Directed evolution of a recombinase for improved genomic integration at a native human sequence

Christopher R. Sclimenti 0 Bhaskar Thyagarajan 0 Michele P. Calos 0 0 Department of Genetics, Stanford University School of Medicine , Stanford, CA 94305-5120, USA We previously established that a unidirectional sitespecific recombinase, the phage C31 integrase, can mediate integration into mammalian chromosomes. The enzyme directs integration of plasmids bearing the phage attB recognition site into pseudo attP sites, a set of native sequences related to the phage attP recognition site. Here we use two cycles of DNA shuffling and screening in Escherichia coli to obtain evolved integrases that possess significant improvements in integration frequency and sequence specificity at a pseudo attP sequence located on human chromosome 8, when measured in the native genomic environment of living human cells. Such integrases represent custom integration tools that will be useful for modifying the genomes of higher eukaryotic cells. - Genetics still lacks tools for efficient site-specific integration of DNA sequences into the genomes of living higher cells. We have been investigating site-specific recombinases that may have the potential to provide such tools. This class of enzymes performs efficient recombination at recognition sites having the correct length to be present at low frequencies in large genomes (1). Phage integrases are site-specific recombinases that perform a unidirectional integration reaction with no competing reverse reaction, and may therefore be optimal for achieving integration at high frequency. The integrase from the Streptomyces phage C31 (2,3) requires no cofactors (4) and performs efficient recombination between its attB and attP recognition sites on extrachromosomal plasmids in the mammalian cell environment (5). Furthermore, the C31 integrase recognizes native sequences in the human and mouse genomes that possess partial sequence identity to attP, called pseudo attP sites, and mediates the integration of plasmids bearing an attB site into such pseudo attP sequences (6). This reaction occurs at a frequency at least 510-fold above that of random integration and involves a hierarchy of endogenous sequences bearing varying degrees of identity to attP (6). The wild-type C31 integrase thus carries out the general type of reaction we are seeking, directing relatively efficient and site-specific integration, compared with random integration. However, there appeared to be an opportunity to create integration reagents with even greater utility by using the wild-type enzyme as the starting material for a directed evolution study in which we might generate enzymes with greater sequence specificity and higher integration frequencies. The requirements for this outcome would be expected to include alteration of the DNA recognition domain of the enzyme so that it avidly recognizes a particular native pseudo attP sequence, while losing ability to react with related sequences. Because the integrase recognizes both attP and attB, although not addressed in the present work, an optimal result might require parallel changes in attB. As well, the overall catalytic efficiency of the enzyme must increase. Gain of these features might require alterations involving both the DNA binding domain of the protein and the active site. We lack detailed structural information about the C31 integrase. Even if we had such data, it would be difficult to predict how to engineer the enzyme. In this situation, non-rational methods are valuable to generate large libraries of variants from which enzymes having the desired properties can be found by screening (7,8). A strikingly effective directed evolution strategy that includes mutagenesis and combinatorial exchanges is the DNA shuffling protocol (911). We have applied DNA shuffling in combination with a genetic screen that is capable of identifying mutant enzymes that possess an increased ability to perform integration at the desired sequence. We report here on such a screen and how we have used it to isolate new integrases that now display improved specificity for an endogenous sequence in the context of the native genome of a living human cell. This type of technology, in conjunction with more powerful screens, is likely to be effective for generating integration reagents customized to act at a wide variety of target sequences in a broad range of species. Such integration tools would have extensive applications across genetics, because they can be used to insert DNA site specifically and efficiently in processes such as functional genomics analysis, gene therapy, modification of the genomes of stem cells and construction of transgenic organisms. MATERIALS AND METHODS The screening assay uses two plasmids, called the resident plasmid, pRES-A, and the cloning plasmid, pINT-T. Both plasmids have compatible origins and can be propagated together in the same bacterial cell. The plasmid pTSK30 (12), which is temperature sensitive for replication, was used as the backbone for pRES-A (Fig. 1A). pTSK30 was cut with DrdI and SmaI to remove the lacZ alpha gene, which was replaced with the components below through a series of cloning steps involving several custom linkers (details available upon request). The phage T5 promoter (13) was added to provide a strong promoter for lacZ. The pseudo attP site A (Fig. 1C) from the human genome (6) (GenBank accession number AF333429) was added downstream of the T5 promoter. The 470-bp A sequence was copied from human genomic DNA by PCR with the primers 5-ATTTGTAGAACTATTATGGG and 5-AAGTCTTCTGGCTATACAGG. A stuffer sequence was then inserted to act as a spacer between the two recombining att sites and to prevent transcriptional read-through. The stuffer sequence was obtained by PCR from human genomic DNA and encompassed a 2.3-kb GC-rich fragment from an intron of the human FGFR3 gene. A 45-bp C31 attB site made from the oligonucleotides 5-CGCGCCTGCGGGTGCCAGGGCGTGCCCTTGGGCTCCCCGGGCGCGTACTCCGG and 5-CGCGCCGGAGTACGCGCCCGGGGAGCCCAAGGGCACGCCCTGGCACCCGCAGG was cloned downstream of the stuffer. The full-length lacZ gene from pCMVSPORTgal (Life Technologies) was added. A temperature-sensitive muNStant of the lac repressor gene (lacI TS) was introduced from the plasmid pNH40lacIqTS (14). To make pINT-T (Fig. 1B), the pInt plasmid (5) was modified. To first make the vector tetracycline resistant (TcR), pINT was cut with DraIII and PflMI and made blunt with T4 polymerase. This step provided a position for the TcR gene and also removed the kanamycin resistance gene from the pInt vector. From pBR322, the TcR gene was removed with EcoRI and PflMI, made blunt, and used to replace the kanamycin resistance gene, resulting in the plasmid pINT-Tc 2nd(+). pINT-Tc 2nd(+) was cut with BstEII and SpeI, which removed the integrase gene, producing pREC. A linker created with the oligonucleotides 5-GTCACGCTCGAGAGATCTGA and 5-CTAGTCAGATCTCTCGAGC was placed into these sites, which introduced unique BglII and XhoI restriction enzyme (...truncated)