Molecular breeding of viruses.

Genetic recombination is a major force driving the evolution of many viruses. Recombination between two copackaged retroviral genomes may occur at rates as high as 40% per replication cycle. This enables genetic information to be shuffled rapidly, leading to recombinants with new patterns of mutations and phenotypes. The in vitro process of DNA shuffling (molecular breeding) mimics this mechanism on a vastly parallel and accelerated scale. Multiple homologous parental sequences are recombined in parallel, leading to a diverse library of complex recombinants from which desired improvements can be selected. Different proteins and enzymes have been improved using DNA shuffling. We report here the first application of molecular breeding to viruses. A single round of shuffling envelope sequences from six murine leukaemia viruses (MLV) followed by selection yielded a chimaeric clone with a completely new tropism for Chinese Hamster Ovary (CHOK1) cells. The composition and properties of the selected clone indicated that this particular permutation of parental sequences cannot be readily attained by natural retroviral recombination. This example demonstrates that molecular breeding can enhance the inherently high evolutionary potential of retroviruses to obtain desired phenotypes. It can be an effective tool, when information is limited, to optimize viruses for gene therapy and vaccine applications when multiple complex functions must be simultaneously balanced.

Directed evolution of proteins by exon shuffling.

Evolution of eukaryotes is mediated by sexual recombination of parental genomes. Crossovers occur in random, but homologous, positions at a frequency that depends on DNA length. As exons occupy only 1% of the human genome and introns about 24%, by far most of the crossovers occur between exons, rather than inside. The natural process of creating new combinations of exons by intronic recombination is called exon shuffling. Our group is developing in vitro formats for exon shuffling and applying these to the directed evolution of proteins. Based on the splice frame junctions, nine classes of exons and three classes of introns can be distinguished. Splice frame diagrams of natural genes show how the splice frame rules govern exon shuffling. Here, we review various approaches to constructing libraries of exon-shuffled genes. For example, exon shuffling of human pharmaceutical proteins can generate libraries in which all of the sequences are fully human, without the point mutations that raise concerns about immunogenicity.

Molecular evolution of an arsenate detoxification pathway by DNA shuffling.

Functional evolution of an arsenic resistance operon has been accomplished by DNA shuffling, involving multiple rounds of in vitro recombination and mutation of a pool of related sequences, followed by selection for increased resistance in vivo. Homologous recombination is achieved by random fragmentation of the PCR templates and reassembly by primerless PCR. Plasmid-determined arsenate resistance from plasmid pl258 encoded by genes arsR, arsB, and arsC was evolved in Escherichia coli. Three rounds of shuffling and selection resulted in cells that grew in up to 0.5 M arsenate, a 40-fold increase in resistance. Whereas the native plasmid remained episomal, the evolved operon reproducibly integrated into the bacterial chromosome. In the absence of shuffling, no increase in resistance was observed after four selection cycles, and the control plasmid remained episomal. The integrated ars operon had 13 mutations. Ten mutations were located in arsB, encoding the arsenite membrane pump, resulting in a fourfold to sixfold increase in arsenite resistance. While arsC, the arsenate reductase gene, contained no mutations, its expression level was increased, and the rate of arsenate reduction was increased 12-fold. These results show that DNA shuffling can improve the function of pathways by complex and unexpected mutational mechanisms that may be activated by point mutation. These mechanisms may be difficult to explain and are likely to be overlooked by rational design.

Directed evolution of thymidine kinase for AZT phosphorylation using DNA family shuffling.

The thymidine kinase (TK) genes from herpes simplex virus (HSV) types 1 and 2 were recombined in vitro with a technique called DNA family shuffling. A high-throughput robotic screen identified chimeras with an enhanced ability to phosphorylate zidovudine (AZT). Improved clones were combined, reshuffled, and screened on increasingly lower concentrations of AZT. After four rounds of shuffling and screening, two clones were isolated that sensitize Escherichia coli to 32-fold less AZT compared with HSV-1 TK and 16,000-fold less than HSV-2 TK. Both clones are hybrids derived from several crossover events between the two parental genes and carry several additional amino acid substitutions not found in either parent, including active site mutations. Kinetic measurements show that the chimeric enzymes had acquired reduced K(M) for AZT as well as decreased specificity for thymidine. In agreement with the kinetic data, molecular modeling suggests that the active sites of both evolved enzymes better accommodate the azido group of AZT at the expense of thymidine. Despite the overall similarity of the two chimeric enzymes, each contains key contributions from different parents in positions influencing substrate affinity. Such mutants could be useful for anti-HIV gene therapy, and similar directed-evolution approaches could improve other enzyme-prodrug combinations.

