An in vitro polysome display system for identifying ligands from very large peptide libraries

Phage display of cDNA repertoires: the pVI display system and its applications for the selection of immunogenic ligands

The selection of phage displayed cDNA repertoires on an immobilised target has been reported to be an efficient way to rapidly identify interacting partners. To date, however, only a few successful applications have been reported. Here, we present a review of the current status of the display and selection of cDNA libraries using phage. As an example, we report the construction of a set of phage display vectors suitable for cDNA display based on fusion to the minor bacteriophage coat protein 6 (pVI) of filamentous phage. We have evaluated these vectors through the display of the C(H)3 domain of human IgG and of the Escherichia coli alkaline phosphatase (PhoA) gene. Both the C(H)3 domain of IgG and PhoA are shown to be displayed on pVI, and PhoA is also shown to be enzymatically active. We have constructed primary colorectal tumor cDNA repertoires in these vectors and performed selections on both a monoclonal antibody to beta2 microglobulin (beta2M) and polyclonal antibody sera to human IgG. In both cases, relevant ligands were recovered from the phage displayed cDNA repertoire. These vectors may be used for selection of phage displayed cDNA libraries with polyclonal sera from patients. This will allow the identifying antigenic cDNA products in such diseases as cancer, viral/bacterial infections or autoimmune disease. Furthermore, by selections with other specific biomolecules, this display system may aid the identification of interacting partners in functional genomics.

Selection of a human anti-progesterone antibody fragment from a transgenic mouse library by ARM ribosome display

In antibody-ribosome-mRNA complex (ARM) ribosome display, stable complexes of nascent protein, mRNA and ribosomes are produced in a eukaryotic in vitro expression system, through coupled transcription and translation of DNA lacking a 3' stop codon. Selection of the protein simultaneously captures the relevant mRNA, which is recovered as DNA by coupled reverse transcription-polymerase chain reaction (RT-PCR) performed on the intact complexes. Here, we describe the use of ARM display to select a specific human antibody fragment from a transgenic mouse library. The mice carry unrearranged gene segments of the human heavy (H) and kappa light (L) chain loci, while the endogenous murine H and kappa loci are functionally silenced; they respond to immunisation by production of fully human IgM antibodies. A library encoding human single-chain (sc) antibody (V(H)/K) fragments, in which V(H) domains and kappa light chains were combined at random by PCR, was prepared from spleen cells of transgenic mice immunised with progesterone-bovine serum albumin (BSA). Library diversity was demonstrated by sequencing. Progesterone-binding fragments were selected over five cycles of ARM display and the selected DNA cloned and expressed in Escherichia coli. Soluble V(H)/K fragments obtained in periplasmic extracts had the same specificity as ribosome-bound V(H)/K, supporting the view that folding and specificity of the displayed and soluble proteins are equivalent. The affinity of the expressed V(H)/K was approximately 10(-8) M. Sequencing showed that ARM display selected a single V(H)/V(L) combination (V(H)1-2, Vkappa4-1) and rearrangement, with a few mutational differences between clones. Monoclonal antibodies against progesterone-BSA obtained from hybridomas were encoded by the same V(H) and V(L) segments and had similar properties to the fragments obtained in vitro. The combination of ribosome display and transgenic mouse technologies is a rapid means of generating fully human antibody fragments in vitro for expression and further manipulation.

Ribosome display: an in vitro method for selection and evolution of antibodies from libraries

Combinatorial approaches in biology require appropriate screening methods for very large libraries. The library size, however, is almost always limited by the initial transformation steps following its assembly and ligation, as other all screening methods use cells or phages and viruses derived from them. Ribosome display is the first method for screening and selection of functional proteins performed completely in vitro and thus circumventing many drawbacks of in vivo systems. We review here the principle and applications of ribosome display for generating high-affinity antibodies from complex libraries. In ribosome display, the physical link between genotype and phenotype is accomplished by a mRNA-ribosome-protein complex that is used for selection. As this complex is stable for several days under appropriate conditions, very stringent selections can be performed. Ribosome display allows protein evolution through a built-in diversification of the initial library during selection cycles. Thus, the initial library size no longer limits the sequence space sampled. By this method, scFv fragments of antibodies with affinities in the low picomolar range have been obtained. As all steps of ribosome display are carried out entirely in vitro, reaction conditions of individual steps can be tailored to the requirements of the protein species investigated and the objectives of the selection or evolution experiment.

