AU4205500A - Method for selecting recombinase variants with altered specificity - Google Patents

Description

WO 00/60091 PCT/USOO/09154 METHOD FOR SELECTING RECOMBINASE VARIANTS WITH ALTERED SPECIFICITY BACKGROUND OF THE INVENTION Recombinases, integrases and resolvases (collectively referred to herein 5 as recombinases) mediate the site-specific recombination of DNA. These recombinases were first identified in phage that integrate into host chromosomes. Such integration allows the phage to remain latent in the cell as a prophage. Site-specific recombinases catalyze conservative DNA rearrangements 10 at specific target sequences. The 38 kDa Cre recombinase (cyclization recombination), derived from the bacteriophage P1, is a well characterized and widely used enzyme of the Integrase family (reviewed by Sauer, Methods, 14:381-392 (1998)). Cre plays two essential roles in the life cycle of Pl: First, it provides a host-independent mechanism for P1's genome cyclization after 15 infection, which can be important when the recombination system of the host is compromised. Second, Cre resolves dimerized P1 prophage plasmids to guarantee proper segregation during cell division. Cre acts on a 34 bp sequence located on both ends of the linear P1 genome, that is called loxP (locus of crossover of P1; Sternberg and Hamilton, 20 J. Mol. Biol., 150:467-486 (1981)), loxP consists of two 13 bp inverted repeats flanking a non-palindromic 8 bp core that defines the assigned direction of the sequence (as shown on the upper part of Figure 1). Depending on this direction, recombination catalyzed by Cre leads to excision of insertion of DNA flanked by loxP sites orientated in the same direction (indicated by loxP 2 ), but leads to 25 inversion when oriented in the opposite direction (Figure 2). In general, Cre-recombination involves the following four events: (i) DNA binding, (ii) synapsis (as defined below), (iii) cleavage, and (iv) strand exchange. To study this process in greater detail, mutants defective for each step have been isolated using several screening procedures (Wierzbicki et al., J. 30 Mol. Biol., 195:785-794 (1987)). In addition, the crystal structure of Cre complexed with an artificial suicide substrate has been recently resolved, providing additional insights into site-specific recombination (Guo et al., 1 WO 00/60091 PCT/USO0/09154 Nature, 389:40-46 (1997)). From these studies, the following has been proposed: Four interacting Cre molecules are necessary for recombination between two lox sites, with each enzyme binding one inverted repeat plus the two outermost bp of the non-symmetric core region (DNA binding). This leads 5 to the formation of a clamp, allowing DNA contacts in the major, as well as in the minor groove. In the step referred to as synapsis, the two lox sites with the bound Cre molecules, are aligned in parallel leading to an approximate 1000 bending of the DNA. In the following step of strand cleavage, one of the two Cre molecules on each lox site causes a staggered cut in the core region, as 10 indicated by the vertical arrows in Figure 1. This leads to a 6 bp 5' overhang and a covalent 3' phosphotyrosine linkage between the catalytic residue tyrosine 324 of Cre and the guanine (position 4) at the cleaving site of loxP. The created phosphotyrosine intermediate is thought to provide the energy for the reaction, thereby explaining why Cre does not require an external energy source. In the 15 next step, the first strand is exchanged between the two nicked lox sites, creating an intermediate, named Holliday structure (Sigal and Alberts, J. Mol. Biol., 71:789-793 (1972)). Of note, this first strand exchange is asymmetric, since the bottom strand (Figure 1) is always exchanged first (Hoess et al., Proc. Natl. Acad Sci. USA, 84:6840-6844 (1987)). During the final step, the second 20 strand is exchanged and Cre released from its substrate. Because of the simplicity and the ability of Cre to function in yeast and mammalian cells (Sauer, B., Mol. Cell. Biol., 7:2087-2096 (1987); Sauer and Henderson, Proc. Natl. Acad Sci. USA, 85:5166-5170 (1988), Sauer and Henderson, Nucl. Acids Res., 17:147-161 (1989), and Sauer and Henderson, The 25 New Biologist, 2:441-449 (1990), Cre assisted site-specific recombination has become an important tool for efficient, specific, and conditional manipulations of eukaryotic genomes (Lakso et al., Proc. Natl. A cad Sci. USA, 89:6232-6236 (1992)): Kilby et al., Genet., 9:413-421 (1993); Sauer, B., Meth. enzymol., 225:890-900 (1993); KUhn et al., Science, 269:1427-1429 (1995); Metzger et 30 al., Pro. Natl. Acad Sci. USA, 92:6991-6995 (1995). However, there are some inconveniences for the successful use of Cre related technologies, that include the following: (i) lox sites need to be 2 WO 00/60091 PCT/USOO/09154 introduced by homologous recombination at the desired region into the genome before Cre can be used, (ii) the frequency of correct site-specific recombination due to Cre expression is not 100%, and consequently, (iii) selectable markers are necessary in most strategies involving Cre for genome manipulation in 5 higher eukaryotes. These markers, e.g. neo or TK, may introduce problems in subsequent studies, particular in those related to animal development. The number of available selectable markers that can be used in limited also. Additional site-specific recombinases that also function efficiently in eukaryotic systems, but recognize different sites from lox would be helpful. Similar 10 inconveniences limit the usefulness of other recombinases. Therefore, it is an object of the present invention to provide a method of identifying variant recombinases that can mediate recombination between variant recombination sites. It is another object of the present invention to provide variant 15 recombinases that can mediate recombination between variant recombination sites. It is another object of the present invention to provide a method of recombining nucleic acid molecules in vitro and in vivo. It is another object of the present invention to provide Cre variants that 20 recognize variant recombination sites. BRIEF SUMMARY OF THE INVENTION Disclosed is a method for identifying variant forms of recombinases that can mediate recombination between variant recombination sites. The method involves producing mutant recombinases and testing the mutant recombinases 25 with specially designed constructs. The constructs contain variant recombination sites that are not recognized by non-mutant recombinase but will undergo recombination in the presence of a mutant recombinase with altered specificity. Recombination at the variant recombination sites can be monitored or detected by any suitable means. It is preferred that recombination is detected 30 by screening or selection based on the expression or lack of expression of a reporter gene. This can be accomplished by using constructs containing a reporter gene associated with the variant recombination sites such that the 3 WO 00/60091 PCT/USOO/09154 reporter gene is rearranged or deleted. or a spacer sequence interrupting the reporter gene is rearranged or deleted. as a result of recombination at the recombination sites. Recombination of such constructs will result in a loss of expression of the reporter gene, where the construct contained a functional 5 reporter gene, or in a gain in expression of the reporter gene, where the construct contained a non-functional reporter gene. The disclosed method also involves determining whether a variant recombinase retains the ability to mediate recombination at recombination sites recognized by non-variant recombinase. This can be accomplished by using 10 constructs containing recombination sites recognized by non-variant recombinase. Recombination at these recombination sites can be monitored or detected by any suitable means. It is preferred that recombination is detected by screening or selection based on the expression or lack of expression of a reporter gene. This can be accomplished by using constructs containing a 15 reporter gene associated with the recombination sites recognized by non-variant recombinase such that the reporter gene is rearranged or deleted, or a spacer sequence interrupting the reporter gene is rearranged or deleted, as a result of recombination at the recombination sites. Recombination of such constructs will result in a loss of expression of the reporter gene, where the construct 20 contained a functional reporter gene, or in a gain in expression of the reporter gene, where the construct contained a non-functional reporter gene. When variant recombinases are tested for activity on both variant recombination sites and recombination sites recognized by non-variant recombinase in the same system or at the same time, it is preferred that two 25 different reporter genes, which can be separately detected or monitored, be used. In this case, the first reporter gene can be associated with the variant recombination sites and the second reporter gene can be associated with recombination sites recognized by non-variant recombinase. Recombination between two recombination sites requires (1) that the 30 recombinase recognize the sites as recombination sites, and (2) that the sequences of the two sites is sufficiently similar. It has been discovered that recombination between two recombination sites (both of which are recognized 4 WO 00/60091 PCT/USOO/09154 by a recombinase) can be substantially reduced or prevented by using different compatibility sequences for the recombination sites (the recognition sequences can also differ if the recombinase can recognize different sequences). Thus, it is also preferred that the variant recombination sites be made incompatible with 5 the recombination sites recognized by non-variant recombinase by using different compatibility sequences for the two sets of recombination sites. Compatibility sequences in a recombination site are those sequences in the recombination site, other than the sequences required for recognition of the site by the recombinase, that must be similar in a pair of recombination sites for 10 recombination to occur between them. Many recombination sites contain repeats of a characteristic sequence separated by spacer sequences. In such recombination sites, the spacer sequences are generally compatibility sequences and the repeats (or parts of the repeats) are recognition sequences. Recombinases require specific recognition sequences but allow wide variation 15 in compatibility sequences. Thus, recombination sites that are recognized by a given recombinase but are incompatible with each other can be readily designed using the disclosed principles. Also disclosed are variant recombinases made or identified by the disclosed method that have broadened specificity for the site of recombination. 20 Specifically, the disclosed variants mediate recombination between sequences other than recombination sites on which the wild type recombinase is active. In general, the disclosed recombinase variants can mediate efficient recombination between recombination sites that wild type recombinase can act on (referred to as wild type recombination sites), between variant recombination sites not 25 efficiently utilized by wild type recombinase (referred to as variant recombination sites), and between a wild type recombination site and a variant recombination site. Also disclosed are methods of recombining nucleic acids using the disclosed variant recombinases. For example, the disclosed variant 30 recombinases can be used in any method or technique where wild type recombinases can be used. In addition, the disclosed variant recombinases allow different alternative recombinations to be performed since the variant 5 WO 00/60091 PCT/US0O/09154 recombinases can allow much more efficient recombination between wild type recombination sites and variant recombination sites. Control of such alternative recombination can be used to accomplish more sophisticated sequential recombinations to achieve results not possible with wild type recombinases. 5 The disclosed variant recombinases also allow recombination at specific genomic sites without the need to first introduce a recombination site. Also disclosed are variants of Cre recombinase that have broadened specificity for the site of recombination. Specifically, the disclosed variants mediate recombination between sequences other than the loxP sequence and 10 other lox site sequences on which wild type Cre recombinase is active. In general, the disclosed Cre variants mediate efficient recombination between lox sites that wild type Cre can act on (referred to as wild type lox sites), between variant lox sites not efficiently utilized by wild type Cre (referred to as variant lox sites), and between a wild type lox site and a variant lox site. Also disclosed 15 are methods of recombining nucleic acids using the disclosed Cre variants. For example, the disclosed Cre variants can be used in any method or technique where Cre recombinase (or other, similar recombinases such as FLP) can be used. In addition, the disclosed Cre variants allow different alternative recombinations to be performed since the Cre variants allow much more 20 efficient recombination between wild type lox sites and variant lox sites. Control of such alternative recombination can be used to accomplish more sophisticated sequential recombinations to achieve results not possible with wild type Cre recombinase. BRIEF DESCRIPTION OF THE DRAWINGS 25 Figure 1 is a comparison of three different lox sites. loxP is the original recombination site for Cre recombinase. loxK1 and loxK2 are variant lox sites. Figure 2 is a diagram of two different forms of construct and the resulting recombination products. Figure 3 is a diagram of an example of a random mutagenesis using 30 DNA shuffling. Figure 4 is a diagram of the selection plasmid for loxK2 recombination, pBS584. Recombination of two loxK2 sites by a potent Cre mutant will result 6 WO 00/60091 PCT/US00/09154 in the excision of EGFP and the transcriptional terminator rrnBT T2. Subsequently, neo transcription can take place, rendering E. coli resistant to kanamycin. Note that the promoter (pRSV) even though of eukaryotic origin was shown to be functional in E. coli (Antonucci et al., J. Biol. Chem., 5 264:17656-17659 (1989)). Figure 5 depicts gels of nucleic acid fragments and PCR products generated during the DNA shuffling process. Figure 6 is a diagram of plasmid pBAD33 used for expression of mutant cre pools. 10 Figure 7 is a diagram of the construction of selection plasmids pBS568 and pBS569. Figure 8 is a diagram of the construction of selection plasmids pBS583 and pBS584. Figure 9 is a diagram of control plasmid pBS613. 15 Figure 10 is a diagram of the construction of screening plasmids pBS601 and pBS602. Figure 11 is a diagram of examples of basic types of constructs useful in the disclosed method. These types of constructs are: (1) interrupted constructs where the gene is interrupted by a nucleic acid segment (which is flanked by 20 recombination sites) that is deleted during recombination, (2) flanked constructs where the gene as a unit is flanked by recombination sites and the gene is deleted by recombination, and (3) inverted constructs where a portion of the gene is on an inverted nucleic acid segment and recombination causes the segment to invert and reconstitute the intact gene. The type of recombination is 25 indicated in parentheses. Figures 12A, 12B, and 12C are diagrams of examples of constructs and their expected recombination when used in the disclosed method. Figure 12A shows examples of deletion constructs (flanked and interrupted). Figure 12B shows examples of inverted constructs. Figure 12C shows examples of 30 constructs that combine through recombination to reconstitute an intact gene. Figure 13 is a diagram showing the identified amino-acid changes in the six selected Cre mutants are listed according to their position in the protein's 7 WO 00/60091 PCT/US0O/09154 secondary structure (silent mutations in parenthesis). Only one amino acid change, E262G, is common to all mutants with remarkably increased loxK2 activity (R3M1, 2, 3, 5, and 6), suggesting that this mutation is essential for the observed phenotype. 5 Figure 14A is a table comparing recombination frequencies in vivo obtained with a variety of lox sites altered at positions 11 and 12. Figure 14B is a table comparing recombination frequencies in vivo obtained with identical and mixed lox sites. Wild type Cre and five different mutant enzymes were tested for their performance on different lox 2 substrates, as indicated. Given are the 10 obtained percentages of recombination in vivo based on the described negative selection. Figure 15 is a table comparing recombination frequencies in vitro obtained with a variety of lox sites altered at positions 11 and 12. Figure 16 is a graph of percent of various Cre recombinases (wt, G, GA, 15 GN, GS, R3M3) bound to various lox sites (loxP, loxK2, loxKl). Figure 17 shows wildtype and target FRT sites. Figure 18 shows the strategy for selection of altered specificity FLP mutants. Figure 19 shows an alternate target mutant FRT site. The design and 20 rationale for design of the target mutant FRT site is as described in Figure 17, but the mutant FRT-M2 site differs from FRT-M by carrying a different mutational alteration in the repeat elements. DETAILED DESCRIPTION OF THE INVENTION Disclosed is a method for identifying variant forms of recombinases that 25 can mediate recombination between variant recombination sites. The method involves producing mutant recombinases and testing the mutant recombinases with specially designed constructs. The constructs contain variant recombination sites that are not recognized by non-mutant recombinase but will undergo recombination in the presence of a mutant recombinase with altered 30 specificity. The disclosed method also involves determining whether a variant recombinase retains the ability to mediate recombination at recombination sites recognized by non-variant recombinase. 8 WO 00/60091 PCT/USOO/09154 When variant recombinases are tested for activity on both variant recombination sites and recombination sites recognized by non-variant recombinase in the same system or at the same time, it is preferred that two different reporter genes which can be separately detected or monitored be used. 5 In this case, a first reporter gene can be associated with the variant recombination sites and a second reporter gene can be associated with recombination sites recognized by non-variant recombinase. It is also preferred that the variant recombination sites be made incompatible with the recombination sites recognized by non-variant recombinase by using different 10 compatibility sequences for the two sets of recombination sites. This allows separate assessment of the ability of a variant recombinase to mediate recombination between variant recombination sites and recombination sites recognized by non-variant recombinase. Also disclosed are variant recombinases made or identified by the 15 disclosed method that have broadened specificity for the site of recombination. Also disclosed are methods of recombining nucleic acids using the disclosed variant recombinases. For example, the disclosed variant recombinases can be used in any method or technique where wild type recombinases can be used. In addition, the disclosed variant recombinases allow different alternative 20 recombinations to be performed since the variant recombinases can allow much more efficient recombination between wild type recombination sites and variant recombination sites. Control of such alternative recombination can be used to accomplish more sophisticated sequential recombinations to achieve results not possible with wild type recombinases. 25 Also disclosed are variants of Cre recombinase that have broadened specificity for the site of recombination. Specifically, the disclosed variants mediate recombination between sequences other than the loxP sequence and other lox site sequences on which wild type Cre recombinase is active. Preferred forms of the disclosed Cre variants have the amino acid sequence 30 SEQ ID NO: 1 (top sequence, Table 11) altered by one or more amino acid substitutions, deletions, or insertions, where the glutamic acid at amino acid 262 has been substituted with an amino acid other than glutamic acid, and where the 9 WO 00/60091 PCT/US0O/09154 Cre variant recognizes (that is, mediates recombination at) a variant lox recombination site. Useful Cre variants include proteins that recognize a variant lox recombination site and have the amino acid sequence SEQ ID NO: 1 altered by substitution of the glutamic acid at amino acid 262 with an amino acid other 5 than glutamic acid and one or more of the following amino acid substitutions: isoleucine at amino acid 16, alanine at amino acid 29, glutamine at amino acid 101, glycine at amino acid 138, asparagine at amino acid 189, serine at amino acid 198, glutamine at amino acid 220, glutamine at amino acid 223, isoleucine at amino acid 227, glycine at amino acid 254, arginine at amino acid 255, 10 glutamine at amino acid 284, leucine at amino acid 307, and serine at amino acid 316. Preferred amino acid substitutions at amino acid position 262 include alanine, tryptophan, or glycine. Examples of preferred Cre variants include proteins having the amino acid sequence SEQ ID NO: 1 altered by substitutions E262G and D1 89N; 15 proteins having the amino acid sequence SEQ ID NO: I altered by substitutions E262G and T316S; proteins having the amino acid sequence SEQ ID NO: 1 altered by substitutions E262G and D29A; proteins having the amino acid sequence SEQ ID NO:1 altered by substitutions E262G, V161, D189N, G198S, R223Q, Q255R, and P307L; proteins having the amino acid sequence SEQ ID 20 NO: 1 altered by substitution E262G; proteins having the amino acid sequence SEQ ID NO: 1 altered by substitution E262A; and proteins having the amino acid sequence SEQ ID NO: 1 altered by substitution E262W. The substitutions above are listed using the convention where the first letter is the original amino acid (in single letter amino acid code), the number is the amino acid position in 25 the protein (in this case, using the positions of wild type Cre (SEQ ID NO: 1)), and the last letter is the new amino acid (in single letter amino acid code). All of these Cre variants recognize both wild type lox sites and variant lox sites with an inverted repeat sequence NNNACNNCGTATA (SEQ ID NO:2). The disclosed Cre variants recognize variant lox recombination sites. 30 Preferred variant lox sites are variant lox sites recognized by the Cre variant but not recognized by wild type Cre. Examples of useful variant lox sites include sites having two 13 base pair inverted repeats flanking 8 base pairs, where one 10 WO 00/60091 PCT/USOO/09154 of the inverted repeats has the sequence NNNACNNCGTATA (SEQ ID NO:2); sites having the sequence N 1

N

2

N

3

ACN

4

N

5 CGTATANNNNNNNNTATA

CGN

5

'N

4

'GTN

3

'N

2

'N

1 ' (SEQ ID NO:3). where NI', N 2 ', N3, N 4 ', and N 5 ' are complementary to NI, N 2 , N 3 , N 4 , and N 5 , respectively; sites having the 5 sequence NiN 2

N

3

ACN

4

N

5

CGTATANNNNNNNNTATACGN

5

'N

4

'GN

3

'N

2 'NI (SEQ ID NO:3), where N 4

N

5 are AA, TC, GT, TG, GG, or CC; and sites having the sequence GATACAACGTATATACCTTTCTATACGTTGTAT (SEQ ID NO:4). Also disclosed is a method for producing site-specific recombination of 10 DNA in cells using the disclosed Cre variants. DNA sequences comprising first and second lox sites are introduced into cells and contacted with a Cre variant, thereby producing recombination at the lox sites. As with wild type Cre, the location and orientation of the lox sites determines the nature of the recombination. 15 As used herein, the expression "site-specific recombination" refers to three different types of recombination events: 1. deletion of a pre-selected DNA segment flanked by recombination sites, 2. inversion of the nucleotide sequence of a pre-selected DNA segment 20 flanked by recombination sites, and 3. reciprocal exchange of DNA segments proximate to recombination sites located on different DNA molecules. It is to be understood that this reciprocal exchange of DNA segments can result in an integration event if one or both of the DNA molecules are 25 circular. "Nucleic acid segment" refers to a linear segment of single- or double stranded nucleic acid, which can be derived from any source. The segment may be a fragment consisting of the segment or a segment within a larger nucleic acid fragment or molecule. The expression "nucleic acid in eukaryotic cells" includes all nucleic acid present in eukaryotic cells. The expression "nucleic 30 acid in yeast" includes all nucleic acid present in yeast cells. "DNA segment" refers to a linear segment of single- or double-stranded deoxyribonucleic acid (DNA), which can be derived from any source. The expression "DNA in 11 WO 00/60091 PCT/USOO/09154 eukaryotic cells" includes all DNA present in eukaryotic cells. The expression "DNA in yeast" includes all DNA present in yeast cells. As used herein, a "gene" is intended to mean a DNA segment which is normally regarded as a gene by those skilled in the art. The expression "regulatory molecule" refers to 5 a polymer of ribonucleic acid (RNA) or a polypeptide which is capable of enhancing or inhibiting expression of a gene. "Regulatory nucleotide sequence," as used herein, refers to a nucleotide sequence located proximate to a gene whose transcription is controlled by the regulatory nucleotide sequence in conjunction with the gene expression 10 apparatus of the cell. Generally, the regulatory nucleotide sequence is located 5' to the gene. The expression "nucleotide sequence" refers to a polymer of DNA or RNA, which can be single- or double-stranded, optionally containing synthetic, non-natural, or altered nucleotides capable of incorporation into DNA or RNA polymers. As used herein, a "regulatory nucleotide sequence" can 15 include a promoter region, as that term is conventionally employed by those skilled in the art. A promoter region can include an association region recognized by an RNA polymerase, one or more regions which control the effectiveness of transcription initiation in response to physiological conditions, and a transcription initiation sequence. "Gene product" refers to a polypeptide 20 resulting from transcription, translation, and, optionally, post-translational processing of a selected DNA segment. Materials A. Recombinases Recombinases suitable for use in the disclosed method include any 25 enzyme that mediates recombination at specific sites. This includes enzymes identified as recombinases as well as other enzymes that function to produce recombination such as integrases and resolvases. As used herein, recombination at specific sites does not refer only to recombination at completely defined sequences. Rather, a recombinase is considered to mediate recombination at 30 specific sites when the sites of recombination are limited in some way by sequence. For example, wild type Cre recombinase mediates recombination between sites having the sequence NiN 2

