, then expression of the protein can be induced from the S 15 transformants and the resulting protein assayed in a ligand *"°'"binding assay by methods known in the art to identify that subset of the mutated population which shows enhanced binding to the ligand. Alternatively, the protein can be expressed in another system to ensure proper processing.

Once a subset of the first recombined specific nucleic acid S. sequences (daughter sequences) having the desired characteristics are identified, they are then subject to a second round of recombination.

In the second cycle of recombination, the recombined S 2' specific nucleic acid sequences may be mixed with the original mutated specific nucleic acid sequences (parent sequences) and the cycle repeated as described above. In this way a set of second recombined specific nucleic acids sequences can be identified which have enhanced characteristics or encode for proteins having enhanced properties. This cycle can be repeated a number of times as desired.

It is also contemplated that in the second or subsequent recombination cycle, a backcross can be performed. A molecular backcross can be performed by mixing the desired specific nucleic acid sequences with a large number of the wild-type sequence, such that at least one wild-type nucleic acid sequence and a mutated nucleic acid sequence are present in the same host cell after transformation. Recombination with the wild-type specific nucleic acid sequence will eliminate those neutral mutations that may affect unselected characteristics such as immunogenicity but not the selected characteristics.

In another embodiment of this invention, it is contemplated that during the first round a subset of the specific nucleic acid sequencescan be fragmented prior to introduction into the host cell. The size of the fragments must be large enough to contain some regions of identity with the other sequences so as to homologously recombine with the other sequences. The size of the fragments will range from 0.03 kb to 100 kb more preferably from 0.2 kb to 10 kb. It is also contemplated that in subsequent rounds, all of the specific nucleic acid sequences other than the sequences selected from the previous round may be cleaved into fragments prior to introduction into the host cells.

S 15 Fragmentation of the sequences can be accomplished by a variety of method known in the art. The sequences can be 9% 0randomly fragmented or fragmented at specific sites in the nucleic acid sequence. Random fragments can be obtained by *e breaking the nucleic acid or exposing it to harsh physical treatment shearing or irradiation) or harsh chemical agents by free radicals; metal ions; acid treatment to go depurinate and cleave). Random fragments can also be obtained, in the case of DNA by the use of DNase or like nuclease. The sequences can be cleaved at specific sites by the use of restriction enzymes. The fragmented sequences can be singlestranded or double-stranded. If the sequences were originally single-stranded they can be denatured with heat, chemicals or 9.

enzymes prior to insertion into the host cell. The reaction conditions suitable for separating the strands of nucleic acid are well known in the art.

The steps of this process can be repeated indefinitely, being limited only by the number of possible mutants which can be achieved. After a certain number of cycles, all possible mutants will have been achieved and further cycles are redundant.

In an embodiment the same mutated template nucleic acid is repeatedly recombined and the resulting recombinants selected for the desired characteristic.

Therefore, the initial pool or population of mutated 37 template nucleic acid is cloned into a vector capable of replicating in a bacteria such as E. coli. The particular vector is not essential, so long as it is capable of autonomous replication in E. coli. In a preferred embodiment, the vector is designed to allow the expression and production of any protein encoded by the mutated specific nucleic acid linked to the vector. It is also preferred that the vector contain a gene encoding for a selectable marker.

The population of vectors containing the pool of mutated nucleic acid sequences is introduced into the E. coli host cells.

The vector nucleic acid sequences may be introduced by transformation, transfection or infection in the case of phage.

The concentration of vectors used to transform the bacteria is such that a number of vectors is introduced into each cell. Once *0e 15 present in the cell, the efficiency of homologous recombination is such that homologous recombination occurs between the various vectors. This results in the generation of mutants (daughters) having a combination of mutations which differ from the original go parent mutated sequences.

The host cells are then clonally replicated and selected for 0: the marker gene present on the vector. Only those cells having eea plasmid will grow under the selection.

0000 The host cells which contain a vector are then tested for the presence of favorable mutations. Such testing may consist of placing the cells under selective pressure, for example, if the gene to be selected is an improved drug resistance gene. If the vector allows expression of the protein encoded by the mutated nucleic acid sequence, then such selection may include allowing expression of the protein so encoded, isolation of the protein and testing of the protein to determine whether, for example, it binds with increased efficiency to the ligand of interest.

Once a particular daughter mutated nucleic acid sequence has been identified which confers the desired characteristics, the nucleic acid is isolated either already linked to the vector or separated from the vector. This nucleic acid is then mixed with the first or parent population of nucleic acids and the cycle is repeated.

It has been shown that by this method nucleic acid sequences having enhanced desired properties can be selected.

In an alternate embodiment, the first generation of mutants are retained in the cells and the parental mutated sequences are added again to the cells. Accordingly, the first cycle of Embodiment I is conducted as described above. However, after the daughter nucleic acid sequences are identified, the host cells containing these sequences are retained.

The parent mutated specific nucleic acid population, either as fragments or cloned into the same vector is introduced into the host cells already containing the daughter nucleic acids.

Recombination is allowed to occur in the cells and the next generation of recombinants, or granddaughters are selected by the methods described above.

This cycle can be repeated a number of times until the @0 @0 o 15 nucleic acid or peptide having the desired characteristics is obtained. It is contemplated that in subsequent cycles, the population of mutated sequences which are added to the preferred

V.

mutants may come from the parental mutants or any subsequent 0 generation.

In an alternative embodiment, the invention provides a method of conducting a "molecular" backcross of the obtained recombinant specific nucleic acid in order to eliminate any_ neutral mutations. Neutral mutations are those mutations which do not confer onto the nucleic acid or peptide the desired 6 properties. Such mutations may however confer on the nucleic acid or peptide undesirable characteristics. Accordingly, it is *feedesirable to eliminate such neutral mutations. The method of e 0s this invention provide a means of doing so.

In this embodiment, after the mutant nucleic acid, having the desired characteristics, is obtained by the methods of the embodiments, the nucleic acid, the vector having the nucleic acid or the host cell containing the vector and nucleic acid is isolated.

The nucleic acid or vector is then introduced into the host cell with a large excess of the wild-type nucleic acid. The nucleic acid of the mutant and the nucleic acid of the wild-type sequence are allowed to recombine. The resulting recombinants are placed under the same selection as the mutant nucleic acid.

Only those recombinants which retained the desired characteristics will be selected. Any silent mutations which do not provide the desired characteristics will be lost through recombination with the wild-type DNA. This cycle can be repeated a number of times until all of the silent mutations are eliminated.

Thus the methods of this invention can be used in a molecular backcross to eliminate unnecessary or silent mutations.

Utility The in vivo recombination method of this invention can be performed blindly on a pool of unknown mutants or alleles of a specific nucleic acid fragment or sequence. However, it is not necessary to know the actual DNA or RNA sequence of the specific 15 nucleic acid fragment.

6* The approach of using recombination within a mixed population of genes can be useful for the generation of any useful proteins, for example, interleukin I, antibodies, tPA, growth hormone, etc. This approach may be used to generate proteins having altered specificity or activity. The approach S. imay also be useful for the generation of mutant nucleic acid So*. sequences, for example, promoter regions, introns, exons, enhancer sequences, 3' untranslated regions or 5' untranslated regions of genes. Thus this approach may be used to generate 2- genes having increased rates of expression. This approach may 00** also be useful in the study of repetitive DNA sequences.

*$Go Finally, this approach may be useful to mutate ribozymes or go aptamers.

Scaffold-like regions separating regions of diversity in proteins may be particularly suitable for the methods of this invention.l The conserved scaffold determines the overall folding by self-association, while displaying relatively unrestricted loops that mediate the specific binding. Examples of such scaffolds are the immunoglobulin beta barrel, and the four-helix bundle. The methods of this invention can be used to create scaffold-like proteins with various combinations of mutated sequences for binding.

The equivalents of some standard genetic matings may also be

I

performed by the methods of this invention. For example, a "molecular" backcross can be performed by repeated mixing of the mutant's nucleic acid with the wild-type nucleic acid while selecting for the mutations of interest. As in traditional breeding, this approach can be used to combine phenotypes from different sources into a background of choice. It is useful, for example, for the removal of neutral mutations that affect unselected characteristics itmmunogenicity). Thus it can be useful to determine which mutations in a protein are involved in the enhanced biological activity and which are not.

Peptide Display Methods The present method can be used to shuffle, by in vitro and/or in vivo recombination by any of the disclosed methods, and in any combination, polynucleotide sequences selected by peptide display methods, wherein an associated polynucleotide encodes a displayed peptide which is screened for a phenotype for affinity for a predetermined receptor (ligand) S 2An increasingly important aspect of biopharmaceutical S 20 drug development and molecular biology is the identification of peptide structures, including the primary amino acid sequences, .of peptides or peptidomimetics that interact with biological macromolecules. One. method of identifying peptides that possess a desired structure or functional property, such as binding to a Spredetermined biological macromolecule a receptor), involves the screening of a large library or peptides for o individual library members which possess the desired structure or functional property conferred by the amino acid sequence of the peptide.

In addition to direct chemical synthesis methods for generating peptide libraries, several recombinant DNA methods also have been reported. One type involves the display of a peptide sequence, antibody, or other protein on the surface of a bacteriophage particle or cell. Generally, in these methods each bacteriophage particle or cell serves as an individual library member displaying a single species of displayed peptide in addition to the natural bacteriophage or cell protein sequences.

Each bacteriophage or cell contains the nucleotide sequence information encoding the particular displayed peptide sequence; thus, the displayed peptide sequence can be ascertained by nucleotide sequence determination of an isolated library member.

A well-known peptide display method involves the presentation of a peptide sequence on the surface of a filamentous bacteriophage, typically as a fusion with a bacteriophage coat protein. The bacteriophage library can be incubated with an immobilized, predetermined macromolecule or small molecule a receptor) so that bacteriophage particles which present a peptide sequence that binds to the immobilized macromolecule can be differentially partitioned from those that do not present peptide sequences that bind to the predetermined macromolecule. The bacteriophage particles library members) which are bound to the immobilized macromolecule are 15 then recovered and replicated to amplify the selected bacteriophage subpopulation for a subsequent round of affinity enrichment and phage replication. After several rounds of affinity enrichment and phage replication, the bacteriophage library members that are thus selected are isolated and the nucleotide sequence encoding the displayed peptide sequence is determined, thereby identifying the sequence(s) of peptides that bind to the predetermined macromolecule receptor). Such methods are further described in PCT patent publication Nos.

91/17271, 91/18980, and 91/19818 and 93/08278.

The latter PCT publication describes a recombinant

DNA

method for the display of peptide ligands that involves the production of a library of fusion proteins with each fusion protein composed of a first polypeptide portion, typically comprising a variable sequence, that is available for potential binding to a predetermined macromolecule, and a second polypeptide portion that binds to DNA, such as the DNA vector encoding the individual fusion protein. When transformed host cells are cultured under conditions that allow for expression of the fusion protein, the fusion protein binds to the DNA vector encoding it. Upon lysis of the host cell, the fusion protein/vector DNA complexes can be screened against a predetermined macromolecule in much the same way as bacteriophage particles are screened in the phage-based display system, with the replication and sequencing of the DNA vectors in the selected fusion protein/vector DNA complexes serving as the basis for identification of the selected library peptide sequence(s) Other systems for generating libraries of peptides and like polymers have aspects of both the recombinant and in vitro chemical synthesis methods. In these hybrid methods, cell-free enzymatic machinery is employed to accomplish the in vitro synthesis of the library members peptides or polynucleotides). In one type of method, RNA molecules with the ability to bind a predetermined protein or a predetermined dye molecule were selected by alternate rounds of selection and PCR amplification (Tuerk and Gold (1990) Science 249: 505; Ellington and Szostak (1990) Nature 346: 818). A similar technique was used to identify DNA sequences which bind a predetermined human S 15 transcription factor (Thiesen and Bach (1990) Nucleic Acids Res.

o 18: 3203; Beaudry and Joyce (1992) Science 257; 635; PCT patent publication Nos. 92/05258 and 92/14843). In a similar fashion, the technique of in vitro translation has been used to synthesize o 0 proteins of interest and has been proposed as a method for o" 20 generating large libraries of peptides. These methods which rely upon in vitro translation, generally comprising stabilized polysome complexes, are described further in PCT patent "ov° publication Nos. 88/08453, 90/05785, 90/07003, 91/02076, 91/05058, and 92/02536. Applicants have described methods in which library members comprise a fusion protein having a first polypeptide portion with DNA binding activity and a second o polypeptide portion having the library member unique peptide sequence; such methods are suitable for use in cell-free in vitro selection formats, among others.

The displayed peptide sequences can be of varying lengths, typically from 3-5000 amino acids long or longer, frequently from 5-100 amino acids long, and often from about 8-15 amino acids long. A library can comprise library members having varying lengths of displayed peptide sequence, or may.comprise library members having a fixed length of displayed peptide sequence. Portions or all of the displayed peptide sequence(s) can be random, pseudorandom, defined set kernal, fixed, or the like. The present display methods include methods for in vitro and in vivo display of single-chain antibodies, such as nascent scFv on polysomes or scFv displayed on phage, which enable largescale screening of scFv libraries having broad diversity of variable region sequences and binding specificities.

The present invention also provides random, pseudorandom, and defined sequence framework peptide libraries and methods for generating and screening those libraries to identify useful compounds peptides, including single-chain antibodies) that bind to receptor molecules or epitopes of interest or gene products that modify peptides or RNA in a desired fashion. The random, pseudorandom, and defined sequence framework peptides are produced from libraries of peptide library members that comprise displayed peptides or displayed singlechain antibodies attached to a polynucleotide template from which 15 the displayed peptide was synthesized. The mode of attachment may vary according to the specific embodiment of the invention selected, and can include encapsidation in a phage particle or incorporation in a cell.

A method of affinity enrichment allows a very large 20 library of peptides and single-chain antibodies to be screened and the polynucleotide sequence encoding the desired peptide(s) or single-chain antibodies to be selected. The polynucleotide can then be isolated and shuffled to recombine combinatorially the amino acid sequence of the selected peptide(s) (or 2 predetermined portions thereof) or single-chain antibodies (or just VH, VL, or CDR portions thereof). Using these methods, one can identify a peptide or single-chain antibody as having a desired binding affinity for a molecule and can exploit the process of shuffling to converge rapidly to a desired highaffinity peptide or scFv. The peptide or antibody-can then be synthesized in bulk by conventional means for any suitable use as a therapeutic or diagnostic agent).

A significant advantage of the present invention is that no prior information regarding an expected ligand structure is required to isolate peptide ligands or antibodies of interest.

The peptide identified can have biological activity, which is meant to include at least specific binding affinity for a selected receptor molecule and, in some instances, will further 44 include the ability to block the binding of other compounds, to stimulate or inhibit metabolic pathways, to act as a signal or messenger, to stimulate or inhibit cellular activity, and the like.