Generation of recombinant antibodies

Recombinant antibody technology is opening new perspectives for the development of novel therapeutic and diagnostic agents. In this review we focus on advances in the generation of both genetically engineered humanized and fully human monoclonal antibodies. Methods for their production in different expression systems are also discussed.

STABLE: protein-DNA fusion system for screening of combinatorial protein libraries in vitro

We have developed a new method that permits the complete in vitro construction and selection of peptide or protein libraries. This method relies on an in vitro transcription/translation reaction compartmentalized in water in oil emulsions. In each emulsion compartment, streptavidin (STA)-fused polypeptides are synthesized and attached to the encoding DNA via its biotin label. The resulting protein-DNA fusion molecules recovered from the emulsion can be subjected to affinity selection based on the properties of the peptide portion, whose sequence can be determined from that of its DNA-tag. This method, named 'STABLE' (STA-biotin linkage in emulsions), should be useful for rapid in vitro evolution of proteins and for ligand-based selection of cDNA libraries.

Comparison of Escherichia coli and rabbit reticulocyte ribosome display systems

Ribosome display is a technology for library selection and simultaneous molecular evolution in vitro. We present here a comparison between an optimized Escherichia coli system and different rabbit reticulocyte ribosome display systems, optimized in a number of parameters, as a coupled eukaryotic system had been suggested to result in high enrichment factors [He and Taussig (1997) Nucleic Acids Res. 25, 5132-5134]. With all systems, antibody scFv fragments, complexed to the ribosomes and the corresponding mRNA, were enriched by binding to their cognate antigen and enrichment was always dependent on the absence of a stop codon and the presence of cognate antigen. However, the efficiency of the E. coli ribosome display system was 100-fold higher than an optimized uncoupled rabbit reticulocyte ribosome display system, with separate in vitro transcription and translation, which was in turn several-fold more efficient than the reported coupled system. Neither the E. coli nor the rabbit reticulocyte ribosome display system was dependent on the orientation of the domains of an antibody scFv fragment or on the spacer sequence. In summary, we could not detect any intrinsic advantage of using a eukaryotic translation system for ribosome display.

Investigation of translation dynamics under cell-free protein biosynthesis conditions using high-resolution two-dimensional gel electrophoresis

A cell-free extract from Escherichia coli, generated through a routine procedure according to Chen and Zubay (Methods Enzymol. 1983, 101, 674-690), was used for an in vitro protein synthesis. High-resolution two-dimensional gel electrophoresis (2-DE) was exploited to investigate the protein composition of the cell-extract and its dynamic development during a 24 h-period of cell-free protein synthesis performed in a membrane reactor device. Green fluorescent protein (GFP) was chosen as a target protein to be produced in a cell-free reactor because of its functional activity, which can easily be monitored by measurement of fluorescence, and because of its high sensitivity. GFP synthesis was observed by a standard fluorescence assay and was correlated to a quantitative assessment of the silver-stained GFP spot appearing on 2-DE gel maps. A constant protein synthesis rate was obtained for at least 8 h of process operation. While declining continuously, protein synthesis stopped entirely after 24 h. Both, the total protein content and total number of detectable spots were found to decrease over the reaction time, due to proteolytic digestion and protein precipitation. Certain proteins taking part in the translation process, such as the elongation factors (EF-Tu, EF-Ts) and the ribosomal protein RP-L9, were identified by Edman N-terminal sequencing and have thus been considered for reaction evaluation. The dynamics obtained during the entire process suggest that these translational factors were likewise affected by proteolytic decay.

Ribosome display efficiently selects and evolves high-affinity antibodies in vitro from immune libraries

Ribosome display was applied for affinity selection of antibody single-chain fragments (scFv) from a diverse library generated from mice immunized with a variant peptide of the transcription factor GCN4 dimerization domain. After three rounds of ribosome display, positive scFvs were isolated and characterized. Several different scFvs were selected, but those in the largest group were closely related to each other and differed in 0 to 5 amino acid residues with respect to their consensus sequence, the likely common progenitor. The best scFv had a dissociation constant of (4 +/- 1) x 10(-11) M, measured in solution. One amino acid residue in complementarity determining region L1 was found to be responsible for a 65-fold higher affinity than the likely progenitor. It appears that this high-affinity scFv was selected from the mutations occurring during ribosome display in vitro, and that this constitutes an affinity maturation inherent in this method. The in vitro-selected scFvs could be functionally expressed in the Escherichia coli periplasm with good yields or prepared by in vitro refolding. Thus, ribosome display can be a powerful methodology for in vitro library screening and simultaneous sequence evolution.