N

3 ACTTCGTATANNNNNNNNT 12 WO 00/60091 PCT/USO0/09154

ATACGAAGTN

3

'N

2 'NI', which includes both specific and non-specific sequences. The sequences ACTTCGTATA and TATACGAAGT (an inverted repeat of the first sequence) are recognized by the Cre recombinase. The non specific sequences (positions with "N" in the recognition sequence), although 5 not limited in sequence, must be compatible with the non-specific sequences of the partner recombination site in order for recombination to be efficient. The recombination sites need not have any particular number of specific nucleotides. All that is required is some constraint on the sequence of the site such that the recombinase is limited to recombination at some set of sites. 10 Examples of recombinases that can be used in the disclosed method include Cre recombinase, FLP recombinase, Beta recombinase of pSM19035 (Diaz et al., JBiol Chem 274: 6634-6640 (1999)), Int recombinases (Nunes Dbby et al., Nucleic Acids Res. 26:391-406 (1998)), and resolvases (Hallet et al., FEMS Microbiol Rev. 21: 157-178 (1997); Oram et al., Curr Biol. 5: 1106 15 1109 (1995); Mondragon, Structure 3: 755-758 (1995)). B. Recombination Sites Recombination sites are locations within a nucleic acid where recombination mediated by a recombinase takes place. Recombination sites generally include specific sequences, referred to as recognition sequences, 20 through which the recombinase recognizes a given nucleotide sequence as a recombination site. Different recombinases generally recognize different recognition sequences. Recombination between two recombination sites requires (1) that the recombinase recognize the sites as recombination sites, and (2) that the sequences of the two sites are sufficiently similar. It has been 25 discovered that recombination between two recombination sites (both of which are recognized by a recombinase) can be substantially reduced or prevented by using different compatibility sequences for the recombination sites (the recognition sequences can also differ if the recombinase can recognize different sequences). Thus, it is also preferred that the variant recombination sites be 30 made incompatible with the recombination sites recognized by non-variant recombinase by using different compatibility sequences for the two sets of recombination sites. Compatibility sequences in a recombination site are those 13 WO 00/60091 PCT/USOO/09154 sequences in the recombination site. other than the sequences required for recognition of the site by the recombinase, that must be similar in a pair of recombination sites for recombination to occur between them. Generally, recombinases require specific recognition sequences but allow wide variation in 5 compatibility sequences. Thus, recombination sites that are recognized by a given recombinase but are incompatible with each other can be readily designed using the disclosed principles. It should be understood that. for a given recombinase site or for a given recombinase, whether a given base position in the recombination site is a 10 recognition sequence base or a compatibility sequence base may depend on other sequences in the recombination site. For example, a particular base may function as a compatibility sequence base in a recombination site having one sequence while the same base may function as a recognition sequence base in a recombination site having a different sequence. It should also be understood 15 that recognition sequences and compatibility sequences do not necessarily occur in blocks within a recombination site. That is, recognition sequence base and compatibility sequence bases may be interspersed in a given recombination site. As discussed below, what is and is not a recognition sequence or a compatibility sequence in a given recombination site is determined functionally. 20 The disclosed variant recombination sites and the variant recombinases that can act on them allow more freedom in the selection of sites of recombination. In particular, the disclosed variant recombinases can allow amino acid changes in a protein of interest while retaining the ability to recombine at a given site. 25 1. Recognition Sequences Recognition sequences are regions within a recombination site that must have a specific sequence, or defined range of sequences, for the cognate recombinase to recognize the recombination site. Recognition sequences in a recombination site need not be contiguous. Thus, required nucleotides 30 dispersed in a recombination site are collectively considered recognition sequences. Nucleic acid segments can be said to have a defined range of sequences when every nucleotide position in the nucleic acid segment(s) is 14 WO 00/60091 PCT/USO0/09154 limited to one, two, or three nucleotide bases. That is, so long as a nucleotide position cannot have one of the possible nucleotide bases, that position has a defined range of sequence. For example, a nucleotide sequence ATRVBYGC has a defined range of sequences since each nucleotide position has at least one 5 limitation. Standard nomenclature for nucleic acid sequences is used herein. Thus, in this example, R represents A or G; V represents A, C, or G; B represents C, G, or T; and Y represents C or T. Recognition sequences for recombinases are known or can be determined through routine analysis. In general, recognition sequences can be 10 determined by varying the sequence of recombination sites and determining if recombination between the sites can still occur. For this purpose, the pair of sites to be recombined should be identical. That is, the same sequence changes should be made to both sites. This eliminates any incompatibility effect between the recombination sites. If recombination is eliminated or significantly 15 reduced when a specific nucleotide is changed, then that nucleotide is required for recognition of the recombination site by the recombinase. Examples of dissection of the critical sequences in recombination sites of recombinases are described by Hoess et al., Nucleic Acids Res. 14:2287-2300 (1986) (involving P1 recombinase); Sauer B., Nucleic Acids Res., 24:4608-4613 20 (1996) (involving Cre recombinase); Lee and Saito, Gene 216(1):55-65 (1998) (involving Cre recombinase); and Umlauf and Cox, EMBO J 7(6):1845-52 (1988) (involving FLP recombinase). Similar techniques can be used to determine the recognition sequences of other recombinases. 2. Compatibility Sequences 25 Compatibility sequences are regions in a recombination site that must be similar in a pair of recombination sites for recombination to occur between them. In general, the sequence of recombination sites must be similar for recombination to occur between them. Examples of compatibility sequences are spacer sequences between repeats in recombination sequences. All or some 30 of the nucleotides in the recognition sequences for a recombination site may be involved in compatibility. For example, where some degeneracy of the recognition sequences is allowed, similar recognition sequences may be 15 WO 00/60091 PCT/USOO/09154 required in a pair of recombination sites for recombination to occur between them. Thus, compatibility between recombination sites can be affected by using different sequences in the compatibility sequences other than the sequences required for recognition of the site by the recombinase (that is, recognition 5 sequences), compatibility sequences that are part of the recognition sequences, or both. It is preferred that compatibility between recombination sites be altered by using different sequences in the compatibility sequences other than the sequences required for recognition of the site by the recombinase. Compatibility sequences for many recombinases are known or can be 10 determined through routine analysis. In general, compatibility sequences can be easily determined by varying the sequence of recombination sites and determining if recombination between the sites can still occur. For this purpose, only one of the sites in the pair of sites to be recombined should be altered. That is, the same sequence changes should not be made to both sites. This 15 isolates incompatibility effect between the recombination sites. Further, only those nucleotide positions that are not a part of the recognition sequence of the site should be altered to avoid recognition problems. If recombination is eliminated or significantly reduced when a specific nucleotide is changed, then that nucleotide is required for compatibility of the recombination site. 20 Examples of dissection of the critical sequences in recombination sites of recombinases are described by Hoess et al., Nucleic Acids Res. 14:2287-2300 (1986) (involving P1 recombinase); Sauer B., Nucleic Acids Res., 24:4608-4613 (1996) (involving Cre recombinase); Lee and Saito, Gene 216(1):55-65 (1998) (involving Cre recombinase); and Umlauf and Cox, EMBO J 7(6):1845-52 25 (1988) (involving FLP recombinase). Similar techniques can be used to determine the compatibility sequences of other recombinases. Recognition and compatibility sequences can be further understood using Cre recombination sites as an example. Wild type Cre recombinase mediates recombination between sites having the sequence 30 NiN 2