Improved green fluorescent protein by molecular evolution using DNA shuffling.

Green fluorescent protein (GFP) has rapidly become a widely used reporter of gene regulation. However, for many organisms, particularly eukaryotes, a stronger whole cell fluorescence signal is desirable. We constructed a synthetic GFP gene with improved codon usage and performed recursive cycles of DNA shuffling followed by screening for the brightest E. coli colonies. A visual screen using UV light, rather than FACS selection, was used to avoid red-shifting the excitation maximum. After 3 cycles of DNA shuffling, a mutant was obtained with a whole cell fluorescence signal that was 45-fold greater than a standard, the commercially available Clontech plasmid pGFP. The expression level in E. coli was unaltered at about 75% of total protein. The emission and excitation maxima were also unchanged. Whereas in E. coli most of the wildtype GFP ends up in inclusion bodies, unable to activate its chromophore, most of the mutant protein is soluble and active. Three amino acid mutations appear to guide the mutant protein into the native folding pathway rather than toward aggregation. Expressed in Chinese Hamster Ovary (CHO) cells, this shuffled GFP mutant showed a 42-fold improvement over wildtype GFP sequence, and is easily detected with UV light in a wide range of assays. The results demonstrate how molecular evolution can solve a complex practical problem without needing to first identify which process is limiting. DNA shuffling can be combined with screening of a moderate number of mutants. We envision that the combination of DNA shuffling and high throughput screening will be a powerful tool for the optimization of many commercially important enzymes for which selections do not exist.

DNA shuffling of subgenomic sequences of subtilisin.

DNA family shuffling of 26 protease genes was used to create a library of chimeric proteases that was screened for four distinct enzymatic properties. Multiple clones were identified that were significantly improved over any of the parental enzymes for each individual property. Family shuffling, also known as molecular breeding, efficiently created all of the combinations of parental properties, producing a great diversity of property combinations in the progeny enzymes. Thus, molecular breeding, like classical breeding, is a powerful tool for recombining existing diversity to tailor biological systems for multiple functional parameters.

Evolution of a cytokine using DNA family shuffling.

DNA shuffling of a family of over 20 human interferon-alpha (Hu-IFN-alpha) genes was used to derive variants with increased antiviral and antiproliferation activities in murine cells. A clone with 135,000-fold improved specific activity over Hu-IFN-alpha2a was obtained in the first cycle of shuffling. After a second cycle of selective shuffling, the most active clone was improved 285,000-fold relative to Hu-IFN-alpha2a and 185-fold relative to Hu-IFN-alpha1. Remarkably, the three most active clones were more active than the native murine IFN-alphas. These chimeras are derived from up to five parental genes but contained no random point mutations. These results demonstrate that diverse cytokine gene families can be used as starting material to rapidly evolve cytokines that are more active, or have superior selectivity profiles, than native cytokine genes.

Breeding of retroviruses by DNA shuffling for improved stability and processing yields.

Manufacturing of retroviral vectors for gene therapy is complicated by the sensitivity of these viruses to stress forces during purification and concentration. To isolate viruses that are resistant to these manufacturing processes, we performed breeding of six ecotropic murine leukemia virus (MLV) strains by DNA shuffling. The envelope regions were shuffled to generate a recombinant library of 5 x 106 replication-competent retroviruses. This library was subjected to the concentration process three consecutive times, with amplification of the surviving viruses after each cycle. Several viral clones with greatly improved stabilities were isolated, with the best clone exhibiting no loss in titer under conditions that reduced the titers of the parental viruses by 30- to 100-fold. The envelopes of these resistant viruses differed in DNA and protein sequence, and all were complex chimeras derived from multiple parents. These studies demonstrate the utility of DNA shuffling in breeding viral strains with improved characteristics for gene therapy.

Protein evolution by molecular breeding.