The present invention also provides a method for shuffling a pool of polynucleotide sequences selected by affinity screening a library of polysomes displaying nascent peptides (including single-chain antibodies) for library members which bind to a predetermined receptor a mammalian proteinaceous receptor such as, for example, a peptidergic hormone receptor, a cell surface receptor, an intracellular protein which binds to other protein(s) to form intracellular protein complexes such as heterodimers and the like) or epitope an immobilized protein, glycoprotein, oligosaccharide, and the like).

15 Polynucleotide sequences selected in a first selection round (typically by affinity selection for binding to a receptor a ligand) by any of these methods are pooled and the pool(s) is/are shuffled by in vitro and/or in vivo recombination to produce a shuffled pool comprising a population of recombined selected polynucleotide sequences. The recombined selected polynucleotide sequences are subjected to at least one subsequent selection round. The polynucleotide sequences selected in the subsequent selection round(s) can be used directly, sequenced, and/or subjected to one or more additional rounds of shuffling and subsequent selection. Selected sequences can also be backcrossed with polynucleotide sequences encoding neutral sequences having insubstantial functional effect on binding), such as for example by backcrossing with a wild-type or naturally-occurring sequence substantially identical to a selected sequence to produce native-like functional *peptides, which may be less immunogenic. Generally, during backcrossing subsequent selection is applied to retain the property of binding to the predetermined receptor (ligand).

Prior to or concomitant with the shuffling of selected sequences, the sequences can be mutagenized. In one embodiment, selected library members are cloned in a prokaryotic vector plasmid, phagemid, or bacteriophage) wherein a collection of individual colonies (or plaques) representing discrete library members are produced. Individual selected library members can then be manipulated by site-directed mutagenesis, cassette mutagenesis, chemical mutagenesis, PCR mutagenesis, and the like) to generate a collection of library members representing a kernal of sequence diversity based on the sequence of the selected library member. The sequence of an individual selected library member or pool can be manipulated to incorporate random mutation, pseudorandom mutation, defined kernal mutation comprising variant and invariant residue positions and/or comprising variant residue positions which can comprise a residue selected from a defined subset of amino acid residues), codon-based mutation, and the like, either segmentally or over the entire length of the individual selected library member sequence. The mutagenized selected library members are then shuffled by in vitro and/or in vivo recombinatorial shuffling as disclosed herein.

The invention also provides peptide libraries comprising a plurality of individual library members of the invention, wherein each individual library member of said plurality comprises a sequence produced by shuffling of a pool of selected sequences, and each individual library member comprises a variable peptide segment sequence or single-chain antibody segment sequence which is distinct from the variablepeptide segment sequences or single-chain antibody sequences of other individual library members in said plurality (although some library members may be present in more than one copy per library due to uneven amplification, stochastic probability, or the *like).

The invention also provides a product-by-process, wherein selected polynucleotide sequences having (or encoding a peptide having) a predetermined binding specificity -are formed by the process of: screening a displayed peptide or displayed single-chain antibody library against a predetermined receptor ligand) or epitope antigen macromolecule) and identifying and/or enriching library members which bind to the predetermined receptor or epitope to produce a pool of selected library members, shuffling by recombination the selected library members (or amplified or cloned copies thereof) which binds the predetermined epitope and has been thereby isolated 46 and/or enriched from the library to generate a shuffled library, and screening the shuffled library against the predetermined receptor ligand) or epitope antigen macromolecule) and identifying and/or enriching shuffled library members which bind to the predetermined receptor or epitope to produce a pool of selected shuffled library members.

Antibody Display and Screening Methods The present method can be used to shuffle, by in vitro and/or in vivo recombination by any of the disclosed methods, and in any combination, polynucleotide sequences selected by antibody display methods, wherein an associated polynucleotide encodes a displayed antibody which is screened for a phenotype for affinity for binding a predetermined antigen (ligand).

15 Various molecular genetic approaches have been devised to capture the vast immunological repertoire represented by the extremely large number of distinct variable regions which can be present in immunoglobulin chains. The naturally-occurring germline immunoglobulin heavy chain locus is composed of separate 20 tandem arrays of variable segment genes located upstream of a tandem array of diversity segment genes, which are *e *themselves located upstream of a tandem array of joining (J) region genes, which are located upstream of the constant (CH) region genes. During B lymphocyte development, V-D-J rearrangement occurs wherein a heavy chain variable region gene (VH) is formed by rearrangement to form a fused D-J segment followed by rearrangement with a V segment to form a V-D-J joined product gene which, if productively rearranged, encodes a functional variable region (VH) of a heavy chain. Similarly, light chain loci rearrange one of several V segments with one of several J segments to form a gene encoding the variable region (VL) of a light chain.

The vast repertoire of variable regions possible in immunoglobulins derives in part from the numerous combinatorial possibilities of joining V and J segments (and, in the case of heavy chain loci, D segments) during rearrangement in B cell development. Additional sequence diversity in the* heavy chain variable regions arises from non-uniform rearrangements of the D 47 segments during V-D-J joining and from N region addition.

Further, antigen-selection of specific B cell clones selects for higher affinity variants having nongermline mutations in one or both of the heavy and light chain variable regions; a phenomenon referred to as "affinity maturation" or "affinity sharpening".

Typically, these "affinity sharpening" mutations cluster in specific areas of the variable region, most commonly in the complementarity-determining regions (CDRs).

In order to overcome many of the limitations in producing and identifying high-affinity immunoglobulins through antigen-stimulated B cell development immunization), various prokaryotic expression systems have been developed that can be manipulated to produce combinatorial antibody libraries which may be screened for high-affinity antibodies to specific 15 antigens. Recent advances in the expression of antibodies in Escherichia coli and bacteriophage systems (see, "Alternative Peptide Display Methods", infra) have raised the possibility that virtually any specificity can be obtained by either cloning antibody genes from characterized hybridomas or by de novo S 20 selection using antibody gene libraries from Ig cDNA).

Combinatorial libraries of antibodies have been generated in bacteriophage lambda expression systems which may be screened as bacteriophage plaques or as colonies of lysogens (Huse et al. (1989) Science 246: 1275; Caton and Koprowski (1990) 2 Proc. Natl. Acad. Sci. 87: 6450; Mullinax et al (1990) Proc. Natl. Acad. Sci. 87: 8095; Persson et al. (1991) Proc. Natl. Acad. Sci. 88: 2432). Various embodiments of bacteriophage antibody display libraries and lambda phage expression libraries have been described (Kang et al. (1991) Proc. Natl. Acad. Sci. 88: 4363; Clackson et a-1. (1991) Nature 352: 624; McCafferty et al. (1990) Nature 348: 552; Burton et al. (1991) Proc. Natl. Acad. Sci. 88: 10134; Hoogenboom et al. (1991) Nucleic Acids Res. 19: 4133; Chang et al. (1991) J. Immunol. 147: 3610; Breitling et al. (1991) Gene 104: 147; Marks et al. (1991) J. Mol. Biol. 222: 581; Barbas et al. (1992) Proc. Natl. Acad. Sci. 89: 4457; Hawkins and Winter (1992) J. Immunol. 22: 867; Marks et al. (1992) Biotechnoloqy 10: 779; Marks et al. (1992) J. Biol. Chem. 267: 16007; Lowman et al (1991) Biochemistry 30: 10832; Lerner et al.

(1992) Science 258: 1313, incorporated herein by reference).

Typically, a bacteriophage antibody display library is screened with a receptor polypeptide, carbohydrate, glycoprotein, nucleic acid) that is immobilized by covalent linkage to a chromatography resin to enrich for reactive phage by affinity chromatography) and/or labeled to screen plaque or colony lifts).

One particularly advantageous approach has been the use of so-called single-chain fragment variable (scFv) libraries (Marks et al. (1992) Biotechnology 10: 779; Winter G and Milstein C (1991) Nature 349: 293; Clackson et al. (1991) op.cit.; Marks et al. (1991) J. Mol. Biol. 222: 581; Chaudhary et al. (1990) Proc. Natl. Acad. Sci. (USA) 87: 1066; Chiswell et al. (1992) 15 TIBTECH 10: 80; McCafferty et al. (1990) op.cit.; and Huston et al. (1988) Proc. Natl. Acad. Sci. (USA) 85: 5879). Various embodiments of scFv libraries displayed on bacteriophage coat proteins have been described.

Beginning in 1988, single-chain analogues of Fv 20 fragments and their fusion proteins have been reliably generated by antibody engineering methods. The first step generally involves obtaining the genes encoding VH and VL domains with desired binding properties; these V genes may be isolated from a specific hybridoma cell line, selected from a combinatorial V-gene library, or made by V gene synthesis. The single-chain Fv is formed by connecting the component V genes with an oligonucleotide that encodes an appropriately designed linker peptide, such as (Gly-Gly-Gly-Gly-Ser) 3 or equivalent linker peptide(s). The linker bridges the C-terminus of the first V region and N-terminus of the second, ordered- as either VH-linker-VL or VL-linker-V

H

In principle, the scFv binding site can faithfully replicate both the affinity and specificity of its parent antibody combining site.

Thus, scFv fragments are comprised of VH and VL domains linked into a single polypeptide chain by a flexible linker peptide. After the scFv genes are assembled, they are cloned into a phagemid and expressed at the tip of the M13 phage (or similar filamentous bacteriophage) as fusion proteins with the 49 bacteriophage pIII (gene 3) coat protein. Enriching for phage expressing an antibody of interest is accomplished by panning the recombinant phage displaying a population scFv for binding to a predetermined epitope target antigen, receptor).

The linked polynucleotide of a library member provides the basis for replication of the library member after a screening or selection procedure, and also provides the basis for the determination, by nucleotide sequencing, of the identity of the displayed peptide sequence or VH and VL amino acid sequence. The displayed peptide(s) or single-chain antibody scFv) and/or its VH and VL domains or their CDRs can be cloned and expressed in a suitable expression system. Often polynucleotides encoding the isolated VH and VL domains will be ligated to polynucleotides encoding constant regions (CH and CL) to form polynucleotides 15 encoding complete antibodies chimeric or fully-human), antibody fragments, and the like. Often polynucleotides encoding the isolated CDRs will be grafted into polynucleotides encoding a suitable variable region framework (and optionally constant regions) to form polynucleotides encoding complete antibodies humanized or fully-human), antibody fragments, and the like. Antibodies can be used to isolate preparative quantities *.e e of the antigen by immunoaffinity chromatography. Various other uses of such antibodies are to diagnose and/or stage disease neoplasia), and for therapeutic application to treat disease, such as for example: neoplasia, autoimmune disease, AIDS, cardiovascular disease, infections, and the like.

.*Various methods have been reported for increasing the combinatorial diversity of a scFv library to broaden the repertoire of binding species (idiotype spectrum). The use of PCR has permitted the variable regions to be rapidly cloned either from a specific hybridoma source or as a gene library from non-immunized cells, affording combinatorial diversity in the assortment of VH and VL cassettes which can be combined.

Furthermore, the VH and VL cassettes can themselves be diversified, such as by random, pseudorandom, or directed mutagenesis. Typically,

V

H and VL cassettes are diversified in or near the complementarity-determining regions (CDRs), often the third CDR, CDR3. Enzymatic inverse PCR mutagenesis has been shown to be a simple and reliable method for constructing relatively large libraries of scFv site-directed mutants (Stemmer et al. (1993) Biotechniques 14: 256), as has error-prone PCR and chemical mutagenesis (Deng et al. (1994) J. Biol. Chem. 269: 9533). Riechmann et al. (1993) Biochemistry 32: 8848 showed semirational design of an antibody scFv fragment using sitedirected randomization by degenerate oligonucleotide PCR and subsequent phage display of the resultant scFv mutants. Barbas et al. (1992) op.cit. attempted to circumvent the problem of limited repertoire sizes resulting from using biased variable region sequences by randomizing the sequence in a synthetic CDR region of a human tetanus toxoid-binding Fab.

CDR randomization has the potential to create approximately 1 x 1020 CDRs for the heavy chain CDR3 alone, and 15 a roughly similar number of variants of the heavy chain CDR1 and CDR2, and light chain CDR1-3 variants. Taken individually or together, the combinatorics of CDR randomization of heavy and/or light chains requires generating a prohibitive number of S: .bacteriophage clones to produce a clone library representing all possible combinations, the vast majority of which will be nonbinding. Generation of such large numbers of primary transformants is not feasible with current transformation technology and bacteriophage display systems. For example, Barbas et al. (1992) op.cit. only generated 5 x 107 transformants, 2 which represents only a tiny fraction of the potential diversity of a library of thoroughly randomized CDRs.

Despite these substantial limitations, bacteriophage display of scFv have already yielded a variety of useful antibodies and antibody fusion proteins. A bispecific single chain antibody has been shown to mediate efficient tumor cell lysis (Gruber et al. (1994) J. Immunol. 152:. 5368).

Intracellular expression of an anti-Rev scFv has been shown to inhibit HIV-1 virus replication in vitro (Duan et al. (1994) Proc. Natl. Acad. Sci. (USA) 91: 5075), and intracellular expression of an anti-p21 r a s scFv has been shown to inhibit meiotic maturation of Xenopus oocytes (Biocca et al. (1993) Biochem. Biophys. Res. Commun. 197: 422. Recombinant scFv which can be used to diagnose HIV infection have also been reported, 51 demonstrating the diagnostic utility of scFv (Lilley et al.

(1994) J. Immunol. Meth. 171: 211). Fusion proteins wherein an scFv is linked to a second polypeptide, such as a toxin or fibrinolytic activator protein, have also been reported (Holvost et al. (1992) Eur. J. Biochem. 210: 945; Nicholls et al. (1993) J. Biol. Chem. 268: 5302).

If it were possible to generate scFv libraries having broader antibody diversity and overcoming many of the limitations of conventional CDR mutagenesis and randomization methods which can cover only a very tiny fraction of the potential sequence combinations, the number and quality of scFv antibodies suitable for therapeutic and diagnostic use could be vastly improved. To address this, the in vitro and in vivo shuffling methods of the invention are used to recombine CDRs which have been obtained (typically via PCR amplification or cloning) from nucleic acids obtained from selected displayed antibodies. Such displayed antibodies can be displayed on cells, on bacteriophage particles, on polysomes, or any suitable antibody display system wherein the antibody is associated with its encoding nucleic acid(s). In a variation, the CDRs are initially obtained from mRNA (or cDNA) from antibody-producing cells plasma cells/splenocytes from an immunized wild-type mouse, a human, or a transgenic mouse capable of making a human antibody as in W092/0391 8 W093/12227, and W094/25585), including hybridomas derived therefrom.