N

3 ACTTCGTATANN NNNNNNTATACGAAGTN 3

'N

2 'Ni', which includes both specific and non-specific sequences (that is, recognition sequences and compatibility sequences, respectively). The sequences 16 WO 00/60091 PCT/USOO/09154 ACTTCGTATA and TATACGAAGT (an inverted repeat of the first sequence) are recognized by the Cre recombinase and are the recognition sequences in Cre recombinase sites. Variant Cre recombinases recognize sites having different recognition sequences. The non-specific sequences (positions with "N" in the 5 recognition sequence), although not limited in sequence, must be compatible with the non-specific sequences of the partner recombination site in order for recombination to be efficient. Thus, the non-specific sequences are the compatibility sequences of a recombinase site. C. Recombination Constructs 10 Recombination constructs are designed to provide an observable change when recombination between recombination sites occurs. Preferred recombination constructs include two pairs of recombination sites, one pair having a variant sequence and another pair having a sequence recognized by non-mutant recombinase (for example, wild type recombinase). Sites in the 15 first pair are referred to as variant recombination sites. Generally, recombination constructs include a first nucleic acid sequence that includes a first reporter gene and first and second recombination sites, where the first and second recombination sites are variant recombination sites, and a second nucleic acid sequence that includes a second reporter gene and third and fourth 20 recombination sites, where the third and fourth recombination sites can be recombined by a non-mutant recombinase. The first and second nucleic acid sequences need not be present on the same vector or on the same nucleic acid molecule (for example, the chromosome), although this is preferred. It is preferred that recombination constructs be embodied in vectors, such as 25 plasmids. In one embodiment of the disclosed recombination constructs, the sequence of the recombination sites in the constructs are chosen such that the recognition sequences of the first and second recombination sites differ from the recognition sequences of the third and fourth recombination sites. The sequence 30 of the recombination sites can also be chosen such that the compatibility sequences of the first and second recombination sites differ from the compatibility sequences of the third and fourth recombination sites such that the 17 WO 00/60091 PCT/USO0/09154 first and second recombination sites cannot recombine with the third and fourth recombination sites. The sequence of the recombination sites can also be chosen such that the compatibility sequences of the first and second recombination sites are sufficiently similar to allow recombination between the 5 first and second recombination sites, and such that the compatibility sequences of the third and fourth recombination sites are sufficiently similar to allow recombination between the third and fourth recombination sites. The above sequence relationships result in constructs where the first and second recombination sites can recombine (in the presence of a recombinase that 10 recognizes the sites), the third and fourth recombination sites can recombine, but where the neither the first nor second recombination site can recombine with either the third or fourth recombination site (since differences in the compatibility sequences prevent recombination). Arriving at recombination sites having relationships as described above 15 is preferably accomplished in the following way. Starting with a given recombination site sequence (which can be recombined by a non-mutant recombinase), parallel changes are made in the compatibility sequences of the first and second recombination sites. These altered recombination sites should then be tested to make sure that the non-mutant recombinase can still mediate 20 their recombination. This helps insure that compatibility sequence changes have not inadvertently affected the function of the recombination sites. Once this is confirmed, changes can be made to the recognition sequences of the first and second recombination sites. These changes result in variant recombination sites for which variant recombinases can be identified using the method 25 disclosed herein. The resulting recombination sites have the desired propertied: incompatibility with the third and fourth recombination sites and variant recognition sequences that extend the range of recombination-competent sites. The recombination sites can have a variety of properties and relationships that make them useful for particular purposes. For example, the 30 recombination sites can be designed such that the first and second recombination sites cannot be recombined by non-mutant recombinase to a significant extent. This allows separate assessment of cleavage by mutant and 18 WO 00/60091 PCT/US00/09154 non-mutant recombinase. It is also useful if the first and second recombination sites have identical sequences, and the third and fourth recombination sites have identical sequences. Recombination between the recombination sites can have a variety of 5 effects that allows detection of recombination. For example, the constructs can be designed such that recombination between the first and second recombination sites alters the expression of the first reporter gene, where recombination between the first and second recombination sites is determined by determining if expression of the first reporter gene is altered; recombination 10 between the third and fourth recombination sites alters the expression of the second reporter gene, where recombination between the third and fourth recombination sites is determined by determining if expression of the second reporter gene is altered; recombination between the first and second recombination sites allows the first reporter gene to be expressed; the first 15 nucleic acid sequence includes a spacer sequence flanked by the first and second recombination sites, where the spacer sequence interrupts the first reporter gene such that the first reporter gene is not expressed, and where recombination of the first and second recombination sites excises the spacer sequence which allows the first reporter gene to be expressed; and/or a portion 20 of the first reporter gene is inverted, wherein the inverted portion of the first reporter gene is flanked by the first and second recombination sites, wherein recombination of the first and second recombination sites inverts the inverted portion of the first reporter gene which allows the first reporter gene to be expressed. 25 The constructs can also be designed such that recombination between the first and second recombination sites prevents expression of the first reporter gene; the first reporter gene is flanked by the first and second recombination sites, where recombination of the first and second recombination sites excises the first reporter gene which prevents expression of the first reporter gene; a 30 portion of the first reporter gene is flanked by the first and second recombination sites, where recombination of the first and second recombination sites inverts the flanked portion of the first reporter gene which prevents 19 WO 00/60091 PCT/USOO/09154 expression of the first reporter gene: recombination between the third and fourth recombination sites allows the second reporter gene to be expressed; and/or the second nucleic acid sequence includes a spacer sequence flanked by the third and fourth recombination sites, where the spacer sequence interrupts the second 5 reporter gene such that the second reporter gene is not expressed, and where recombination of the third and fourth recombination sites excises the spacer sequence which allows the second reporter gene to be expressed. The constructs can also be designed such that a portion of the second reporter gene is inverted, where the inverted portion of the second reporter gene 10 is flanked by the third and fourth recombination sites, and where recombination of the third and fourth recombination sites inverts the inverted portion of the second reporter gene which allows the second reporter gene to be expressed; recombination between the third and fourth recombination sites prevents expression of the second reporter gene to be expressed; the second reporter gene 15 is flanked by the third and fourth recombination sites, where recombination of the third and fourth recombination sites excises the second reporter gene which prevents expression of the second reporter gene; and/or a portion of the second reporter gene is flanked by the third and fourth recombination sites, where recombination of the third and fourth recombination sites inverts the flanked 20 portion of the second reporter gene which prevents expression of the second reporter gene. Expression of a reporter gene can include transcription of the gene, translation of the transcript (if the gene encodes a protein), and/or production of an active protein. As used herein, whether a reporter gene is expressed depends 25 on the context. In general, a gene is considered to be expressed if it produces the expression product to be detected. Such expression products include full or partial transcripts of the gene, full or partial proteins, including active or inactive forms of the proteins, translated from the transcript. Since the goal in using reporter genes in the disclosed method is the detection of expression, any 30 of these forms of expression product can be the object of detection. For example, if the gene's transcript is to be detected, the gene will be considered to be expressed if it produces the transcript, regardless of whether the transcript is 20 WO 00/60091 PCT/USOO/09154 translated or whether the resulting protein is active. If the an active protein encoded by the gene is to be detected, the gene is not expressed unless active protein is produced--mere transcription of the gene, or even translation to produce an inactive protein, will not be enough in this context. As a 5 consequence, the expression product to be detected will influence the manner in which reporter genes should be interrupted or invented in the disclosed constructs. For example, nearly any interruption of a reporter gene would prevent expression of an active protein encoded by the gene. On the other hand, an interruption of the coding region will usually not prevent production of a 10 transcript. The structure of the disclosed constructs should be designed with these principles in mind. As used herein, an inactive expression product refers to an expression product that does not have an activity exhibited by the active form of the expression product where that activity is required for detection of expression in the assay scheme being used. 15 The constructs can be designed such that the first nucleic acid sequence is a first nucleic acid construct and the second nucleic acid sequence is on a second nucleic acid construct; the first nucleic acid construct is an extrachromosomal vector and the second nucleic acid construct is in the genome of a host cell; and/or the first and second nucleic acid constructs are on the same 20 nucleic acid construct. D. Reporter Genes Reporter genes are used to monitor whether recombination occurs in the disclosed constructs. Reporter genes can be any gene the expression of which can be detected either directly or indirectly. These include genes encoding 25 enzymes, such as p-galactosidase, luciferase, and alkaline phosphatase, that can produce specific detectable products, and genes encoding proteins that can be directly detected. Virtually any protein can be directly detected by using, for example, specific antibodies to the protein. A preferred reporter protein that can be directly detected is the green fluorescent protein (GFP). GFP, from the 30 jellyfish Aequorea victoria, produces fluorescence upon exposure to ultraviolet light without the addition of a substrate (Chalfie et al., Science 263:802-5 (1994)). A number of modified GFPs have been created that generate as much 21 WO 00/60091 PCT/USO0/09154 as 50-fold greater fluorescence than does wild type GFP under standard conditions (Cormack et al., Gene 173:33-8 (1996); Zolotukhin et al., J. Virol 70:4646-54 (1996)). This level of fluorescence allows the detection of low levels of expression in cells. 5 Reporter genes encoding proteins producing a fluorescent signal are useful since such a signal allows cells to be sorted using FACS. Another way of sorting cells based on expression of the reporter gene involves using the reporter protein as a hook to bind cells. For example, a cell surface protein such as a receptor protein can be bound by a specific antibody. Cells expressing such a 10 protein can be captured by, for example, using antibodies bound to a solid substrate, using antibodies bound to magnetic beads, or capturing antibodies bound to the reporter protein. Many techniques for the use of antibodies as capture agents are known and can be used with the disclosed method. The reporter gene can also encode an expression product that regulates 15 the expression of another gene. This allows detection of expression of the reporter gene by detecting expression of the regulated gene. For example, a repressor protein can be encoded by the reporter gene. Loss of expression of the reporter gene (via recombination) would then result in derepression of the regulated gene. This type of indirect detection allows positive detection of loss 20 of the expression of the reporter gene by the affector RNA molecule. One preferred form of this type of regulation is the use of an antibiotic resistance gene regulated by a repressor protein encoded by the reporter gene. By exposing the host cells to the antibiotic, only those cells in which expression of the reporter gene has been inhibited will grow since expression of the antibiotic 25 resistance gene will be derepressed. E. Expression Sequences The reporter genes can be expressed using any suitable expression sequences. Numerous expression sequences are known and can be used for expression of the reporter genes. Expression sequences can generally be 30 classified as promoters, terminators, and, for use in eukaryotic cells, enhancers. Expression in prokaryotic cells also requires a Shine-Dalgarno sequence just upstream of the coding region for proper translation initiation. Inducible 22 WO 00/60091 PCT/USO0/09154 promoters are preferred for use with the first reporter gene since it is preferred that expression of the first reporter gene be adjustable. Promoters suitable for use with prokaryotic hosts illustratively include the p-lactamase and lactose promoter systems, tetracycline (tet) promoter, 5 alkaline phosphatase promoter, the tryptophan (trp) promoter system and hybrid promoters such as the tac promoter. However, many other functional bacterial promoters are suitable. Their nucleotide sequences are generally known. Suitable promoting sequences for use with yeast hosts include the promoters for 3-phosphoglycerate kinase, enolase, glyceraldehyde-3 -phosphate 10 dehydrogenase, hexokinase, pyruvate decarboxylase, phosphofructokinase, glucose-6-phosphate isomerase, 3-phosphoglycerate mutase, pyruvate kinase, triosphosphate isomerase, phosphoglucose isomerase, and glucokinase. Examples of inducible yeast promoters suitable for use in the disclosed vectors include the promoter regions for alcohol dehydrogenase 2, isocytochrome C, 15 acid phosphatase, degradative enzymes associated with nitrogen metabolism, metallothionein, glyceraldehyde-3-phosphate dehydrogenase, and enzymes responsible for maltose and galactose utilization. Yeast enhancers also are advantageously used with yeast promoters. Preferred promoters for use in mammalian host cells include promoters 20 from polymoma virus, Simian Virus 40 (SV40), adenovirus, retroviruses, hepatitis B virus, herpes simplex virus (HSV), Rous sarcoma virus (RSV), mouse mammary tumor virus (MMTV), and most preferably cytomegalovirus (CMV), or from heterologous mammalian promoters such as the P actin promoter. Particularly preferred are the early and late promoters of the SV40 25 virus and the immediate early promoter of the human cytomegalovirus, MMTV LTR, RSV-LTR, and the HSV thymidine kinase promoter. Transcription of the reporter gene by higher eukaryotes can be increased by inserting an enhancer sequence into the vector. Many enhancer sequences are now known from mammalian genes (globin, elastase, albumin, 30 and insulin). Typically, however, one will use an enhancer from a eukaryotic cell virus. Examples include the SV40 enhancer on the late side of the 23 WO 00/60091 PCT/USOO/09154 replication origin, the cytomegalovirus early promoter enhancer, the polyoma enhancer on the late side of the replication origin, and adenovirus enhancers. The disclosed vectors preferably also contain sequences necessary for accurate 3' end formation of both reporter and affector RNAs. In eukaryotic 5 cells, this would be a polyadenylation signal. In prokaryotic cells, this would be a transcription terminator. Method A. Identification of Variant Recombinases The disclosed method involves producing mutant recombinases and 10 testing the mutant recombinases with specially designed constructs. The constructs contain variant recombination sites that are not recognized by non mutant recombinase but will undergo recombination in the presence of a mutant recombinase with altered specificity. The disclosed method also involves determining whether a variant recombinase retains the ability to mediate 15 recombination at recombination sites recognized by non-variant recombinase. This can be accomplished by using constructs containing recombination sites recognized by non-variant recombinase. Recombination at these recombination sites can be monitored or detected by any suitable means. It is preferred that recombination is detected by screening or selection based on the expression or 20 lack of expression of a reporter gene. This can be accomplished by using constructs containing a reporter gene associated with the recombination sites such that the reporter gene is rearranged or deleted, or a spacer sequence interrupting the reporter gene is rearranged or deleted, as a result of recombination at the recombination sites. Recombination of such constructs 25 will result in a loss of expression of the reporter gene, where the construct contained a functional reporter gene, or in a gain in expression of the reporter gene, where the construct contained a non-functional reporter gene. 1. Production of Mutant Recombinases Mutant recombinases can be produced by any suitable technique. In 30 general, all that is required is a method of generating a variety of recombinase proteins having a variety of amino acid sequences. the most preferred way of doing this is to mutagenize or alter nucleic acid encoding the recombinase and 24 WO 00/60091 PCT/USO0/09154 then expressing the mutant recombinases. Numerous techniques for introducing alterations into nucleic acid sequences are known and can be used in the disclosed method. For example, alterations can be made by chemical mutagenesis, introduction of degenerate nucleic acid fragments into the base 5 nucleic acid molecule, and low fidelity PCR. The goal of this mutagenesis or alteration will be the generation of a population or set of mutant recombinases having a variety of sequences. The broader the range of variants, the more raw material for the identification process. 2. Identification of Variant Recombinases That Recognize Variant 10 Recombination Sites Variant recombinases that can mediate recombination at variant recombination sites are identified in the disclosed method by selecting for, screening for, or otherwise detecting recombination of specially designed constructs having variant recombination sites. Recombination at variant 15 recombination sites can be monitored or detected by any suitable means. It is preferred that recombination is detected by screening or selection based on the expression or lack of expression of a reporter gene. This can be accomplished by using constructs containing a reporter gene associated with the variant recombination sites such that the reporter gene is rearranged or deleted, or a 20 spacer sequence interrupting the reporter gene is rearranged or deleted, as a result of recombination at the recombination sites. Recombination of such constructs will result in a loss of expression of the reporter gene, where the construct contained a functional reporter gene, or in a gain in expression of the reporter gene, where the construct contained a non-functional reporter gene. 25 3. Identification of Variant Recombinases That Recognize Non Variant Recombination Sites Variant recombinases that can mediate recombination at recombination sites recognized by non-variant recombinase (non-variant recombination sites) are identified in the disclosed method by selecting for, screening for, or 30 otherwise detecting recombination of specially designed constructs having recombination sites recognized by non-variant recombinase. Recombination at these recombination sites can be monitored or detected by any suitable means. 25 WO 00/60091 PCT/USOO/09154 It is preferred that recombination is detected by screening or selection based on the expression or lack of expression of a reporter gene. This can be accomplished by using constructs containing a reporter gene associated with the recombination sites recognized by non-variant recombinase such that the 5 reporter gene is rearranged or deleted, or a spacer sequence interrupting the reporter gene is rearranged or deleted, as a result of recombination at the recombination sites. Recombination of such constructs will result in a loss of expression of the reporter gene, where the construct contained a functional reporter gene, or in a gain in expression of the reporter gene, where the 10 construct contained a non-functional reporter gene. It is preferred that the ability of a variant recombinase to mediate recombination at both variant recombination sites and recombination sites recognized by non-variant recombinase be assessed in the same system (such as a cell strain) either sequentially or simultaneously. When variant recombinases 15 are tested for activity on both variant recombination sites and recombination sites recognized by non-variant recombinase in the same system or at the same time, it is preferred that two different reporter genes which can be separately detected or monitored be used. In this case, a first reporter gene can be associated with the variant recombination sites and a second reporter gene can 20 be associated with recombination sites recognized by non-variant recombinase. B. Use of Variant Recombinases Variant recombinases produced in the disclosed method can be used for any purpose that unmodified recombinases can be used. The advantage is that the variant recombinases have a different or broader site specificity. In general, 25 the disclosed variant recombinases can be used to mediate recombination of any nucleic acid in any setting, including in vitro, in cell culture, and in vivo. Recombination can be obtained in single celled organisms, such as bacterial cells, fungal cells, yeast cells, prokaryotic cells, and archae bacterial cells, the cells of multicellular organisms, including plants and animals, both in the 30 organism and in culture. The disclosed variant recombinases can also be used in combination with other recombinases (including other variant recombinases) having a different site specificity. Such combinations allow more complex 26 WO 00/60091 PCT/USO0/09154 recombination schemes to be used. Examples of such schemes are discussed below. For some uses of the disclosed recombinases, first, second, and fourth DNA sequences comprising a first recombination site, a second recombination 5 site, and a third recombination site, respectively, are introduced into cells. As used herein the expression "recombination site" means a nucleotide sequence at which a recombinase or variant recombinase can catalyze a site-specific recombination. Methods for introducing a DNA sequence into cells are known in the art. 10 These methods typically include the use of a DNA vector to introduce the sequence into the DNA of a single or limited number of eukaryotic cells and then growing such cell or cells to generate a suitable population of cells. As used herein, the term "vector" includes plasmids, viruses, and viral vectors. Preferably, the DNA sequences are introduced by a plasmid capable of 15 transforming a selected cell while carrying a DNA sequence. The particular vector which is employed to introduce the DNA sequence into a selected cell is not critical. In the present method, the recombination sites are contacted with a variant recombinase, thereby producing the site specific recombination. A 20 preferred means of contacting the DNA to be recombined with a variant recombinase is to place the DNA to be recombined into a cell expressing nucleic acid encoding the variant recombinase. Preferably, expression of the variant recombinase is under the control of a regulatory nucleotide sequence. Suitable regulatory nucleotide sequences are known in the art. The regulatory 25 nucleotide sequence which is employed with a selected eukaryotic cell is not critical to the method. A partial list of suitable regulatory nucleotide sequences includes the long terminal repeat of Moloney sarcoma virus described by Blochlinger and Diggelmann, Mol. Cell Bio., 4:2929-293 1 (1984); the mouse metallothionein-I promoter described by Pavlakis and Hamer, Proc. Natl. A cad. 30 Sci USA, 80:397-401 (1983); the long terminal repeat of Rous sarcoma virus described by Gorman et al., Proc. Natl. Acad Sci USA, 79:6777-6781 (1982); 27 WO 00/60091 PCT/USO0/09154 and the early region promoter of SV40 described by Southern and Berg, J. Mol. Apple. Genet., 1:327-341 (1982). In an embodiment where the cells are yeast, suitable regulatory nucleotide sequences include GAL1, GAL10, ADH1, CYC1, and TRP5 5 promoters. GALl and GAL 10 promoters are present on plasmid pBM150 which is described by Johnston and Davis, Molec. Cell. Biol., 4:1440 (1984). The ADHI promoter, also called ADC1, is present on plasmid pAAH5 which is described by Ammer, Methods Enzymol., 101:192 (1983). The CYCI promoter is described by Stiles et al., Cell, 25:277 (1981). The TRP5 promoter is 10 described by Zalkin and Yanofsky, J. Biol. Chem., 257:1491 (1982). Preferably, the regulatory nucleotide sequence is a GAL 1 promoter. In one embodiment where the cell is yeast, the first, second, and optionally, third and fourth DNA sequences are introduced into one strain of yeast. Alternatively, the DNA sequences are introduced into two different 15 strains of yeast of opposite mating types which are subsequently mated to form a single strain having all three or four DNA sequences. Preferably, the plasmid contains either (1) a nucleotide sequence of DNA homologous to a resident yeast sequence to permit integration into the yeast DNA by the yeast's recombination system or (2) a nucleotide sequence of DNA which permits 20 autonomous replication in yeast. One nucleotide sequence which permits autonomous replication in yeast is an ARS sequence described by Stinchcomb et al., Nature, 282:39 (1979). A partial list of plasmids capable of transforming yeast includes YIP5, YRP17 and YEP24. These plasmids are disclosed and described by Botstein and Davis, The Molecular Biology of the Yeast 25 Saccharomyces, Metabolism and Gene Expression (ed. Strathern et al.), (Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y., 1982), at page 607. Since most recombination sites are asymmetrical nucleotide sequences, two recombination sites on the same DNA molecule can have the same or opposite orientations with respect to each other. Recombinations between 30 recombination sites in the same orientation result in a deletion of the DNA segment located between the two recombination sites and a connection between the resulting ends of the original DNA molecule. The deleted DNA segment 28 WO 00/60091 PCT/US00/09154 forms a circular molecule of DNA. The original DNA molecule and the resulting circular molecule each contain a single recombination site (see Figure 2). Recombination between recombination sites in opposite orientations on the same DNA molecule result in an inversion of the nucleotide sequence of the 5 DNA segment located between the two recombination sites (see Figure 2). In addition, reciprocal exchange of DNA segments proximate to recombination sites located on two different DNA molecules can occur. All of these recombination events are catalyzed by recombinases, including the disclosed variants and wild type recombinases. 10 Recombination using the disclosed variant recombinases can be used in vitro to produce site-specific recombination of nucleic acid molecules. This is useful for a wide variety of manipulations that currently employ wild type recombinases or involve traditional restriction enzyme cleavage followed by ligation. Examples include recombination of libraries of DNA fragments into 15 vectors or in desired structures, and labeling of DNA via recombination. Recombined DNA formed by in vitro recombination can then be introduced into cells. For example, constructs formed in vitro can be introduced into cells to resolve the structures formed in vitro or to select active structures. In particular, large concatemers of subject DNA and spacer/vector fragments can be made, 20 introduced into cells, and circularized into vector units in the cells. Such recombination could also be performed in vitro if desired. The disclosed variant recombinases can be used to label DNA by recombining a DNA molecule of interest with a labeled DNA molecule. Use of a recombinase for labeling is advantageous since it involves fewer steps than 25 traditional labeling via DNA synthesis or ligation. These considerations are particularly important when large DNA molecules (over 20 kb) are to be labeled since such large molecules will fragment more the more they are manipulated. Recombination mediated by the disclosed variant recombinases and variant recombination sites can be used to manipulate a host cell genome as 30 desired and simultaneously introduce a marker gene flanked by the recognition sites of a second recombinase. After selection, leading to an accumulation of cells carrying the desired genomic alteration, one could simply remove the 29 WO 00/60091 PCT/USOO/09154 marker gene by expression of the second site-specific recombinase. A large number of recombinases suitable for this purpose exists in nature, including k Integrase (Int), yeast Flp, etc. (Nunes-Dbby et al., Nucl. Acids Res., 26:391-406 (1998)). Variant recombinases having different site specificity can also be used. 5 Since the disclosed variant recombinases recognize both wild type recombination sites and variant recombination sites that are not recognized by wild type recombinase, wild type recombinase and variant recombinases can be used to mediate sequential recombination between nucleic acids containing a combination of wild type recombination sites and variant recombination sites. 10 For example, generation of knockout animals and plants can be made more efficient by using a structure wild type recombination site-selectable marker wild type recombination site-endogenous gene-variant recombination site (rather than the conventional wild type recombination site-selectable marker wild type recombination site-endogenous gene-wild type recombination site). 15 Such a structure allows the selectable marker to be removed by the action of wild type recombinase without disturbing the gene since wild type recombinase will not recognize the variant recombination site to any significant degree. The endogenous gene can then be removed later by the action of a variant recombinase since the disclosed variant recombinases recognize both wild type 20 and variant recombination sites. In a preferred embodiment of the disclosed method, the first and second DNA sequences are introduced into cells connected by a pre-selected DNA segment. The segment can be a gene or any other sequence of deoxyribonucleotides of homologous, heterologous or synthetic origin. 25 Preferably, the pre-selected DNA segment is a gene for a structural protein, an enzyme, or a regulatory molecule. If the first and second recombination sites have the same orientation, activation of the regulatory nucleotide sequence produces a deletion of the pre-selected DNA segment. If the first and second recombination sites have opposite orientation, activation of the regulatory 30 nucleotide sequence produces an inversion of the nucleotide sequence of the pre-selected DNA segment. 30 WO 00/60091 PCT/US0O/09154 If a fourth DNA sequence (containing the third recombination site) is also introduced into cells, it is preferred that the second and fourth DNA sequences be introduced into cells connected by a second pre-selected DNA segment. The second segment can be a gene or any other sequence of 5 deoxyribonucleotides of homologous, heterologous or synthetic origin. Preferably, the second pre-selected DNA segment is a gene for a structural protein, an enzyme, or a regulatory molecule. If the second and third recombination sites have the same orientation, activation of the regulatory nucleotide sequence produces a deletion of the second pre-selected DNA 10 segment. If the second and third recombination sites have opposite orientation, activation of the regulatory nucleotide sequence produces an inversion of the nucleotide sequence of the second pre-selected DNA segment. Combinations of wild type and variant recombination sites, and combinations of different orientations of the recombination sites, in DNA 15 introduced into cells can multiply recombination options. For example, if the first and second recombination sites are wild type recombination sites and the third recombination site is a variant recombination site (all in the same orientation) then wild type recombinase can produce a deletion of the first pre selected DNA segment (but not the second) and a variant recombinase can 20 produce a deletion of the first, second, or both pre-selected DNA segments. This arrangement allows sequential deletion of the first and second pre-selected DNA segments. If the first and second recombination sites are wild type recombination sites and the third recombination site is a variant recombination site, and the 25 first recombination site has the opposite orientation from the second and third recombination sites (which, of course, have the same orientation) then wild type recombinase can produce an inversion of the first pre-selected DNA segment and a variant recombinase can produce a deletion of the second pre-selected DNA segment (and/or produce an inversion of the first pre-selected DNA 30 segment or the entire section spanning the first, second, and third recombination sites). 31 WO 00/60091 PCT/USOO/09154 If the first and third recombination sites are wild type recombination sites and the second recombination site is a variant recombination site, and the second recombination site has the opposite orientation from the first and third recombination sites (which, of course, have the same orientation) then wild type 5 recombinase can produce a deletion of the entire section spanning the first, second, and third recombination sites, and a Cre variant can produce an inversion of the first, second, or both pre-selected DNA segments. If the first and third recombination sites are wild type recombination sites and the second recombination site is a variant recombination site, and the 10 first recombination site has the opposite orientation from the second and third recombination sites (which, of course, have the same orientation) then wild type recombinase can produce an inversion of the entire section spanning the first, second, and third recombination sites, and a variant recombinase can produce a deletion of the second pre-selected DNA segments and an inversion of the first 15 pre-selected DNA segment. Many more combinations of wild type and variant recombination sites and or recombination site orientations are possible. For example, the variant recombinase can also be used with a different variant recombinase having a different site specificity rather than wild type recombinase. The above 20 examples illustrate the general principles involved in designing specific recombinations that may be desired. It should be understood that the above combinations of recombination sites can be extended to the use of more recombination sites (that is more than three) and more intervening, pre-selected DNA segments. 