Natural evolution has guided the development of 'molecular breeding' processes used in the laboratory for the rapid modification of subgenomic sequences including single genes. The most significant recent development has been the in vitro permutation of natural diversity. Homologous recombination of multiple related sequences produced high-quality libraries of chimeric sequences encoding proteins with functions that differ dramatically from any of the parents. Increasingly powerful screening methods are also being developed, allowing these libraries to be screened for novel biocatalysts.

Affinity selective isolation of ligands from peptide libraries through display on a lac repressor "headpiece dimer".

DNA binding by the Escherichia coli lac repressor is mediated by the approximately 60 amino acid residue 'headpiece' domain. The dimer of headpiece domains that binds to the lac operator is normally formed by association of the much larger approximately 300 amino acid residue C-terminal domain. We have used in vitro selection to isolate 'headpiece dimer' molecules containing two headpiece domains connected via a short peptide linker. These proteins bind plasmid molecules with sufficient stability to allow association of a peptide epitope displayed at the C terminus of the headpiece dimer with the plasmid encoding that peptide. Libraries of peptides displayed on the C terminus of a headpiece dimer can be screened for specific receptor ligands by affinity enrichment of peptide-headpiece dimer-plasmid complexes using an immobilized receptor. After each round of enrichment, transformation of E. coli with recovered plasmids permits amplification of the selected population. After several rounds of enrichment, sequencing of individual clones reveals the structure of the selected peptides. Headpiece dimer libraries allow selection of peptide ligands of higher average affinity than similar libraries based on the intact lac repressor. Interestingly, the presence of the lac operator is not required for plasmid binding by the headpiece dimer protein.

Novel enzyme activities and functional plasticity revealed by recombining highly homologous enzymes.

BACKGROUND: Directed evolution by DNA shuffling has been used to modify physical and catalytic properties of biological systems. We have shuffled two highly homologous triazine hydrolases and conducted an exploration of the substrate specificities of the resulting enzymes to acquire a better understanding of the possible distributions of novel functions in sequence space. RESULTS: Both parental enzymes and a library of 1600 variant triazine hydrolases were screened against a synthetic library of 15 triazines. The shuffled library contained enzymes with up to 150-fold greater transformation rates than either parent. It also contained enzymes that hydrolyzed five of eight triazines that were not substrates for either starting enzyme. CONCLUSIONS: Permutation of nine amino acid differences resulted in a set of enzymes with surprisingly diverse patterns of reactions catalyzed. The functional richness of this small area of sequence space may aid our understanding of both natural and artificial evolution.

Applications of DNA shuffling to pharmaceuticals and vaccines.

DNA shuffling is a practical process for directed molecular evolution which uses recombination to dramatically accelerate the rate at which one can evolve genes. Single and multigene traits that require many mutations for improved phenotypes can be evolved rapidly. DNA shuffling technology has been significantly enhanced in the past year, extending its range of applications to small molecule pharmaceuticals, pharmaceutical proteins, gene therapy vehicles and transgenes, vaccines and evolved viruses for vaccines, and laboratory animal models.

Increased antibody expression from Escherichia coli through wobble-base library mutagenesis by enzymatic inverse PCR.

We tested the value of a new library mutagenesis approach, called library enzymatic inverse PCR (LEIPCR), for expression-level enhancement of antibody Fv fragments produced in Escherichia coli. The production level of active, metal chelate-specific antibody from our constructs is limited by a low expression level of the second, heavy-chain cistron. To increase the production level, LEIPCR was applied to the wobble bases of the second cistron leader peptide. In LEIPCR mutagenesis, the entire plasmid is amplified using mutagenic primers with class-IIS restriction endonuclease (ENase) sites at their 5' ends. The PCR product is digested with the class-IIS ENase (here, BsaI; GGTCTCN[symbol: see text]NNNN[symbol: see text]), which removes its own recognition sequence, and the ends are self-ligated. Thus, LEIPCR can be used to make plasmid mutant libraries regardless of the nucleotide sequence, and independent of available ENase sites. The resulting library of 10(7) wobble mutants was screened for active Fv by a colony filter lift. A selected mutant was shown to produce between four- and elevenfold more active Fv than the wild type (wt), and fivefold more heavy chain. Mutations outside of the leader peptide were shown not to be involved. The mutated areas of the mRNAs of two different up-mutants may have less secondary structure than the wt. Thus, the sequence of the mRNA of the second leader peptide was limiting to the expression level of heavy-chain and active Fv.

Single-step assembly of a gene and entire plasmid from large numbers of oligodeoxyribonucleotides.