Polynucleotide sequences selected in a first selection round (typically by affinity selection for displayed antibody binding to an antigen a ligand) by any of these methods are pooled and the pool(s) is/are shuffled by in vitro and/or in vivo recombination, especially shuffling of CDRs (typically shuffling heavy chain CDRs with other heavy chain CDRs and light chain CDRs with other light chain CDRs) to produce a shuffled pool comprising a population of recombined selected polynucleotide sequences. The recombined selected polynucleotide sequences are expressed in a selection format as a displayed antibody and subjected to at least one subsequent selection round. The polynucleotide sequences selected in the subsequent selection round(s) can be used directly, sequenced, and/or subjected to one or more additional rounds of shuffling and 52 subsequent selection until an antibody of the desired binding affinity is obtained. Selected sequences can also be backcrossed with polynucleotide sequences encoding neutral antibody framework sequences having insubstantial functional effect on antigen binding), such as for example by backcrossing with a human variable region framework to produce human-like sequence antibodies. Generally, during backcrossing subsequent selection is applied to retain the property of binding to the predetermined antigen.

Alternatively, or in combination with the noted variations, the valency of the target epitope may be varied to control the average binding affinity of selected scFv library members. The target epitope can be bound to a surface or substrate at varying densities, such as by including a competitor 15 epitope, by dilution, or by other method known to those in the art. A high density (valency) of predetermined epitope can be used to enrich for scFv library members which have relatively low o. affinity, whereas a low density (valency) can preferentially enrich for higher affinity scFv library members.

For generating diverse variable segments, a collection of synthetic oligonucleotides encoding random, pseudorandom, or a defined sequence kernal set of peptide sequences can be inserted by ligation into a predetermined site a CDR) Similarly, the sequence diversity of one or more CDRs of the 2 single-chain antibody cassette(s) can be expanded by mutating the CDR(s) with site-directed mutagenesis, CDR-replacement, and the like. The resultant DNA molecules can be propagated in a host for cloning and amplification prior to shuffling, or can be used directly may avoid loss of diversity which may occur upon propagation in a host cell) and the selected library members subsequently shuffled.

Displayed peptide/polynucleotide complexes (library members) which encode a variable segment peptide sequence of interest or a single-chain antibody of interest are selected from the library by an affinity enrichment technique. This is accomplished by means of a immobilized macromolecule or epitope specific for the peptide sequence of interest, such as a receptor, other macromolecule, or other epitope species.

Repeating the affinity selection procedure provides an enrichment of library members encoding the desired sequences, which may then be isolated for pooling and shuffling, for sequencing, and/or for further propagation and affinity enrichment.

The library members without the desired specificity are removed by washing. The degree and stringency of washing required will be determined for each peptide sequence or singlechain antibody of interest and the immobilized predetermined macromolecule or epitope. A certain degree of control can be exerted over the binding characteristics of the nascent peptide/DNA complexes recovered by adjusting the conditions of the binding incubation and the subsequent washing. The temperature, pH, ionic strength, divalent cations concentration, .and the volume and duration of the washing will select for 15 nascent peptide/DNA complexes within particular ranges of affinity for the immobilized macromolecule. Selection based on slow dissociation rate, which is usually predictive of high affinity, is often the most practical route. This may be done either by continued incubation in the presence of a saturating S 20 amount of free predetermined macromolecule, or by increasing the volume, number, and length of the washes. In each case, the rebinding of dissociated nascent peptide/DNA or peptide/RNA complex is prevented, and with increasing time, nascent .peptide/DNA or peptide/RNA complexes of higher and higher affinity are recovered.

.Additional modifications of the binding and washing procedures may be applied to find peptides with special characteristics. The affinities of some peptides are dependent on ionic strength or cation concentration. This is a useful characteristic for peptides that will be used in affinity purification of various proteins when gentle conditions for removing the protein from the peptides are required.

One variation involves the use of multiple binding targets (multiple epitope species, multiple receptor species), such that a scFv library can be simultaneously screened for a multiplicity of scFv which have different binding specificities.

Given that the size of a scFv library often limits the diversity of potential scFv sequences, it is typically desirable to us scFv libraries of as large a size as possible. The time and economic considerations of generating a number of very large polysome scFv-display libraries can become prohibitive. To avoid this substantial problem, multiple predetermined epitope species (receptor species) can be concomitantly screened in a single library, or sequential screening against a number of epitope species can be used. In one variation, multiple target epitope species, each encoded on a separate bead (or subset of beads), can be mixed and incubated with a polysome-display scFv library under suitable binding conditions. The collection of beads, comprising multiple epitope species, can then be used to isolate, by affinity selection, scFv library members. Generally, subsequent affinity screening rounds can include the same mixture S: of beads, subsets thereof, or beads containing only one or two 15 individual epitope species. This approach affords efficient screening, and is compatible with laboratory automation, batch processing, and high throughput screening methods.

A variety of techniques can be used in the present invention to diversify a peptide library or single-chain antibody 20 library, or to diversify, prior to or concomitant with shuffling, around variable segment peptides or VH, VL, or CDRs found in early rounds of panning to have sufficient binding activity to the predetermined macromolecule or epitope. In one approach, the positive selected peptide/polynucleotide complexes (those Sidentified in an early round of affinity enrichment) are sequenced to determine the identity of the active peptides.

Oligonucleotides are then synthesized based on these active peptide sequences, employing a low level of all bases incorporated at each step to produce slight variations of the primary oligonucleotide sequences. This mixture of (slightly) degenerate oligonucleotides is then cloned into the variable segment sequences at the appropriate locations. This method produces systematic, controlled variations of the starting peptide sequences, which can then be shuffled. It requires, however, that individual positive nascent peptide/polynucleotide complexes be sequenced before mutagenesis, and thus is useful for expanding the diversity of small numbers of recovered complexes and selecting variants having higher binding affinity and/or higher binding specificity. In a variation, mutagenic PCR amplification of positive selected peptide/polynucleotide complexes (especially of the variable region sequences, the amplification products of which are shuffled in vitro and/or in vivo and one or more additional rounds of screening is done prior to sequencing. The same general approach can be employed with single-chain antibodies in order to expand the diversity and enhance the binding affinity/specificity, typically by diversifying CDRs or adjacent framework regions prior to or concomitant with shuffling. If desired, shuffling reactions can be spiked with mutagenic oligonucleotides capable of in vitro recombination with the selected library members can be included.

Thus, mixtures of synthetic oligonucleotides and PCR fragments (synthesized by error-prone or high-fidelity methods) can be 15 added to the in vitro shuffling mix and be incorporated into resulting shuffled library members (shufflants).

The present invention of shuffling enables the generation of a vast library of CDR-variant single-chain antibodies. One way to generate such antibodies is to insert 20 synthetic CDRs into the single-chain antibody and/or CDR randomization prior to or concomitant with shuffling. The sequences of the synthetic CDR cassettes are selected by_ referring to known sequence data of human CDR and are selected in the discretion of the practitioner according to the following guidelines: synthetic CDRs will have at least 40 percent positional sequence identity to known CDR sequences, and preferably will have at least 50 to 70 percent positional sequence identity to known CDR sequences. For example, a collection of synthetic CDR sequences can be generated by synthesizing a collection of oligonucleotide sequences on the basis of naturally-occurring human CDR sequences listed in Kabat et al. (1991) op.cit.; the pool(s) of synthetic CDR sequences are calculated to encode CDR peptide sequences having at least percent sequence identity to at least one known naturallyoccurring human CDR sequence. Alternatively, a collection of naturally-occurring CDR sequences may be compared to generate consensus sequences so that -amino acids used at a residue position frequently in at least 5 percent of known CDR sequences) are incorporated into the synthetic CDRs at the corresponding position(s). Typically, several 3 to about known CDR sequences are compared and observed natural sequence variations between the known CDRs are tabulated, and a collection of oligonucleotides encoding CDR peptide sequences encompassing all or most permutations of the observed natural sequence variations is synthesized. For example but not for limitation, if a collection of human VH CDR sequences have carboxy-terminal amino acids which are either Tyr, Val, Phe, or Asp, then the pool(s) of synthetic CDR oligonucleotide sequences are designed to allow the carboxy-terminal CDR residue to be any of these amino acids. In some embodiments, residues other than those which naturally-occur at a residue position in the collection of CDR sequences are incorporated: conservative amino S 15 acid. substitutions are frequently incorporated and up to residue positions may be varied to incorporate non-conservative amino acid substitutions as compared to known naturally-occurring CDR sequences. Such CDR sequences can be used in primary 0 library members (prior to first round screening) and/or can be S 20 used to spike in vitro shuffling reactions of selected library *o o:member sequences. Construction of such pools of defined and/or 00.0degenerate sequences will be readily accomplished by those of ordinary skill in the art.

S• The collection of synthetic CDR sequences comprises at S least one member that is not known to be a naturally-occurring CDR sequence. It is within the discretion of the practitioner to .:include or not include a portion of random or pseudorandom sequence corresponding to N region addition in the heavy chain CDR; the N region sequence ranges from 1 nucleotide to about 4 nucleotides occurring at V-D and D-J junctions. A collection of synthetic heavy chain CDR sequences comprises at least about 100 unique CDR sequences, typically at least about 1,000 unique CDR sequences, preferably at least about 10,000 unique CDR sequences, frequently more than 50,000 unique CDR sequences; however, usually not more than about 1 x 106 unique CDR sequences are included in the collection, although occasionally 1 x 107 to 1 x 108 unique CDR sequences are present, especially if conservative amino acid substitutions are permitted at positions where the conservative amino acid substituent is not present or is rare less than 0.1 percent) in that position in naturallyoccurring human CDRs. In general, the number of unique CDR sequences included in a library should not exceed the expected number of primary transformants in the library by more than a factor of 10. Such single-chain antibodies generally bind to a predetermined antigen the immunogen) with an affinity of about at least 1 x 107 M preferably with an affinity of about at least 5 x 107 more preferably with an affinity of at least 1 x 108 M- 1 to 1 x 109 M- 1 or more, sometimes up to 1 x 101 0 M- or more.. Frequently, the predetermined antigen is a human protein, such as for example a human cell surface antigen CD4, CD8, IL-2 receptor, EGF receptor, PDGF receptor), other human biological macromolecule thrombomodulin, 15 protein C, carbohydrate antigen, sialyl Lewis antigen, Lselectin), or nonhuman disease associated macromolecule bacterial LPS, virion capsid protein or envelope glycoprotein) :and the like.

High affinity single-chain antibodies of the desired specificity can be engineered and expressed in a variety of systems. For example, scFv have been produced in plants (Firek eo** et al. (1993) Plant Mol. Biol. 23: 861) and can be readily made in prokaryotic systems (Owens RJ and Young RJ (1994) J. Immunol.

S* Meth. 168: 149; Johnson S and Bird RE (1991) Methods Enzymol.

203: 88). Furthermore, the single-chain antibodies can be used *as a basis for constructing whole antibodies or various fragments thereof (Kettleborough et al. (1994) Eur. J. Immunol. 24: 952).

The variable region encoding sequence may be isolated by PCR amplification or subcloning) and spliced to a sequence encoding a desired human constant region to encode a human sequence antibody more suitable for human therapeutic uses where immunogenicity is preferably minimized. The polynucleotide(s) having the resultant fully human encoding sequence(s) can be expressed in a host cell from an expression vector in a mammalian cell) and purified for pharmaceutical formulation.

The DNA expression constructs will typically include an expression control DNA sequence operably linked to the coding sequences, including naturally-associated or heterologous 58 promoter regions. Preferably, the expression control sequences will be eukaryotic promoter systems in vectors capable of transforming or transfecting eukaryotic host cells. Once the vector has been incorporated into the appropriate host, the host is maintained under conditions suitable for high level expression of the nucleotide sequences, and the collection and purification of the mutant "engineered" antibodies.

As stated previously, the DNA sequences will be expressed in hosts after the sequences have been operably linked to an expression control sequence positioned to ensure the transcription and translation of the structural gene). These expression vectors are typically replicable in the host organisms either as episomes or as an integral part of the host chromosomal DNA. Commonly, expression vectors will contain selection markers, tetracycline or neomycin, to permit detection of those cells transformed with the desired DNA sequences (see, U.S. Patent 4,704,362, which is incorporated herein by reference).

In addition to eukaryotic microorganisms such as yeast, 20 mammalian tissue cell culture may also be used to produce the polypeptides of the present invention (see, Winnacker, "From Genes to Clones," VCH Publishers, N.Y. (1987), which is incorporated herein by reference). Eukaryotic cells are actually preferred, because a number of suitable host cell lines capable S of secreting intact immunoglobulins have been developed in the S art, and include the CHO cell lines, various COS cell lines, HeLa cells, myeloma cell lines, etc, but preferably transformed Bcells or hybridomas. Expression vectors for these cells can include expression control sequences, such as an origin of replication, a promoter, an enhancer (Queen et- al. (1986) Immunol. Rev. 89: 49), and necessary processing information sites, such as ribosome binding sites, RNA splice sites, polyadenylation sites, and transcriptional terminator sequences.

Preferred expression control sequences are promoters derived from immunoglobulin genes, cytomegalovirus, SV40, Adenovirus, Bovine Papilloma Virus, and the like.

Eukaryotic DNA transcription can be increased by inserting an enhancer sequence into the vector. Enhancers are cis-acting sequences of between 10 to 300bp that increase transcription by a promoter. Enhancers can effectively increase transcription when either 5' or 3' to the transcription unit.

They are also effective if located within an intron or within the coding sequence itself. Typically, viral enhancers are used, including SV40 enhancers, cytomegalovirus enhancers, polyoma enhancers, and adenovirus enhancers. Enhancer sequences from mammalian systems are also commonly used, such as the mouse immunoglobulin heavy chain enhancer.

Mammalian expression vector systems will also typically include a selectable marker gene. Examples of suitable markers include, the dihydrofolate reductase gene (DHFR), the thymidine kinase gene or prokaryotic genes conferring drug resistance. The first two marker genes prefer the use of mutant 15 cell lines that lack the ability to grow without the addition of thymidine to the growth medium. Transformed cells can then be oeooo identified by their ability to grow on non-supplemented media.

Examples of prokaryotic drug resistance genes useful as markers include genes conferring resistance to G418, mycophenolic acid 20 and hygromycin.

The vectors containing the DNA segments of interest can be transferred into the host cell by well-known methods, depending on the type of cellular host. For example, calcium chloride transfection is commonly utilized for prokaryotic cells, whereas calcium phosphate treatment. lipofection, or electroporation may be used for other cellular hosts. Other methods used to transform mammalian cells include the use of Polybrene, protoplast fusion, liposomes, electroporation, and microinjection (see, generally, Sambrook et al., supra) Once expressed, the antibodies, individual mutated immunoglobulin chains, mutated antibody fragments, and other immunoglobulin polypeptides of the invention can be purified according to standard procedures of the art, including ammonium sulfate precipitation, fraction column chromatography, gel electrophoresis and the like (see, generally, Scopes, Protein Purification, Springer-Verlag, N.Y. (1982)). Once purified, partially or to homogeneity as desired, the polypeptides may then be used therapeutically or in developing and performing assay procedures, immunofluorescent stainings, and the like (see generally, Immunological Methods, Vols. I and II, Eds. Lefkovits and Pernis, Academic Press, New York, N.Y. (1979 and 1981)).