25 For some uses of the disclosed recombinases, first, second, and fourth DNA sequences comprising a first lox site, a second lox site, and a third lox site, respectively, are introduced into cells. As used herein the expression "lox site" means a nucleotide sequence at which the gene product of the cre gene, referred to herein as Cre, and/or the disclosed Cre variants, can catalyze a site 30 specific recombination. LoxP site is a 34 base pair nucleotide sequence (Figure 1) which can be directly synthesized or isolated from bacteriophage P1 by methods known in the art. The Lox P site is an example of a wild type lox site. 32 WO 00/60091 PCT/USO0/09154 One method for isolating a LoxP site from bacteriophage P1 is disclosed by Hoess et al., Proc. Nati. Acad. Sci. USA, 79:3398-3402 (1982). The LoxP site consists of two 13 base pair inverted repeats separated by an 8 base pair spacer region. The nucleotide sequences of the insert repeats and the spacer region are 5 as follows. ATAACTTCGTATAATGTATGCTATACGAAGTTAT Other wild type lox sites include LoxB, LoxL and LoxR sites which are nucleotide sequences isolated from E. coli. These sequences are disclosed and described by Hoess et al., Proc. Vatl. Acad. Sci. USA, 79:3398-3402 (1982). 10 Preferred wild type lox sites are LoxP or LoxC2. Lox sites can also be produced by a variety of synthetic techniques which are known in the art. For example, synthetic techniques for producing lox sites are disclosed by Ito et al., Nuc. Acid Res., 10:1755 (1982) and Ogilvie et al., Science, 214:270 (1981). The gene product of the cre gene is a recombinase herein designated 15 "Cre" which effects site-specific recombination of DNA at lox sites. As used herein, the expression "cre gene" means a nucleotide sequence which codes for a gene product which effects site-specific recombination of DNA in cells at lox sites. One cre gene (the wild type cre gene) can be isolated from bacteriophage P1 by methods known in the art. One method for isolating a cre gene from 20 bacteriophage P1 is disclosed by Abremski et al., Cell, 32:1301-1311 (1983). Genes engineered into cells for producing a foreign protein are often placed under the control of a highly active promoter. The activity of the promoter can result in an overproduction of the protein which interferes with the growth of the engineered cell. This overproduction of the protein can make it 25 difficult to grow the engineered cell in sufficient quantity to make protein production economically feasible. The present invention provides a method whereby engineered cells can be grown to a desired density prior to expressing the engineered gene. The engineered gene is expressed, as desired, by activating a regulatory nucleotide sequence responsible for controlling 30 expression of DNA encoding a variant recombinase. Methods of controlling the expression of an engineered gene include the following: 33 WO 00/60091 PCT/USOO/09154 (1) A DNA segment flanked by recombination sites in the same orientation is introduced into DNA in a cell between a promoter and an engineered gene to render the promoter incapable of expressing the gene. A second DNA sequence comprising a regulatory nucleotide sequence and DNA 5 encoding a variant recombinase is also introduced in the DNA. After the engineered cells are grown to a desired density, the regulatory nucleotide sequence is activated thereby effecting expression of the variant recombinase and producing a deletion of the DNA segment. The engineered gene would then be expressed. 10 (2) A gene for a regulatory molecule flanked by recombination sites in the same orientation is introduced into DNA in a cell. The regulatory molecule inhibits expression of an engineered gene. A second DNA sequence comprising a regulatory nucleotide sequence and DNA encoding a variant recombinase is also introduced into the DNA. After the engineered cells are grown to a desired 15 density, the regulatory nucleotide sequence is activated thereby effecting expression of the variant recombinase and producing a deletion of the gene for the regulatory molecule. The engineered gene would then be expressed. (3) An engineered gene lacking a promoter and flanked by two recombination sites in opposite orientations is introduced into DNA in a cell 20 such that the 3' end of the gene lies adjacent to the transcription start site of a regulatory nucleotide sequence. A second DNA sequence comprising a regulatory nucleotide sequence and DNA encoding a variant recombinase is also introduced into the DNA. Since the engineered gene would be transcribed in the antisense direction, no engineered protein would be produced. After the 25 engineered cell is grown to a desired density, the regulatory nucleotide sequence is activated thereby effecting expression of the variant recombinase and producing an inversion of the desired gene. The engineered gene could then be transcribed in the proper direction and expressed. Numerous methods and techniques have been developed for the use of 30 Cre recombinase and other, similar recombinases such as FLP. The disclosed variant recombinases can also be used in any of these methods. Adaptation of these methods to the use of the disclosed variant recombinases is 34 WO 00/60091 PCT/USO0/09154 straightforward. Generally, all that is required is substitution of a variant recombinase (or a gene expressing a variant recombinase) for the original recombinase (or recombinase gene) and, if appropriate, substitution of variant (or wild type) recombination sites for the original recombination sites used in 5 the method. Examples of methods involving wild type recombinases and wild type recombination sites that can be adapted for use with the disclosed variant recombinases and recombination sites include recombination of DNA in phage packaging systems, recombination of DNA to form phage display libraries (for 10 example, Fisch et al., Proc Natl Acad Sci USA 93(15):7761-6 (1996), and Waterhouse et al., Nucleic Acids Res 21(9):2265-6 (1993), and other uses (for example, (Sauer et al., Proc. Natl. A cad. Sci. USA 84: 9108-9112 (1987), Mullins et al., Nucleic Acids Res 25(12):2539-40 (1997), Aoki et al., Mol Med 5(4):224-31 (1999)). 15 Other examples of specific methods in which the disclosed variant recombinases can be used or substituted include methods disclosed in U.S. Patent No. 5,888,981, U.S. Patent No. 5,888,732, U.S. Patent No. 5,885,836, U.S. Patent No. 5,885,793, U.S. Patent No. 5,885,779, U.S. Patent No. 5,885,776, U.S. Patent No. 5,882,893, U.S. Patent No. 5,882,888, U.S. Patent 20 No. 5,877,400, U.S. Patent No. 5,871,907, U.S. Patent No. 5,866,755, U.S. Patent No. 5,866,361, U.S. Patent No. 5,859,310, U.S. Patent No. 5,858,657, U.S. Patent No. 5,854,067, U.S. Patent No. 5,851,808, U.S. Patent No. 5,849,995, U.S. Patent No. 5,849,989, U.S. Patent No. 5,849,708, U.S. Patent No. 5,849,572, U.S. Patent No. 5,849,571, U.S. Patent No. 5,849,553, U.S. 25 Patent No. 5,844,079, U.S. Patent No. 5,843,744, U.S. Patent No. 5,843,742, U.S. Patent No. 5,843,694, U.S. Patent No. 5,840,540, U.S. Patent No. 5,837,844, U.S. Patent No. 5,837,242, U.S. Patent No. 5,834,202, U.S. Patent No. 5,830,729, U.S. Patent No. 5,830,698, U.S. Patent No. 5,830,461, U.S. Patent No. 5,817,492, U.S. Patent No. 5,814,618, U.S. Patent No. 5,814,500, 30 U.S. Patent No. 5,814,300, U.S. Patent No. 5,807,995, U.S. Patent No. 5,807,708, U.S. Patent No. 5,801,030, U.S. Patent No. 5,800,998, U.S. Patent No. 5,795,734, U.S. Patent No. 5,795,726, U.S. Patent No. 5,792,833, U.S. 35 WO 00/60091 PCT/USO0/09154 Patent No. 5.792.632, U.S. Patent No. 5,789,156, U.S. Patent No. 5,777,194, U.S. Patent No. 5,776,449, U.S. Patent No. 5,773,697, U.S. Patent No. 5,770,384, U.S. Patent No. 5,767,376, U.S. Patent No. 5,763,240, U.S. Patent No. 5,756,671, U.S. Patent No. 5,744,343, U.S. Patent No. 5,744,336, U.S. 5 Patent No. 5,736,377, U.S. Patent No. 5,733,744, U.S. Patent No. 5,733,743, U.S. Patent No. 5,733,733, U.S. Patent No. 5,731,182, U.S. Patent No. 5,723,765, U.S. Patent No. 5,723,333, U.S. Patent No. 5,723,287, U.S. Patent No. 5,721,367, U.S. Patent No. 5,721,118, U.S. Patent No. 5,700,470, U.S. Patent No. 5,686,595, U.S. Patent No. 5,679,523, U.S. Patent No. 5,677,177, 10 U.S. Patent No. 5,658,772, U.S. Patent No. 5,656,438, U.S. Patent No. 5,654,182, U.S. Patent No. 5,654,168, U.S. Patent No. 5,650,491, U.S. Patent No. 5,650,308, U.S. Patent No. 5,650,298, U.S. Patent No. 5,643,727, U.S. Patent No. 5,641,866, U.S. Patent No. 5,641,748, U.S. Patent No. 5,639,726, U.S. Patent No. 5,635,381, U.S. Patent No. 5,629,179, U.S. Patent No. 15 5,629,159, U.S. Patent No. 5,614,389, U.S. Patent No. 5,612,205, U.S. Patent No. 5,596,089, U.S. Patent No. 5,591,609, U.S. Patent No. 5,589,362, U.S. Patent No. 5,539,094, U.S. Patent No. 5,530,191, U.S. Patent No. 5,527,695, U.S. Patent No. 5,510,099, U.S. Patent No. 5,478,731, U.S. Patent No. 5,441,884, U.S. Patent No. 5,434,066, U.S. Patent No. 5,378,618, U.S. Patent 20 No. 5,354,668, U.S. Patent No. 5,334,515, U.S. Patent No. 5,300,431, and U.S. Patent No. 4,959,317. 1. Use of Variants Recombinases in Plants and Plant Cells Methods for introducing a DNA sequence into plant cells are known in the art. Nucleic acids can generally be introduced into plant protoplasts, with or 25 without the aid of electroporation, polyethylene glycol, or other processes known to alter membrane permeability. Nucleic acid constructs can also be introduced into plants using vectors comprising part of the Ti- or Ri-plasmid, a plant virus, or an autonomously replicating sequence. Nucleic acid constructs can also be introduced into plants by microinjection or by high-velocity 30 microprojectiles, also termed "particle bombardment" or "biolistics" (Sanford, J. C., Tibtech 6: 299 (1988)), directly into various plant parts. The preferred means of introducing a nucleic acid fragment into plant cells involves the use of 36 WO 00/60091 PCT/US00/09154 A. tuniefaciens containing the nucleic acid fragment between T-DNA borders either on a disarmed Ti-plasmid (that is, a Ti-plasmid from which the genes for tumorigenicity have been deleted) or in a binary vector in trans to a disarmed Ti-plasmid. The Agrobacterium can be used to transform plants by inoculation 5 of tissue explants, such as stems. roots, or leaf discs, by co-cultivation with plant protoplasts, or by inoculation of seeds or wounded plant parts. Foreign genes can be introduced into a wide range of crop species. Thus, the disclosed variant recombinases and method are applicable to a broad range of agronomically or horticulturally useful plants. The particular method 10 which is employed to introduce the DNA sequence into a selected plant cell is not critical. In a preferred embodiment, DNA sequences are introduced into plant cells by co-cultivation of leaf discs with A. tumefaciens essentially as described by Horsch et al., Science, 227: 1229-1231 (1985) omitting the nurse cultures. 15 In the present method, the recombination sites are contacted with a variant recombinase, thereby producing the site specific recombination. In one embodiment, a variant recombinase, or messenger RNA encoding a variant recombinase, is introduced into the cells directly by micro-injection, biolistics, or other protein or RNA introduction procedure. In a preferred embodiment, 20 DNA encoding the variant recombinase is introduced into the plant cell under the control of a promoter that is active in plant cells. Suitable regulatory nucleotide sequences are known in the art. The promoter which is employed with a selected plant cell is not critical to the method of the invention. A partial list of suitable promoters include the 35S promoter of cauliflower mosaic virus 25 described by Odell et al., Nature, 313: 810-812 (1985); the promoter from the nopaline synthase gene of A. tumefaciens described by Depicker et al., J. of Mol. Apple. Genet., 1: 561-573 (1982); the promoter from a Rubisco small subunit gene described by Mazur and Chui, Nucleic Acids Research 13: 2373 2386 (1985); the 1' or 2' promoter from the TR-DNA of A. tumefaciens 30 described by Velten et al., EMBO J. 12: 2723-2730 (1984); the promoter of a chlorophyll a/b binding protein gene described by Dunsmuir et al., J. Mol. Apple. Genet. 2: 285-300 (1983); the promoter of a soybean seed storage protein gene 37 WO 00/60091 PCT/USOO/09154 described by Chen et al., Proc. Natl. A cad. Sci. USA, 83: 8560-8564 (1986); and the promoter from the wheat EM gene described by Marcotte et al., Nature 335: 454-457 (1988). Variant recombinases can be expressed throughout the plant generally in all cells at all stages of development, or expression of variant 5 recombinases can be more specifically controlled through the use of promoters or regulatory nucleotide sequences having limited expression characteristics. Variant recombinases can be expressed in a tissue specific manner, for example only in roots, leaves, or certain flower parts. Variant recombinases can be expressed in a developmentally specific time period, for example only during 10 seed formation or during reproductive cell formation. Expression of variant recombinases can also be placed under the control of a promoter that can be regulated by application of an inducer. In this case expression is off or very low until the external inducer is applied. Promoters active in plant cells have been described that are inducible by heat shock (Gurley et al., Mol. Cell. Biol. 6: 559 15 565 (1986)), ethylene (Bfoglie et al., Plant Cell 1: 599-607 (1989)), auxin (Hagan and Guilfoyle, Mol. Cell. Biol. 5: 1197-1203 (1985)), abscisic acid (Marcotte et al., Nature 335: 454-457 (1988)), salicylic acid (EPO 332104A2 and EPO 337532A1), and substituted benzenesulfonamide safeners (WO 90/11361). Control of expression of variant recombinases by the safener 20 inducible promoter 2-2, or its derivatives, allows the expression to be turned on only when the inducing chemical is applied and not in response to environmental or phytohormonal stimuli. Thus expression can be initiated at any desired time in the plant life cycle. Preferably, the regulatory nucleotide sequence is a 35S promoter or a 2-2 promoter. The above techniques and 25 materials can also be used to express wild type recombinase in plant cells if required by the particular recombination pattern to be accomplished. One application of the disclosed variant recombinases is in controlling male fertility in a method for producing hybrid crops. Hybridization of a crop involves the crossing of two different lines to produce hybrid seed from which 30 the crop plants are grown. Hybrid crops are superior in that more of the desired traits can be introduced into the production plants. For instance, quality traits such as oil content, herbicide resistance, disease resistance, adaptability to 38 WO 00/60091 PCT/USOO/09154 environmental conditions, and the like, can be hybridized in offspring so that the latter are invested with the most desirable traits of its parents. In addition, progeny from a hybrid cross may possess new qualities resulting from the combination of the two parental types, such as yield enhancement resulting 5 from the phenomenon known as heterosis. Controlled cross-fertilization to produce hybrid seeds has been difficult to achieve commercially due to competing self-fertilization, which occurs in most crop plants. Hybrid seed production is typically performed by one of the following means: (a) mechanically removing or covering the male organs to prevent self 10 fertilization followed by exposing the male-disabled plants to plants with male organs that contain the trait(s) desired for crossing; (b) growing genetically male-sterile plants in the presence of plants with fertile male organs that contain the trait that is desired for crossing; or (c) treating plants with chemical hybridizing agents (CHA) that selectively sterilize male organs followed by 15 exposing the male-disabled plants to plants with fertile male organs that contain the trait that is desired for crossing. Some disadvantages to each of these methods include: (a) applicability only to a few crops, such as corn, where the male and female organs are well separated; and it is labor intensive and costly; (b) genetically male sterile lines are cumbersome to maintain, requiring crosses 20 with restorer lines; (c) all CHAs exhibit some degree of general phytotoxicity and female fertility reduction. Also CHAs often show different degrees of effectiveness toward different crop species, or even toward different varieties within the same species. A molecular genetic approach to hybrid crop production applicable to a 25 wide range of crops and involves genetic male sterility is described in EPA 89 344029. This system involves the introduction of a cell disruption gene that is expressed only in the tapetal tissue of anthers thereby destroying the developing pollen. The resulting genetically male sterile plants serve as the female parents in the cross to produce hybrid seed. This system could be highly effective and 30 desirable. However one disadvantage is that since the male sterile parent is heterozygous for the sterility gene which acts as a dominant trait, only 50% of the plants grown from the hybrid seed are fertile, the rest retain the sterility 39 WO 00/60091 PCT/US0O/09154 gene. This situation will result in reduced pollen shed in the production field which may lead to reduced seed set and yield. Addition of recombinase technology to this hybrid scheme allows restoration of fertility to a much higher percentage of plants in the production field. as well as elimination of the cell 5 disruption gene. Placing the male sterility gene between recombination sites allows it to be deleted following introduction of a variant recombinase into the hybrid from the male parent. Another application of the disclosed variant recombinases is in making seedless produce. Seedlessness is desirable in consumed produce for 10 convenience and taste. Currently "seedless" watermelon is sold that actually contains some developed seed and a large number of immature seed that varies in size up to that of fully mature seed. To produce these watermelon first a hybrid cross is made between a tetraploid maternal parent and a diploid pollinator. The resulting triploid seed produces self-infertile plants that are 15 crossed with a diploid pollinator to produce seedless fruit (Kihara, Proc. Soc. Hort. Sci., 58: 217-230, (1951)). This production scheme suffers the following problems: (i) Creating a tetraploid plant, which is accomplished by a chromosome duplication method, is difficult. Also the number of seeds per fruit on this tetraploid plant must be low since this has a positive correlation with 20 seed number in the final product (Andrus, Production of Seedless Watermelons, USDA Tech. Bull. No. 1425 (1971)); (ii) good combining ability of the diploid pollinator and the tetraploid plant is difficult to achieve (Henderson, J. Amer. Soc. Hort. Sci., 102: 293-297 (1977)); (iii) the triploid seeds are much inferior to regular diploid seeds in vigor and germinability (Maynard, Hort. Sci., 24: 25 603-604 (1989)). These problems, together with incomplete seedlessness in the final product, make the development of seedless watermelon slow and difficult. This ploidy-based approach to seedlessness is possible only in those few species where unusual euploidy plants (tetraploid and triploid for watermelon, for example) are viable. 30 A molecular genetic approach to seedlessness involving the disclosed variant recombinases is much more efficient, resulting in a more reliably seedless product and does not involve changes in ploidy. Thus it is more 40 WO 00/60091 PCT/USO0/09154 generally applicable to a wider range of species. A recombination site/polyA inactivated cell disruption gene regulated by a seed-specific promoter is introduced into a plant. When this plant is crossed to a plant expressing a variant recombinase,. the disruption gene is activated and expressed in the seed, 5 thereby disrupting seed development. The certainty of endosperm failure (caused by the cell disruption gene product) leading to the abortion of the whole seed is very high. In most dicots, the endosperm supplies the nutrients needed for early embryo development. Endosperm abortion invariably leads to seed abortion (Brink and Cooper, Bot. Rev. 8: 423-541 (1947)). 10 The seed-specific promoter used can be selected from the group of promoters known to direct expression in the embryo and/or the endosperm of the developing seed, most desirable in the endosperm. Examples of seed specific promoters include but are not limited to the promoters of seed storage proteins. The seed storage proteins are strictly regulated, being expressed 15 almost exclusively in seeds in a highly tissue-specific and stage-specific manner (Higgins et al., Ann. Rev. Plant Physiol. 35: 191-221 (1984); Goldberg et al., Cell 56: 149-160 (1989)). Also, different seed storage proteins may be expressed at different stages of seed development and in different parts of the seed. 20 There are numerous examples of seed-specific expression of seed storage protein genes in transgenic dicotyledonous plants. These include genes from dicotyledonous plants for bean p-phaseolin (Sengupta-Goplalan et al., Proc. Natl. Acad. Sci. USA 82: 3320-3324 (1985) and Hoffman et al., Plant Mol. Biol. 11: 717-729 (1988)), bean lectin (Voelker et al., EMBO J6: 3571 25 3577 (1987), soybean lectin (Ocamuro et al., Proc. Natl. Acad. Sci. USA 83: 8240-8344 (1986)). soybean kunltz trypsin inhibitor (Perez-Grau and Goldberg Plant Cell 1: 1095-1109 (1989)), soybean P-conglycinin (Beachy et al., EMBO J 4: 3047-3053 (1985), Barker et al., Proc. Natl. Acad Sci. 85: 458-462 (1988), Chen et al., EMBO J7: 297-302 (1988), Chen et al., Dev. Genet, 10: 112-122 30 (1989), Naito et al., Plant Mol. Biol. 11: 683-695 (1988)), pea vicillin (Higgins et al., Plant Mol. Biol. 11: 109-123 (1988)), pea convicilllin (Newbigin et al., Planta 180: 461 (1990)), pea legumin (Shirsat et al., Mol. Gen. Genetics 215: 41 WO 00/60091 PCT/USOO/09154 326 (1989)), rapeseed napin (Radke et al., Theor. Apple. Genet. 75: 685-694 (1988)), as well as genes from monocotyledonous plants such as for maize 15 kd zein (Hoffman et al., EMBOJ6: 3213-3221 (1987)), barley p-hordein (Marris et al., Plant Mol. Biol. 10: 359-366 (1988)), and wheat glutenin (Colot 5 et al., EMBOJ6: 3559-3564 (1987)). Moreover, promoters of seed-specific genes operably linked to heterologous coding regions in chimeric gene constructions also maintain their temporal and spatial expression pattern in transgenic plants. Such examples include Arabidopsis thaliana 2S seed storage protein gene promoter to express enkephalin peptides in Arabidopsis and 10 Brassica napus seeds (Vandekerckhove et al., Bio/Technology 7: 929-932 (1989)), bean lectin and bean p-phaseolin promoters to express luciferase (Riggs et al., Plant Sci. 63: 47-57 (1989)), and wheat glutenin promoters to express chloramphenicol acetyl transferase (Colot et al., EMBO J. 6: 3559-3564 (1987)). Promoters highly expressed early in endosperm development are most 15 effective in this application. Of particular interest is the promoter from the a subunit of the soybean p-conglycinin gene (Walling et al., Proc. NatL. Acad, Sci. USA 83: 2123-2127 (1986)) which is expressed early in seed development in the endosperm and the embryo. The cell disruption gene used can be selected from a group of genes 20 encoding products that disrupt normal functioning of cells. There are many proteins that are toxic to cells when expressed in an unnatural situation. Examples include the genes for the restriction enzyme EcoRI (Barnes and Rine, Proc. Natl. Acad. Sci. USA 82: 1354-1358 (1985)), diphtheria toxin A (Yamaizumi et al., Cell 15: 245-250 (1987)), streptavidin (Sano and Cantor, 25 Proc. Natl. A cad. Sci. USA 87: 142-146 (1990)), and barnase (Paddon and Hartley, Gene 53: 11-19 (1987)). Most preferred for this system is the coding region of barnase which has been shown to be highly effective in disrupting the function of plant cells (EPA 89-344029). A highly desirable seedless system is one in which fully fertile F 1 seed 30 develops, that can then be grown into plants that produce only seedless fruit. This system is economically favorable in that for each cross pollination, a large number of seedless fruits result: the number of F 1 seed from one cross X the 42 WO 00/60091 PCT/US00/09154 number of fruits produced on an Fl plant. Also incorporated in this scheme are the advantages of growing a hybrid crop, including the combining of more valuable traits and hybrid vigor. This is accomplished in the same manner as described above except that the recombination site/polyA-inactivated disruption 5 gene is expressed from a seed maternal tissue (seed coat or nucellus)-specific promoter. For example, the seed coat is the outgrowth of the integuments, a strictly maternal tissue. Therefore the hybrid cross that brings the recombination site/polyA-inactivated disruption gene together with the recombinase gene does not involve this seed coat tissue. The seed coat of the F1 10 seed has either recombination sites or recombinase, depending on which is used as the female parent, and thus F1 seed develop normally. After the Fl seed gives rise to a fruit-bearing F1 plant, all vegetative cells (including seed coat cells) inherit both recombination sites and recombinase from the embryo. Thus the seed coat of the F1 plant has an activated cell disruption gene. 15 The seed coat is an essential tissue for seed development and viability. When the seed is fully matured, the seed coat serves as a protective layer to inner parts of the seed. During seed development, the seed coat is a vital nutrient-importing tissue for the developing embryo. The seed is nutritionally parasitici" to the mother plant. All raw materials necessary for seed growth 20 must be imported. In seeds of dicotyledonous plants, the vascular tissue enters the seed through the funiculus and then anastamoses in the seed coat tissue. There is no vascular tissue connection or plasmodesmata linkage between the seed coat and the embryo. Therefore, all nutrient solutes delivered into the developing seed must be unloaded inside the seed coat and then move by 25 diffusion to the embryo. Techniques have been developed to study the nutrient composition in the seed coat (Hsu et al., Plant Physiol. 75: 181 (1984); Thorne & Rainbird, Plant Physiol. 72: 268 (1983); Patrick, J. Plant Physiol. 115: 297 (1984); Wolswinkel & Ammerlaan, J. Exp. Bot. 36: 359 (1985)), and also the detailed cellular mechanisms of solute unloading (Offler & Patrick, Aust. J. 30 Plant Physiol. 11: 79 (1984); Patrick, Physiol. Plant 78: 298 (1990)). It is obvious that the destruction of this vital nutrient-funnelling tissue causes seed abortion. 43 WO 00/60091 PCT/USOO/09154 The disclosed tissue-specific and site-directed DNA recombination can be used to obtain seedless fruit production. This method is useful for the production of seedless watermelon. for example. A combination of gene expression specific for maternally inherited seed tissue and the disclosed 5 recombinase system can be used for the production of seedless watermelon. The system can be universally applied to any horticultural crop in which the presence of seeds is undesirable and difficult to be eliminated through conventional breeding methods. The system also allows the normal production of F 1 seeds. The ability to maintain heterosis is an advantage of producing F2 10 seedless fruits. The existing production of seedless watermelon indicates that seed development is not essential for the watermelon fruit development. However, conventional production of seedless watermelon using the ploidy imbalance trick has never been very popular due to the difficulty of overcoming the yield 15 and production problems. Creating and maintaining the tetraploid (4n) female germline, and producing the triploid (3n) seeds have made the seed cost high. Cross-pollination is needed for the production of triploid seeds (4n X 2n) and seedless fruits (3n X 2n). Also triploid seed germination is usually poor due to ploidy imbalance. 20 The present approach eliminates the dependence on polyploid germlines and provides an efficient system for producing seedless fruit. The products of double fertilization of higher plants are the embryo and endosperm. The seed coat (including the integumentary tapetum) and nucellus (the tissue encompassing the embryo sac) are the remaining seed tissues that are 25 maternally inherited. In addition to general protection, the seed coat and nucellus also play an important role in importing nutrients into the developing embryo and endosperm. Seed development will be aborted if this vital nutrient importing mechanism of the seed coat/nucellus is debilitated. This will be accomplished by using the recombinase system to activate a cell-damaging gene 30 only in these tissues. Controlling the gene activation in a maternal tissue specific manner allows production of normal F 1 seed, but abortion of F2 seed. A seed coat or nucellus promoter is coupled to a tissue-destructive (lethal) gene 44 WO 00/60091 PCT/USOO/09154 in order to prevent seeds from forming. The destructive gene is inactive in the seed parent due to the presence of a blocking transcription terminator. The terminator is flanked by recombination sites for subsequent excision by a recombinase-mediated recombination event. Expression of the recombinase is 5 also controlled by the seed coat/nucellus-specific promoter. When plants carrying the separate recombinase and recombination site constructs are crossed, the Fl seed will be viable because seed coat/nucellus is maternal tissue, and in that tissue recombinase and recombination sites are not combined. When the F 1 seed is used as planting seed, the self-pollinated or out-crossed plants 10 will produce seedless fruits or vegetables, since in seed coat/nucellus tissues recombinase and recombination sites are combined, and the lethal gene is activated. 2. Use of Variant Recombinases for Phage Packaging The disclosed variant recombinases can also be used to aid in phage 15 packaging. The cloning system described herein utilizes a headful in vitro packaging system to clone foreign DNA fragments as large as 95 kb which permits the isolation of DNA fragments that are at least twice the size of those that can be obtained by lambda cosmid cloning. This increased cloning capacity has the following utility: 20 (1) Genes in the 45-95 kb size range and, more particularly, in the 70-95 kb size range can now be directly cloned and genes in the 25-45 kb size range can be cloned more easily. (2) Chromosomal "walking" and "jumping" techniques can be speeded up by a factor of at least two and should be more accurate because of the 25 reduced number of contiguous segments that need to be linked together. (3) The cloning system of the invention is useful as a means for the delivery of DNA efficiently to bacteria which otherwise do not take up DNA from solution well. Specifically, the headful packaging system of this invention for cloning 30 foreign DNA fragments as large as 95 kb comprises: (a) modifying vector DNA by inserting a stuffer fragment into a blunt end producing site which is proximal to a pac site; 45 WO 00/60091 PCT/USOO/09154 (b) digesting the product of step (a) to produce two vector arms each of which contains (i) a blunt end, (ii) another end which is compatible with the foreign DNA fragment which is to be cloned, and (iii) a recombination site; (c) ligating the foreign DNA to the product of step (b) without 5 generating concatemers; (d) reacting the product of step (c) with pac cleavage proficient extract and head-tail proficient extract wherein the ratio of large heads to small heads in the head-tail extract is at least 5:1; (e) infecting a bacterial strain expressing a variant recombinase with the 10 product of step (d); and (f) recovering the cloned DNA. The term pac is a generic name which refers to the site needed to initiate packaging of DNA. The pac cleavage proficient extract contains the recognition proteins necessary to cleave the pac site and, thus, initiate 15 packaging. The head-tail proficient extract contains the heads and tails needed to package the cloned DNA into a virus particle. The term concatemer means a DNA molecule consisting of repeating units arranged in a head-to-tail configuration. The term stuffer fragment refers to a DNA fragment which is inserted into the vector DNA at a unique site, and within which headful 20 packaging is terminated. The terms bacteriophage and phage are used interchangeably herein. Although many of the elements described herein pertain to the P1 bacteriophage cloning system, those skilled in the art will appreciate that, with the exception of the components needed to package DNA (pac and packaging 25 extracts), many of the elements discussed below, such as plasmid replicon and a multicopy or lytic replicon, pertain to the recovery of packaged DNA and can be used to recover DNA in bacteria, such as E. coli, with other cloning systems, for example, bacteriophage, yeast, etc. Bacteriophages which are suitable to practice the invention must have a 30 large head capacity and the elements necessary for packaging DNA must be defined. For example, for phages P22 and TI, which utilize headful packaging, the necessary packaging elements are defined. However, P22 and TI do not 46 WO 00/60091 PCT/USOO/09154 have a very large head capacity. On the other hand, for phage T4, which has a large head capacity, the necessary packaging elements have not been defined. The elements necessary for packaging DNA (i.e., an in vitro headful packaging system) are the following: 5 (1) a unique site, pac, which is cleaved by recognition proteins; it is the pac cleavage proficient extract which contains the recognition proteins necessary to cleave the pac site; and (2) empty phage heads into which the DNA is packaged until the head has been completely filled, then a cleavage event is triggered (the "headful" cut) 10 which separates the packaged DNA away from the remaining components; it is the head-tail proficient extract which contains the heads and tails needed to package the cloned DNA into virus particle. Although initiation of packaging is site-specific (cleavage of pac site initiates packaging), termination of packaging is not site-specific. In other 15 words, no unique site is recognized, as packaging will terminate at whatever point the head has been filled. In the case of P1, the DNA substrate used in the packaging reaction during the viral life cycle is a concatemer consisting of individual units of the P1 chromosome arranged in a head-to-tail manner. Headful packaging, using 20 either P1 phage or any other phage, is a four step process: (1) In the first step a unique site, pac, is recognized and cleaved by the pac recognition proteins (PRPs); (2) DNA on one side of the cleavage is packaged into an empty phage head until the head has been completely filled; (3) a second cleavage event is then triggered (the "headful" cut) that separates the packaged DNA away from 25 the rest of the concatemer; and (4) initiation of a second round of DNA packaging from the free end generated by the previous "headful" cut--hence the term processive headful cutting. However, if a concatemer is not generated then processive headful packaging does not occur. The ends of the packaged P1 DNA do not contain complementary 30 single-stranded sequences, as do the ends of packaged bacteriophage lambda DNA, and consequently after P1 DNA is injected into a bacterium its cyclization does not occur by strand annealing but rather by recombination 47 WO 00/60091 PCT/US0O/09154 between homologous sequences present at the ends of the molecule. Because of this circumstance, any vector that uses P1 packaging, or for that matter any headful packaging mechanism, must devise a means of cyclizing the linear packaged DNA by recombination. Cyclizing is accomplished by incorporating 5 recombination sites into the vector and using a disclosed variant recombinase to cyclize the DNA after injection into gram-negative bacterial strains expressing the variant recombinase. P1 produces two head sizes, a big head that can accommodate 105-110 kb of DNA, and a small head that can accommodate no more than 45 kb of 10 DNA. Normally the ratio of big to small heads in a P1 wild-type infection is 10:1, however, in the cm-2 mutant of P1 used to prepare some of the packaging lysates described herein, the radio of head sizes is 1:1. The head-tail packaging lysate prepared from the cm mutant of P1 contained the usual ratio of big to small heads which is about 10:1. This is the preferred lysate for preparing head 15 tail packaging extract. To ensure packaging of DNA exclusively into the big phage heads, the DNA must be bigger than that which can be accommodated by the small heads. It is generally desired that there be a large excess of big heads. However, the ratio of large heads to small heads should not fall below a ratio of about 5:1. 