Reproducing the natural evolution of protein structural features with the selectively infective phage (SIP) technology. The kink in the first strand of antibody kappa domains

The beta-sandwich structure of immunoglobulin variable domains is characterized by a typical kink in the first strand, which allows the first part of the strand to hydrogen bond to the outer beta-sheet (away from the VH-VL interface) and the second part to the inner beta-sheet. This kink differs in length and sequence between the Vkappa, Vlambda and VH domains and yet is involved in several almost perfectly conserved interactions with framework residues. We have used the selectively infective phage (SIP) system to select the optimal kink region from several defined libraries, using an anti-hemagglutinin single-chain Fv (scFv) fragment as a model system. Both for the kink with the Vkappa domain length and that with the Vlambda length, a sequence distribution was selected that coincides remarkably well with the sequence distribution of natural antibodies. The selected scFv fragments were purified and characterized, and thermodynamic stability was found to be the prime factor responsible for selection. These data show that the SIP technology can be used for optimizing protein structural features by evolutionary approaches.

Recent advances in producing and selecting functional proteins by using cell-free translation

Prokaryotic and eukaryotic in vitro translation systems have recently become the focus of increasing interest for tackling fundamental problems in biochemistry. Cell-free systems can now be used to study the in vitro assembly of membrane proteins and viral particles, rapidly produce and analyze protein mutants, and enlarge the genetic code by incorporating unnatural amino acids. Using in vitro translation systems, display techniques of great potential have been developed for protein selection and evolution. Furthermore, progress has been made to efficiently produce proteins in batch or continuous cell-free translation systems and to elucidate the molecular causes of low yield and find possible solutions for this problem.

RNA-binding proteins tamed

Novel RNA-binding proteins with customized specificities can be isolated by genetic selection from combinatorial libraries. Such proteins have great potential as agents for targeted manipulation of gene expression.

Antibody engineering

Among the most important advances in antibody engineering of this past year is the advent of new tools to study the relationship between protein (including antibody) structure and function. Very rapid large-scale mutational analysis of antibodies is now possible by using in vitro transcription and translation. Ribosome display is a rapidly evolving technology for modifying antibody function that offers several potential advantages over phage display.

Phage-displayed peptide libraries

Over the past year, significant advances have been achieved through the use of phage-displayed peptide libraries. A wide variety of bioactive molecules, including antibodies, receptors and enzymes, have selected high-affinity and/or highly-specific peptide ligands from a number of different types of peptide library. The demonstrated therapeutic potential of some of these peptides, as well as new insights into protein structure and function that peptide ligands have provided, highlight the progress made within this rapidly-expanding field.

Man-made cell-like compartments for molecular evolution

Cellular compartmentalization is vital for the evolution of all living organisms. Cells keep together the genes, the RNAs and proteins that they encode, and the products of their activities, thus linking genotype to phenotype. We have reproduced this linkage in the test tube by transcribing and translating single genes in the aqueous compartments of water-in-oil emulsions. These compartments, with volumes close to those of bacteria, can be recruited to select genes encoding catalysts. A protein or RNA with a desired catalytic activity converts a substrate attached to the gene that encodes it to product. In other compartments, substrates attached to genes that do not encode catalysts remain unmodified. Subsequently, genes encoding catalysts are selectively enriched by virtue of their linkage to the product. We demonstrate the linkage of genotype to phenotype in man-made compartments using a model system. A selection for target-specific DNA methylation was based on the resistance of the product (methylated DNA) to restriction digestion. Genes encoding HaeIII methyltransferase were selected from a 10(7)-fold excess of genes encoding another enzyme.

Antibody phage display technology and its applications

In recent years, the use of display vectors and in vitro selection technologies has transformed the way in which we generate ligands, such as antibodies and peptides, for a given target. Using this technology, we are now able to design repertoires of ligands from scratch and use the power of phage selection to select those ligands having the desired (biological) properties. With phage display, tailor-made antibodies may be synthesized and selected to acquire the desired affinity of binding and specificity for in vitro and in vivo diagnosis, or for immunotherapy of human disease. This review addresses recent progress in the construction of, and selection from phage antibody libraries, together with novel approaches for screening phage antibodies. As the quality of large naïve and synthetic antibody repertoires improves and libraries becomes more generally available, new and exciting applications are pioneered such as the identification of novel antigens using differential selection and the generation of receptor a(nta)gonists. A combination of the design and generation of millions to billions of different ligands, together with phage display for the isolation of binding ligands and with functional assays for identifying (and possibly selecting) bio-active ligands, will open even more challenging applications of this inspiring technology, and provide a powerful tool for drug and target discovery well into the next decade.