The antibodies generated by the method of the present invention can be used for diagnosis and therapy. By way of illustration and not limitation, they can be used to treat cancer, autoimmune diseases, or viral infections. For treatment of cancer, the antibodies will typically bind to an antigen expressed preferentially on cancer cells, such as erbB-2, CEA, CD33, and many other antigens and binding members well known to those skilled in the art.

Yeast Two-Hybrid Screening Assays Shuffling can also be used to recombinatorially 15 diversify a pool of selected library members obtained by screening a two-hybrid screening system to identify library members which bind a predetermined polypeptide sequence. The selected library members are pooled and shuffled by in vitro S" and/or in vivo recombination. The shuffled pool can then be screened in a yeast two hybrid system to select library members which bind said predetermined polypeptide sequence and SH2 domain) or which bind an alternate predetermined polypeptide sequence an SH2 domain from another protein species) An approach to identifying polypeptide sequences which bind to a predetermined polypeptide sequence has been to use a so-called "two-hybrid" system wherein the predetermined polypeptide sequence is present in a fusion protein (Chien et al.

(1991) Proc. Natl. Acad. Sci. (USA) 88: 9578). This approach identifies protein-protein interactions in vivo through reconstitution of a transcriptional activator (Fields S and Song O (1989) Nature 340: 245), the yeast Gal4 transcription protein.

Typically, the method is based on the properties of the yeast Gal4 protein, which consists of separable domains responsible for DNA-binding and transcriptional activation. Polynucleotides encoding two hybrid proteins, one consisting of the yeast Gal4 DNA-binding domain fused to a polypeptide sequence of a known protein and the other consisting of the Gal4 activation domain fused to a polypeptide sequence of a second protein, are 61 constructed and introduced into a yeast host cell.

Intermolecular binding between the two fusion proteins reconstitutes the Gal4 DNA-binding domain with the Gal4 activation domain, which leads to the transcriptional activation of a reporter gene lacZ, HIS3) which is operably linked to a Gal4 binding site. Typically, the two-hybrid method is used to identify novel polypeptide sequences which interact with a known protein (Silver SC and Hunt SW (1993) Mol. Biol. Rep. 17: 155; Durfee et al. (1993) Genes Devel. 7; 555; Yang et al. (1992) Science 257: 680; Luban et al. (1993) Cell 73: 1067; Hardy et al.

(1992) Genes Devel. 6; 801; Bartel et al. (1993) Biotechniques 14: 920; and Vojtek et al. (1993) Cell 74: 205). However, variations of the two-hybrid method have been used to identify mutations of a known protein that affect its binding to a second 15 known protein (Li B and Fields S (1993) FASEB J. 7: 957; Lalo et al. (1993) Proc. Natl. Acad. Sci. (USA) 90: 5524; Jackson et al.

(1993) Mol. Cell. Biol. 13; 2899; and Madura et al. (1993) J.

Biol. Chem. 268: 12046). Two-hybrid systems have also been used to identify interacting structural domains of two known proteins 20 (Bardwell et al. (1993) med. Microbiol. 8: 1177; Chakraborty et S0.. al. (1992) J. Biol. Chem. 267: 17498; Staudinger et al. (1993) J.

Biol. Chem. 268: 4608; and Milne GT and Weaver DT (1993) Genes Devel. 7; 1755) or domains responsible for oligomerization of a single protein (Iwabuchi et al. (1993) Oncoqene 8; 1693; Bogerd et al. (1993) J. Virol. 67: 5030). Variations of two-hybrid systems have been used to study the in vivo activity of a proteolytic enzyme (Dasmahapatra et al. (1992) Proc. Natl. Acad.

Sci. (USA) 89: 4159) Alternatively, an E. coli/BCCP interactive screening system (Germino et al. (1993) Proc. Natl. Acad. Sci.

90: 933; Guarente L (1993) Proc. Natl.- Acad. Sci.

90: 1639) can be used to identify interacting protein sequences protein sequences which heterodimerize or form higher order heteromultimers). Sequences selected by a twohybrid system can be pooled and shuffled and introduced into a two-hybrid system for one or more subsequent rounds of screening to identify polypeptide sequences which bind to the hybrid containing the predetermined binding sequence. The sequences thus identified can be compared to identify consensus sequence(s) 62 and consensus sequence kernals.

As can be appreciated from the disclosure above, the present invention has a wide variety of applications. Accordingly, the following examples are offered by way of illustration, not by way of limitation.

In the examples below, the following abbreviations have the following meanings. If not defined below, then the abbreviations have their art recognized meanings.

ml milliliter l4 microliters MM micromolar nM nanomolar PBS phosphate buffered saline ng nanograms Ag micrograms IPTG isopropylthio-3-D-galactoside bp basepairs kb kilobasepairs dNTP deoxynucleoside triphosphates 20 PCR polymerase chain reaction X-gal 5-bromo-4-chloro-3-indolyl-3-D-galactoside DNAseI deoxyribonuclease PBS phosphate buffered saline CDR complementarity determining regions MIC minimum inhibitory concentration scFv single-chain Fv fragment of an antibody In general, standard techniques of recombination DNA technology are described in various publications, e.g. Sambrook et al., 1989, Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory; Ausubel et al., 1987, Current Protocols in Molecular Biology, vols. 1 and 2 and supplements, and Berger and Kimmel, Methods in Enzymoloqy, Volume 152, Guide to Molecular Cloning Techniques (1987), Academic Press, Inc., San Diego, CA, each of which is incorporated herein in their entirety by reference. Restriction enzymes and polynucleotide modifying enzymes were used according to the manufacturers recommendations.

Oligonucleotides were synthesized on an Applied Biosystems Inc.

Model 394 DNA synthesizer using ABI chemicals. If desired, PCR amplimers for amplifying a predetermined DNA sequence may be selected at the discretion of the practitioner.

EXAMPLES

Example 1. LacZ alpha gene reassembly 63 1) Substrate preparation The substrate for the reassembly reaction was the dsDNA polymerase chain reaction product of the wild-type LacZ alpha gene from pUC18. (Fig. 2) (28; Gene Bank No. XO2514) The primer sequences were 5'AAAGCGTCGATTTTTGTGAT3' (SEQ ID NO:1) and 5'ATGGGGTTCCGCGCACATTT3' (SEQ ID NO:2). The free primers were removed from the PCR product by Wizard PCR prep (Promega, Madison WI) according to the manufacturer's directions. The removal of the free primers was found to be important.

2) DNAseI digestion About 5 gg of the DNA substrate was digested with 0.15 units of DNAsel (Sigma, St. Louis MO) in 100 Al of [50 mM Tris-HC1 pH 7.4, 1 mM MgCl 2 for 10-20 minutes at room temperature. The digested DNA was run on a 2% low melting point agarose gel.

15 Fragments of 10-70 basepairs (bp) were purified from the 2% low melting point agarose gels by electrophoresis onto DE81 ion exchange paper (Whatman, Hillsborough OR). The DNA fragments were eluted from the paper with 1 M NaCl and ethanol precipitated.

20 3) DNA Reassembly The purified fragments were resuspended at a concentration •O°0 of 10 30 ng/Al in PCR Mix (0.2 mM each dNTP, 2.2 mM MgCl 2 mM KC1, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100, 0.3 ~l Taq DNA polymerase, 50 il total volume). No primers were added at this Spoint. A reassembly program of 94 0 C for 60 seconds, 30-45 cycles of [94 0 C for 30 seconds, 50-55 0 C for 30 seconds, 72 0 C for seconds] and 5 minutes at 72 0 C was used in an MJ Research (Watertown MA) PTC-150 thermocycler. The PCR reassembly of small fragments into larger sequences was followed by taking samples of the reaction after 25, 30, 35 ,40 and 45 cycles of -reassembly (Fig. 2).

Whereas the reassembly of 100-200 bp fragments can yield a single PCR product of the correct size, 10-50 base fragments typically yield some product of the correct size, as well as products of heterogeneous molecular weights. Most of this size heterogeneity appears to be due to single-stranded sequences at the ends of the products, since after restriction enzyme digestion a single band of the correct size is obtained.

4) PCR with primers After dilution of the reassembly product into the PCR Mix with 0.8 iM of each of the above primers (SEQ ID Nos: 1 and 2) and about 15 cycles of PCR, each cycle consisting of [94 0 C for seconds, 50 0 C for 30 seconds and 72 0 C for 30 seconds], a single product of the correct size was obtained (Fig. 2).

Cloning and analysis The PCR product from step 4 above was digested with the terminal restriction enzymes BamHI and EcoO109 and gel purified as described above in step 2. The reassembled fragments were ligated into pUC18 digested with BamHI and EcoO109. E. coli were transformed with the ligation mixture under standard conditions as recommended by the manufacturer (Stratagene, San Diego CA) and S: plated on agar plates having 100 ig/ml ampicillin, 0.004% X-gal 15 and 2mM IPTG. The resulting colonies having the HinDIII-NheI fragment which is diagnostic for the recombinant were identified because they appeared blue.

This Example illustrates that a 1.0 kb sequence carrying the LacZ alpha gene can be digested into 10-70 bp fragments, and that 20 these gel purified 10-70 bp fragments can be reassembled to a single product of the correct size, such that 84% (N=377) of the resulting colonies are LacZ+ (versus 94% without shuffling; Fig.

2).

SThe DNA encoding the LacZ gene from the resulting LacZcolonies was sequenced with a sequencing kit (United States Biochemical Co., Cleveland OH) according to the manufacturer's SS*: instructions and the genes were found to have point mutations due to the reassembly process (Table 11/12 types of substitutions were found, and no frameshifts.

TABLE 1 Mutations introduced by mutagenic shuffling Transitions Frequency Transversions Frequency G- A 6 A -T 1 A G 4 A-C 2 C T 7 C A 1 T C 3 C G 0 G C 3 G-T 2 T A 1 T -G 2 A total of 4,437 bases of shuffled lacZ DNA were sequenced.

The rate of point mutagenesis during DNA reassembly from bp pieces was determined from DNA sequencing to be 0.7 (N=4,473), which is similar to error-prone PCR. Without being limited to any theory it is believed that the rate of point mutagenesis may be lower if larger fragments are used for the Sreassembly, or if a proofreading polymerase is added.

When plasmid DNA from 14 of these point-mutated LacZcolonies were combined and again reassembled/shuffled by the method described above, 34% (N=291) of the resulting colonies 25 were LacZ and these colonies presumably arose by recombination of the DNA from different colonies.

The expected rate of reversal of a single point mutation by error-prone PCR, assuming a mutagenesis rate of 0.7% would be expected to be Thus large DNA sequences can be reassembled from a random S* mixture of small fragments by a reaction that is surprisingly efficient and simple. One application of this technique is the recombination or shuffling of related sequences based on homology.

Example 2. LacZ gene and whole plasmid DNA shuffling 1) LacZ gene shuffling Crossover between two markers separated by 75 bases was measured using two LacZ gene constructs. Stop codons were inserted in two separate areas of the LacZ alpha gene to serve as negative markers. Each marker is a 25 bp non-homologous sequence with four stop codons, of which two are in the LacZ gene reading frame. The 25 bp non-homologous sequence is indicated in Figure 3 by a large box. The stop codons are either boxed or underlined. A 1:1 mixture of the two 1.0 kb LacZ templates containing the and versions of the LacZ alpha gene (Fig. 3) was digested with DNAseI and 100-200 bp fragments were purified as described in Example 1. The shuffling program was conducted under conditions similar to those described for reassembly in Example 1 except 0.5 pl of polymerase was added and the total volume was 100 l.

After cloning, the number of blue colonies obtained was 24%; (N=386) which is close to the theoretical maximum number of blue colonies indicating that recombination between the two markers was complete. All of the 10 blue colonies contained the expected HindIII-NheI restriction fragment.

2) Whole plasmid DNA shuffling Whole 2.7 kb plasmids (pUC18-+ and pUC18+-) were also tested. A 1:1 mixture of the two 2.9 kb plasmids containing the and versions of the LacZ alpha gene (Fig. 3) was digested with DNAsel and 100-200 bp fragments were purified as described in Example 1. The shuffling program was conducted under 20 conditions similar to those described for reassembly in step (1) above except the program was for 60 cycles [94 0 C for 30 seconds, 0 C for 30 seconds, 72 0 C for 30 seconds]. Gel analysis showed that after the shuffling program most of the product was greater than 20 kb. Thus, whole 2.7 kb plasmids (pUC18 and pUC18 were efficiently reassembled from random 100-200 bp fragments without added primers.

After digestion with a restriction enzyme having a unique o. site on the plasmid (EcoO109), most of the product consisted of a single band of the expected size. This band was gel purified, S" 30 religated and the DNA used to transform E. coli. The transformants were plated on 0.004% X-gal plates as described in Example 1. 11% 328) of the resulting plasmids were blue and thus recombinants.

3) Spiked DNA Shuffling Oligonucleotides that are mixed into the shuffling mixture can be incorporated into the final product based on the homology of the flanking sequences of the oligonucleotide to the template DNA (Fig. The LacZ- stop codon mutant (pUC18 described above was used as the DNAseI digested template. A oligonucleotide, including 18 bases of homology to the w LacZ gene at both ends was added into the reaction at molar excess to correct stop codon mutations present original gene. The shuffling reaction was conducte conditions similar to those in step 2 above. The re product was digested, ligated and inserted into E.

described above.

Table 2 blue colonies Control 0.0 (N>1000) Top strand spike 8.0 (N=855) Bottom strand spike 9.3 (N=620) Top and bottom strand spike 2.1 (N=537) ssDNA appeared to be more efficient than dsDNA, pre 20 due to competitive hybridization. The degree of incorp can be varied over a wide range by adjusting the molar annealing temperature, or the length of homology.

Example 3. DNA reassembly in the complete absence of prii Plasmid pUC18 was digested with restriction enzymes EcoO109, XmnI and AlwNI, yielding fragments of approximate 460, 770 and 1080 bp. These fragments were electrophore separately purified from a 2% low melting point agarose g 370 and 460 basepair bands could not be separated), yie large fragment, a medium fragment and a mixture of twc fragments in 3 separate tubes.

Each fragment was digested with DNAseI as descri Example 1, and fragments of 50-130 bp were purified from a melting point agarose gel for each of the original fragme PCR mix (as described in Example 1 above) was added purified digested fragments to a final concentration of 1, of fragments. No primers were added. A reassembly react.

performed for 75 cycles [94 0 C for 30 seconds, 60 0

C

seconds] separately on each of the three digested DNA fi mixtures, and the products were analyzed by agaros electrophoresis.

The results clearly showed that the 1080, 770 and 460 bp bands reformed efficiently from the purified fragments, demonstrating that shuffling does not require the use of any primers at all.