20 Illustration The following illustration describes an example of how the disclosed method can be used to generate variant FLP recombinases with altered site specificity. As with other recombinase, the method preferably uses the following components: 25 1. An in vitro mutagenesis system; 2. A recombinase expression plasmid that allows varying levels of expression by a simple environmental control (for example, by the presence of varying amounts of an inducer substance in the growth media, by temperature, or by osmolarity); 30 3. An indicator/selector bacterial strain. The strain carries both an indicator recombination substrate for detection of recombination at the wild type recombination site and a second recombination substrate that allows 48 WO 00/60091 PCT/USO0/09154 selection for recombinase mutants that have gained the ability to recognize and perform recombination at a target mutant recombination site (that is, a variant recombination site). Importantly, the wild type and target mutant sites are designed so that recombination between the mutant and wild type sites is 5 blocked even with a mutant recombinase that can recognize both the wildtype site and also the target mutant site. This design prevents unwanted recombination between the wildtype and target mutant recombination sites that could interfere with either selection or detection of desired recombinational outcomes. The block is imposed by designing the wildtype and mutant sites to 10 have different spacer regions (that is, different compatibility sequences), for example, the normal "wt" spacer for the wildtype recombination site, and an alternative spacer "Al" for the other recombination substrate. In an otherwise nonmutant recombination site DNA recombination proceeds efficiently both for recombination sites having the wt spacer (that is, a recombination between two 15 wt sites) and also for sites having the Al spacer (that is, a recombination between two Al sites). Yet, recombination between the Al site and the wt site is blocked (that is, recombination between a wt site and an Al does not occur). This strategy is applicable to all recombinases that have a recombination target site displaying one or more recombinase binding sites (repeat elements) on each 20 side of a spacer region in which recombination occurs (Nunes-D6by et al., Nucl. Acids Res., 26:391-406 (1998)). Such sites display a requirement for homology in the spacer elements for optimal recombination activity and has been shown to be the case for members of the Int family of recombinases, including Cre, lambda Int, and FLP (Craig, Ann. Rev. Genet. 22:77-105 (1988)). 25 The preferred in vitro mutagenesis system is that of Stemmer (Stemmer, Nature 370:389-391 (1994)), or a variant of that strategy. After mutagenesis and assembly of fragments into a full-length FLP gene, it is cloned into the expression vector. The expression plasmid to be used can be any of the "inducible" 30 expression plasmids available in bacteria. For this illustration one of the pBAD plasmids for E. coli was chosen that allows expression of recombinase by growth on arabinose (Guzman et al., J. Bacteriol. 177:4121-4130 (1995)), and 49 WO 00/60091 PCT/USOO/09154 which can be turned off by growth on glucose (and no arabinose). For this illustration the expression plasmid carries the replication origin of pACYC , and the FLP recombinase gene is under the control of the E. coli ara promoter region. In addition, the plasmid carries the selectable marker Cm' which 5 confers resistance to the antibiotic chloramphenicol. Because the pACYC replicon is low copy, its use may be advantageous in preventing excessive expression of FLP. Alternatively, a higher copy replicon could be used, such as that of ColE 1. In that case the expression level of FLP must also be carefully controlled using the inducer substance arabinose. 10 The indicator/selector bacteria carries two different reporter constructs for FLP-mediated recombination. The first reporter construct consists of two FRT sites (ELP recombination target; that is, the recombination site recognized by FLP recombinase) in direct orientation (an excision substrate) and resides on a low copy replicon that is compatible with the FLP expression construct. In 15 this example the first substrate is integrated into the E. coli genome. This can be done by incorporating the FRT substrate onto phage lambda and then constructing a lambda lysogen. Alternatively the FRT substrate could reside on a low-copy replicon that is compatible with that of the FLP expression vector and which has an additional selectable marker, for example resistance to 20 bleomycin. This FRT substrate carries two FRT sites in direct orientation flanking a gene whose presence can be easily monitored. In this illustration a constitutively expressing lacZ gene was used whose presence can be determined simply by growing colonies plates containing X-gal, upon which they will become blue in color. Loss of the lacZ gene by FLP-mediated recombination 25 results in white colony formation on X-gal plates. The indicator/selector bacterial strain also carries a second FRT-like substrate. This is a plasmid element having two FRT-M sites in direct orientation flanking DNA sequences (STOP) that disallow expression of a downstream selectable marker. In this example, nonexpression of the selectable 30 marker is achieved by placing genetic elements in the following order: constitutive promoter - FRT-M - STOP - FRT-M - 'neo, where 'neo indicates the promoterless neo gene of Tn5. Hence this cassette cannot express neo and 50 WO 00/60091 PCT/USO0/09154 cells are sensitive to the antibiotic kanamycin. Excision of STOP by recombination at FRT-M is designed to permit expression of neo so that cells now become resistant to kanamycin. The plasmid carries an additional selectable marker, Ap' conferring resistance to ampicillin, to maintain presence 5 of the plasmid in E. coli. The STOP sequence here is the strong transcriptional terminator rrnBT1T2 (Liebke et al.. Nucleic Acids Res. 13:5515-5525 (1985)). Figure 17 shows wt FRT, FRT-Al, and FRT-M sites used in this illustration. Although the wildtype FRT site displays three inverted repeat elements, recombination proceeds efficiently with sites carrying two of these 10 repeats in the inverted configuration shown (Jayaram, Proc. Nati. A cad. Sci. USA 82:5875-5879 (1985)). Either the full or minimal site can be used since both are recombinationally functional. The FRT-Al site is designed to have an altered spacer but which is still functional for self X self recombination (Senecoff et al., . Biol. Chem. 261:7380-7386 (1986)). The target FRT-M site 15 is designed to carry symmetrical mutations in the repeat elements that disallow efficient FLP-mediated recombination (Senecoff et al., J. Mol. Biol. 201:405 421 (1988)), and also the spacer mutation of FRT-Al. Importantly, the FRT-M site differs from the FRT site in two ways. First, both of the 13 bp inverted repeat elements (that is the recognition 20 sequences) are mutated in a symmetrical manner such that the wt FLP enzyme does not catalyze recombination between two FRT-M sites, or does so only extremely poorly (< 0.1%). Second, the spacer region (that is, the compatibility sequence; the 8 bp region between the 13 bp inverted repeats) is replaced with an alternate spacer Al. The alternate spacer when present in an otherwise wt 25 FRT site, which we will call FRT-A1, is permissive for FLP-mediated recombination between two FRT-Al sites, but does not permit recombination between FRT-Al and FRT. Use of the FRT-M site which contains the heterologous spacer prevents FLP-mediated recombination between FRT and FRT-M by FLP or a mutant FLP protein that might otherwise catalyze 30 recombination between the wt FRT substrate and the mutant target FRT-M substrate. Unwanted recombination between the wildtype and target mutant recombination sites would decrease efficiency of the selection procedure by (a) 51 WO 00/60091 PCT/USOO/09154 not limiting recombination at the target mutant site specifically to these sites and thus compromising the selection (at FRT-M sites) for mutant FLP recombinases, (b) affecting the accuracy of the specific indication of activity at the wt FRT sites, and (c) decreasing either the plasmid stability of the FRT-M 5 selector substrate or the integrity of bacterial chromosome (or compatible plasmid) carrying the wt FRT sites. Procedure (Figure 18): the FLP gene is mutagenized in vitro and then cloned into the inducible expression vector, in this case a pBAD derivative that places FLP under the control of the arabinose-inducible pBAD promoter. The 10 pool of mutagenized FLP genes is transformed into the FRT indicator/selector strain which is pre-induced with arabinose and/or induced with arabinose during DNA transformation. Bacterial colonies are then selected to be simultaneously resistant to chloramphenicol (to retain the FLP expression plasmid), ampicillin or carbenicillin (to retain the selector plasmid) and kanamycin (to select for 15 cells in which FRT-M X FRT-M recombination has occurred) on agar plates containing either arabinose (for continued FLP expression) or glucose (to prevent prolonged FLP expression). In some instances it may be advantageous to limit FLP expression to better enrich for those FLP mutants that either have more avidity for recombination at the FRT-M sites or to better exclude those 20 FLP mutants that retain activity at the wildtype FRT sites. This is because prolonged or high-level FRT expression can lead to inefficient but detectable recombination at mutant sites. Either all Kan' colonies or only those that are blue on X-gal plates are then pooled and harvested for DNA preparation. A second round of FRT gene 25 mutagenesis and selection is then initiated by PCR amplification. Multiple rounds of mutagenesis and selection are used to obtain FRT mutants with altered site-specificity. Comparison of various individual isolates allows determination of critical amino acid residues that contribute to the desired mutant phenotype. 30 The same rationale and procedure can be used to generate a second class of altered FLP recombinases. The target mutant FRT site used is, however, 52 WO 00/60091 PCT/USOO/09154 different. In this case the target is the FRT-M2 site (Figure 19) which carries a different binding site mutation(s) than does the FRT-M site as described above. Examples Example 1: Selection of Variant Cre Recombinases 5 The following example describes the production and analysis of some examples of the disclosed variant recombinases. Cre mutants characterized by a wider substrate recognition were created, applying a technique called directed molecular evolution: Multiple rounds of a random mutagenesis procedure (DNA shuffling; Stemmer, W. P. C., Proc. Nati. Acad. Sci. USA, 91:10747 10 10751 (1994)) and a sensitive selection for the desired phenotypes allow to accumulate candidate mutants within the generated pools of mutated sequences. The Cre mutants created in this example showed wt-like activity on loxP sites. In addition, they performed on an altered substrate, called loxK2, that is no recognized by the wt enzyme. Two transversions from adenine (loxP) to 15 thymine (loxK2) at positions 11' and 12' of the lox sequence are the barriers that inhibit wt Cre from recognizing loxK2: The two thymines are believed to cause repulsive forces with the acidic side chain of a glutamate residue in the J helix of wt Cre (position 262). This glutamate was found to be replaced by a glycine in all mutants with remarkably increased activity on loxK2. Additional 20 site-directed mutagenesis experiments, confined to the glutamate at position 262 of Cre, could confirm that E262G but also E262W mutations alone are sufficient to increase loxK2 activity by a factor of 103 without affecting loxP recognition. Other point mutations identified in the analyzed mutants may however be responsible for increasing the newly obtained specificity even 25 further (10 fold compared to E262G alone). MATERIALS AND METHODS General Procedures Standard Reagents The following reagents were used in all experiments: 1 Ox' TBE (Tris 30 Borate-EDTA, pH 8.3) was purchased from Biofluids, Inc. (Rockville, MD) and diluted to 1x with deionized water prior to use. TE (Tris-HC 110 mM, EDTA 1 mM, pH 8.0 and pH 7.5) and Tris-HCI (1 M, pH 7.5 and 8.0) came from 53 WO 00/60091 PCT/USOO/09154 Quality Biological, Inc. (Gaithersburg, MD), as well as autoclaved LB (Luria Bertani) and SOC broth. L-(+)-arabinose (>99%) was ordered from Sigma Aldrich Fine Chemicals (St. Louis, MO) and anyhdrous D-glucose from Mallinckrodt Laboratory Chemicals (Phillipsburg, NJ). 5 Gel Electrophoresis For DNA electrophoresis, 0.8% agarose TBE gels were used (GTG Sea Kem Agarose (FMC, Rockland, ME)). Gels were prestained with 0.25 Ig/ml EtBr (Ethidium Bromide, 10 mg/ml (Life Technologies, Inc., Grand Island, NY)). The used electrophoresis apparatus was a DNA SUB CELLTM (BioRad, 10 Hercules, CA) with an OSP 105 (OWL, Woburn, MA) powersupply. Gels were run at 60 V (5 V/cm) as recommended by Sambrook et al., Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press (Second Edition) (1989). Occasionally, for small amounts of samples, 50 ml minigels were used under similar conditions (Hoefer HE33, Hoefer Scientific Instruments, San Francisco, 15 CA). Molecular weight standards were k / Hind III digest (Research Genetics, Huntsville, AL) and Ready-LoadTM 100 bp DNA ladder (Life Technologies), providing a standard size range from 100 bp to 23130 bp (Figure 4). For standard fragment purification from gel, the Geneclean II@ Kit (BIO 101, Inc., La Jolla, CA (11/98) was used, following the manufacturer's instructions. 20 Minipreps and Plasmids Plasmids for diagnostics, cloning, and sequence analysis were prepared using the WizardTM Minipreps Plus Kit (Promega, WizardTM Minipres Plus DNA Purification System. Instruction Manual (Madison, WI) (1/96)). Useful ones were assigned a pBS number and stored in TE pH 8.0 at +4*C. 25 Oligonucleotides All oligonucleotides used as PCR primers, for plasmid construction, or in the mutagenesis procedure, were ordered from Midland, Inc. (Midland, TX) in gel filtration (GF) quality. The lyophilized oligonucleotides were assigned a BSB number, suspended in HPLC grad water (Sigma-Aldrich) at a final 30 concentration of 300 ptM, and stored at -20*C. 54 WO 00/60091 PCT/USOO/09154 DNA Digests and Ligations All enzymes used for DNA manipulations (restriction enzymes, T4 ligase, etc.) were purchased from New England Biolabs, Inc., Catalog. (Beverly, MA) (1998/99) and used as recommended in the manufacturer's 5 catalog (1998/99). Briefly, for restriction enzyme digests the total reaction volume was 20 pl with approximately 10 units (U) of enzyme. For DNA fragment ligations, 10 pd with 200 U of T4 DNA ligase were used. E. coli Strains All E. coli strains, except otherwise mentioned, were derived originally 10 from DH5a: endAl hsdR17 (rKmK+) supE44 thi-1 recAl gyrA (NaIR) relA1 A(lacIZYA-argF) U169 deoR (M80 dlac A(lacZ)M 15) (Woodcock et al., Nucl. Acids Res., 17:3469-3478 (1989); Raleigh et al., In Current Protocols in Molecular Biology, eds. Ausubel, F.M. et. al. (New York: Publishing Associates and Wiley Interscience). Unit 1.4 (1989). After modification with k prophages 15 or plasmids, strains were catalogued by assigning them a BS number and stored at -80*C with 10% DMSO (Dimethylsulfoxide) after overnight culture in appropriate selection medium. Transformation of E. coli For all plasmid transformations of E. coli strains, electroporation was 20 preferred over chemical protocols. Electorcompetent cells were made and used for electroporation as described by Smith et al., Focus, 12:38-40 (1990). The appropriate cell porator and cuvettes were from Life Technologies. Depending on the selection procedure after electroporation, the time in SOC medium (Smith et al., Focus, 12:38-40 (1990) at 37 0 C under agitation (Lab-Line® Orbit 25 Environ-Shaker, Lab Line Instruments, Inc., Melrose Park, IL) prior to plating on selection medium was 1 h for ampicillin (Ap) and 2 h or more h for kanamycin (Kan) selection. For induction of cre expression (as described below), the transformants were cultivated in SOB (Smith et al., Focus, 12:38-40 (1990)) supplemented with 0.2% of L-(+)-arabinose (Sigma-Aldrich) plus 30 20mM of MgCl 2 (referred to as induction medium) for 2.5 h and 4 h before plating on the appropriate selection media (see below). Resulting colony 55 WO 00/60091 PCT/USOO/09154 numbers were counted after overnight incubation at 37'C in a gravity convection incubator (Precision Scientific, Chicago, IL). E. coli Cultures LB (as mentioned above) was used as the standard medium for all E 5 coli cultures (liquid or solid). For selection and screening, the appropriate reagents at the following concentrations were added: Table 1: List of reagents used for selection and screening of E. coli cultures. Reagent Concentration Stock Solution 10 Ampicillin (Ap) 100 pLg/ml 50 mg/ml in H 2 0 Chloramphenicol (Cm) 27 tg/ml 34 mg/ml in EtOH Kanamycin (Kan) 16 p.g/ml 10 mg/ml in H 2 0 X-gal 0.003% 2% in DMF (w/v) Z-ara 0.006% 2% in DMF (w/v) 15 The concentration of stock solutions, stored at -20'C and their dilutions in liquid LB medium or LB-agar plates is given. X-gal stands for 5-bromo-4 chloro-3-indolyl-p-D-galactopyranoside, and Z-ara for 5-bromo-3-indolyl-a-L arabinofuranoside. All reagents, except Z-ara, were purchased from Life Technologies and aliquoted in the desired stock concentration for storage. Z-ara 20 (Berlin and Sauer, Anal. Biochem., 243:171-175 (1996)) was a generous gift from W. Berlin. Ready-to-use solid LB-agar plates (2%) plain, or supplemented with Ap (100 pg/ml) were purchased from Digene, Inc. (Beltsville, MD). For all other reagent combinations in solid medium, plates were poured according to the 25 needs using autoclaved 2% LB agar purchased from Biofluids. Polymerase Chain Reaction (PCR) Standard PCR reactions were carried out in 50 ptl total volume with the following reagents (all, except noted, from Perkin Elmer, Foster City, CA): 1x PE buffer II (without MgCl 2 ), 2 mM MgCl 2 , 250 pM of each dNTP, 0.8 pM of 30 each primer, ca. 50 ng of template DNA, qsp. H 2 0 (HPLC grade, Sigma Aldrich) to 49.5 pl. For mutagenic PCR reactions (also referred to as error prone PCR), the amount of each dNTP was reduced to 20 pM and 0.25 mM of 56 WO 00/60091 PCT/US00/09154 MnCl 2 added. After denaturation at 95'C for 5 min, 0.5 tl of 5 U/ l PE Amlpi Taq Polymerase was added at approximately the annealing temperature of the primers. After mixing, the appropriate thermal cycles were carried out (as indicated individually below), using a PTC 200 thermal cycler (DNA Engine, 5 MJ Research, Cambridge, MA). When finished, all PCR products immediately were loaded on an agarose gel, or separated from enzyme, nucleotides and primers by applying the Wizard TM PCR Preps Kit (Promega, WizardTM Minipres Plus DNA Purification System. Instruction Manual (Madison, WI) (1/96)). PCR products were recovered using deionized water and stored frozen at -20*C. 10 Sequence Analysis Sequence analysis of plasmid constructions and cre mutants were carried out on a PE ABI PRISM T M 310 Genetic Analyzer (Perkin Elmer) according to recommendations in the manufacturer's protocol P/N 402078 Revision A (1995) for the ABI PRISM T M Dye Terminator Cycle Sequencing Kit 15 (Perkin Elmer). Briefly, a cycle sequencing reaction contained ca. 50 ng of template DNA in miniprep quality, 4 pmol of primer, and 8 pl of the ABI Terminator Ready Reaction Mix (Perkin Elmer), in a total volume of 20 pl, and subjected to the following conditions: (96*C, 10 s; melting temperature of primer, 15 s; 60*C., 4 min) 26 times on a PTC 200 thermal cycler (MJ 20 Research). After removal of residual primers and dye by ethanol precipitation, the DNA was resuspended in 25 pl of ABI Template Suppression Reagent (Perkin Elmer) and denatured at 95*C for 5 min before loading the ABI Genetic Analyzer. The obtained data files were examined using ABI PRISM T M Sequencing Analysis Version3.0 Software (1996, Perkin Elmer). Gene Jockey 25 II (1996, Biosoft, Cambridge, UK) software was used for sequence comparison, translation, and alignments. Mutagenesis Procedure Substrate Preparation by PCR The cre gene for the following DNase I shuffling reaction was amplified 30 by PCR using 5' forward primer BSB436 (5' AAATAATCTAGACTGAGTGTGAAATGTCC 3') and 3' reverse primer BSB376 (5' ATATATAAGCTTATCATTTACGCGTTAATGG 3'), 57 WO 00/60091 PCT/US0O/09154 introducing an Xba I and Hind III cloning site, respectively (underlined). Mutagenic and non-mutagenic PCR's were carried out: (94'C, 30 s; 52'C, 30 s; 72'C, 90 s) 45 or 30 times, respectively. The 5' primer was designed to include the endogenous Shine-Dalgarno (SD) of cre, whereas its three endogenous 5 promoters were excluded (position -17, 6; positions refer to the adenine of the start codon of the cre coding sequence as position 1). Thus, after introducing the resulting cre genes into pBAD33 (see below), expression was exclusively under control of the pBAD promoter without interference or background expression due to endogenous promoters. Including the SD sequence of cre 10 was necessary, since pBAD33 does not contain this sequence 5' of its multiple cloning-site (MCS). The 3' reverse primer was designed to be homologous to the 3' untranslated region (UTR) of cre (position 1057, 1032). Mutagenic events were therefore permitted in 1020 bp of the entire 1026 bp cre coding sequence, excluding the first two codons. For the first round of the directed 15 evolution procedure, the wt cre expression plasmid pBS 185 (Sauer and Henderson, The New Biologist, 2:441-449 (1990)) served as template. In following cycles, the pool of mutated cre genes from the previous round was used. In all experiments, both, mutagenic and non-mutagenic PCR's were carried out in parallel using the appropriate template. 20 Homologous Recombination in Vitro DNase 1 digest Approximately 5 ptg of the cre PCR product (ca. 1.1 kb) wee digested with 0.03 U of DNase I, type IV (Sigma-Aldrich) in 20 pl total volume of 50 mM Tris-HCI pH 7.5 plus 1 mM MgCl 2 for 2 to 3 minutes at room temperature. 25 After digestion, samples immediately were loaded on a 2% minigel to separate the generated fragments (Figure 5). Fragments of 25 bp to 300 bp were purified from the gel by DE81 (Whatman, Maidstone, GB) extraction and ethanol precipitation (Sambrook et al., Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press (Second Edition) (1989)), before suspending in 5 pl of 30 TE pH 8.0. 58 WO 00/60091 PCT/US0O/09154 Self-Printing PCR A 60 cycle non-mutagenic PCR (as described above) was carried out without added primers, allowing the fragments to prime themselves and thereby to undergo shuffling while reassembling. Conditions for PCR were: 94'C, 90 5 s; (94'C, 30 s; 45'C, 30 s; 72'C, 90 s) 60 times; 72"C, 10 min. Reassembling of cre Since the self-priming step never yielded a single size product but rather a range of fragments between 300 bp to 2000 bp (Figure 5), the self-priming PCR mixture was diluted 1/40 in a non-mutagenic PCR mix with primers 10 BSB376 and BSB436 (see above), and subjected to an additional 20 cycles (94"C, 30 s; 52'C, 30 s; 72'C, 90 s). This additional step lead to one product of 1.1 kb size (Figure 5). cre Expression After digesting the linkers of the generated mutant cre pool with Xba I 15 and Hind III, the fragments were ligated into the identical sites of the MCS of the cre expression vector pBAD33 (Figure 6). Two features favored the choice of pBAD33 as the vector to express the mutant cre pool for the selection procedure (see below): First, its pACYC184 derived origin of replication is compatible with the ColE 1 derived ones of the plasmids used in the selection 20 procedure. Second, pBAD33 contains the promoter of the arabiniose operon (pBAD), as well as expresses the regulatory protein AraC. It is therefore possible to regulate the expression of a gene cloned into the MCS and under PBAD control, from moderately high levels to nearly complete regression, by simply changing 0.2% L-(+)-arabinose in the medium to 0.2% D-glucose 25 (Guzman et al., J. Bacteriol., 177:4121-4130 (1995) and Miyada et al., Proc. Nati. Acad. Sci. USA, 81:4120-4124 (1984). As indicated above, the primers for cre PCR were designed in order to include the endogenous SD sequence but to exclude the three cre promoters. cre expression therefore will be under complete control of the pBAD promoter. This is important for the selection 30 procedure (see below) that was intended for few Cre molecules acting on different lox sites. High concentrations or long term background expression of 59 WO 00/60091 PCT/USOO/09154 cre could eventually defeat the selection since wt Cre also catalyzes at very low frequencies recombination events between altered lox sites. Plasmids and E. coli Strains Used for Selection and Screen Mutant lox sites 5 Figure 1 compares the original loxP site to the two mutant sites, loxK1 and loxK2, used during the described experiments. The lox sites with 5' Sal I and Xho I compatible, and 3' Xba I and Nhe I compatible ends were received as single stranded oligonucleotides from Midland and annealed by heating the appropriate ones together at 70'C, followed by a gradual cool down. 10 Plasmids for Selection and Screening Plasmid pBS561 was constructed using three fragments: (i) the 5' modified neo gene derived from pBS398 (Sauer et al., Methods, 4:143-149 (1992), (ii) the RSVneo (Gorman et al., Science, 221:551-553 (1983) backbone without the neo gene, and (iii) the oligonucleotide-derived MCS (Figure 7). 15 The EGFP gene derived from pEGFP-N 1 (Clontech, Palo Alto, CA) was then inserted into the MCS along with 5' and 3' lox sites orientated in the same direction to produce plasmids pBS568 (loxK12) and pBS569 (loxK2 2 ). Figure 8 summarizes the procedures used to construct plasmids pBS583 and pBS584. To restore the original neo reading frame without the 5' extension, the loxK 2 20 cassettes containing MIu I/Kpn I-fragments from pBS568 and pBS569 were ligated into the RSV neo backbone that contains Mlu I and Bgl II sites. The Bgl II - Kpn I junction was achieved by filling the 3' recessed end of Bgl II with Klenow (NEB) followed by a blunt-end ligation to the Kpn I end. This junction also was checked by sequencing and found to be correct. A transcriptional 25 terminator, rrnBTIT 2 , derived from pBAD33 by non-mutagenic PCR with primers BSB425 (5' ATAAGCGGCCGCTGAGCTTGGCTGTT TTGGCGG 3') and BSB426 (5' GCCGTCTCGAGAGAGTTTGTAG AAACGCAAAAAGGC 3'), was inserted into the loxK2 cassette 3' of the EGFP gene after digest of the its Not I and Xho I linkers (underlined). With this 30 construct, it could be predicted that a catalyzed recombination by K1* or K2* Cre mutants between the lox sites would result in the excision of EGFP and the transcriptional terminator, and thereby permit the transcription of the neo gene 60 WO 00/60091 PCT/US0O/09154 due to the RSV promoter, located 5'. Expression of neo would no longer be impaired. because it is placed under control of 5' promoter elements present in RSVneo (a KanR rendering plasmid in E. coli). A similar loxP 2 cassette selection plasmid also was designed (pBS613, 5 Figure 9), to be used as a control. Using this plasmid, the frequency of loxP recombination by Cre mutants could be determined in the same manner as used to evaluate loxKl or loxK2 recombination by pBS583 and pBS584. Finally, a completely different set of loxK 12/K2 2 cassette plasmids was created. These plasmids were no longer used for selecting mutants that 10 recognize loxKl or loxK2, but rather were used to screen for these mutants in conjunction with a different bacterial background (see below). Figure 10 summarizes the construction process for pBS601 (loxK12) and pBS602 (loxK2 2): In a first step, the neo resistance marker of pBS581 and pBS582 (intermediates in the construction of pBS583 and pBS584) was removed by 15 deleting the Pvu II fragment and thus restoring the possibility to use neo for a different selection procedure. Following this, the EGFP gene between the lox sites was replaced by the pUC19 (Yanish-Perron, et. al., 1985) derived lac promoter with its 3' MCS. This pUC19 fragment was obtained by non mutagenic PCR with primers BSB448 (5' GTCAAGCTAGCTAGCAGGTTT 20 CCCGACTGG 3') and BSB449 (5' ACATTGCGGCCGCAGATCTCCTCTA GAGTCGACCTG 3'). An Nhe I site 5', and a Bgl II and a Not I site 3' (underlined) were introduced thereby, which made it possible to replace the Nhe I - Not I EGFP fragment after linker digestion. The newly generated polylinker between the two lox sites permitted insertion of the Xba I - BanH I fragment of 25 pBS481 (that carries the abfA.st marker gene) into its Xba I and Bgl II sites. As shown recently, E. coli strains expressing a recombinant aX-L arabinofuranosidase gene from Streptomyces lividans (abfA.st), can be detected by eye on LB plates containing 5-bromo-3-indolyl-ca-L-arabinofuranoside (Z ara) Berlin and Sauer, Anal. Biochein., 243:171-175 (1996). This leads to the 30 formation of an indigo blue pigment, that is similar to the classical lacZ/5 bromo-4-chloro-3-indolyl-p-D-galactopyranoside (X-gal) marker system. It could therefore be expected that E. coli strains expressing mutant Cre 61 WO 00/60091 PCT/USO0/09154 recombinases which allow recombination of loxK 1 or loxK2 should lose the abfA.st gene and should form white colonies on Z-ara plates. All other clones, however, should be dark blue. Using abfA.st instead of the well established lacZ screen was necessary, because the E. coli strain (see below) used for this 5 construct expresses p-galactosidase endogenously. E. coli Strains for Selection and Screen The E. coli strain BS583, DH5 Alac (kD69 loxP 2 [lacZ LEU2]), was chosen as the bacterial background for the selection procedure for Kl or K2+ Cre mutants by plasmids pBS583 or pBS584. Due to the loxP 2 [lacZ] containing 10 k prophage, Cre activity on loxP can be evaluated simply by using X-gal plates. The selection strains BS1493 and BS1494 were made by introducing the selection plasmids pBS583 and pBS584 into BS583 (Table 2). The loxP2 plasmid pBS613, to be used as a control, needed to be in BS583 cells as well, becoming the strain BS1541 (Table 2). 15 For the screening plasmids pBS601 and pBS602, E coli strain NS2300 (Sternberg et al., J. Mol. Biol., 187:197-212 (1986)) was selected as host: K12 recA::Tn10 (Xi 434 loxP2[neo]). This strategy combines a kanamycin selection for Cre enzymes, that are no longer active on loxP (P~), with a screen for K1* or K2* enzymes. By transforming pBS601 and pBS602 into NS23 00, the P~ 20 selection strains BS 1523 and BS 1524 were formed (Table 2). Selection and Screen for CRE Mutants Selection for K]*/K2* and Screenfor P After ligation of the generated mutant cre pool into pBAD33 for 3 h, BS1493 or BS1494 electrocompetent cells were transformed with 2 pl of the 25 microdialyzed reaction mixture (VS membrane, MilliporeTM' Bedford, MA). To induce expression of the cre pool, the transformed cells were incubated at 37'C in induction medium for 2.5 h and/or 4 h under agitation (as mentioned before). Cultures were diluted 1/500 or 1/5000 and grown on LB plates with the following formulation: Ap, Cm, glucose, and X-gal for determining the 30 transformation efficiency, referred to as non-selection plates. Dilutions of 1/5 and occasionally 1/50 were grown on plates with addition of Kan, used to select for Kl or K2* mutants and called selection-plates. The formulation of the 62 WO 00/60091 PCT/US0O/09154 plates served the following purposes: (i) Ap and Cm were added to assure that all clones contained both, selection and expression plasmid, (ii) X-gal to distinguish between P* and P~ clones (Table 2). After overnight incubation at 37 0 C, blue and white colonies were counted and pools prepared for the next 5 round of DNA shuffling. Alternatively, certain mutants were chosen for further analysis (see below). Selectionfor P and Screenfor KJ' or K2' After 2.5 h and/or 4 h of expression of the mutant cre pool in the transformed BS1523 and BS1524 cells, usually dilutions of 10-3 or 104 were 10 grown on LB agar plates supplemented with the same reagents as listed above, except, that Z-ara replaced X-gal to allow the K1/K2+ screen. Non-selection plates were used for determining the transformation efficiency, and Kan containing plates for the P~ selection (Table 2). Mutant Analysis 15 wt cre Expression Plasmid With fewer cycles (15) of non-mutagenic PCR on the cre expression plasmid pBS1 85 and after linker digestion, the cre pool obtained was cloned into pBAD33 and transformed into BS583 cells. After 2 h of cre expression, the transformed cells were grown on X-gal plates. After overnight incubation at 20 37 0 C, two white colonies (indicating loxP recombination) were picked for plasmid preparation and complete sequencing. No point mutation was found in either one, so that each could be used as a control plasmid for wt Cre expression. One of the two was selected for further use and named pBS606. Functional Testing 25 In order to determine the frequency of lox recombination of isolated mutant Cre enzymes by the described selection procedure, it is necessary to separate the cre candidate expression plasmid (pBAD33) from the selection plasmid of the chosen KanR candidate. Then, the cleaned expression vector can be used to retransform the appropriate selection and screening strain BS 1494, as 30 well as BS1493 and BS1 541 to determine the candidate's capacity for loxK2, loxK1 and loxP recombination under identical conditions. By comparing the resulting frequencies of KanR of different Cre mutants and wt Cre, all treated 63 WO 00/60091 PCT/US00/09154 identically, one can determine quantitatively how well each chosen mutant candidate really performs on the altered lox sites. Therefore, overnight cultures of candidates were grown in liquid LB supplemented with Cm and Kan for plasmid minipreps yielding a mixture of 5 both the mutant Cre expressing plasmid pBAD33 and the newly KanR selection plasmid. In order to eliminate the latter, minipreps were digested with the restriction enzyme Aat II which only cuts the selection plasmid. After transformation of BS583 cells with this digestion mixture and approximately 2 h cre expression, different dilutions were grown on LB agar plates 10 supplemented with Cm and X-gal to select for pBAD33. Plates with Ap plus Cm were used to determine the background of contamination with uncut selection plasmid. The next day, clones obtained by the Cm selection were tested for Aps and Kans to confirm the elimination of the selection plasmid. A final overnight culture, followed by a miniprep procedure, yielded the unique 15 plasmid for functional testing, as described above. Sequencing To obtain the DNA sequence of candidate cre genes in pBAD33, eight primers (BSB454 to BSB461, Table 3), four for each strand, were designed so that the entire gene could be sequenced in both directions. 