Targeting growth factor and cytokine receptors with recombinant peptide libraries

The single-transmembrane-spanning receptors of cytokines and growth factors have historically proven resistant to the small-molecule screening efforts of the pharmaceutical industry. Advances in combinatorial library approaches to ligand discovery have begun to show success with these targets. There are several recent reports of peptides, derived from randomly assembled collections of L-peptides expressed in recombinant display vectors, that are high-affinity antagonists and even agonists of these receptors. These results indicate that molecules much smaller than the natural protein factors can interact effectively with these receptors, and this may lead the way to the discovery of even smaller nonpeptidic agents.

Creating and engineering human antibodies for immunotherapy

Targeting in immunotherapy has traditionally been achieved by using monoclonal rodent antibodies. Despite gene-engineering, there are many problems and limitations associated with the non-human origin, the targeting specificity and the binding strength of these molecules. Now these issues may be addressed in a more rational way, by designing and then shaping, in vitro, the desired human antibodies. This review addresses how this may be achieved by the selection of monoclonal human antibodies from phage display libraries and the engineering of affinity and specificity thereafter. Phage display of antibody fragments has allowed access to large collections of different phage antibodies, created by cloning antibody V-genes from B-cells. Antibodies against any type of antigen may be derived from such repertoires, by rounds of enrichment on antigen and re-amplification. This review presents the state of the art in rational antibody design and creation. It will highlight the strengths of this increasingly important field, which will aid in the generation of tailor-made targeting entities for immunotherapy.

Antibody-ribosome-mRNA (ARM) complexes as efficient selection particles for in vitro display and evolution of antibody combining sites

We describe a rapid, eukaryotic, in vitro method for selection and evolution of antibody combining sites using antibody-ribosome-mRNA (ARM) complexes as selection particles. ARMs carrying single-chain (VH/K) binding fragments specific for progesterone were selected using antigen-coupled magnetic beads; selection simultaneously captured the genetic information as mRNA, making it possible to generate and amplify cDNA by single-step RT-PCR on the ribosome-bound mRNA for further manipulation. Using mutant libraries, antigen-binding ARMs were enriched by a factor of 10(4)-10(5)-fold in a single cycle, with further enrichment in repeated cycles. While demonstrated here for antibodies, the method has the potential to be applied equally for selection of receptors or peptides from libraries.

RNA-peptide fusions for the in vitro selection of peptides and proteins

Covalent fusions between an mRNA and the peptide or protein that it encodes can be generated by in vitro translation of synthetic mRNAs that carry puromycin, a peptidyl acceptor antibiotic, at their 3' end. The stable linkage between the informational (nucleic acid) and functional (peptide) domains of the resulting joint molecules allows a specific mRNA to be enriched from a complex mixture of mRNAs based on the properties of its encoded peptide. Fusions between a synthetic mRNA and its encoded myc epitope peptide have been enriched from a pool of random sequence mRNA-peptide fusions by immunoprecipitation. Covalent RNA-peptide fusions should provide an additional route to the in vitro selection and directed evolution of proteins.

In vitro virus: bonding of mRNA bearing puromycin at the 3'-terminal end to the C-terminal end of its encoded protein on the ribosome in vitro

Adequate means for genotype assignment to phenotype is essential in evolutionary molecular engineering. In this study, construction of 'in vitro virus' was carried out in which a genotype molecule (mRNA) covalently binds to the phenotype molecule (protein) through puromycin on the ribosome in a cell-free translation system. Bonding efficiency was approximately 10%, thus indicating a population of the in vitro virus to have approximately 10(12) protein variants, this number being 10(4) that in the phage display. The in vitro virus is useful for examining protein evolution in a test tube and the results may possibly serve as basis for a general method for selecting proteins possessing the most desirable functions.

3D structural information as a guide to protein engineering using genetic selection

A great variety of protein systems have been investigated in the past year using structure-guided evolutionary strategies. On the basis of available 3D structural information, critical regions of proteins have been targeted for randomizing mutagenesis and active variants of the corresponding genes have been selected. These approaches help characterize structural and mechanistic features of proteins and have important implications for design.