Example 4. IL-1B gene shuffling This example illustrates that crossovers based on homologies of less than 15 bases may be obtained. As an example, a human and a murine IL-1B gene were shuffled.

A murine IL1-B gene (BBG49) and a human IL1-8 gene with E.

coli codon usage (BBG2; R&D Systems, Inc., Minneapolis MN) were used as templates in the shuffling reaction. The areas of complete homology between the human and the murine IL-1B sequences are on average only 4.1 bases long (Fig. 5, regions of heterology are boxed).

Preparation of dsDNA PCR products for each of the genes, removal of primers, DNAseI digestion and purification of 10-50 bp fragments was similar to that described above in Example 1. The sequences of the primers used in the PCR reaction were 5'TTAGGCACCCCAGGCTTT3' (SEQ ID NO:3) and 5'ATGTGCTGCAAGGCGATT3' 20 (SEQ ID NO:4).

The first 15 cycles of the shuffling reaction were performed with the Klenow fragment of DNA polymerase I, adding 1 unit of fresh enzyme at each cycle. The DNA was added to the PCR mix of Example 1 which mix lacked the polymerase. The manual program was 94 0 C for 1 minute, and then 15 cycles of: [95 0 C for 1 minute, 1 0 seconds on dry ice/ethanol (until frozen), incubate about seconds at 25 0 C add 1U of Klenow fragment and incubate at 25 0

C

for 2 minutes]. In each cycle after the denaturation step, the tube was rapidly cooled in dry ice/ethanol and reheated to the S 30 annealing temperature. Then the heat-labile polymerase was added. The enzyme needs to be added at every cycle. Using this approach, a high level of crossovers was obtained, based on only a few bases of uninterrupted homology (Fig. 5, positions of cross-overs indicated by After these 15 manual cycles, Taq polymerase was added and an additional 22 cycles of the shuffling reaction [94 0 C for seconds, 35 0 C for 30 seconds] without primers were performed.

The reaction was then diluted 20-fold. The following primers were added to a final concentration of 0.8 MM: 5'AACGCCGCATGCAAGCTTGGATCCTTATT3' (SEQ ID NO:5) and 5'AAAGCCCTCTAGATGATTACGAATTCATAT3I (SEQ ID NO:6) and a PCR reaction was performed as described above in Example i. The second primer pair differed from the first pair only because a change in restriction sites was deemed necessary.

After digestion of the PCR product with XbaI and SphI, the fragments were ligated into XbaI-SphI-digested pUC18. The sequences of the inserts from several colonies were determined by a dideoxy DNA sequencing kit (United States Biochemical Co., Cleveland OH) according to the manufacturer's instructions.

A total of 17 crossovers were found by DNA sequencing of nine colonies. Some of the crossovers were based on only 1-2 bases of uninterrupted homology.

It was found that to force efficient crossovers based on short homologies, a very low effective annealing temperature is required. With any heat-stable polymerase, the cooling time of the PCR machine (94 0 C to 25 0 C at 1-2 degrees/second) causes the effective annealing temperature to be higher than the set annealing temperature. Thus, none of the protocols based on Taq polymerase yielded crossovers, even when a ten-fold excess of one of the IL1-8 genes was used. In contrast, a heat-labile polymerase, such as the Klenow fragment of DNA polymerase I, can be used to accurately obtain a low annealing temperature.

Example 5. DNA shuffling of the TEM-1 betalactamase gene The utility of mutagenic DNA shuffling for directed molecular evolution was tested in a betalactamase model system.

TEM-1 betalactamase is a very efficient enzyme, limited in its 30 reaction rate primarily by diffusion. This example determines whether it is possible to change its reaction specificity and obtain resistance to the drug cefotaxime that it normally does not hydrolyze.

The minimum inhibitory concentration (MIC) of cefotaxime on bacterial cells lacking a plasmid was determined by plating 10 Al of a 10- 2 dilution of an.overnight bacterial culture (about 1000 cfu) of E. coli XLl-blue cells (Stratagene, San Diego CA) on plates with varying levels of cefotaxime (Sigma, St. Louis MO), followed by incubation for 24 hours at 37 0

C.

Growth on cefotaxime is sensitive to the density of cells, and therefore similar numbers of cells needed to be plated on each plate (obtained by plating on plain LB plates). Platings of 1000 cells were consistently performed.

1) Initial Plasmid Construction A pUCl8 derivative carrying the bacterial TEM-1 betalactamase gene was used The TEM-1 betalactamase gene confers resistance to bacteria against approximately 0.02 pg/ml of cefotaxime. Sfil restriction sites were added 5' of the promoter and 3' of the end of the gene by PCR of the vector sequence with two primers: P r i m e r A (S E Q I D N 7) 5'TTCTATTGACGGCCTGTCAGGCCTCATATATACTTTAGATTGATTT3' and Primer B S E Q I D N 0 8 5'TTGACGCACTGGCCATGGTGGCCAAAAATAAACAAATAGGGGTTCCGCGCACATTT3' and by PCR of the betalactamase gene sequence with two other primers: Primer C (SEQ ID NO:9): 20 5'AACTGACCACGGCCTGACAGGCCGGTCTGACAGTTACCAATGCTT and Primer D (SEQ ID The two reaction products were digested with SfiI, mixed, ligated and used to transform bacteria.

The resulting plasmid was pUC182Sfi. This plasmid contains an Sfil fragment carrying the TEM-1 gene and the P-3 promoter.

The minimum inhibitory concentration of cefotaxime for E.

coli XLl-blue (Stratagene, San Diego CA) carrying this plasmid 0 was 0.02 g/ml after 24 hours at 37 0

C.

30 The ability to improve the resistance of the betalactamase gene to cefotaxime without shuffling was determined by stepwise replating of a diluted pool of cells (approximately 107 cfu) on 2-fold increasing drug levels. Resistance up to 1.28 Ag/ml could be obtained without shuffling. This represented a 64 fold increase in resistance.

2) DNAseI digestion The substrate for the first shuffling reaction was dsDNA of 0.9 kb obtained by PCR of pUC182Sfi with primers C and D, both of 71 which contain a SfiI site.

The free primers from the PCR product were removed by Wizard PCR prep (Promega, Madison WI) at every cycle.

About 5 Ag of the DNA substrate(s) was digested with 0.15 units of DNAseI (Sigma, St. Louis MO) in 100 ,l of 50 mM Tris-HCl pH 7.4, 1 mM MgCl 2 for 10 min at room temperature. Fragments of 100-300 bp were purified from 2% low melting point agarose gels by electrophoresis onto DE81 ion exchange paper (Whatman, Hillsborough OR), elution with 1 M NaCl and ethanol precipitation by the method described in Example 1.

3) Gene shuffling The purified fragments were resuspended in PCR mix (0.2 mM each dNTP, 2.2 mM MgCl 2 50 mM KC1, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100), at a concentration of 10 30 ng/gl. No primers were added at this point. A reassembly program of 94 0 C for seconds, then 40 cycles of [94 0 C for 30 seconds, 50-550C for seconds, 72 0 C for 30 seconds] and then 72 0 C for 5 minutes was used in an MJ Research (Watertown MA) PTC-150 thermocycler.

4) Amplification of Reassembly Product with primers 20 After dilution of the reassembly product into the PCR mix with 0.8 gM of each primer (C and D) and 20 PCR cycles [94 0 C for seconds, 50 0 C for 30 seconds, 72 0 C for 30 seconds] a single product 900 bp in size was obtained.

5) Cloning and analysis After digestion of the 900 bp product with the terminal restriction enzyme SfiI and agarose gel purification, the 900 bp product was ligated into the vector pUC182Sfi at the unique SfiI site with T4 DNA ligase (BRL, Gaithersburg MD). The mixture was electroporated into E. coli XLl-blue cells and plated on LB 30 plates with 0.32-0.64 Ag/ml of cefotaxime (Sigma, St..Louis

MO)

The cells were grown for up to 24 hours at 37 0 C and the resulting colonies were scraped off the plate as a pool and used as the PCR template for the next round of shuffling.

6) Subsequent Reassembly Rounds The transformants obtained after each of three rounds of shuffling were plated on increasing levels of cefotaxime. The colonies (>100, to maintain diversity) from the plate with the highest level of cefotaxime were pooled and used as the template for the PCR reaction for the next round.

A mixture of the cefotaximer colonies obtained at 0.32-0.64 ig/ml in Step above were used as the template for the .next round of shuffling. 10 ul of cells in LB broth were used as the template in a reassembly program of 10 minutes at 99 0 C, then cycles of [94°C for 30 seconds, 52 0 C for 30 seconds, 72 0 C for seconds] and then 5 minutes at 72 0 C as described above.

The reassembly products were digested and ligated into pUC182Sfi as described in step above. The mixture was electroporated into E. coli XLl-blue cells and plated on LB plates having 5-10 ig/ml of cefotaxime.

Colonies obtained at 5-10 ig/ml were used for a third round similar to the first and second rounds except the cells were plated on LB plates having 80-160 jg/ml of cefotaxime. After the third round, colonies were obtained at 80-160 ig/ml, and after replating on increasing concentrations of cefotaxime, colonies could be obtained at up to 320 jg/ml after 24 hours at 37 0

C

(MIC=320 jg/ml).

Growth on cefotaxime is dependent on the cell density, 20 requiring that all the MICs be standardized (in our case to about 1,000 cells per plate). At higher cell densities, growth at up to 1280 Ag/ml was obtained. The 5 largest colonies grown at 1,280 jig/ml were plated for single colonies twice, and the Sfil inserts were analyzed by restriction mapping of the colony PCR products.

2 One mutant was obtained with a 16,000 fold increased resistance to cefotaxime (MIC=0.02 ig/ml to MIC=320 jg/ml) After selection, the plasmid of selected clones was transferred back into wild-type E. coli XLl-blue cells 0 (Stratagene, San Diego CA) to ensure that none of the measured 30 drug resistance was due to chromosomal mutations.

Three cycles of shuffling and selection yielded a 1.6 x 104fold increase in the minimum inhibitory concentration of the extended broad spectrum antibiotic cefotaxime for the TEM-1 betalactamase. In contrast, repeated plating without shuffling resulted in only a 16-fold increase in resistance (error-prone PCR or cassette mutagenesis).

7) Sequence analysis All 5 of the largest colonies grown at 1,280 jg/ml had a restriction map identical to the wild-type TEM-1 enzyme. The SfiI insert of the plasmid obtained from one of these colonies was sequenced by dideoxy DNA sequencing (United States Biochemical Co., Cleveland OH) according to the manufacturer's instructions. All the base numbers correspond to the revised pBR322 sequence and the amino acid numbers correspond to the ABL standard numbering scheme The amino acids are designated by their three letter codes and the nucleotides by their one letter codes. The term G4205A means that nucleotide 4205 was changed from guanidine to adenine.

Nine single base substitutions were found. G4205A is located between the -35 and -10 sites of the betalactamase P3 promoter (31) .The promoter up-mutant observed by Chen and Clowes (31) is located outside of the Sfil fragment used here, and thus could not have been detected. Four mutations were silent (A3689G, G3713A, G3934A and T3959A), and four resulted in an amino acid change (C3448T resulting in Gly238Ser, A3615G S*resulting in Metl82Thr, C3850T resulting in Glul04Lys, and G4107A resulting in Alal8Val).

20 8) Molecular Backcross Molecular backcrossing with an excess of the wild-type DNA was then used in order to eliminate non-essential mutations.

Molecular backcrossing was conducted on a selected plasmid from the third round of DNA shuffling by the method identical to normal shuffling as described above, except that the DNAseI digestion and shuffling reaction were performed in the presence of a 40-fold excess of wild-type TEM-1 gene fragment. To make the backcross more efficient, very small DNA fragments (30 to 100 bp) were used in the shuffling reaction. The backcrossed mutants 30 were again selected on LB plates with 80-160 pg/ml .of .cefotaxime (Sigma, St. Louis MO).

This backcross shuffling was repeated with DNA from colonies from the first backcross round in the presence of a fold excess of wild-type TEM-1 DNA. Small DNA fragments (30-100 bp) were used to increase the efficiency of the backcross. The second round of backcrossed mutants were again selected on LB plates with 80-160 ig/ml of cefotaxime.

The resulting transformants were plated on 160 4g/ml of 74 cefotaxime, and a pool of colonies was replated on increasing levels of cefotaxime up to 1,280 Ag/ml. The largest colony obtained at 1,280 fg/ml was replated for single colonies.

This backcrossed mutant was 32,000 fold more resistant than wild-type. (MIC=640 Ag/ml) The mutant strain is 64-fold more resistant to cefotaxime than previously reported clinical or engineered TEM-1-derived strains. Thus, it appears. that DNA shuffling is a fast and powerful tool for at least several cycles of directed molecular evolution.

The DNA sequence of the SfiI insert of the backcrossed mutant was determined using a dideoxy DNA sequencing kit (United States Biochemical Co., Cleveland OH) according to the manufacturer's instructions (Table The mutant had 9 single base pair mutations. As expected, all four of the previously identified silent mutations were lost, reverting to the sequence of the wild-type gene. The promoter mutation (G4205A) as well as three of the four amino acid mutations (Glul04Lys, Metl82Thr, and Gly238Ser) remained in the backcrossed clone, suggesting that they are essential for high level cefotaxime resistance.

20 However, two new silent mutations (T3842C and A3767G), as well as three new mutations resulting in amino acid changes were found (C3441T resulting in Arg241His, C3886T resulting in Gly92Ser, and G4035C resulting in Ala42Gly). While these two silent mutations do not affect the protein primary sequence, they may influence protein expression level (for example by mRNA structure) and possibly even protein folding (by changing the codon usage and therefore the pause site, which has been implicated in protein folding).

*o Table 3 Mutations in Betalactamase Mutation Type Non-Backcrossed Backcrossed amino acid Alal8Lys change Glul04Lys Glul04Lys Metl82Thr Metl82Thr Gly238Ser Gly238Ser -Ala42Gly Gly92Ser silent T3959A G3934A G3713A A3689G T3842C A3767G promoter G4205A G4205A Both the backcrossed and the non-backcrossed mutants have a promoter mutation (which by itself or in combination results in a 2-3 fold increase in expression level) as well as three common 25 amino acid changes (Glul04Lys, Metl82Thr and Gly238Ser).

Glul04Lys and Gly238Ser are mutations that are present in several cefotaxime resistant or other TEM-1 derivatives (Table 4) 9) Expression Level Comparison S: The expression level of the betalactamase gene in the wild- 3. 0 type plasmid, the non-backcrossed mutant and in the backcrossed mutant was compared by SDS-polyacrylamide gel electrophoresis (4- Novex, San Diego CA) of periplasmic extracts prepared by osmotic shock according to the method of Witholt, B. (32).

Purified TEM-1 betalactamase (Sigma, St. Louis MO) was used 35 as a molecular weight standard, and E. coli XLl-blue cells lacking a plasmid were used as a negative control.