20 Site-Directed Mutagenesis After identifying one essential mutation for the decrease in substrate specificity, the Stratagene QuickChange TM Site-Directed Mutagenesis Kit (Stratagene Cloning Systems, La Jolla, CA) was used to create cre mutants with mutations at the determined location, only. Using three different mutant primer 25 sets (BSB 465 to BSB 470, Table 4), all steps were carried out as detailed in the manufacturer's instruction manual, except that electrocompetent BS 1494 cells were used for transformation and mutant selection, replacing the provided XL 1 blue cells. The made mutant candidates were subjected to functional testing and sequencing as detailed before. 30 In a different experiment, the DNA shuffling mutagenesis procedure was repeated on wt cre by adding one 5' phosphorylated strand of each set of mutant oligonucleotides (BSB465, BSB467 and BSB469) to the pool of small 64 WO 00/60091 PCT/US00/09154 fragments prior to reassembly. This allowed to incorporate them into the resulting cre pool. The desired mutations should consequently be introduced at much higher frequency than without the addition of oligonucleotides. RESULTS 5 Establishment of the Selection Procedure Selection Plasmids (pBS583 and pBS584) To test the generated plasmids pBS583 and pB584, a recombination event between their lox sites was mimicked by digesting pBS566 and pBS567 (intermediates of the pBS583/584 construction, containing only the 3' lox site) 10 with Sal I and Xho I, followed by religation. This lead to excision of the EGFP gene and terminators. After deletion of EGFP and rmBTiT 2 , the KanR phenotype was observed as anticipated. In addition, the frequency of spontaneously occurring KanR clones carrying the original plasmids was approximately 10-7. This background is inconsequential, since the 15 transformation efficiency of BS583 cells was determined as 10 7 per pg of pBAD33. The equivalent loxP control plasmid pBS613 was tested directly with the wt cre expression plasmid pBS606. After 2.5 h of cre expression, 94% of all clones were determined KanR and about 6% showed blue color. Without cre 20 expression, no KanR and no white colonies were observed. This confirms that the control cell line BS1541 (Table 2) permits the combined P* selection and P screen. Screening Plasmids pBS601 and pBS602 Only pBS602 was tested before use by expression of wt Cre and a 25 K2*/P* Cre mutant (see below). On non-selection medium, wt Cre expression resulted in more than 95%, whereas expression of the mutant Cre resulted in less than 3% of blue colonies. This indicates that excision of the loxK2 flanked abfA.st marker by K2+ Cre is possible. On selection medium, very few colonies could be found, since both types of cre have shown activity on loxP before (see 30 below). 65 WO 00/60091 PCT/USOO/09154 Mutagenic vs. Non-Mutagenic PCR The frequency of P~ Cre mutants obtained after non-mutagenic and after error-prone PCR was determined by the following experiment: After one mutagenic or one non-mutagenic PCR on the wt Cre expression plasmid 5 pBS 185, the resulting cre pools were inserted into the expression vector pBAD33 and transformed into BS583 cells. After 2.5 h of arabinose-mediated induction or glucose-mediated repression (by SOC medium) or cre expression, dilutions were transferred to LB plates with Ap, Cm, glucose and X-gal. The results are presented in Table 5: Under glucose repression, exclusively blue 10 colonies could be identified (first line in Table 5), indicating that cre expression is insufficient for loxP recombination and excision of lacZ of BS583. Induction with L-(+)-arabinose, however, lead to the formation of white colonies at the presented frequencies (second line). indicating that (i) the described control of cre expression by pBAD33 is functioning, and (ii) the mutagenic PCR 15 conditions cause three times more impaired Cre enzymes for loxP recombination than the non-mutagenic conditions (60%) blue colonies vs. 30%). It is worth mentioning that ligation reactions lacking cre insertion resulted in 50 to 100 times less blue colonies than obtained with the ligations with cre insertion. This phenotypically blue background of empty pBAD33 was 20 subtracted before calculating the presented data. Leung et al., Technique, 1:11-15 (1989) reported that the frequency of point mutations created by error-prone PCR is about 0.5%. If this is true, in average five point mutations should occur in each 1 kb cre coding sequence subjected to an error-prone PCR. By extrapolating this data to the three times 25 less P enzymes after a non-mutagenic PCR, one can conclude that the frequency for point mutations should also be reduced by a factor of three to 0.18%. Experiments made by Zhou et. al. (1991) showed 11% of a 633 bp marker gene phenotypically impaired after non-mutagenic PCR. About 37% of all genes in the pool, however, carried at least one point mutation. Even though 30 the conditions for the non-mutagenic PCR were similar, the observed discrepancy between 11% and 20% of phenotypical mutants may be due to a variety of reasons, among which: (i) the size difference between the two genes 66 WO 00/60091 PCT/USOO/09154 (633 bp vs. 1020 bp), (ii) different elongation times during PCR, and (iii) different sensibility of the two proteins for disabling point mutations. Testing wt Cre on loxK1 2 vs. loxK2 2 Substrates The level of loxK1 and loxK2 recombination due to wt cre expression 5 was determined using the wt Cre control plasmid pBS606. After transformation of the cell lines BS1493 and BS1494 with pBS606 and 2.5 h and 4 h of cre expression, cells were grown on selection and non-selection plates (as described previously). The recombination frequency between the altered lox2 sites was considered equal to the observed frequency of KanR phenotype: for loxKl, it 10 was about 10- after 2.5 h and 2 X 10- after 4 h of wt cre expression, for loxK2, it changed from about 2 X 10-' to 2 X I0 3 . All colonies found were white, indicating effective loxP recombination by wt Cre within the allowed expression time. This result shows that long term expression of the wt enzyme permits a slight increase in recombination between the altered lox sites. The use of 15 pBAD33 to avoid background lox recombination by suppressing cre expression is therefore justified. Because loxK2 was eventually 100 fold better recognized by wt Cre than loxKl (but still at low frequency), first the creation of novel Cre recombinases with loxK2 specificity was attempted. Mutagenesis on loxK2 20 First Rounds of Directed Evolution The result of the first four rounds of the described mutagenesis procedure on loxK2 with selection plasmid pBS584 are presented in Table 6. The following symbols are used to describe the status of the DNA shuffling procedure for cre: "o" indicates a non-mutagenic PCR, "in" a mutagenic PCR, 25 and "x" stands for the in vitro reassortment event. For example, mxoxox cre represents a cre pool subjected to three rounds of the directed evolution process, with a mutagenic PCR followed by in vitro shuffling in the first, and non mutagenic PCRs and shuffling as mutagenic and recombinogenic events in the following two rounds. The phenotypically blue background due to empty 30 pBAD33 was subtracted from all results by control ligations without cre insertion. In every round, error-prone and non-mutagenic PCR served as the necessary mutagenic event on the template pool of the previous round. After in 67 WO 00/60091 PCT/US00/09154 vitro reassortment and selection, always the larger KanR population of the two parallel experiments was chosen as template for the next round (as indicated in the last column of Table 6). Only in the first round the error-prone PCR could lead to more candidates, whereas in all following rounds the reduced mutagenic 5 frequency of the non-mutagenic PCR turned out to be more beneficial. The density of point mutations resulting out of two mutagenic PCR's was obviously too high to allow efficient elimination of deleterious mutations from advantageous ones during the in vitro shuffling step. This is confirmed by the high frequencies of blue colonies found within any pool in any round subjected 10 to mutagenic PCR's twice. Error-prone PCR in the context of the applied selection therefore appears to be useful in the first round only, where its three times higher mutagenic frequency increases the amount of beneficial mutations compared to a non-mutagenic PCR. With increasing cycle numbers non mutagenic PCR's should be preferred to avoid high densities of deleterious 15 mutations. The established directed molecular evolution process allows effective evolution of cre. Column five of Table 6 shows that with every round the number of KanR colonies is increasing, while the time for cre expression could be lowered from 4 h to 2.5 h (column three). After only three rounds, Cre 20 mutants capable for loxK2 recombination at decreased concentrations due to the reduced expression time were found. However, it was not possible to isolate any blue colony on the selection plates. All identified K2+ mutants therefore are also P+. As mentioned above, high densities of deleterious mutations in the cre pool subjected to error-prone PCR's twice could explain why no blue colonies 25 were seen on selection plates, even with over 90% of P~ candidates on non selection plates. Evaluation of Six K2*/P* cre Mutants Functional Test As indicated in Table 6, 36 white KanR colonies could be isolated from 30 the mxoxox cre pool after only 2.5 h of cre expression with an 1/5 dilution grown on selection plates. This result indicated that competent Cre mutant capable of loxK2 recombination were produced. Six were selected for further 68 WO 00/60091 PCT/USOO/09154 analysis. After elimination of the selection plasmid from the minipreps (as described in Materials and Methods), all six of them, as well as wt Cre (pBS606) were subjected to the described functional test on loxK2, but also on loxK1 and loxP recombination with plasmids pBS584, pBS583 and pBS613. 5 The results are presented in Table 7. Briefly, by selecting for loxK2 recombination, all mutants except mxoxox 4, showed significantly increased percentages of KanR (between 3% and nearly 70%), compared to wt Cre (0.002%), as indicated in column three of Table 7. This indicates a 10 3 to over 104 fold increase in activity on loxK2. 10 On loxP, all (including mxoxox 4) showed recombination frequencies between 80% and 100% after 2.5 h of cre expression (column 4). This was expected from the results obtained with the X-gal screen for loxP recombination during the selection procedure. 2.5 h of induction for wt cre expression is therefore sufficient for almost complete loxP recombination, justifying 2.5 h of 15 expression of mutant cre pools for selecting competent K2+ Cre mutants. The observed slight decrease in loxP recombination with the mutants mxoxox 3 to 6 either derived from usual variations during experiments, or may indicate a slightly reduced loxP activity. With BS1541 blue colonies on both selection and non-selection plates were found in approximately the same frequency (2% 20 to 20%) as kanamycin sensitivity (Kans). This indicates competition between the loxP 2 [lacZ] site on the genome and the loxP 2 [EGFP-rrnBTIT 2 ] sites on pBS613. Since the same Cre mutants never resulted in blue colonies during selection for loxK2 recombination in cell line BS1494, it is possible to conclude that loxP is still preferred over loxK2. This argument is supported by higher 25 frequencies of loxP recombination, close to 100%, compared to the loxK2 recombination frequencies of 3% to 70%. All frequencies found for loxK1 recombination, determined by using cell line BS1493, lie below 0.01% after 2.5 h, as well as after 4 h of induction of cre expression (column five). This indicates that no analyzed mutant developed an 30 increased activity on lox KI compared to wt Cre. 69 WO 00/60091 PCT/USOO/09154 To summarize, five out of the six analyzed mutants showed a significant decrease in specificity, resulting in the possibility for loxP and loxK2 recombination. Sequencing 5 The six described mutants have been completely sequenced in both directions to determine the mutations which lead to the observed decrease in specificity. The resulting aligned cre coding and aa sequences of all mutants and wt Cre are represented in Tables 10 and 11. Each mutant showed between 3 and 8 point mutations, altogether 31, as listed in column two of Table 8. The 10 overall mutagenic frequency can therefore be calculated at 0.5% (31 mutations in 6 clones of 1020 bp), which is similar to only one round of error-prone PCR (Leung et al., Technique, 1:11-15 (1989)). The reason for the low frequency of point mutations after three rounds of the mutagenesis procedure (i.e. nine PCR's) is the applied selection after each round, cre mutants with low density of 15 point mutations seem to be favored by the stringent kanamycin selection. 26 of the 31 identified point mutations resulted in aa changes compared to the wt sequence, as indicated in column three. No deletions or frame-shift mutations, as well as no codons affected by more than one point mutation at the same time could be identified. All possible transition events could be observed, 20 but only half of all possible transversion events (shown in Table 12). Adenine to guanine and vice versa transition events represented almost 30% of all identified point mutations. All other events occurred less frequent (10%, 7%, or never). The represented statistic may however be biased, either since only six mutants were analyzed, or due to the directed molecular evolution technique 25 itself: The types of point mutations observed less often, may more frequently be deleterious in cre and were consequently removed from the pools. With more mutants to be sequenced, this question could be addressed further. Only one point mutation, a transition event from adenine to guanine at position 785 in the cre coding sequence is common for all 5 mutants with 30 remarkably increased loxK2 activity. This mutation leads to a replacement of a glutamate residue at position 262 in the J helix of wt Cre by a glycine (indicated in column five of Table 8). This glutamate is believed to contact the loxP 70 WO 00/60091 PCT/USOO/09154 sequence at positions 11 or 12 with its acidic side chain (Guo et al., Nature, 389:40-46 (1997)). Another point mutation, resulting in a conservative threonine to serine exchange at position 316 was identified in three mutants. Five point mutations were found independently in two of the six mutants, 5 among which two types of silent mutations. Finally, eleven point mutations occurred only once (indicated in column four). Therefore, the critical mutation for loxK2 activity appears to be E262G. Some of the additional mutations could be responsible for the observed ten fold difference in loxK2 recombination frequency among the five E262G carrying mutants. 10 Site-Directed Mutagenesis to Verify Results Site-Directed Mutagenesis Procedures To determine whether the E262G mutation alone is responsible for the increase of loxK2 activity of the Cre mutants by at least a factor of 103, two different experiments were made: 15 First, the described directed evolution procedure was repeated in three different sets on wt cre, by adding three 5' phosphorylated mutant oligonucleotides prior to reassembly: Incorporation of the first oligonucleotide (BSB465) into cre should lead to the E262G mutation, incorporation of the second one (BSB467) to a E262A mutation, and the equimolar mixture of 20 random oligonucleotides (represented by BSB469) to 203 (= 8000) possible aa combinations at positions 261 to 263 of Cre. According to Stemmer, W. P. C., Proc. Natl. Acad Sci. USA, 91:10747-10751 (1994), these oligonucleotides should be incorporated during reassembly and cause the desired mutations at a frequency of about 8%. After insertion of the resulting cre pools into pBAD33 25 and 2.5 h of expression, 0.8% of white KanR colonies were found with BSB465. Therefore, the frequency of the KanR phenotype due to loxK2 recombination is ten times lower than the expected frequency for the E262G mutation to occur. This indicates that the E262G mutation increases the specificity for loxK2 from 0.002% of wt Cre (see above) to approximately 10% of recombination during 30 the standard expression experiment (2.5 h of cre expression prior to selection). By using the second oligonucleotide (BSB467) only 0.09% of the total white colonies showed KanR. Thus, E262A still favors lox K2 recombination but by 71 WO 00/60091 PCT/US0O/09154 almost a factor ten less efficiently than the E262G mutation. With the random oligonucleotide mixture (BSB469). the frequency of loxK2 recombination shrunk to 0.02%. Compared to a control experiment with no oligonucleotides added, resulting in 0.003% KanR (consistent with the results obtained below), 5 still some of the possible 8000 aa combinations at positions 261 to 263 of Cre are expected to favor loxK2 recombination. Since the frequency of blue colonies on non-selection plates was about 30%, a clear indication for additional mutations, it cannot be completely excluded that some of the occurred KanR mutants of this experiment have been carrying some additional beneficial 10 mutations. Nevertheless, this experiment indicated that the E262G mutation is probably the basis for significant increase in loxK2 activity. For better defined mutations, a second experiment was done: The same oligonucleotides, now in sets for both DNA strands, were used with the Stratagene QuckChangeTM Site-Directed Mutagenesis Kit as described in 15 Materials and Methods. After transformation of the loxK2 selection strain BS1494, the percentage of white KanR colonies in the three different experiments could be determined as 6.2% for E262G, 0.8% for E262A and 0.6% for the 8000 different ten (10) days combinations as positions 261 to 263. Due to the QuickChangeTM procedure, which eliminates parental DNA, one 20 could expect that the desired mutations occurred in at least 50% of all clones of the three different pools. This assumption indicates that the E262G mutation alone results in approximately 12% or less recombination frequency on loxK2 (6% out of 50% or more carrying the mutation) and the E262A exchange in 1.6% or less (under standard conditions). This calculation is not valid for the 25 third set of oligonucleotides, because no defined mutation is introduced but rather a mixture of 8000 different ones. Blue colonies were not found during this experiment, confirming that the frequency of additional mutations altering Cre activity on loxP was very low. The results of both experiments indicate consistently, that the E262G 30 mutation alone is sufficient to increase loxK2 recognition by approximately a factor of 103 compared to wt Cre. The ten times higher frequency observed with three of the six analyzed mutants after three rounds of the mutagenesis 72 WO 00/60091 PCT/USOO/09154 procedure can only be explained with additional beneficial mutations. The E262A mutation increases the frequency of loxK2 recombination also, but approximately by a factor of ten less effectively than E262G. Sequencing and Elimination of Possible Additional Mutations 5 Two white KanR colonies of each pool derived from the site-directed mutagenesis procedure were selected and the entire cre gene was sequenced. Both E262G and E262A candidates showed the desired sequence with no additional mutations. The random candidates surprised: One of them did not reveal any point mutation, representing an artifact which managed to survive the 10 selection, whereas the other one showed nucleotide alterations from position 783 to 786. The wt sequence ( 781 CTG GAA 7 s 6 ) was found to be changed to (783 CTT TGG 7 86 ), resulting in one silent mutation, conserving L261, and a E262W exchange. To exclude possible mutations in pBAD33 due to the PCR involved site-directed mutagenesis procedure, the three defined cre mutants (E262G, 15 E262A, and E262W) were excised by Hind III and Xba I and reinserted into the MCA of fresh pBAD33. Functional Testing The three defined Cre mutants for the amino acid position 262 were subjected to a functional test for loxK2, loxK1 and loxP recombination activity, 20 as mentioned before. The results are summarized in Table 9. First, the results described previously for the E262G and E262A mutations were confirmed: As indicated in column three, the loxK2 recombination frequency increased 10 3 fold with the E262G mutant compared to the wt enzyme, whereas the E262A mutant shows only an increase of 200 fold. Surprisingly, the E262W mutant 25 also achieved a similar activity on loxK2 as seen with the E262G mutant. The test for loxP recombination frequency with cell line B S 1541 (column four) showed that the ability for loxP recognition is at best slightly impaired by the three different mutations (as already seen with the analyzed mxoxox mutants). Again, blue colonies could only be found with this cell line, indicating that loxP 30 is preferred over loxK2 (as described before). As expected, none of the three mutants performed significantly better on loxK1 than wt Cre (column five). To conclude, this final experiment provides the necessary evidence that the E262G 73 WO 00/60091 PCT/USO0/09154 mutation presents the basis for the observed decrease in specificity of Cre. However, additional mutations seem to be helpful to increase the newly obtained activity further. In addition, glycine at position 262 is not the only possible residue to permit a remarkable increase in loxK2 activity. 5 DISCUSSION K2*/P* Cre Mutants The chosen random mutagenesis procedure linked to the described selection in E. coli allowed the identification of Cre mutants, characterized by a wider substrate recognition. The evolved enzymes showed wt-like activity on 10 loxP sites and in addition had almost the same activity on altered lox2 sites, referred to as loxK2. By contrast, the wt enzyme showed only marginal activity on loxK2 2 substrates. lox Sites loxP and loxK2 differ at several locations as illustrated in Figure 1: 15 First, the three outermost bp of the inverted repeats are altered, facilitating the construction of the various plasmids used for selection. Second, the entire non canonical 8 bp spacer is completely exchanged, and third, two transversion events (thymine to adenine) are introduced at positions 11 and 12 of the lox site. Only the two transversion events are considered important for inhibiting wt Cre 20 from recognizing the site. Other investigations have shown before, that the two alterations of loxP mentioned first are without inhibiting effect on wt Cre (B. Sauer, unpublished results). The design of loxK2 was based in part on so-called cryptic lox sites, that were identified previously in the yeast genome (Sauer, B., J. Mol. Biol., 233:911-928 (1992)). Another consideration was the choice of a 25 good starting sequence for the described mutagenesis procedure. Starting with sites that contained several and/or widespread alterations of loxP was avoided, because the greater the number of alterations in the substrate, the more the enzyme would have to be altered. Consequently, to most effectively use the mutagenesis procedure, the two sites presented in Figure 1, loxK1 and loxK2, 30 were designed to have only two critical alterations. In initial experiments, wt Cre was found to recombine loxK2 2 substrate pBS584 slightly better than loxK12 pBS583. This difference may depend on the fact that in loxK2 the two 74 WO 00/60091 PCT/US00/09154 alternations are located next to each other. while in loxKl they are separated by 3 bp (positions 14 and 10). Thus, loxK1 could interfere with wt Cre binding at two distinct DNA-protein interaction sites as compared to loxK2, where only one location of incompatibility is available. For this reason, loxK2 was chosen 5 for the initial set of experiments, described in this work. Cre Mutants after Three Rounds of Directed Evolution Three iterations of the in vitro evolution procedure were necessary to identify 36 candidates, expressing Cre mutants that could process loxK2 (based on the applied selection procedure in E. coli). Tests showed that five out of six 10 selected ones had 103 and 104 fold increased activity on loxK2 when compared to wt Cre. On loxP and loxKl, however, there was almost no difference between wt and the mutant enzymes. The mutants therefore had developed an increased tolerance from transversions at positions 11 and 12 (loxK2) of the lox sequence, but not for other positions like 10 and 14 (loxK 1). To indicate this 15 phenotype they were referred to as K2*/P. The E262G Mutation Sequence analysis identified that the five mutants with remarkably enhanced loxK2 activity (103 to 104 fold compared to wt Cre) had in common only one point mutation, leading to the aa change E262G. Site-directed 20 mutagenesis experiments confirmed, that the E262G mutation is sufficient to increase loxK2 activity by a factor of 103 compared to wt Cre. Based on the recently described crystal structure (Guo et al., Nature, 389:40-46 (1997)) glutamate at position 262, located in the J helix of the enzyme, may be a DNA contacting residue and permit the formation of a 25 hydrogen bond between the carboxyl group of its side chain and an amino group of one of the two adenines at positions 11' or 12' in the loxP sequence (Figure 1). However, changing these two bases to thymines in the loxK2 sequence should lead to an electrostatic repulsion between their oxygens and the acidic side chain of glutamate. This could explain the observation that wt Cre is 30 unable to catalyze a recombination between two loxK2 sites. Exchanging glutamate for a glycine, that does not have a side chain, should remove the electrostatic repulsion and thereby permit loxK2 binding. On the other hand, 75 WO 00/60091 PCT/US0O/09154 this alteration could affect loxP binding because an electropositive DNA-protein interaction may be lost. Results from the mentioned experiments support this proposition: The E262G mutation alone lead to an increase in loxK2 recombination from 0.002% to about 2% and to a slight decrease in loxP 5 recombination from 94% to about 80%. The 50 times lower frequency of loxK2 compared to loxP recombination may depend on the second thymine or more likely on the complementary adenine (position 11 in loxK2) that could contribute to sterical repulsion between the J helix and the loxK2 site. To prove this hypothesis, the role of arginine at position 258, located one helix turn away 10 from the glutamate, should be further investigated by site-directed mutagenesis. As proposed by Guo et al., Nature, 389:40-46 (1997), R258 is a DNA contacting residue that forms hydrogen bonds with the guanine at position 10' of loxP and may also interact with the bp at position 11. There is yet no confirmatory experimental evidence for this proposal. 15 Results from three initially isolated mutants (mxoxox 1 to 3) indicated about 50% of recombination frequency on loxK2. This is about ten fold higher than that obtained with the E262G mutation alone. It is therefore likely that some of the additional point mutations identified in these three mutants account for this increase in activity. Table 8 lists all point mutations that were found. If 20 silent and conservative mutations are considered not to influence specificity, only a limited number of candidates to account for the phenotype remains. Among these, S254G and Q255R of mxoxox 2 and 3, because of their location close to the amino-terminus of the J helix, could be expected to influence DNA contacts with positions 11 or 12 of the lox site. The other mutations are 25 scattered in the N- and C-terminal domain of Cre. All, except R1O1Q of mxoxox 5, affect aa that are not located within the proximity of DNA contacting areas. Some appear independently in two mutants, e.g. D29A or D189N, that could influence protein folding or the interactions among the four Cre enzymes necessary for recombination. Such alterations could influence for example the 30 orientation of the J helix and thereby reduce remaining interference between the loxK2 site and the enzyme. Alternatively, some mutations, also silent ones, could influence protein expression, leading to a faster accumulation of enzymes 76 WO 00/60091 PCT/US0O/09154 and consequently to higher recombination frequency. This possibility should however also influence loxP recombination. In fact, after 2.5 h of cre expression, the mutants mxoxox 1 and 2 showed a slightly higher frequency of loxP recombination compared to wt cre (98% vs. 94%). This difference, on the 5 other hand, may be attributed to the variations that occur normally within experiments. To address this question further, shorter cre expression times on loxP would be required. When the coordinates of the crystal structure (Guo et al., Nature, 389:40-46 (1997)) are available (protein data bank, Brookhaven National 10 Laboratory), it will be possible to confirm many of the tenets discussed. Finally, the increase in loxK2 tolerance between the E262G mutation alone and the isolated mutants carrying additional point mutations justifies the use of the DNA shuffling procedure linked to selection: Not only has it permitted the elimination of deleterious mutations from the sequence pool, but it 15 helped to accumulate various more or less beneficial aa alterations as well. The E262A and E262W Mutations The mentioned site-directed mutagenesis procedure was used to generate two more defined Cre mutants, E262A and E262W. Compared to the E262G mutati8on, E262A permitted loxK2 recombination ten fold less effectively. The 20 E262W mutation however resulted in similar activity on loxK2. The aliphatic side chain (a methyl group) of A262 could be the reason for slight sterical interference. This would explain the observed reduced frequency of loxK2 activity with E262A. loxP recognition, however, could not be found to be affected compared to E262G. The lowered loxK2 activity 25 explains why no E262A mutation was identified in the small pool of six analyzed mutants: With a ten fold decrease in activity, one would expect to encounter the corresponding mutation ten times less often during selections as well. In contrast to the small side chain of alanine, the large aromatic side 30 chain of W262, was expected to inhibit recombination between any lox sites due to sterical interference. Surprisingly, this seems not to be true. A possible explanation for the observed activity on loxP and loxK2 could be that the 77 WO 00/60091 PCT/USOO/09154 aromatic and planar structure of the tryptophan side chain, fits better into the J helix - lox interface than does a methyl group. Different influences of A262 and W262 on folding of the J helix could also contribute to the observed phenotypes. The reason why the E262W mutation was not identified among the 5 six analyzed mutants is the genetic code. Whereas for the E262G and E262A mutations only one bp in the glutamate encoding GAA codon needs to be mutated to GGA or GCA, for E262W the whole codon must be exchanged to TGG. This is unlikely to occur during the applied random mutagenesis procedure with a mutation frequency of 0.5%. Other amino acid changes due to 10 two or three mutations of the glutamate encoding codon therefore cannot be considered to have occurred during the random mutagenesis procedure. When the coordinates of the crystal structure are public, it will be interesting to confirm and further investigate the discussed hypothesis. ABBREVIATIONS 15 5-BI-ara a.k.a Z-ara 5-bromo-3-indoyl-c-L-arabinofuranoside aa amino acid Ap(ApR , Aps) ampicillin bp base pairs Cm(CmR, CmS) chloramphenicol (resistant, sensitive) 20 cre - Cre cyclization recombination (gene - Protein) Da Dalton DMSO dimethylsulfoxide ds double stranded GF gel filtration 25 Kl+ or K2* Cre mutant capable for loxK1 or loxK2 recombination R S Kan (Kan , Kan kanamycin (resistant, sensitive) lox locus of crossover MCS multiple cloning site 30 mRNA messenger RNA neo - Neo neophosphotransferase (gene - Protein) P or P* Cre mutant defective or capable for loxP recombination 78 WO 00/60091 PCT/US00/09154 PCR Polymerase Chain Reaction SD Shine-Dalgarno sequence ss single stranded TK thymidine kinase 5 wt wildtype X-gal 5-bromo-4-chloro-3-indolyl- -D-galactopyranoside Example 2: Analysis of Variant Cre Recombinases This example describes analysis the activity of several specific variant 10 Cre recombinases. Vectors pBS606, 614, 626, 627, 628 and 650: pBAD33 with wt, E262G, E262G/D29A, E262G/D189N, E262G/T316S, and R3M3 cre insertion used for expression of the corresponding Cre proteins in DH5x for in vivo testing. 15 pBS632 to pBS641: pUC 19 based plasmids for in vivo tests of different Cre mutants on a variety of different lox sites and combination of lox sites, all bearing the FAS1 spacer (Sauer B., Nucleic Acids Res., 24:4608-4613 (1996)). Recombination between two lox sites leads to excision of a neo cassette to give kanamycin sensitive E coli. The same plasmids were also used for in vitro 20 recombination experiments. pRH200: wt Cre expression plasmid (a generous gift from Ron Hoess, DuPont, Wilmington, DE) used to overexpress wt Cre in BL21(DE3) (Novagen, Madison, WI) strain. pBS654 to pBS658: wt cre of pRH200 was replaced with different cre 25 mutants (E262G, E262G/D29A, E262G/D189N, E262G/T316S, and R3M3) using Age I and A4lu I restriction sites. E. coli Strains BS583: The E. coli strain BS583, DH5 Alac (2D69 loxP 2 [lacZ LEU2]), was chosen as the bacterial background for the selection procedure using 30 plasmid pBS584. Due to the loxP 2 [lacZ] containing k prophage, Cre activity on loxP can be evaluated simply by using X-gal plates. 79 WO 00/60091 PCT/USOO/09154 BS1494: The E. coli strain for selection was established by introducing the selection plasmid pBS584 into BS583. Thus, BS1494 allows a kanamycin selection for loxK2 and in parallel a blue/white-screen for loxP recombination with 5-bromo-4chloro-3-indolyl-p-D-galactopyranoside (X-gal). Note that the 5 spacer region of the loxK2 site (FAS1) is different from the original one of loxP. Thus, recombination events between loxP of the k prophage and loxK2 of the selection plasmid pBS584 catalyzed by potent Cre mutants are excluded in BS1494. BS1576 to BS1581: For the in vivo recombination experiments wt and 10 mutant Cre expressing strains were generated by introducing plasmids pBS606, 614, 626, 627, 628, and 650 into DH5a. Transformation of E. coli For plasmid transformations of E coli strains, electroporation was preferred over chemical protocols. Electrocompetent cells were made and used 15 for electroporation as described by Smith et al., Focus, 12:38-40 (1990). The appropriate cell porator and cuvettes were from Life Technologies (Bethesda, MD). Site-Directed Mutagenesis The QuickChange T M Site-Directed Mutagenesis Kit (Stratagene Cloning 20 Systems, La Jolla, CA) (1997) was used to generate defined single and double mutations in the cre gene. Overexpression and Purification of Candidate Mutants Wild-type and five different mutant Cre proteins were overexpressed using plasmids pRH200, and pBS654 to pBS658 in conjunction with Novagen 25 B121 (DE3) cells. After induction for 2.5 h cells were harvested, sonicated, and Cre partially purified after DNase I digest with a single step Whatman@ P11 resin (Whatman Inc., Fairfield, NJ) as described before by Wierzbicki et al., J. Mol. Bio., 195:785-794 (1987). The obtained Cre preps were about 80% pure and protein concentrations ranged between 100 and 200 ng/pl. 30 Mutant Analysis In Vivo: Plasmids pBS632 to pBS641 were transformed into Cre expressing E. coli strains BS1576 (wt), BS1577 (E262G), BS1578 80 WO 00/60091 PCT/USO0/09154 (E262G/D29A), BS1579 (E262G/D189N), BS1580 (E262G/T3 16S), and BS1581 (R3M3). After 1 hour of induction of cre expression with 0.2% L-(+) arabinose, 10- dilutions were plated on non-selection medium containing 0.2% D-glucose. After overnight incubation at 37 0 C, colonies were transformed to 5 kanamycin-selection plates for the described negative selection for neo excision. In Vitro: Purified Cre mutants were used for both in vitro recombination and gel retardation experiments as described before by Sauer B.. Nucleic Acids Res., 24:4608-4613 (1996) and Wierzbicki et al., J. Mol. Biol., 195:785-794 (1987), respectively. For the recombination reactions plasmids pBS632 to 10 pBS641 served as substrates, whereas for the DNA binding reactions y[ 33