Peptide agonist of the thrombopoietin receptor as potent as the natural cytokine

Two families of small peptides that bind to the human thrombopoietin receptor and compete with the binding of the natural ligand thrombopoietin (TPO) were identified from recombinant peptide libraries. The sequences of these peptides were not found in the primary sequence of TPO. Screening libraries of variants of one of these families under affinity-selective conditions yielded a 14-amino acid peptide (Ile-Glu-Gly-Pro-Thr-Leu-Arg-Gln-Trp-Leu-Ala-Ala-Arg-Ala) with high affinity (dissociation constant approximately 2 nanomolar) that stimulates the proliferation of a TPO-responsive Ba/F3 cell line with a median effective concentration (EC50) of 400 nanomolar. Dimerization of this peptide by a carboxyl-terminal linkage to a lysine branch produced a compound with an EC50 of 100 picomolar, which was equipotent to the 332-amino acid natural cytokine in cell-based assays. The peptide dimer also stimulated the in vitro proliferation and maturation of megakaryocytes from human bone marrow cells and promoted an increase in platelet count when administered to normal mice.

In vitro selection and evolution of functional proteins by using ribosome display

We report here a system with which a correctly folded complete protein and its encoding mRNA both remain attached to the ribosome and can be enriched for the ligand-binding properties of the native protein. We have selected a single-chain fragment (scFv) of an antibody 10(8)-fold by five cycles of transcription, translation, antigen-affinity selection, and PCR. The selected scFv fragments all mutated in vitro by acquiring up to four unrelated amino acid exchanges over the five generations, but they remained fully compatible with antigen binding. Libraries of native folded proteins can now be screened and made to evolve in a cell-free system without any transformation or constraints imposed by the host cell.

High-affinity peptide ligands to prostate-specific antigen identified by polysome selection

We have used the polysome-selection method to isolate peptide ligands that bind with high affinity to Prostate-Specific Antigen (PSA), an important prostate-cancer marker. Two random libraries, each encoding approximately 10(12) random peptides, were transcribed into RNA and translated in vitro. Polysomes were panned by affinity selection of the nascent peptides against immobilized PSA. Over 30% of the selected species had significant affinity for PSA; the dissociation constant of the complex formed by the best isolate with PSA was < 10(-9) M. Formation of streptavidin conjugates of selected peptides improved their affinities and, in one case, virtually eliminated non-specific binding. The polysome-selection method can be used to produce high-affinity peptide ligands of potential use in diagnostic and therapeutic procedures.

Structural and mechanistic determinants of affinity and specificity of ligands discovered or engineered by phage display

The scope and utility of phage display is reviewed with emphasis on medical applications and structure-based ligand and drug design, from literature mostly after 1994. General principles by which phage-displayed peptides achieve affinity and selectivity for targets are described, along with selected structural or mechanistic studies of the binding of peptides or proteins discovered or engineered by phage display. Such engineered proteins whose wild-type or mutant crystal or 2D-NMR structures yield insight about the basis for enhanced affinity or altered specificity include antibodies, zinc fingers, human growth hormone, protein A, and atrial natriuretic peptide. Structures of complexes of de novo phage-discovered peptide ligands with targets such as the Src SH3 domain, streptavidin, and erythropoietin receptor reveal the structural basis for receptor-peptide recognition in these systems.

Codon-reading specificity of an unmodified form of Escherichia coli tRNA1Ser in cell-free protein synthesis

Unmodified tRNA molecules are useful for many purposes in cell-free protein biosynthesis, but there is little information about how the lack of tRNA post-transcriptional modifications affects the coding specificity for synonymous codons. In the present study, we prepared an unmodified form of Escherichia coli tRNA1Ser, which originally has the cmo5UGA anticodon (cmo5U = uridine 5-oxyacetic acid) and recognizes the UCU, UCA and UCG codons. The codon specificity of the unmodified tRNA was tested in a cell-free protein synthesis directed by designed mRNAs under competition conditions with the parent tRNA1Ser. It was found that the unmodified tRNA with the UGA anti-codon recognizes the UCA codon nearly as efficiently as the modified tRNA. The unmodified tRNA recognized the UCU codon with low, but detectable efficiency, whereas no recognition of the UCC and UCG codons was detected. Therefore, the absence of modifications makes this tRNA more specific to the UCA codon by remarkably reducing the efficiencies of wobble reading of other synonymous codons, without a significant decrease in the UCA reading efficiency.