The mutant and the backcrossed mutant appeared to produce a 2-3 fold higher level of the betalactamase protein compared to the wild-type gene. The promoter mutation appeared to result in a 2-3 times increase in betalactamase.

Example 6. Construction of mutant combinations of the TEM-1 betalactamase gene To determine the resistance of different combinations of mutations and to compare the new mutants to published mutants, 76 several mutants were constructed into an identical plasmid background. Two of the mutations, GlUlO4Lys and Gly238Ser, are known as cefotaxime mutants. All mutant combinations constructed had the promoter mutation, to allow comparison to selected mutants. The results are shown in Table 4.

Specific combinations of mutations were introduced into the wild-type pUCl82Sfi by PCR, using two o ligonucleot ides per mutation.

The oligonucleotides to obtain the following mutations were: Ala42Gly (SEQ ID NO:l11) AGTTGGGTGGACGAGTGGGTTACATCGAACT and (SEQ ID NO: 12)

AACCCACTCGTCCACCCAACTGATCTTCAGCAT;

Gln39Lys: (SEQ ID NO:13) AGTAAAAGATGCTGAAGATAAGTTGGGTGCAC GAGTGGGTT and (SEQ ID NO:14) ACTTATCTTCAGCATCTTTTACTT; Gly92Ser: (SEQ ID NO: 15) AAGAGCAACTCAGTCGCCGCATACACTATTCT and (SEQ ID NO: 16) ATGGCGGCGACTGAGTTGCTCTTGCCCGGCGTCAAT; GluIO4Lys: (SEQ ID NO:17) TATTCTCAGAATGACTTGGTTAAGTACTCACCAGT CACAGAA and (SEQ ID NO:18) TTAACCAAGTCATTCTGAGAAT; et182Thr: ID NO:19) AACGACGAGCGTGACACCACGACGCCTGTAGCAATG and (SEQ ID 2' NO:20) TCGTGGTGTCACGCTCGTCGTT; Gly23Ber alone: (SEQ ID NO:21) TTGCTGATAAATCTGGAGCCAGTGAGCGTGGGTCTC GCGGTA and (SEQ ID NO:22) TGGCTCCAGATTTATCAGCAA; Gly23BSer and Arg24lHis (combined): 30 (SEQ ID NO:23) ATGCTCACTGGCTCCAGATTTATCAGCAAT and (SEQ ID NO:24) TCTGGAGCCAGTGAGCATGGGTCTCGCGGTATCATT; G420SA: (SEQ ID NO: 25) AACCTGTCCTGGCCACCATGGCCTAAATACAATCAAA

TATGTATCCGCTTATGAGACAATAACCCTGATA.

These separate PCR fragments were gel purified away from the synthetic oligonucleotides. 10 ng of each fragment were combined and a reassembly reaction was performed at 940C for 1 minute and then 25 cycles; [94 0 C for 30 sec, 500C for 30 seconds and 72 0

C

for 45 seconds] PCR was performed on the reassembly product for 77 cycles in the presence of the SfiI-containing outside pri mers (primers C and D from Example The DNA was digested with Sf11 and inserted into the wild-type pUCl82Sfi vector. The following mutant combinations were obtained (Table 4).

Table 4 Name Genotype MIC Source of MIC TEM-l Wild-type 0.02 GlulO4Lys 0.08 ______Gly238Ser 016 TEM-i5 Gll4y/Gy3Sr TEM-3 Glul04Lys/Gly238Ser/Gln39Lys 10 37, 2-32 ST-4 Glu104Lys/Gly23Ser/Met182 Thr ST-i Glu104Lys/Gly238Ser/Met182 320 Thr/Alal8Val/T3959AG37iJA, G3934A/A3689G* ST-2 GlulO4Lys/Gly238Ser/Metl82Thr 640 /Aa 2 Gly/Gly92Ser/Arg24 lHis/ T3842C/A3767G* 15 ST-3 GlulO4Lys/Gly238Ser/Metl82Thr 640 _______/Ala42Gly/Gly92Ser/Arg24 lHis*, All of these mutants additionally contain the G4205A promoter mutation.

It was concluded that conserved mutations account for 9 of z0 15 doublings in the MIC.

GlulO4Lys alone was shown to result only in a doubling of the MIC to 0.08 iig/ml, and Gly238Ser (in several contexts with one additional amino acid change) resulted only in a MIC of 0.16 ug/ml The double mutant Glu104Lys/Gly238Ser -ha a MIC of 10 ilg/ml. This mutant corresponds to These same GlulO4Lys and Gly238Ser mutations in combination with Gln39Lys (TEM-3) or Thr263Met (TEM-4) result in a high level of resistance (2-32 iig/ml for TEM-3 and 8-32 Igg/ml for TEM-4 (34, A mutant containing the three amino acid changes that were conserved after the backcrosq (Glu1o4Lys/Met182Thr/Gly238Ser) also had a MIC of 10 pg/mi. This meant that the mutations that 1 78 each of the new selected mutants had in addition to the three known mutations were responsible for a further 32 to 64-fold increase in the resistance of the gene to cefotaxime.

The naturally occurring, clinical TEM-1-derived enzymes (TEM-1-19) each contain a different combination of only 5-7 identical mutations (reviews). Since these mutations are in well separated locations in the gene, a mutant with high cefotaxime resistance cannot be obtained by cassette mutagenesis of a single area. This may explain why the maximum MIC that was obtained by the standard cassette mutagenesis approach is only 0.64 gg/ml For example, both the Glul04Lys as well as the Gly238Ser mutations were found separately in this study to have MICs below 0.16 gg/ml. Use of DNA shuffling allowed combinatoriality and thus the Glul04Lys/Gly238Ser combination was found, with a MIC of 10 Ag/ml.

An important limitation of this example is the use of a Ssingle gene as a starting point. It is contemplated that better combinations can be found if a large number of related, naturally occurring genes are shuffled. The diversity that is present in 20 such a mixture is more meaningful than the random mutations that are generated by mutagenic shuffling. For example, it is contemplated that one could use a repertoire of related genes from a single species, such as the pre-existing diversity of the immune system, or related genes obtained from many different species.

Example 7. Improvement of antibody AO1B by DNA shuffling of a library of all six mutant CDRs.

The A 1 0 B scFv antibody, a mouse anti-rabbit IgG, was a gift from Pharmacia (Milwaukee WI) The commercially available Pharmacia phage display system was used, which uses the pCANTABS phage display vector.

The original A10B antibody reproducibly had only a low avidity, since clones that only bound weakly to immobilized antigen (rabbit IgG), (as measured by phage ELISA (Pharmacia assay kit) or by phage titer) were obtained. The concentration of rabbit IgG which yielded 50% inhibition of the A10B antibody binding in a competition assay was 13 picomolar. The observed 79 low avidity may also be due to instability of the A1OB clone.

The A10B scFv DNA was sequenced (United States Biochemical Co., Cleveland OH) according to the manufacturer's instructions.

The sequence was similar to existing antibodies, based on comparison to Kabat (33).

1) Preparation of phage DNA Phage DNA having the A10B wild-type antibody gene (10 ul) was incubated at 99 0 C for 10 min, then at 72 0 C for 2.min. PCR mix (50 mM KC1, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100, 200 iM each dNTP, 1.9 mM MgCl), 0.6 Am of each primer and 0.5 p1 Taq DNA Polymerase (Promega, Madison WI) was added to the phage DNA. A PCR program was run for 35 cycles of [30 seconds at 94 0 C, seconds at 45 0 C, 45 seconds at 72 0 The primers used were: ATGATTACGCCAAGCTTT 3' (SEQ ID NO:26) and 5' TTGTCGTCTTTCCAGACGTT 3' (SEQ ID NO:27).

The 850 bp PCR product was then electrophoresed and purified from a 2% low melting point agarose gel.

2) Fragmentation 0:60 o 300 ng of the gel purified 850 bp band was digested with 20 0.18 units of DNAse I (Sigma, St. Louis MO) in 50 mM Tris-HCl pH 7.5, 10 mM MgC1 for 20 minutes at room temperature. The digested DNA was separated on a 2% low melting point agarose gel and bands "between 50 and 200 bp were purified from the gel.

3) Construction of Test Library The purpose of this experiment was to test whether the insertion of the CDRs would be efficient.

The following CDR sequences having internal restriction enzyme sites were synthesized. "CDR H" means a CDR in the heavy "chain and "CDR L" means a CDR in the light chain of the antibody.

CDR Oligos with restriction sites: CDR HI (SEQ ID NO:34) 5'TTCTGGCTACATCTTCACAGAATTCATCTAGATTTGGGTGAGGCAGACGCCTGAA3, CDR H2 (SEQ ID CCAAGTACTTAAAGGGCCACACTGAGTGTA 3' CDR H3 (SEQ ID NO:36) ACCAGGTCA 3' CDR LI (SEQ ID NO:37) GAAGCCTGGAT 3' CDR L2 (SEQ ID NO:38)

'TCCCCCAGACTCCTGATTTATTAAGGGAGATCTAAACAGCTGTTGGTCCCTTTTCGCTTCAGT

3' CDR L3 (SEQ ID NO:39) CAAGCT 3' The CDR oligos were added to the purified A10B antibody DNA fragments of between 50 to 200 bp from step above at a fold molar excess. The PCR mix (50 mM KC1, 10 mM Tris-HCI pH 9.0, 0.1% Triton x-100, 1.9 mM MgCl, 200 pm each dNTP, 0.3 A1 15 Taq DNA polymerase (Promega, Madison WI), 50 il total volume) was added and the shuffling program run for 1 min at 94 0 C, 1 min at 72 0 C, and then 35 cycles: 30 seconds at 94 0 C, 30 seconds at 55 0

C,

30 seconds at 72 0

C.

1 p1 of the shuffled mixture was added to 100 p1 of a PCR mix (50 mM KC1, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100, 200 pm each dNTP, 1.9 mM MgCl, 0.6 pM each of the two outside primers (SEQ ID NO:26 and 27, see below), 0.5 1l Taq DNA polymerase) and the PCR program was run for 30 cycles of [30 seconds at 94 0 C, seconds at 45 0 C, 45 seconds at 72 0 The resulting mixture of DNA fragments of 850 basepair size was phenol/chloroform extracted and ethanol precipitated.

The outside primers were: Outside Primer 1: SEQ ID NO:27 TTGTCGTCTTTCCAGACGTT 3' Outside Primer 2: SEQ ID NO:26 ATGATTACGCCAAGCTTT 3' The 850 bp PCR product was digested with the restriction enzymes SfiI and NotI, purified from a low melting point agarose gel, and ligated into the pCANTAB5 expression vector obtained from Pharmacia, Milwaukee WI. The ligated vector was electroporated according to the method set forth by Invitrogen (San Diego CA) into TG1 cells (Pharmacia, Milwaukee WI) and plated for single colonies.

81 The DNA from the resulting colonies was added to 100 gl of a PCR mix (50 mM KC1, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100, 200 pm each dNTP, 1.9 mM MgCl, 0.6 gM of Outside primer 1 (SEQ ID No. 27; see below) six inside primers (SEQ ID NOS:40-4 5 see below), and 0.5 il Taq DNA polymerase) and a PCR program was run for 35 cycles of [30 seconds at 94 0 C, 30 seconds at 45 0 C, seconds at 72 0 CJ. The sizes of the PCR products were determined by agarose gel electrophoresis, and were used to determine which CDRs with restriction sites were inserted.

CDR Inside Primers: H 1 (SEQ ID NO:40) 5' AGAATTCATCTAGATTTG 3' H 2 (SEQ ID NO:41) 5' GCTTATCCTTTATCTCAGGTC 3', H 3 (SEQ ID NO:42) 5' ACTGCAGTCTTATACGAGGAT 3' L 1 (SEQ ID NO:43) 5' GACGTCTTTAAGCGATCG 3', 15 L 2 (SEQ ID NO:44) 5' TAAGGGAGATCTAAACAG 3', L 3 (SEQ ID NO:45) 5' TCTGCGCGCTTAAAGGAT 3' The six synthetic CDRs were inserted at the expected locations in the wild-type A10B antibody DNA (Figure These studies showed that, while each of the six CDRs in a specific clone has a small chance of being a CDR with a restriction site, most of the clones carried at least one CDR with a restriction site, and that any possible combination of CDRs with restriction sites was generated.

4) Construction of Mutant Complementarity Determining Regions 25 ("CDRs") Based on our sequence data six oligonucleotides corresponding to the six CDRs were made. The CDRs (Kabat definition) were synthetically mutagenized at a ratio of (existing base) :10:10:10, and were flanked on the 5' and 3' sides by about 20 bases of flanking sequence, which provide the homology for the incorporation of the CDRs when mixed into a mixture of unmutagenized antibody gene fragments in a molar excess. The resulting mutant sequences are given below.

Oligos for CDR Library CDR H1 (SEQ ID NO:28) 3 CDR H2 (SEQ ID NO:29) 82 AAGTTCAAGGGCAGGGCCACACTGAGTGTA 3' CDR H3 (SEQ ID 'TGTCTATTTCTGTGCTAGAGGGGACTACTATAGGCGCTACTTTGACTTGTGGGGCCkAGGG ACCACGGTCA 3' CDR Li (SEQ ID NO:31)

'AGAGGGTCACCATGACCTGCAGTGCCAGCTCAGGTATACGTTACATATATTGGTACCACA

GAAGCCTGGAT 3' CDR L2 (SEQ ID NO:32)

'TCCCCCAGACTCCTGATTTATGACACATCCAACGTGGCTCCTGGAGTCCCTTTTCGCTTCAGT

31 C D R L 3 (S E Q I D N 0 3 3 I ATG CTG C CACTTATTACTTG CCAGGAGTGGAGTGGTTATCCGTACACGTTCGGAGGGGGGA CCAAGCT 3'.

Bold and underlined sequences were the mutant sequences synthesized using a mixture of nucleosides of 70:10:10:10 where was the wild-type nucleoside.

10 fold molar excess of the CDR mutant oligos were added 0.to the purified AlOB antibody DNA fragments between 50 to 200 bp

S.

.in length f rom step above. The PCR mix (50 mM KCl, 10 mM Tris-HCl pH 9.0, 0.1% Triton x-100, 1.9 mM MgCl, 200 Aim each dNTP, 0.3 jil Taq DNA polymerase (Promega, Madison WI) 50 Ail total volume) was added and the shuffling program run for 1 min at 94 0 C, 1 min at 72 0 C, and then 35 cycles: [30 seconds at 94-C, seconds at 55 0 C, 30 seconds at 72 0

C].

1 jil of the shuffled mixture was added to 100 41l of a PCR mix (50 mM KCl, 10 mM Tris-HCl pH 9.0, 0.1% Triton X-100, 200 jim each dNTP, 1.9 mM MgCl, o.6 jiM each of the two outside primers **(SEQ ID NO: 26 and 27, see below), 0. 5 Ail Taq DNA polymerase) and the PCR program was run for 30 cycles of [30 seconds at 94 0 C. seconds at 45 0 C, 45 seconds at 72 0 The resulting mixture of DNA fragments of 850 basepair size was phenol/chloroform extracted and ethanol precipitated.