P]

dATP (Amersham Pharmacia Biotech, Piscataway, NJ) end-labeled 35 bp oligonucleotides were used, each encoding a lox-halfsite and one half of the FAS 1 spacer. Evaluation of Six K2*/P* cre Mutants 15 Six of the 36 identified single KanR colonies of the third round were chosen for further analysis and referred to as R3M1 to 6 (Round 3 Mutants 1 to 6). Retesting them in the indicator strain revealed that all but one (R3M4) show significant loxK2 recombination and all are unbiased in their activity on loxP (Table 14). Sequencing analysis revealed one amino acid change common to all 20 five mutants having increased loxK2 activity: a glutamate to glycine exchange at position 262 (E262G) in the J helix of the Cre protein (Figure 13). A second point mutation, a conservative threonine to serine exchange at position 316 (T316S), was identified in three of the mutants with enhanced loxK2 activity. Two non-conservative mutations (D29A and D1 89N) were found in two of the 25 five mutants. In addition, ten mutations occurred only once. Therefore, the critical mutation for loxK2 activity appears to be E262G. Site-Directed Mutagenesis To address the question of the influence of the different point mutations further, the QuickChange TM Site-Directed Mutagenesis Kit (Stratagene) was 30 applied to generate the following Cre mutants, each confirmed by sequencing: E262G, E262G/D29A, E262G/D189N, and E262G/T316S. 81 WO 00/60091 PCT/USO0/09154 In Vivo Characterization To elucidate the contribution of specific amino acid changes in conferring altered recombinational specificity to Cre, recombination assays with different lox sites were carried out with the following Cre enzymes: the wt 5 enzyme, one of the originally sequenced third round mutants (R3M3), and the generated single and double mutants. All lox sites used for the in vivo tests were designed to have the same 8 bp spacer region (FAS 1) so that recombinational specificity was completely dependent on Cre's recognition of the symmetrical inverted repeats of the lox sites. Note that wt Cre-dependent 10 recombination between loxP sites bearing the FAS 1 spacer does not differ from recombination between original loxP sites (Sauer B., Nucleic Acids Res., 24:4608-4613 (1996)). Figure 14 presents the recombination frequencies of various mutant lox 2 substrates and combination of sites (loxP with loxK2 and as control with loxK]) 15 from the marker excision assay. Mutant lox sites with symmetric nucleotide substitutions at positions 11 and 12 of the loxP sequence were tested with the wt enzyme and the five variant Cre mutants, including the multiple mutant R3M3(A). All enzymes showed a maximum in recombination (close to 100%) with thymines at these positions, i.e. with loxP 2 . Adenines, i.e. loxK2 2 , lowered 20 the recombination frequencies drastically for the wt enzyme, whereas the single and double mutants performed approximately 50% to 70% less effectively. R3M3, however, showed nearly LoxP-like activity on the loxK22 substrate, as seen before with the selection strain (Table 14). Altering the two thymines to guanines resulted in similar recombination frequencies as seen with adenines at 25 these positions. Cytosines, on the other hand, did not result in similar recombination frequencies as seen with thymines, but were surprisingly the least efficiently recognized substitutions of positions 11 and 12. To conclude, it appears that the E262G mutation is necessary and sufficient to significantly increase recombination frequencies on lox sequences which are symmetrically 30 altered at positions 11 and 12. Of the additional mutations tested, D29A seems to be slightly beneficial, whereas Dl 89N and T3 16S appear indifferent or even slightly deleterious for recombination on the variant sites. Thus, additional 82 WO 00/60091 PCT/US00/09154 mutations of R3M3 (Figure 13) must be responsible for its further increased performance on the mutant substrates in vivo. Figure 14B shows the observed recombination frequencies on mixed substrates (e.g. loxP with loxK2). For both loxK2 and the control substrate, 5 loxK], recombination with loxP by wt enzyme was substantially less than for loxP 2 recombination. This recombination frequency was increased dramatically with all of the mutant Cre protein. These results hint that not only a cooperativity in binding of two Cre molecules to one lox site exists but moreover also cooperativity between Cre molecules binding to different sites 10 which then are synapsed and recombined. This finding is especially useful for genomic targeting: It suggests that a targeting vector carrying a loxP site will be effectively recombined with the endogenous lox-like site by the Cre mutants as long as the spacers are compatible. Noteworthy also is R3M3-Cre's increased ability to recombine loxK] by 15 itself, whereas all the mutants, like wt, did not. Again, some additional mutations found in R3M3 (Figure 13) seem to be responsible not only for increased recombination frequencies on loxK2 but also on loxK1 compared to Cre bearing the E262G mutation alone or in conjunction with the D29A mutation. 20 In Vitro Characterization In Vitro Recombination Assays. For the in vitro recombination experiments the same substrates as in vivo--that is, pBS632 to pBS641--were used. An in vitro recombination experiment using wt, E262G, and R3M3 Cre preps on loxP 2 , loxK2 2 , and loxK12 substrate showed that, as seen in vivo before, 25 wt Cre is capable to recombine loxP 2 substrates only. No recombination products are visible when loxK2 2 or loxK1 2 substrates were used. By introducing the E262G mutation into Cre, however, recombination of loxK2 2 substrates becomes possible at elevated frequencies, and even for loxK1 2 substrate recombination products become weakly visible in vitro. E262G-Cre 30 activity on the loxK1 control site in vitro but not in vivo probably derives from differences in ionic strength and/or enzyme concentration between the assays. Finally, with R3M3 Cre the ability for both loxK2 2 and loxK1 2 recombination is 83 WO 00/60091 PCT/USOO/09154 further increased, as expected from the in vivo assays. As mentioned in the in vivo results before, guanines at positions 11 and 12 of the lox sequence were recognized at similar frequencies as seen with adenines (i.e. loxK2), whereas cytosines were clearly less tolerated by all Cre mutants. Slight differences 5 between in vivo and in vitro recombination frequencies are probably due to differences in ionic strength, temperature, DNA condensation, enzyme concentration, etc. In general, the pattern of the in vitro recombination frequencies of the different Cre enzymes on the different lox sites mirrors the one seen in vivo. 10 Gelshifit Experiments. Gelshift experiments were applied to address the question of in vitro DNA-affinities of the different Cre mutants. As expected from the previous results, all three Cre enzymes bind with similar efficiency to loxP, whereas to loxK2, only E262G and R3M3 show binding affinity. Surprisingly, R3M3 binding appears less efficient on the loxK2 half-site than 15 binding of E262G, and on the loxK1 half-site only E262G does bind weakly whereas R3M3 does not. DISCUSSION lox sites loxP and loxK2 differ at several locations as illustrated in Figure 1: 20 First, the three outermost bp of the inverted repeats are altered (positions 15, 16, 17, 15', 16' and 17'). Second, the entire non-canonical 8 bp spacer is completely exchanged (positions 4 to 4'), and third, two transversion events (thymine to adenine) are introduced at positions 11 and 12 (and mirrored at 11' and 12') of the lox site, mimicking potential recombination targets in 25 eukaryotes. Only the two mutations at positions 11 and 12 are considered important for inhibiting wt Cre from recognizing the site. Thus, they were the only alterations in the loxi sites used for the in vivo and in vitro characterization experiments. Other investigations have shown that the two alterations of loxP mentioned first are without inhibiting effect on wt Cre (Sauer, B., Mol. Cell. 30 Biol., 7:2087-2096 (1987) and Sauer B., Nucleic Acids Res., 24:4608-4613 (1996)). Noteworthy is however that the altered 8 bp spacer region (FASI spacer) does not allow loxP - loxK2 recombination, since the regions where the 84 WO 00/60091 PCT/USOO/09154 single-strand cleavages and exchanges take place are not compatible. To allow simultaneous monitoring of Cre-mediated recombination both at the wt loxP 2 and at a mutant lox 2 substrate (loxK2 2 , incompatible spacer elements were used to prevent recombination between the two types of lox sites by a candidate Cre 5 mutant with altered specificity. Such recombination might easily compromise ready detection of desired Cre specificity mutants. Incompatible spacers (original loxP and FAS 1) formed the basis for the simultaneous selection for loxK2 recombination and screen for loxP recombination with E. coli strain BS1494 which led to the disclosed variant Cre recombinases. 10 LoxK1, the other lox sequence used in this study, bears two critical bp exchanges per arm as well, however at different positions: 10 and 14. It was used as a control lox site, addressing the question whether the generated Cre mutants with novel specificity for loxK2 can also tolerate adjacent alterations within the lox sequence. 15 Cre Mutants after Three Rounds of Directed Evolution Three iterations of the in vitro evolution procedure were used to identify 36 candidates which express Cre mutants that could process loxK2 (based on the selection), as well as loxP (based on the simultaneous screen). wt Cre, on the other hand, only shows marginal activity on loxK2 when expressed at very high 20 levels. Characterization Sequence analysis revealed that five mutants with significant loxK2 activity had in common only one point mutation, leading to the amino acid change E262G. However, several other point mutations occurred independently 25 twice or thrice, among which D29A, D189N, and T316S. To investigate the influence of the mentioned mutations on the observed phenotype, specific single and double mutants were generated by site-directed mutagenesis. In vivo and in vitro assays were then carried out with wt-Cre and five different mutants (R3M3, E262G, E262G/D29A, E262G/D189N, E262G/T316S) to compare 30 their performance on a variety of different alterations of the loxK2 site. The in vivo and in vitro recombination assays showed a similar pattern in recombination frequencies for the different enzymes on the different sites 85 WO 00/60091 PCT/USOO/09154 tested. In general. recombination frequencies on mutant substrates were distinctively the highest with R3M3. The double mutant E262G/D29A was about half as effective as R3M3, whereas the other double mutants and the single mutant E262G showed slightly further decreased recombination 5 frequencies on the altered sites. The wt enzyme did not recombine any of the mutant substrates presented here, neither in vivo nor in vitro. With previous results showing that single D29A, Dl 89N, and T316S mutants of Cre perform like the wt enzyme on loxK2 and loxK1 in vivo and the fact that the E262G point-mutation was the only one found independently in all the originally 10 isolated Cre mutants with loxK2 specificity, it is clear that E262G is a critical mutation that allows Cre to recognize loxK2. However, in combination with E262G, D29A permits still higher recombination frequencies on lox sites altered at positions 11 and 12. Since R3M3 shows even further increased recombination frequencies on the altered sites compared to E262G/D29A, other 15 of the point mutations identified in this Cre mutant (see Figure 14) must account for this increase in activity. Because of its location close to the amino-terminus of the J helix, the Q255R mutation of R3M3 could be expected to influence DNA contacts. Other mutations may influence protein folding or protein protein interactions which could result in a higher flexibility within the Cre-lox 20 interface and thus allowing a better tolerance of alterations of the lox sequence. This hypothesis is also supported by the observation that R3M3 recognizes the loxK1 site at frequencies similar to E262G recognizing loxK2. The double and single Cre mutants, on the other hand, did not show activity on loxK1. In addition, the gel-shift experiments showed that R3M3-Cre's binding affinity for 25 loxK2 and loxK1 half-sites is less than E262G-Cre's and the three double mutant's. Taking these results together, other mutations of R3M3 must further influence Cre-lox interactions to allow enhanced recombination on loxK2. On the one hand, this results in less efficient binding to lox half-sites, on the other hand, when complete lox sites are available, the cooperativity phenomenon 30 between Cre enzymes binding to the same and different lox sits may compensate for this loss in binding activity. Then the postulated increased flexibility between DNA and protein seems to become advantageous for 86 WO 00/60091 PCT/USOO/09154 recognizing and recombining altered lox sites, as seen with the in vivo and in vitro recombination assays. Alternatively, some mutations, also silent ones, could influence protein expression, leading to a faster accumulation of enzymes and consequently to 5 higher recombination frequency. An E. coli codon usage table suggested, however, that none of the identified mutations should improve Cre expression in E. coli remarkably. Modeling With the published crystal structure of four Cre molecules bound to two 10 synapsed loxA sites after the first single strand cleavage (Guo et al., Nature, 389:40-46 (1997)) the identified point mutations were analyzed for being involved in DNA and/or protein interactions. All of them, including the E262 position in the J-helix of Cre were found to be not involved in either interactions in this state of the reaction. This observation indicates that the mutations which 15 were found to account for the described novel substrate recognition in vivo and in vitro lead to this new phenotype in a less direct and obvious manner. As mentioned earlier, they may influence protein folding, resulting in a higher flexibility within the Cre-lox interface. This hypothesis is especially well supported with the described differences between binding to lox halfsites and 20 recombination. Alternatively, they may still be involved in protein-protein (D29A) or protein-DNA (E262G) contacts before or after the formation of the clamp-like strand-exchange state. In Vivo and In Vitro Recombination Some additional variations of the loxK2 site were tested also. The list in 25 Table 17 shows all the sequences of lox sites which were tested and assigns them a name, as well as their plasmid (pBS) number. In Figure 15 the obtained in vivo recombination frequencies on all the variants of loxK2 are indicated. The additional results indicate that alterations at position 12 of the lox halfsite are of more importance for Cre-based recombination than ones at position 11. 30 The in vitro recombination frequencies of all six Cre enzymes tested on the lox sites listed above are given in table 15. The frequencies were calculated after quantitation of the brightness of fluorescence of the Ethidium-Bromide 87 WO 00/60091 PCT/USOO/09154 stained DNA fragments on agarose-gels. Differences in temperature, ionic strength, medium composition, and enzyme concentration probably account for the observed differences between in vivo and in vitro recombination results. Most strikingly, loxP is no longer recognized with the highest frequencies in 5 vitro. However, when the ionic strength in the in vitro assay was increased results began to resemble the ones seen in vivo. Thus, efficiency of recombination with variant lox sites by each of the Cre mutant and the wt enzyme can be further controlled in vitro by adjusting ionic strength and other in vitro conditions. 10 Qualitatively, however, the in vivo and in vitro recombination frequencies mirror each other. These novel Cre mutants thus possess a specificity for substrates (loxK2 and its derivatives) which are not recognized by the wt enzyme. Gelshift Assays 15 DNA binding (gelshift) experiments were also done with the generated Cre double-mutants (E262G with D29A, D1 89N, or T316S) to analyzing binding of the variant Cre recombinases to various recombination sites. In table 16 the observed mean percentages of binding to loxP, loxK2, and loxK] halfsites with the five different Cre mutants and the wt enzyme are given. As 20 shown, all mutants - in contrast to the wt enzyme - do bind to loxK2 with similar frequencies, except R3M3 which shows surprisingly low retardation. As discussed above, this phenomenon may be explained with an increased tolerance, i.e., flexibility, of R3M3 for altered lox sites. On halfsites which precludes the cooperativity between Cre molecules in binding R3M3 cannot 25 bind as tight as wt or the single and double mutants. With the loxK1 halfsite this 'binding versus recombination' difference is even more strikingly. Whereas wt and the R3M3 mutant cannot bind to the halfsite, all the other mutants can. Yet, recombination of loxK1 substrates was seen with R3M3 Cre, only. These results show clearly that simple DNA affinity does not correlate in 30 a one-to-one fashion with recombination. Thus, inappropriate DNA binding by recombinases likely can lead to a block in recombination. 88 WO 00/60091 PCT/US00/09154 Conclusions The in vivo data show that: 1. The E262G mutation confers a generally elevated level of recombination at a number of variant lox sites, those having any of a large number of alterations at positions 11 and 12 (and mirroring 11' and 12' alterations). 5 2. The D1 89N mutation in conjunction with the E262G mutation appears to fine tune the broadened specificity of the E262G mutation by reducing recombination at the loxK2 variants 'GG', 'TC' and 'CC' without decreasing recombination at loxK2 and the 'GT' variant. This mutation is thus useful to limit the broadened specificity of E262G. 10 3. The T316S mutation when in conjunction with E262G provides a slight boost in recombination frequency with the loxK2 variants 'GT' and 'CC', and has no deleterious effect on recombination at the other variant loxK2 sites. 4. The D29A mutation together with E262G boosts recombination at loxK2 and the variants 'CC' and 'GG'. D29A does not reduce recombination at the 15 other variant lox sites or at loxP. 5. Some of the additional mutations of R3M3 must account for the further increased recombination frequency on any of the tested mutant lox sites, including loxK1 but do not compromise loxP recognition. The disclosed variant recombinases have a number of useful features 20 and applications. By recognizing an altered, user-defined target site, they were designed to allow both genetic targeting events in prokaryotes and eukaryotes like wt Cre but on different sites and in vitro recombination strategies. With a wider variety of possible target sequences being now accessible, multiple and defined genomic alterations will now become feasible. This opens more 25 possibilities in designing genomic manipulations in all DNA-based organisms by site-specific recombination. With the various genome projects under way today, there will be more and more applications for site-specific recombination to study the impact of genes or genetic control elements by genomic engineering. In addition, genome 30 manipulations are also more frequently used to express recombinant proteins within the organism of choice. 89 WO 00/60091 PCT/USO0/09154 Table 13. Presented are the results of the first four rounds of the described mutagenesis procedure for loxK2 specificity. Column one indicates the round of the mutagenesis procedure (round 0 indicates that no mutagenesis event has taken place yet, wt enzyme, only) and column two the allotted time 5 for cre expression in induction medium. Column three presents the observed frequencies of KanR, indicating loxK2 recombination by a mutant candidate. Of note, all KanR colonies found during this experiment were white, indicating effective loxP recombination. In column four the actual number of KanR colonies found in each round of 1000 to 10 4 plated is given, and the last column 10 presents the frequency of white colonies on non-selection plates which is decreasing with every round. Table 14. Presented are the frequencies of KanR of six chosen candidate mutants from round three of the mutagenesis procedure (R3M1 to 6) when retesting them in the indicator strain. Also, they were tested for their 15 performance on loxP sites in identical fashion. The obtained KanR frequencies indicate the percentages of recombination within the allotted induction time of 2.5 hours. 