Here, we describe assembly PCR as a method for the synthesis of long DNA sequences from large numbers of oligodeoxyribonucleotides (oligos). The method, which is derived from DNA shuffling [Stemmer, Nature 370 (1994a) 389-391], does not rely on DNA ligase but instead relies on DNA polymerase to build increasingly longer DNA fragments during the assembly process. A 1.1-kb fragment containing the TEM-1 beta-lactamase-encoding gene (bla) was assembled in a single reaction from a total of 56 oligos, each 40 nucleotides (nt) in length. The synthetic gene was PCR amplified and cloned in a vector containing the tetracycline-resistance gene (TcR) as the sole selectable marker. Without relying on ampicillin (Ap) selection, 76% of the TcR colonies were ApR, making this approach a general method for the rapid and cost-effective synthesis of any gene. We tested the range of assembly PCR by synthesizing, in a single reaction vessel containing 134 oligos, a high-molecular-mass multimeric form of a 2.7-kb plasmid containing the bla gene, the alpha-fragment of the lacZ gene and the pUC origin of replication. Digestion with a unique restriction enzyme, followed by ligation and transformation in Escherichia coli, yielded the correct plasmid. Assembly PCR is well suited for several in vitro mutagenesis strategies.

Enzymatic inverse PCR: a restriction site independent, single-fragment method for high-efficiency, site-directed mutagenesis.

A new method is described for rapid site-directed mutagenesis of plasmid DNA. The new method, termed enzymatic inverse polymerase chain reaction (EIPCR), uses inverse PCR to amplify the entire plasmid. The key step to EIPCR is the incorporation of identical class 2s restriction sites in both primers. Class 2s restriction enzymes have a recognition site that is located 5' of the cut site (e.g., BsaI: GGTCTCN'NNNN,). Thus, after completing PCR, the ends of the full-length linearized plasmid are digested with the class 2s enzyme incorporated into the primers. The enzyme cuts off its entire recognition site and leaves the plasmid with compatible overhangs on both ends. Thus, in the ligation the only part that becomes part of the plasmid is the NNNN overhang, which can be made to be the native sequence. We have used the method for many plasmids and several class 2s enzymes. As an example, we report here the use of EIPCR for an insertion into pUC19 containing an inactive lacZ alpha-peptide, causing a frameshift that restores lacZ alpha-activity. Of 300 colonies evaluated, greater than 95% had the expected blue phenotype. The BsaI overhangs were correctly combined in all of the 35 blue colonies analyzed by restriction digestion and in all four clones that were sequenced. EIPCR is compared with four related PCR-based mutagenesis techniques. The major advantage of EIPCR over the other methods is the combination of greater than 95% correctly mutated clones with the need for only two PCR primers.

Selection of an active single chain Fv antibody from a protein linker library prepared by enzymatic inverse PCR.

Enzymatic inverse PCR mutagenesis was developed as a simple and reliable method for the construction of large libraries of site-directed mutants. Enzymatic inverse PCR library mutagenesis uses a single PCR fragment and is restriction-site independent. The usefulness of the technique was demonstrated by the design of a single chain linker for an antibody Fv fragment without computer modeling. The Fv fragment of an antibody specific for a metal chelate was expressed in active form in the periplasm of E. coli. The light and the heavy chains of the Fv are expressed as a bicistronic mRNA. Enzymatic inverse PCR mutagenesis was used to construct a library of 3 x 10(5) Fv mutants, in which the C-terminus of the light chain was connected to the N-terminus of the heavy chain by a 15-amino acid peptide linker of variable composition. After plating, active mutant colonies were identified by screening colony filter lifts with a radiolabeled hapten, N'-(2-hydroxyethyl)-p-thioureidobenzyl EDTA. About 0.2% of the mutants were positive, and a selected sFv clone was shown to have the same affinity as the Fv (9 x 10(9)) and was similar to the whole antibody (11 x 10(9)). This example compares favorably with both of the other approaches to constructing sFv's; namely, molecularly modeled linkers as well as universal linkers, which have often yielded significantly lower affinities than whole antibodies or Fabs. The enzymatic inverse PCR library mutagenesis approach is simple and reliable and can be used to obtain linkers for the great majority of antibodies for which no structural data are available. More generally, it can be used to modify DNA coding for any structural protein or regulatory element.