The outside primers were: Outside Primer 1: SEQ ID NO:27 5' TTGTCGTCTTTCCAGACGTT 3' Outside Primer 2: SEQ ID ND:26 5' ATGATTACGCCAAGCTTT 3' Cloning of the scFv antibody DNA into The 850 bp PCR product was digested with the restriction enzymes Sf11 and NotI, purified from a low melting point agarose gel, and ligated into the pCANTAB5 expression vector obtained from Pharmacia, Milwaukee WI. The ligated vector was electroporated according to the method set forth by Invitrogen (San Diego CA) into TG1 cells (Pharmacia, Milwaukee WI) and the phage library was grown up using helper phage following the guidelines recommended by the manufacturer.

The library that was generated in this fashion was screened for the presence of improved antibodies, using six cycles of selection.

6) Selection of high affinity clones wells of a 96 well microtiter plate were coated with Rabbit IgG (Jackson Immunoresearch, Bar Harbor ME) at 10 gg /well for 1 hour at 37 0 C, and then blocked with 2% non-fat dry milk in PBS for 1 hour at 37 0

C.

15 100 Ll of the phage library (1x1010 cfu) was blocked with 100 il of 2% milk for 30 minutes at room temperature, and then added to each of the 15 wells and incubated for 1 hour at 37 0 c.

Then the wells were washed three times with PBS containing 0.5% Tween-20 at 37 0 C for 10 minutes per wash. Bound phage was eluted with 100 il elution buffer (Glycine-HC1, pH followed by immediate neutralization with 2M Tris pH 7.4 and transfection for phage production. This selection cycle was repeated six times.

After the sixth cycle, individual phage clones were picked and the relative affinities were compared by phage ELISA, and the specificity for the rabbit IgG was assayed with a kit from Pharmacia (Milwaukee WI) according to the methods recommended by the manufacturer.

The best clone has an approximately 100-fold improved expression level compared with the wild-type A10B when tested by the Western assay. The concentration of the rabbit IgG which yielded 50% inhibition in a competition assay with the best clone was 1 picomolar. The best clone was reproducibly specific for rabbit antigen. The number of copies of the antibody displayed by the phage appears to be increased.

Example 8. In vivo recombination via direct repeats of partial genes A plasmid was constructed with two partial, inactive copies 84 of the same gene (beta-lactamase) to demonstrate that recombination between the common areas of these two direct repeats leads to full-length, active recombinant genes.

A pUC18 derivative carrying the bacterial TEM-1 betalactamase gene was used (Yanish-Perron et al., 1985, Gene 33:103-119). The TEM-1 betalactamase gene confers resistance to bacteria against approximately 0.02 ug/ml of cefotaxime. Sfil restriction sites were added 5' of the promoter and 3' of the end of the betalactamase gene by PCR of the vector sequence with two primers: Primer A (SEQ ID NO: 46) TTCTATTGACGGCCTGTCAGGCCTCATATATACTTTAGATTGATTT 3' PRIMER B (SEQ ID NO: 47)

TTGACGCACTGGCCATGGTGGCCAAAAATAAACAAATAGGGGTTCCGCGCAC

S 15 ATTT 3' o* "and by PCR of the beta-lactamase gene sequence with two other primers: Primer C (SEQ ID NO: 48) 5' AACTGACCACGGCCTGACAGGCCGGTCTGACAGTTACCAATGCTT 3' Primer D (SEQ ID NO: 49) 5' AACCTGTCCTGGCCACCATGGCCTAAATACATTCAAATATGTAT 3' The two reaction products were digested with Sfil, mixed, ligated and used to transform competent E. coli bacteria by the procedure described below. The resulting plasmid was pUC182Sfi- 2 Bla-Sfi. This plasmid contains an Sfil fragment carrying the Bla gene and the P-3 promoter.

*ec* The minimum inhibitory concentration of cefotaxime for E.

coli XL1-blue (Stratagene, San Diego CA) carrying pUC182Sfi-Bla- Sfi was 0.02 ug/ml after 24 hours at 37 0

C.

The tetracycline gene of pBR322 was cloned into pUC18Sfi- Bla-Sfi using the homologous areas, resulting in pBR322TetSfi- Bla-Sfi. The TEM-1 gene was then deleted by restriction digestion of the pBR322TetSfi-Bla-Sfi with SspI and FspI and blunt-end ligation, resulting in pUC322TetSfi-Sfi.

Overlapping regions of the TEM-1 gene were amplified using standard PCR techniques and the following primers: Primer 2650 (SEQ ID NO: 50) 5' TTCTTAGACGTCAGGTGGCACTT 3' Primer .2493 (SEQ ID NO: 51) 5' TTT TAA ATC AAT CTA AAG TAT 3' P r i mer 2 6 5 1 SEQ ID NO 5 2) TGCTCATCCACGAGTGTGGAGAAGTGGTCCTGCAACTTTAT 3,'and Primer 2652 (SEQ ID NO: 53)

ACCACTTCTCCACACTCGTGGATGAGCACTTTTAAAGTT

The two resulting DNA fragments were digested with Sfil and BstXl and ligated into the Sfi site of pBR322TetSfi-sfi. The resulting plasmid was called pBR322Sfi-BL-LA-Sfi. A.map of the plasmid as well as a schematic of intraplasmidic recombination and reconstitution of functional beta-lactamase is shown in Figure 9.

The plasmid was electroporated into either TG-1 or JC8679 E.

coli cells. E. coli JC8679 is RecBC sbcA (Oliner et al., 1993, NAR 21:5192). The cells were plated on solid agar plates containing tetracycline. Those colonies which grew, were then S* 15 plated on solid agar plates containing 100 Ag/ml ampicillin and S*the number of viable colonies counted. The beta-lactamase gene inserts in those transformants which exhibited ampicillin resistance were amplified by standard PCR techniques using Primer 2650 (SEQ ID NO: 50) 5' TTCTTAGACGTCAGGTGGCACTT 3' and Primer 2493 (SEQ ID NO: 51) 5' TTTTAAATCAATCTAAAGTAT 3' and the length of the insert measured. The presence of a 1 kb insert indicates that the gene was successfully recombined, as shown in Fig. 9 and Table TABLE Cell Tet Colonies Amp colonies Colony PCR TG-1 131 21 3/3 at 1 kb JC8679 123 31 4/4 at 1 kb vector 51 0 control About 17-25% of the tetracycline-resistant colonies were also ampicillin-resistant and all of the Ampicillin resistant colonies had correctly recombined, as determined by colony PCR.

Therefore, partial genes located on the same plasmid will successfully recombine to create a functional gene.

Example 9. In vivo recombination via direct repeats of full-length genes.

A plasmid with two full-length copies of different alleles of the beta-lactamase gene was constructed. Homologous recombination of the two genes resulted in a single recombinant full-length copy of that gene.

The construction of pBR322TetSfi-Sfi and pBR322TetSfi-Bla- Sfi was described above.

The two alleles of the beta-lactamase gene were constructed as follows. Two PCR reactions were conducted with pUC18Sfi-Bla- Sfi as the template. One reaction was conducted with the following primers.

Primer 2650 (SEQ ID NO: 50) 5' TTCTTAGACGTCAGGTGGCACTT 3' Primer 2649 (SEQ ID NO: 51) 5' ATGGTAGTCCACGAGTGTGGTAGTGACAGGCCGGTCTGACAGTTA CCAATGCTT 3' The second PCR reaction was conducted with the following primers: Primer 2648 (SEQ ID NO: 54) 5' TGTCACTACCACACTCGTGGACTACCATGGCCTAAATACATTCAAA TATGTAT 3' Primer 2493 (SEQ ID NO: 51) 5' TTT TAA ATC AAT CTA AAG TAT 3' This yielded two Bla genes, one with a 5' Sfil site and a 3' BstXl site, the other with a 5' BstXl site and a 3' Sfil site.

After digestion of these two genes with BstXl and Sfil, and 2 ligation into the Sfil-digested plasmid pBR322TetSfi-Sfi, a plasmid (pBR322-Sfi-2BLA-Sfi) with a tandem repeat of the Bla gene was obtained. (See Figure The plasmid was electroporated into E. coli cells. The cells were plated on solid agar plates containing 15 ig/ml tetracycline. Those colonies which grew, were then plated on solid agar plates containing 100 Ag/ml ampicillin and the number of viable colonies counted. The Bla inserts in those transformants which exhibited ampicillin resistance were amplified by standard PCR techniques using the method and primers described in Example 8. The presence of a 1 kb insert indicated that the duplicate genes had recombined, as indicated in Table 6.

87 TABLE 6 Cell Tet Colonies Amp Colonies Colony PCR TG-1 28 54 7/7 at 1kb JC8679 149 117 3/3 at Ikb vector 51 0 control Colony PCR confirmed that the tandem repeat was efficiently recombined to form a single recombinant gene.

Example 10. Multiple cycles of direct repeat recombination Interplasmidic In order to determine whether multiple cycles of recombination could be used to produce resistant cells more 15 quickly, multiple cycles of the method described in Example 9 were performed.

The minus recombination control consisted of a single copy of the betalactamase gene, whereas the plus recombination experiment consisted of inserting two copies of betalactamase as 20 a direct repeat. The tetracycline marker was used to equalize the number of colonies that were selected for cefotaxime resistance in each round, to compensate for ligation S efficiencies.

In the first round, pBR322TetSfi-Bla-Sfi was digested with 25 EcrI and subject to PCR with a 1:1 mix (1 ml) of normal and Cadwell PCR mix (Cadwell and Joyce (1992) PCR Methods and Applications 2: 28-33) for error prone PCR. The PCR program was 0 C for 2 minutes initially and then 30 cycles of 94 0 C for seconds, 52 0 C for 30 second and 72 0 C for 3 minutes and 6 seconds per cycle, followed by 72 0 C for 10 minutes.

The primers used in the PCR reaction to create the one Bla gene control plasmid were Primer 2650 (SEQ ID NO: 50) and Primer 2719 (SEQ ID NO: 55) 5' TTAAGGGATTTTGGTCATGAGATT This resulted in a mixed population of amplified DNA fragments, designated collectively as Fragment #59. These fragments had a number of different mutations.

88 The primers used in two different PCR reactions to create the two Bla gene plasmid were Primer 2650 (SEQ ID NO: 50) and Primer 2649 (SEQ ID NO: 51) for the first gene and Primers 2648 (SEQ ID NO: 54) and Primer 2719 (SEQ ID NO: 55) for the second gene. This resulted in a mixed population of each of the two amplified DNA fragments: Fragment #89 (amplified with primers 2648 and 2719) and Fragment #90 (amplified with primers. 2650 and 2649). In each case a number of different mutations had been introduced the mixed population of each of the fragments.

After error prone PCR, the population of amplified

DNA

fragment #59 was digested with Sfil, and then cloned into pBR322TetSfi-Sfi to create a mixed population of the plasmid pBR322Sfi-Bla-Sfil.

.After error prone PCR, the population of amplified DNA 15 fragments #90 and #89 was digested with SfiI and BstXI at 50 0

C,

and ligated into pBR322TetSfi-Sfi to create a mixed population of the plasmid pBR322TetSfi-2Bla-Sfil (Fig. The plasmids pBR322Sfi-Bla-Sfi 1 and pBR322Sfi-2Bla-Sfil were electroporated into E. coli JC8679 and placed on agar plates having differing concentrations of cefotaxime to select for resistant strains and on tetracycline plates to titre.

An equal number of colonies (based on the number of colonies growing on tetracycline) were picked, grown in LB-tet and DNA extracted from the colonies. This was one round of the 2 recombination. This DNA was digested with Ecrl and used for a second round of error-prone PCR as described above.

After five rounds the MIC (minimum inhibitory concentration) for cefotaxime for the one fragment plasmid was 0.32 whereas the MIC for the two fragment plasmid was 1.28. The results show that after five cycles the resistance obtained with recombination was four-fold higher in the presence of in vivo recombination.

Example 11. In vivo recombination via electroporation of fragments Competent E. coli cells containing pUC18Sfi-Bla-Sfi were prepared as described. Plasmid pUC18Sfi-Bla-Sfi contains the standard TEM-1 beta-lactamase gene as described, supra.

A TEM-1 derived cefotaxime resistance gene from pUCl8Sficef-Sfi, (clone ST2) (Stemmer WPC (1994) Nature 370: 389-91, 89 incorporated herein by reference) which confers on E. coli carrying the plasmid an MIC of 640 Ag/ml for cefotaxime, was obtained. In one experiment the complete plasmid pUC18Sfi-cef- Sfi DNA was electroporated into E. coli cells having the plasmid pUC18Sfi-Bla-Sfi.

In another experiment the DNA fragment containing the cefotaxime gene from pUC18Sfi-cef-Sfi was amplified by PCR using the primers 2650 (SEQ ID NO: 50) and 2719 (SEQ ID NO: 55). The resulting 1 kb PCR product was digested into DNA fragments of <100 bp by DNase and these fragments were electroporated into the competent E. coli cells which already contained pUC18Sfi-Bla-sfi.

The transformed cells from both experiments were then assayed for their resistance to cefotaxime by plating the transformed cells onto agar plates having varying concentrations S* 15 of cefotaxime. The results are indicated in Table 7.

e*e TABLE 7 Colonies/ Cefotaxime Concentration C. 0.16 0.32 1.28 5 0 10.0 5. 10.0 no DNA control 14 20 ST-2 mutant, whole 4000 2000 800 400 ST-2 mutant, fragments 1000 120 22 7 Wildtype, whole 27 Wildtype, fragments 18 25 From the results it appears that the whole ST-2 Cef gene was inserted into either the bacterial genome or the plasmid after electroporation. Because most insertions are homologous, it is expected that the gene was inserted into the plasmid, replacing the wildtype gene. The fragments of the Cef gene from St-2 also inserted efficiently into the wild-type gene in the plasmid. No sharp increase in cefotaxime resistance was observed with the introduction of the wildtype gene (whole or in fragments) and no DNA. Therefore, the ST-2 fragments were shown to yield much greater cefotaxime resistance than the wild-type fragments.

It was contemplated that repeated insertions of fragments, prepared from increasing resistant gene pools would lead to increasing resistance.

Accordingly, those colonies that produced increased cefotaxime resistance with the St-2 gene fragments were isolated and the plasmid DNA extracted. This DNA was amplified using PCR by the method described above. The amplified DNA was digested with DNase into fragments (<100 bp) and 2-4 Ag of the fragments were electroporated into competent E. coli cells already containing pUC322Sfi-Bla-Sfi as described above. The transformed cells were plated on agar containing varying concentrations of cefotaxime.