Identification of D-peptide ligands through mirror-image phage display

Genetically encoded libraries of peptides and oligonucleotides are well suited for the identification of ligands for many macromolecules. A major drawback of these techniques is that the resultant ligands are subject to degradation by naturally occurring enzymes. Here, a method is described that uses a biologically encoded library for the identification of D-peptide ligands, which should be resistant to proteolytic degradation. In this approach, a protein is synthesized in the D-amino acid configuration and used to select peptides from a phage display library expressing random L-amino acid peptides. For reasons of symmetry, the mirror images of these phage-displayed peptides interact with the target protein of the natural handedness. The value of this approach was demonstrated by the identification of a cyclic D-peptide that interacts with the Src homology 3 domain of c- SRC. Nuclear magnetic resonance studies indicate that the binding site for this D-peptide partially overlaps the site for the physiological ligands of this domain.

Alternate protein frameworks for molecular recognition

In an effort to determine whether proteins with structures other than the immunoglobulin fold can be used to mimic the ligand binding properties of antibodies, we generated a library from the four-helix bundle protein cytochrome b562 in which the two loops were randomized. Panning of this library against the bovine serum albumin (BSA) conjugate of N-methyl-p-nitrobenzylamine derivative 1 by phage display methods yielded cytochromes in which residues Trp-20, Arg-21, and Ser-22 in loop A and Arg-83 and Trp-84 in loop B were conserved. The individual mutants, which fold into native-like structure, bind selectively to the BSA-1 conjugate with micromolar dissociation constants (Kd), in comparison to a monoclonal antibody that binds selectively to 1 with a Kd of 290 nM. These and other antibody-like receptors may prove useful as therapeutic agents or as reagents for both intra- and extracellular studies.

Protein-peptide interactions analyzed with the yeast two-hybrid system

The yeast two-hybrid system was used to screen a library of random peptides fused to a transcriptional activation domain in order to identify peptides capable of binding to the retinoblastoma protein (Rb). Seven peptides were identified, all of which contain the Leu-X-Cys-X-Glu motif found in Rb-binding proteins, although their activity in the yeast assay varied over a 40-fold range. Mutagenesis of the DNA encoding two of these peptides followed by screening in the two-hybrid system allowed the delineation of residues apart from the invariant Leu, Cys and Glu that affect binding to Rb. Binding affinities of a peptide and one of its variants to Rb, determined by surface plasmon resonance, correlated with results from the two-hybrid assay. This method offers several advantageous features compared to existing technology for screening peptide libraries: in vivo detection of protein-peptide interactions, high sensitivity, the capacity for rapid genetic screening to identify stronger and weaker binding peptide variants, and the use of a simple assay (transcriptional activity) as a means to assess binding affinity.

Combinatorial drug discovery: which methods will produce the greatest value?

Combinatorial strategies are important new approaches to drug discovery, and it seems quite likely that they will result in the discovery of interesting potential pharmaceuticals. However, it is less clear whether combinatorial approaches will result in quantum advances in therapeutics. Nor is there general agreement about the factors most important in defining how combinatorial strategies will provide value to the discovery of lead and therapeutic compounds. In this review, we propose criteria that define the value of combinatorial strategies and categorize the various approaches by: (a) the type of chemical space to be searched, (b) the tactics employed to synthesize and screen libraries, and (c) the structures of individual molecules in libraries. We evaluate the strengths and weaknesses of the various strategies and suggest milestones that can help to track their success.

DNA shuffling by random fragmentation and reassembly: in vitro recombination for molecular evolution

Computer simulations of the evolution of linear sequences have demonstrated the importance of recombination of blocks of sequence rather than point mutagenesis alone. Repeated cycles of point mutagenesis, recombination, and selection should allow in vitro molecular evolution of complex sequences, such as proteins. A method for the reassembly of genes from their random DNA fragments, resulting in in vitro recombination is reported. A 1-kb gene, after DNase I digestion and purification of 10- to 50-bp random fragments, was reassembled to its original size and function. Similarly, a 2.7-kb plasmid could be efficiently reassembled. Complete recombination was obtained between two markers separated by 75 bp; each marker was located on a separate gene. Oligonucleotides with 3' and 5' ends that are homologous to the gene can be added to the fragment mixture and incorporated into the reassembled gene. Thus, mixtures of synthetic oligonucleotides and PCR fragments can be mixed into a gene at defined positions based on homology. As an example, a library of chimeras of the human and murine genes for interleukin 1 beta has been prepared. Shuffling can also be used for the in vitro equivalent of some standard genetic manipulations, such as a backcross with parental DNA. The advantages of recombination over existing mutagenesis methods are likely to increase with the numbers of cycles of molecular evolution.