As a control, competent E. coli cells having the plasmid pUC18Sfi-Kan-Sfi were also used. DNA fragments from the digestion of the PCR product of pUC18Sfi-cef-Sfi were 15 electroporated into these cells. There is no homology between the kanamycin gene and the beta-lactamase gene and thus recombination should not occur.

This experiment was repeated for 2 rounds and the results are shown in Table 8.

TABLE 8 Round Cef conc. KAN control Cef resistant colonies 1 0.16-0.64 lawn lawn replate 0.32 10 small 1000 10 10 400 25 Replate 100sm 2.5 50 3 40 100 sm 1280 100 sm Example 12 Determination of Recombination Formats This experiment was designed to determine which format of recombination generated the most recombinants per cycle.

In the first approach, the vector pUC18Sfi-Bla-Sfi was amplified with PCR primers to generate a large and small fragment. The large fragment had the plasmid and ends having 91 portions of the Bla gene, and the small fragment coded for the middle of the Bla gene. A third fragment having the complete Bla gene was created using PCR by the method in Example 8. The larger plasmid fragment and the fragment containing the complete Bla gene were electroporated into E. coli JC8679 cells at the same time by the method described above and the transformants plated on differing concentrations of cefotaxime.

In approach 2, the vector pUC18Sfi-Bla-Sfi was amplified to produce the large plasmid fragment isolated as in approach 1 above. The two fragments each comprising a portion of the complete Bla gene, such that the two fragments together spanned the complete Bla gene werealso obtained by PCR. The large plasmid fragment and the two Bla gene fragments were all electroporated into competent E. coli JC8679 cells and the 0 15 transformants plated on varying concentrations of cefotaxime.

In the third approach, both the vector and the plasmid were electroporated into E. coli JC8679 cells and the transformants were plated on varying concentrations of cefotaxime.

In the fourth approach, the complete Bla gene was electroporated into E. coli JC8679 cells already containing the vector pUCSfi-Sfi and the transformants were plated on varying concentrations of cefotaxime. As controls, the E. coli JC8679 cells were electroporated with either the complete Bla gene or the vector alone.

The results are presented in Figure 11. The efficiency of the insertion of two fragments into the vector is 100 X lower than when one fragment having the complete Bla gene is used.

Approach 3 indicated that the efficiency of insertion does depend on the presence of free DNA ends since no recombinants were obtained with this approach. However, the results of approach 3 were also due to the low efficiency of electroporation of the vector. When the expression vector is already in the competent cells, the efficiency of the vector electroporation is not longer a factor and efficient homologous recombination can be achieved even with uncut vector.

Example 12. Kit for cassette shuffling to optimize vector performance In order to provide a vector capable of conferring an optimized phenotype maximal expression of a vector-encoded sequence, such as a cloned gene), a kit is provided comprising a variety of cassettes which can be shuffled, and optimized shufflants can be selected. Figure 12 shows schematically one embodiment, with each loci having a plurality of cassettes. For example, in a bacterial expression system, Figure 13 shows example cassettes that are used at the respective loci. Each cassette of a given locus all promoters in this example) are flanked by substantially identical sequences capable of overlapping the flanking sequence(s) of cassettes of an adjacent locus and preferably also capable of participating in homologous recombination or non-homologous recombination lox/cre or flp/frt systems), so as to afford shuffling of cassettes within a locus but substantially not between loci.

S: 15 Cassettes are supplied in the kit as PCR fragments, which each cassette type or individual cassette species packaged in a separate tube. Vector libraries are created by combining the contents of tubes to assemble whole plasmids or substantial portions thereof by hybridization of the overlapping flanking sequences of cassettes at each locus with cassettes at the adjacent loci. The assembled vector is ligated to a **predetermined gene of interest to form a vector library wherein each library member comprises the predetermined gene of interest and a combination of cassettes determined by the association of 6 2' cassettes. The vectors are transferred into a suitable host cell and the cells are cultured under conditions suitable for t. expression, and the desired phenotype is selected.

While the present invention has been described with reference to what are considered to be the preferred examples, it is to be understood that the invention is not limited to the disclosed examples. To the contrary, the invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

93

REFERENCES

The following references are cited in this application at the relevant portion of the application.

1. Holland, J. H. (1992) Sci. Am. July, 66-72.

2. Holland, J. H. (1992) "Adaptation in natural and artificial systems". Second edition, MIT Press, Cambridge.

3. Joyce, G. F. (1992) Scientific American, December, 90-97.

4. Kauffman, S. A. (1993) "The origins of order". Oxford University Press, New York.

Stormo, G. D. (1991) Methods Enzymol. 208:458-468.

6. Schneider, T. D. et al., (1986) J. Mol. Biol. 188:415-431.

7. Reidhaar-Olson, J. F and Sauer, R.T. (1988) Science 241:53- 57.

8. Stemmer, W. P. C. et al., (1992) Biotechniques 14:256-265.

9. Yockey, H. P. (1977) J. Theor. Biol. 67:345-376.

25 10. Yockey, H. P. (1974) J. Theor. Biol. 46:369-380.

11. Leung, D. W. et al., (1989) Technique 1:11-15.

12. Caldwell, R. C. and Joyce, G. F. (1992) PCR Methods and *i 30 Applications 2:28-33.

13. Bartel, D. and Szostak, J. W. (1993) Science 261:1411- 1418.

35 14. Bock, L. C. et al., (1992) Nature 355:564-566.

Scott, J. K. and Smith, G. P. (1990) Science 249:386-390.

16. Cwirla, S. E. et al. (1990) Proc. Natl. Acad. Sci. USA .i 40 87:6378-6382.

17. McCafferty, J. et al. (1990) Nature 348:552-554.

18. Cull, M. G. et al., (1992) Proc. Natl. Acad. Sci. USA 89:1865-1869.

19. Gramm, H. et al., (1992) Proc. Natl. Acad. Sci. USA 89:3576- 3580.

20. Arkin, A. and Youvan, D. C. (1992) Proc. Natl. Acad. Sci.

USA 89:7811-7815.

21. Oliphant, A. R. et al., (1986) Gene 44:177-183.

22. Hermes, J. D. et al., (1990) Proc. Natl. Acad. Sci. USA 87:696-700.

23. Meyerhans, A. et al., (1990) Nucleic Acids Res. 18:1687- 1691.

24. Osterhout, J. J. et al. (1992) J. Am. Chem. Soc. 114:331- 337.

Cano, R. J. et al., (1993) Nature 363:536-538.

26. Palzkill and Botstein, (1992) J. Bacteriol. 174:5237-5243.

27. Marton et al., Nucleic Acids Res. 19:2423.

28. Yanish-Perron et al., [1985] Gene 33:103-119.

29. Watson (1988) Gene 70:399-403.

Ambler et al. (1991) Biochem J. 276:269-272.

31. Chen and Clowes, (1984) Nucleic Acid Res. 12:3219-3234.

32. Witholt, B. ([1987] Anal. Biochem. 164(2):320-330 Kabat et al., (1991) "Sequences of Proteins of Immunological Interest" U.S. Department of Health and Human Services,

NIH

Publication 91-3242.

34. Philippon et al., (1989) Antimicrob Agents Chernother :33:1131-1136.

35. 'Jacoby and Medeiros (1991) Arltiricrob. Agents Chemother.

35:167-1704.

36. Coelhosampaio (1993) Biochem. 32:10929-10935 37. Tuerk, C. et al., (1992) Proc. Natl. Acad. Sci. USA 89: 6988- 6992.

38. United States Patent No. 4,683,195 39. United States Patent No. 4,683,202 *40. Delagrave et al. (1993) Pr-otein Engineering 6: 327-331 41. Deigrave et al. (1993) Bio/Technology 11: 1548-1552 42. Goldman, ER and Youvan DC (1992) Rio/Technology 10:1-557-1561 43. Nissim et al. (1994) EMBO J. 13: 692-698 44. Winter et al. (1994) Ann. Rev. Imnzunol. 12: 433-55 Caren et al. (1994) Bio/Technology 12: 517-520 46. Calogero et al. (1992) FEMS Microbiology Lett. 97: 41-44 47. Galizzi et al. W091/01087 48. Hayashi et al. (1994) Biotechniques 17: 310-315 49. Radman et al. W090/07576 All publications, patents and patent applications are herein incorporated by reference in their entirety to the same extent as if each individual publication, patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety.

*rO S S 0 0

Claims (38)

1. A population of polynucleotides resulting from shuffling sequences from substrate polynucleotides encoding antibody chains, whereby the population of polynucleotides comprises recombinant polynucleotides encoding antibody chains having different combinations of heavy chain CDRs and/or light chain CDRs than are encoded by the substrate polynucleotides.

2. The population of claim 1, wherein at least 10% of the population of polynucleotides are recombinant polynucleotides encoding different combinations of heavy chain CDRs and/or light chain CDRs than are encoded by the substrate polynucleotides.

3. The population of claim 1 or claim 2, wherein the population of polynucleotides are encoded by a bacteriophage so that antibody chains expressed from the polynucleotides are displayed from the bacteriophage.

4. The population of polynucleotides of any one of claims 1 to 3, wherein the 15 substrate polynucleotides encode CDR sequences generated by synthesizing a collection of sequences based on naturally occurring human CDR sequences.

5. The population of polynucleotides of claim 4, wherein the substrate polynucleotides encode consensus sequences of naturally occurring CDR sequences.

6. The population of polynucleotides of any one of claims 1 to 5, wherein the substrate polynucleotides encode at least 100, optionally, at least 1,000, and optionally at least 10,000, unique heavy chain CDR sequences.

7. The population of polynucleotides of any one of claims 1 to 3, wherein the substrate polynucleotides include CDRs with other than naturally occurring residues.

8. The population of polynucleotides of any one of claims 1 to 3, wherein the substrate polynucleotides encode CDRs having at least 40 percent positional sequence identity to known CDR sequences.

9. The population of polynucleotides of claim 8, wherein the substrate polynucleotides further encode diverse adjacent framework regions.

The population of polynucleotides of any one of claims 1 to 9, wherein the recombinant polynucleotides encode antibody chains with a plurality of epitope specificities.

11. The population of polynucleotides of any one of claims 1 to 10, wherein the substrate polynucleotides are prepared by PCR amplification of mRNA or cDNA from antibody-producing cells. [I:\DayLib\LIBFF]73935spec.doc:gcc 97

12. The population of polynucleotides of any one of claims 1 to 11, wherein the substrate polynucleotides comprise oligonucleotides synthesized on an oligonucleotide synthesizer.

13. The population of polynucleotides of any one of claims 1 to 12, wherein the substrate polynucleotides comprise a plurality of oligonucleotides encoding different forms of each of CDR HI, CDR H2, CDR H3, CDR L1, CDR L2 and CDR L3.

14. The population of polynucleotides of any one of claims 1 to 12, wherein the recombinant nucleic acids encode single chain antibodies in which VL and VH regions are linked by a peptide linker.

15. An antibody obtainable by expression of polynucleotides in a population according to any one of claims 1 to 14, said antibody being capable of specifically binding to a predetermined antigen and comprising a novel combination of heavy CDR regions and/or a novel combination of light chain CDRs.

16. The antibody of claim 15, comprising a novel combination of heavy chain 15 CDRs and a novel combination of light chain CDRs.

17. The antibody of claim 15 or 16 that specifically binds to a predetermined antigen with an affinity of at least 10" 9 M 1 optionally at least 1010 M'.

18. The antibody of any one of claims 15 to 17 that is a whole antibody.

19. The antibody of any one of claims 15 to 18, wherein the predetermined antigen is selected from the group consisting of a human cell surface antigen, other human biological macromolecules and a nonhuman disease associated macromolecule.

20. An antibody comprising a heavy chain comprising three nonnaturally occurring CDR regions, and a light chain comprising three nonnaturally occurring CDR regions, wherein the antibody specifically binds to a predetermined antigen.

21. The antibody of claim 20 that binds to a human antigen.

22. The antibody of claim 21 that binds to a human antigen selected from the group consisting of CD4, CD8, IL-2 receptor, EGF receptor and PDGF receptor.

23. The antibody of claim 21 that binds to a human thrombomodulin, protein C, carbohydrate antigen, sialyl Lewis antigen, or selectin.

24. The antibody of claim 21 that binds to an antigen selected from the group consisting of bacterial LPS, virion capsid protein and envelope glycoprotein.

The antibody of any one of claims 20 to 24, wherein the heavy and light chains are linked as a single chain via a linker peptide.

26. The antibody of any one of claims 20 to 25 that binds with an affinity of at least 108 M [I:\DayLib\LIBFF]73935spec.doc:gcc 98

27. A method of producing a polynucleotide encoding an antibody chain that specifically binds to a predetermined antigen, the method comprising: denaturing a population of polynucleotide segments encoding CDR regions; annealing the denatured population of CDR encoding polynucleotides; incubating the annealed CDR encoding polynucleotides in the presence of a polymerase, thereby forming a population of polynucleotides comprising recombinant polynucleotides encoding antibodies with novel combinations of CDR regions, which novel combinations of CDR regions are not present in an antibody o0 encoded by the population of polynucleotide segments provided in step repeating steps through and screening or selecting the population of polynucleotides to identify at least one polynucleotide that encodes an antibody chain that specifically binds to a predetermined antigen. 15

28. The method of claim 27, wherein at least 10% of the population of polynucleotides comprise recombinant polynucleotides with novel combinations of CDR regions.

29. The method of claim 27, comprising screening or selecting using a bacteriophage display system. 20

30. The method of claim 27, further comprising expressing the at least one polynucleotide that encodes an antibody chain that specifically binds to a predetermined antigen in a host cell.

31. The method of claim 30, wherein the antibody specifically binds with an affinity of at least about 10- 9 M.

32. The method of claim 30, wherein the antibody specifically binds with an affinity of at least about 10'1 M.

33. The method of claim 27, further comprising synthesizing a whole antibody based on the sequence of the at least one polynucleotide that encodes an antibody chain that specifically binds to a predetermined antigen.

34. The method of claim 27, wherein the polynucleotide segments encoding CDR regions are generated by synthesizing a collection of sequences based on naturally occurring human CDR sequences.

The method of claim 27, wherein the population of polynucleotide segments encoding CDR regions comprises consensus sequences of naturally occurring CDR sequences. [I:\DayLib\LBFF]73935spec.doc:gcc 99

36. A population of polynucleotides resulting from shuffling sequences from substrate polynucleotides encoding antibody chains, substantially as hereinbefore described with reference to any one of the examples.

37. An antibody obtainable by expression of the polynucleotide of claim 36.

38. An antibody comprising a heavy chain comprising three nonnaturally occurring CDR regions, and a light chain comprising three nonnaturally occurring CDR regions, substantially as hereinbefore described with reference to any one of the examples. Dated 19 March, 2002 Affymax Technologies N.V. *Patent Attorneys for the Applicant/Nominated Person SPRUSON FERGUSON o [I:\DayLib\LIBFF]73935spec.doc:gcc