90 WO 00/60091 PCT/USOO/09154 Table 2: List of Strains Used for Selection and Screen BS # Description Substrates Selection and Screen 1493 BS583 [pBS583] IoxK 1 2 /loxP 2 ieo selection and lacZ screen, respectively 1494 BS583 [pBS584] loxK2 2 / oxP 2 ieo selection and lacZ screen, respectively 1541 BS583 [pBS613] IoxP 2 / loxP 2 neo selection and lacZ screen, respectively 1523 NS2300 [pBS601] loxK12/IoXp 2 abfA.st screen and neo selection, respectively 1524 NS2300 [pBS602] loxK2 2 /IoXp 2 abfA.st screen and )ieo selection, respectively List of the E. coli strains used for the described selection and screening procedures for the desired Cre mutants. BS1493 and BS1494 served as selection strains for KI' or K2' Cre mutants, respectively. BS1541 was used as a control strain to determine IoxP activity. BS1523 and BS1524 were used to select for P- mutants. Construction and application of the different selection procedures are explained in detail in the text. 91 WO 00/60091 PCT/USOO/09154 Table 3: Sequencing Primers for cre Primer BSB# Sequence 5' - 3' Position (5', 3') in cre 454 TTT GGG CTA GCG AAT TCG AG -55, -36 455 TTT GGG CCA GCTAAA CAT GC 273,292 456 CGG TGG GAG AAT GTT AAT CC 567, 586 457 GGA CAC AGT GCC CGT GTC - 862, 879 458 TCT GCG TTC TGA TTT AAT CTG 1117, 1097 459 CCA GGC CAG GTA TCT CTG 858,841 460 GTA CGT GAG ATA TCT TTA ACC C 563,542 461 TTG CTG GAT AGT T TAC TGC C 270,249 Presented are the eight sequencing primers used for sequencing of the entire cre gene in both directions and their positions within the cre coding sequence. Primers BSB 454 to BSB457 allow sequencing of the coding strand, whereas the four remaining primers have been designed for the non-coding strand. In order to achieve a complete sequence information in both directions, about 350 bases have to be read out of each sequencing reaction. The positions given for each primer refer to its 5' and 3' end, with position I referring to the adenine of the ATG start codon of cre. Table 4: Primers Used in the QuickChange TM Site-Directed Mutagenesis Kit (Stratagene) Primer Sequence.5', - 3' BSB# 35 in cre) 465 GCTATCAACTCGCGCCCTGGGAGGGAT1TITGAAGCAACTCATCG 765, 809 466 GAGTTGCTTCAAAAATCCCTCCCAGGGCGCGAGTTGATAGCTGGC 805, 761 467 GCTATCAACTCGCGCCCTGGCAGGGAFTTGAAGCAACTCATCG 765, 809 48 GAGYI7GCITCAAAAATCCCTGCCAGGGCGCGAG1TGATAGCTGGC 805, 761 469 GCTATCAACTCGCGCCNNNNNNNNNATTTTTGAAGCAACTCATCG 765, 809 470 GAGTTGCTTCAAAAATNNNNNNNNNGGCGCGAGTTGATAGCTGGC 805, 761 Illustrated are the mutant oligonucleotides used in the QuickChange T M Site-Directed Mutagenesis Kit (Stratagene) to introduce single point mutations (mismatches with the wt sequence are highlighted by bold letters). With BS5465/466 a E262G mutation, and with BSB467/468 a E262A mutation is introduced in Cre. The last set of primers (BSB469/470), as indicated by the bold N, represent an equimolar mixture of all possible bases at the assigned positions. This mixture results in all possible aa combinations at positions 261 to 263 of Cre. As mentioned before, the indicated positions refer to 5' and 3' end of the oligonucleotides in the cre coding sequence. 92 WO 00/60091 PCT/USOO/09154 Table 5: Comparison of Mutagenic and Non-Mutagenic PCR on wt cre Induction of Non-Mutagenic PCR , :Mutagenic PCR:-' cre Expression Blue White Blue White SOC, 2.5h > 99.6% < 0.4% > 99.5% < 0.5% -SOB+ara, 2.5h 19.7% 81.3% 60.8% 39.2% The experiments presented here served two purposes: (i) to compare the influence of glucose (SOC medium) and arabinose on cre expression with pBAD33, (ii) to compare the frequencies of deleterious point mutations occurring during non-mutagenic and error-prone PCR. The first column indicates the medium used for repression (SOC), or induction of cre expression (SOB+ara). The second column compares the observed frequencies of blue and white colonies after a non-mutagenic PCR, the third column after a mutagenic PCR. Blue colonies indicate, that the IoxP 2 flanked lacZ gene of BS583 is not excised by Cre, white ones its Cre-mediated excision. Note that the number of blue colonies resulting from empty pBAD33 vector (approximately 1% in all experiments) was subtracted before calculating the presented values. This blue background was determined with a control ligation lacking cre insertion. With glucose in the medium, only blue colonies are found, indicating that cre is not expressed. With arabinose induction, however, the frequencies of disabled Cre mutants (blue) can be determined, increasing three times under mutagenic conditions. 93 WO 00/60091 PCT/USOO/09154 Table 6: The First Four Rounds on loxK2 2 Round Status Induction % White % Kan" Nb. KanR Nb. Pooled for (Non-Selection) (White) (White) the Next Round 1 ox 4 h 83.9% 0.01% I none mx 4 h 42.4% 0.16% 6 6 2 mxox 2.5 h 88.3% 0.02% I none 4 h 82.5% 0.56% 47 47 mxmx 2.5 h 13% < 0.01% 0 none 4 h 24% 0.3% 5 none 3 mxoxox 2.5 h 80.7% 0.2% 36 30 4 h 85% 25% 104 none mxoxmx 2.5 h 25.7% 0.02% 1 none 4 h 21.8% 5% 250 none 4 mxoxoxox 2.5 h 75% 4.6% 102 none mxoxoxmx 2.5 h < 7.7% 0.3% 3. none Presented are the results of the first four rounds of the described mutagenesis procedure for IoxK2 specificity. The following symbols for the status of the cre pools have been used: "o" indicates a non-mutagenic, "m" a mutagenic PCR, and "x" stands for the shuffling step by DNase I digest and reassembly. Column one indicates the round of the mutagenesis procedure, column two its status as mentioned above, and column three the time permitted for cre expression in induction medium. In column four, the calculated percentages of white colonies on non-selection plates is indicated, decreasing with every round, especially when the pools were subjected to mutagenic PCR's twice. Column five presents the observed frequencies of Kan', indicating loxK2 recombination by a mutant candidate. Of note, that all KanR colonies found during this experiment were white, indicating effective IoxP recombination. The two last columns give the actual numbers of KanR colonies found in each round (1/5 dilution) and mark the ones used as starting material for the next round. 94 WO 00/60091 PCT/USOO/09154 Table 7: Functional Evaluation of Six Selected mxoxox K2* Mutants and wt Cre Cre Induction %,Kan" with Candidate Time BS1494 (loxK2') BS1541 (loxP 2 ) BS1493 (loxKI) wt 2.5 h 0.002% 93.9% 0.001% 4h n.d. n.d.. 0.002% mxoxox 1 2.5 h 66.4% 98.2% 0.004% 4 h n.d. n.d. 0.005% mxoxox 2 2.5 h 53.6% 98% 0.007% 4 h n.d. n.d. 0.005% mxoxox 3 2.5 h 66.5% 80.5% 0.007% 4 h n.d. n.d. 0.01% mxoxox 4 2.5 h 0.05% 80.3% 0.008% 4 h n.d. n.d. 0.006% IXoXX 5 2.5 h 15.2% 81.2% 0.006% ~ -4n.d. n.d. 0.009% 'x6 2.5h 3.1% 83.7% 0.003% S4 h n.d. n.d. 0.01% The results of functional testing of the six chosen K2*/P* Cre candidates compared to wt Cre are listed. All enzymes were checked on loxK2, loxP and loxKl recombination in the selection strains BS1494, BS1493 and the control strain BS1541. In the first column the Cre candidate is indicated, in the second the allowed time for cre expression. The three following columns show the calculated percentages of Kan' when subjecting the different Cre mutants and wt Cre to the three selection strains. The frequency of the observed KanR phenotype is considered as an indicator for the frequency of lox recombination: On loxK2, all mutants, except mxoxox 4, show remarkably increased KanR frequencies compared to wt Cre. On loxP, all (including mxoxox 4), show very similar results as wt Cre, indicating, that loxP recombination is at best slightly affected. On loxK1 neither mutant nor wt Cre show more than background activity, even after four hours of cre expression. The obtained mutants are therefore characterized by a broader substrate recognition compared to the wt enzyme. 95 WO 00/60091 PCT/USOO/09154 Table 8: Sequence Analysis of the Six mxoxox K2*/P* Mutants Crc Codon Change in cre aa Change Nb. of Position in Candidate Position w t - i mut Isolates Cre nxoxox 1 45 CCG CCT (P I5P) 2 (N-terminal) 86 GAC GCC D29A 2 A 565 GAC AAC D189N 2 loop 1 - 2 642 AGC AGT (S214S) I (N-term. of 1) 679 GTC ATC V2271 I 785 GAA GGA E262G* 5 j mxoxox 2 45 CCG CCI (P15P) 2 (N-terminal) 413 GAA GGA E138G I N-term. of F 760 AGC GGC S254G I loop 5 - J 785 GAA GGA E262G* 5 1 861 TCT TCC (S287S*) 2 N-term. of K 946 ACC TCC T316S 3 loop L - M mxoxox 3 46 GTC ATC V161 1 N-terminal 565 GAC AAC D189N 2 loop 1 - 2 592 GGC AGC G 198S I C-term. of 2 668 CGA CAA R223Q I l 764 CAG CGG Q255R 1 loop 5 - J 785 GAA GGA E262G* 5 J 861 TCT TCC (S287S*) 2 N-term. of K 920 CCG CIG P307L I L mxoxox 4 230 TAT TGT Y77C I C 412 GAA AAA E138K I N-term. of F 851 CTG CAG L284Q 2 loop J - K mxoxox 5 86 GAC GCC D29A 2 A 302 CGG CAG RIOIQ* I D 659 CTG CAG L220Q I 785 GAA GGA E262G* 5 j 946 ACC TCC T316S 3 loop L - M mxoxox 6 785 GAA GA E262G* 5 J 851 CTG CAG L284Q 2 loop J - K 946 ACC CC T316S 3 loop L - M The point mutations identified by sequence analysis in the six selected K2*/P* mutants are listed. Column one identifies the mutants, column two the observed point mutations and their position in the c-e coding sequence. Column three indicates the resulting aa changes and their positions in the Cre enzyme (silent mutations in parenthesis). DNA contacting residues, according to the crystal structure (Guo, et. al., 1997), are marked by an asterisk. Column four indicates how often the different point mutations were independently found in the pool of the six mutants, and column five shows where the aa changes are located in the secondary structure of the protein. Of the altogether 31 point mutations listed, five are silent. Only one aa change, E262G, is common to all mutants with remarkably increased loxK2 activity (mxoxox 1, 2, 3, 5, and 6), suggesting that this mutation is the essential one for the observed phenotype. Nine point mutations occurred twice or three times in the pool, whereas eleven were found in one mutant only. For further inloriation, see text and appendices A, B, and C. 96 WO 00/60091 PCT/USO0/09154 Table 9: Functional Evaluation of Three Defined Cre Mutants for Position 262 and wt Cre Cre Mutant Induction % Kan" with Time BS1494 (loxK2 2 ) BS1541 (loxP 2 ) BS1493 (loxKl 2 ) wt 2.5 h 0.002% 93.9% 0.001% E262G 2.5 h 2.2% 84.2% 0.004% E262A 2.5 h 0.4% 82.3% 0.006% E262W 2.5 h 2.9% 81.7% 0.003% The results of functional testing of three defined Cre mutants for position 262 compared to wt Cre are listed. All enzymes were checked on loxK2, loxP and loxKl recombination in the selection strains BS1494, BS1493 and the control strain BS 1541. In the first column the Cre candidate is indicated, in the second the allowed time for cre expression. The three following columns show the calculated percentages of KanR when subjecting the different Cre mutants and wt Cre to the three selection strains. The frequency of the observed KanR phenotype is considered as an indicator for the frequency of lox 2 recombination: On loxK2, E262G and E262W show approximately an 10- fold increased Kan' frequency compared to wt Cre, whereas E262A results in a 200 fold increase. On loxP, all show very similar results as wt Cre, indicating, that loxP recombination is at best slightly affected. On loxKl neither mutant nor wt Cre show more than background activity. This indicates, that E262G and E262W alone are sufficient to remarkably increase Cre's activity on loxK2, that these mutations do not increase loxKl recognition, and that loxP activity may be slightly affected. 97 WO 00/60091 PCT/USO0/09154 ~ 10 In Vitro Evolution of the Gre recombinase wtCre TFX3GGCTAG CGAATrCGAG CTCGrGTACCC GGGGATCCT TAGACTGAGr m xoxox I 1TRGGCTAG CGAATTCGAG CFCGGTACCC GGGGATCCWC TAGACTIGAGr mxoxox2 TGGGCrAG CGAATTCGAG CTCGGTACCC GGGGATCCTC TAGACTGAGT mxoxox3 1TfGGCT-AG CGAATIX2GAG CTCGGTACCC GGGGATCCT TAGACTGAGT mxoxox4 TITGGGAG CGAATTCGAG CTCGGTACCC GGGGATCCIC TAGACTGAGT mxoxox5 TGGGCTAG CGAATTCGAG CTCGGTACCC GGGGATCCU TAGACTGACJ7 m xoxox6 T'1TGGGC7AG CGAATTCGAG CTCGGTACCC GGGGATCCIC TAGACIYJ7 -55 wtCre GTGAA -7 : TC CAATTTACrG ACCGTACACC AAAXTTTGCC TGCATT7ACCG mxoxox I GTGAA.-c 3TC CAATrACTG ACCGTACACC AAAATFIGC TGCAYI'ACCI' mxoxox2 GTGAA-2C TC CAA-FITACTG ACCGTACACC AAAATI1XJCC TGCATTACCT mxoxox3 GTGAA-J -TC CAXITTTACXJ ACCGTACACC AAATI-MCC TGCA-17ACCG mxoxox4 GTGAA -J--TC CAATITACTG ACCGTACACC AAAAT[GCC TGCATTACCG rnxoxox5 GTGAA--ciTC CAAT-ITACrG ACCGTACACC AAAATFrGCC flJCANIACCG mxoxox6 GTGAA--2K3TC CAAT1TACT7J ACCGTACACC AAAA'TrGCC TOCATTACCO -5 wtCre GTCGATGCAA CGAGT)GATGA GGrCUGCAAG AACCTGAT)GG ACATTTIAG mxoxoxl GTCGATGCAA CGAGTGKIXJA CGGTTCGCAAG AACCTGATGG CCATGTTCAG moo2GTCGATGCAA CGAGTGATGA GGTrCGCAAG AACrrGATGG ACATrTIAG mx oxox3 AW)GAT)GCAA CGAGTrGATGA GGTTCGCAG AACTGATGG ACATFRTAG Gxxx4: GTGATGCAA CGAGTGAIX3A GGFCGCAAG AACCTGATGG ACATGTITCAG mxoxox5- GTCGATGCAA CGAGTXJA1XA GGTTX3CAAG AACCIGAR3G CCATGTTCAG X IaX0 ;y 16" GT=ATGCAA CGAGTGAflJA GGFrIXCPCAG AACCTGATGG ACAIT=AG 4Cr~ GGAWCGCCAG GCGFTTG= AGCATAcCPJG GAAAATOCIT CT3TICC=fl iiflxxi6i1' GGATCGCCAG GCGTF=P AGCATACrG GAAAATGCIT CTGTRCfl :rxox ox2l, GYGATCGCCAG GCG-rTflC AGCATACCX3 GAAAAWGCIT CTuIXcG1T mxox6o 3 - GGATCGCCAG GCGITITCX3 AGCATACCXJ GAAAA'3CIT CIXJTccGFL ~mxoxoi4.: GGATCGCCAG GCGTITICTG AGCATACCTC GAAAATGCI c-rTICG -mxoxox5. GGAWCGCCAG GCGFTTCTG AGCATACCTG GAAAATGCIT CTcGGI1 * mxoxox6 GGATCGCCAG GCC1TT= AGCATACCT'G GAAAATGCTT CTGTCCGTT 96 wtCre GCCGGTCGTG GiGCGjGCATCG TCAAGTMXA ATAACCGGAA ATGGTFICC mxoxox 1 GCCGGTCGPJ GGCGGCNIXJG PJCAAGTIUA ATAACCGGAA ATGGT]CC mxoxox2 GCCGGTCGIG GGCGGCATGG TUCAAGITTA ATAACCGGAA ATGGFFTCCC mxoxox3 GCCGTCGTX3 GGCGGCAIIJG IUCAAGTPJA ATAACCGGAA ATGGTTTCCC mxoxox4 GCCGGTrCGIXJ GGCGGCATGG TGCAAGTRGA ATAACCGGAA ATGGTCCC mxoxox5 GCCGGTCGTG GGrCGGCATGG TGCAAGVIX3A ATAACCGGAA ATGGTTTICCC mxoxox6 GCCGGTCGTG GGCGGCAT)GC 1GCAAGWPA ATAACCGGAA ATGGTFICC 146 wtCre GCAGAACCTG AAGATGYICG CGATTATCIT CTATATCTTC AGGCGCGCGG mxoxox I GCAGAACCTG AGATG=IX CGATFATCF CrATATXJVIC AGGCGCGCGG; mxoxox2 GCAGAACCTG AAGATGTTCG CGATTATCI CrATATCYIC AGGCGCGCGG rnxoxox3 GCAGAACCU3 AAGATUFPY3 CGATTATCI CTATATUTTC AGGCGvCGCGG m xoxox4 GCAGkACCM AAGATGJV1CG CGATTATC= CTATGTCTTC AGGCGCGCGC mxoxox5 GCAGAACCTC AAGATGTVICG CGATAW CFATATCVR AGGCGCGCGG m xoxo x6 GCAGAACCTt3 AAGATGC1TCG CGATTAT=T CFATATCFU AGGCGCGCG 196 98 WO 00/60091 PCT/USOO/09154 in Vitro Evolution of the Cre recombinase wtCre TCTGGCAGTA AAAACTATCC AGCAACATIT GGGCCAGCTA AACATGCTT mxoxox i TCTGGCAGTA AAAACTATCC AGCAACATIT GGGCCAGCFA AACATGCTTC mxoxox2 TTGCA(:rA AAAAcTATCC AGCAACATITT GGGCCAGCTA AACAIGCFRI mxoxox3 WI'GGCAGTA AAAACTATCC AGCAACA1TU GGGCCAGCTA AACATGCTT mxoxox 4 TCTGGCAGTA AAAACTATCC AGCAACAT1T GGGCCAGCT7A AACATGCI mxoxox5 TCTUGCAGTA AAAACTATCC ACCAACATI GGGCCAGCT'A AACATGCY mxoxox6 WCTGGCAGTA AAAACI'ATCC AGCAACATV GGGCCAGCUFA AACATCCFT 246 wtCre ATCGTCGGTC CGGGCTGCCA CGACCAAGTG ACAGCAATGC TGT~CACTG mxoxox I ATCGTCGGTC CGGGCTGCCA CGACCXAGTG ACAGCAATGC TJITrTACTG mxoxox2 ATCGTCGGTC CGGGCT'GCCA CGACCAAGTG ACAGCAAGC TGTTITCACTG mxoxox3 ATCGTCGGTC CGrGGCTGCCA CGACCAAGTG ACAGCAATGC TGTTrCACTri mxoxox4 ATCGTCGGTIC CGGGCPJCCA CGACCAAGTG ACAGCAAT)GC TG=IACrG mxoxoxS ATCGWCAGTC CGGGCTGjCCA CGACCAAUTG ACAGCAATGC TGTiWCACIG mxoxox6 ATCGTCGGTC CGGGCTGCCA CGACCAAGTG ACAGCAATGC TG-1TCACTG 296 wtCre GTTATGjCGGC GGATCCGAAA AGAAAACGYF GATGCCGGTG AACGIX3CAA mxoxox i GITrATGCGGC GGAWCCGAAA AGAAAACGIT GATGCCGG~ri AACGIXJCAAA mxoxox2. G-17ATGCGGC GGATCCGAA AGAAAACGFF GATGCCGGTG AACGTUCAAA m xo xox3 GrTATG-CGGC GGATCCGAAA AGAMAACGTT GATGCCGGT' AACGIVYCAAA ii10-o-X4 G-ITATGCGGC GGATCGAAA AGAAAACGIT GAIXJCCGGTXJ AACGIrGCAAA -n; ,X0XOX5~- GTATGCGGC GGATCCGAAA AGAAAACGFI' GATGCCGGYG AACGTGCAAA m x xo x 6~ GIFTA1XCGGC GGA'ICCGAAA AGPAAACGFF GATXJCGGGTG AACGMCAAA :fidokel ACAGGCICA GCGTrCGAAC GCACIUAMT CGACCAGG1T CG'ITCACI'CA -~xxc~ ACAGGCPIA GCGTrCGAAC GCACrGATF CGACCAGGIT CGFICACTFCA nmxoxoO3 ACAGGCWA GCGFCGAAC GCACTOATIT CGACCAGG1T C(TrWAC~r-A - xxo4ACAGGCIXA GCG1TTCAAAC GCACTGATIT CGACCAGGTT CGTIWCA mxo x ox5 ACAGGCTA GCGTCGAAC GCACTt3ATTT CGACCAGGIT CGTCACrrA m oxox6 ACAGGCTCTA GCGFICGAAC GCACTUATIT CGACCAGGIT CGTVICACTCA 396 *wtCre TGGAAAATAG CGATCGCrGC CAGGATATAC GIT7AATC NITIrCGGG mxoxox I. TGGAAAATAG CGATCGCTG-C CAGGATATAC GTAATCIUGGC ATTT~flGGG rnxaxo .x2 1XJGAAAATAG CGATCGC CAGGATATAC G7AATCrGGC ATFIL-CGGG mxoxox3 TGGAAAATAG CGATCGCIGC CAGGATATAC (j7AATCPC3GC ATICTOGG mxoxox4 TGGAAAATAG CGATCGCX3C CAGGATATAC GTAATCTGGC ATFICIXGGG mxoxox5 TGGAAPATAG CGATCGCT)GC CAGGATATAC GT7AATCTGGC ATICTOGGG mxoxox6 TUGAAAATAG CGATCGCTGC CAGGATATAC GrAATCTGGC ATTICTGGGG 446 wtCre ATIXJCITATA ACACCCI'GTT ACGTATAGCC GAAATrGCCA GGATCAGGGT mxoxox 1 ATMCIATA ACACCCI'GTT ACCTATAGCC GAAAT1XJCCA GGATCAGGGT7 mxoxox2 ATTUCITATA ACACCCT(TF ACGTATAGCC GAAATTGCCA GGATCAGGGT mxoxox3 ATrUCIATA ACACCCTGTT ACGTATAGCC GAAATrGCCA GGA'ICAGGGT mxoxox4 ATTGCTTATA ACACCCTGT ACGTATAGCC GAAATTGCCA GGATCAGGFT mxoxox5 ATRTIATA ACACCCTGT ACGTATAGCC GAJ.J(TGCCA GGATCAGGGT mxoxox6 ATIXJC1TATA ACACCCT=F ACGATAG;CC GAAAT-rUCCA GGATCAGGGT 496 99 WO 00/60091 PCT/USOO/09154 Tau,~i In Vitro Evolution of the Cre recombinase wtCre TAAAGATATC TCACGTACTG ACGGTGGGAG kATGTAATC CATATrGGCA mooi TAAAGATATC TCACGTACFA ACGG1T3GGAG AATTAATC CATATTGGCA mxoxox 2 TAAAGATATIC TCACGTACTG ACGGTGGAG AATGTTAATC CATATrGGCA mxoxox 3 TAAAGATATC TCACGTACI'A ACGGTG3GGAG AATGTTAATC CATA1TAGCA mxoxo x4 TAAAGATATC TCACGTACTG ACGGTGGGAG AATG'ITAATC CATATIGGCA mxoxox5 TAAAGATATC TCACGTACTG ACGGTGGGAG AATG-ITAAT)C CATATTGGCA mxoxox 6 TAAAGATATC TCACGTACTG ACGGTGGGAG AATITrAATC CATATIX3GCA 546 wtCre GAACGAAAAC GGGTFAGC ACCGCAGGTG TAGAGAAGGC ACITAGCCTG mxoxox I GAACGAAAAC CCGGITAGC ACCGCAGG'RI TAGAGAAGGC ACTTAGTCXJ mxoxox2 GAACGAAAAC CYTAGC ACCGCAGCTG TAGAGAAGGC ACTAGCCTG mxoxox 3 GAACGAAAAC CCITAGC ACCGCAGGTG TAGAGAAGGC AC1TAGCCPJ mxoxox4 GAACGAAAAC GCTGGITAGC ACCGCAGGJiX TAGAGAAGGC ACITAGCCTU mxoxox5 GAACGAAAAC GCTITAGC AcCGCAGGTG TAGAGAAGGC A=IAGCCTG inxoxox6 GAACGAAAAC GCTGGFI'AGC ACCGCAGGTGi TAGAGAAGGC ACrTAGCCTG 596 wtCre GGGG7AACrA AACTGTA GCOGGATF TCCGTC7CTG G=UAGCrGA rnxoxox I GGGGTAACTA AACrXGGTX3A GCGAIGGA'T TCCATCTC7G GTUAGCIUA mxoxox2 GGGGTAACFA AACTGTGA GCGATGGATT TCCGWCTCTG GflJTAGCTGA ,Mxoxox3. GGGGrAACrA ArGLA GCTGT TCCGTCTG mAGC m xoxox4 GGGGTAACTA AACTGGWCGA GOGATGGATT TX2CGTCTICTG GRAGCTMA 6 mxoxox5 GGGGAACTA AACAGGTCGA GCGATGGA'T TCCGTCTCX GRTAGCIIGA -. ixoox6'- GGGGTAACTA AACTGGTCA GCGATGGATF TCOGTrcTC GrIAGCMA ~4 '~C TG ATCCGAAT AACTACCTG TlTnUccG= CAGAAAAAAT GG1UTGC 4mlxoxoxl,. TGATCCGAAT AACTACCI'G7 Tf7GGG7 CAGAAAAAAT GGRTX3GC .xoxox2 TGATCCGAAT AACrACCM7T -TlrjCmGGGT CAGAAAAAAT GG1UI1X3O2CG xnxoxox3, TIGATCCGAAT AACTACCIUI TPJCCGGGr CAGAAAAAAT CGGPJC m Ixoxo Ix 4 TGAICCGAAT AACrACCRT TTTGCCGGG7 CAGAAAAAAT GGIU1TGC= xxox5 TGATCCGAAT AACrACCTG TIGCCGGGT CAGAAAAAAT CGTTGCCG mxoxox6 TGATCCGAAT AACTrACCTGT 'TTGCCGGGT CAGAAAAAAT GGTTGC 696 wtCre CGCCATCTGiC CACCAGCCAG CTATCAACTC GCGCCCTFGGA AGOGATITI 1xxx CGCCATCT)GC CACCAGCCAG CTATCAACTC GCGCCCTGGG AGGGATITI mxoxox2 CGCCATCTIGC CACCGGCCAG CTATCAACI'C GCGCCCI'GGG AGOGATITI mxoxox3 CGCCATCTG3C CACCAGCCGG C1TATCAACTC GCGCCCrGGG AGGGATITI mxoxox4 CGCCATCr)GC CACCAGCCAG CrATC-AACTC GCGiCCCTYGGA AGOGAITI mxoxox5 CGCCATCTGC CACCAGCCAG CTATCAACTC GCGCC47rGGG AGGGATITI mxoxox6 CGCCATCTGC CACCAGCCAG CTATCAACIC GCGCCCTGGG AGGOATrTI 746 _ wiCre GAAGCAACIC AICGATTGAT TTACGGCGCVF AAGGATGACT CTGGTCAGAG Mxoxox2 GAAGCAACIC A'ICGAThJAT TACGGCGC-T AAGGATGACT CTGGIC-AGAG mxoxox3 GAAGCAACXI AThX3A=IAT TTACGGCGCr AAGGATGACT CTGGTCAGAG mxoxox4 GAAGCAACTC ATCGATIGAT TTACGGCGC AAGGA1X3MIT CTXGWTAGAG mxoxox5 GAAGCAACTC ATCGATIX3AT 1TACGGCGCT AAGGATGACT CIXGITCAGAG mx oxox 6 GAAGCAACTC ATCGA'TTGAT TTACGjGCGCr AAGGATGACF CTGGTCAGAG 796 100 WO 00/60091 PCT/USO0/09154 fi, Vitro Evolution of the Cre recombinase wtCre ATACCrGGCC TGGTCTGGAC ACAGTGCCCG TGTCGGAGCC GCGCGAGATA mxoxox I ATACCrGGCC TGTCT'GGAC ACAGTGCCCG TGTCGrGAGCC GCGCGAGATA mxoxox2 ATACCTGjGCC TGGTCCGGAC ACAGTGCCCG TCTCGGAGCC GCGCQGAGATA mxoxox3 ATACCTGGCC TGGTCCGGAC ACAGTGCCCG TGTCGGAGCC GCGCGAGATA mxoxox4 ATACCAGGCC TGCT-GGAC ACAGTGCCCG TcJICGGAGCC GCGCGAGATA mxoxox5 ATACCTCGGCC ThXGrCTGGAC ACAGTGCCCG TGCAGCC GCGCGAGATA m xoxox6 ATACCAGGCC TGGTCTGAC ACAGTGCCCC TGTCGGAGCC GCGCGAGATA 846 wtCre TGGCCCGCGC TGGAGTTIA ATACCGGAGA T'CATGCAAGC TGGTGGCTGG mxoxox I TGGCCCGCGC TGGAG=TA ATACCOGAGA TCATGCAAGC TGGTrGGCTG mxoxox2 TGGCCCGCGC TGGAGTITA ATACCGGAGA TCATGCA.AGC TGGrGGCTG mxoxox 3 TGGCCCGCGC TIGGAGITICA ATACTGAGA TCATGCAAGC TXGTG-CRX mxoxox4 TGGCCCGCGC TGGAGITrCA ATACCGGAGA TCATGCAAGC TUGTGGCTIGG mxoxox5 TtGrCCCGCGC TGGAG1TTCA ATACCGGAGA TCATGCAAGC TGG-TGGCTXG mxoxox6 TGrGCCCGCGC ThXJAGTIA ATACCGGAGA T)CA'DJCAAGC TGGTR3CRFG 896 wtCre ACCAAT7AA ATATIUIAT GAAGrATATC CGTAACCT*GG ATAGTGAAAC mxoxox I ACCANIUIAA ATATIGIAT GAACTATATC CGTAACCXX3 ATAGTGAAAC mxoxox2, TCCAATGTAA ATATTUICAT GAACVATATC CGTAACCRIG ATACTRMAAAC mxoxox3 ACCAATGTAA ATATh~rCAT GAACFATATC CGTAACCIXG ATAGTAAAC mxoxox4 ACCAATGTAA ATATTGTflAT GAACTATATC CGTAACCIX3G ATAGTGAAAC inxoxox5 KT)CCAATGrAA ATATLI-CAT GAACrATATC CGTAACCPJYG ATAGTGAAAC rnxoxox6 TCCAATGTAA ATATRTICAT GAACTATATC CGAACCTGG ATAGTAAAC .wcS AGGGGCAATG GTIXCGCCrGC TrGGAAGATGG CGA~rA-GCCA TITAACGCGrA 1,aax~ AGGGGCAATG GTGCGCCTGC TGGAAGATGG CGATTrAGCCA TTAACGCGrA rnoo2KAGGGGCAATG GTGXCGCCrGC TGGAAGATGG CGATrA*3CCA 'ITAACGcCGA ..inxcxox3 AGGGGCAATG GTGCGCCTGC TGGAAGAflJG CGA 7 PU'l-CCA 7TAACGCGTA : xo ,x o x4 AGGGGCAATG GTGCGC=rG TGGAAGATGG CGATrAGCCA TI7AACGCGT7A mxoxox5 -AGGGGCAATG GTGCGCCrGC TGGAAGATJG CGATTAOCCA 1TAACGCGrA mxoxox6 AGGGGCAATG GTGCGCCTGC TGGAAGATGG CGA7Tj-hGCCA TI'AACGCGTA 996 13 wtCre . AATGATAAGC TIX3jG7fTT TGGCGGATGA GAGAAGATIT TCAGCCTGAT mnxoxox 1 AATUATAAGC TIUGCTI TGGCGGATGA GAGAAGATTF TCAGCCTGJAT mxox ox2 AATGATAAGC T[GGCrtIT TGGCGGAICA GAGAAGATIT TC-AGCCTGAT rnxoxox3 AAIXJATAAGC TFGGCVT[ TGGCGGATGA GAGAAGATrT TCAGCCTGAT mxoxox4 fx1TJATAAGC TIYJGCIUIT TGGCGGKIXJA GAGAAGATr TICAGCCTGAT mxoxox5 AATGATAAGC TFUGCITr TGGCGGATGA GAGAAGA1TFF TCAGCCT)GAT mxoxox6 AA1XJATAAGC TTGGCTiT TGGCGGA1X3A GAGAAGAITI TCAGCCXZAT 1046 wtCre ACAGATFAAA TCAGAACGCA GA mxoxoxl ACAGATFAAA TCAGAACGCA GA -- Start Codon mxoxox2 ACAGAITFAAA T)CAGAACGCA G% MXOXOX3 ACAGATAAA TCAGAACGCA GA - -Stop Codon mxoxox4 ACAGA1TAAA TCAGAACGCA GA rnxoxox5 ACAGATAAA TCAGAACGCA GA T -Point Mutation mx ox ox6 ACAGTFAX-\ TCAGAACGCA GA 1096 101 WO 00/60091 PCT/USOO/091 54 In Vitro Evolution of the Cre recombinase wtcre MSNLLT-VHQN LPALPVDATS DEVRKNLMDM F7RDRQAF3EH TWKMLL jVCR mxoxox I MSNLLT-VHQN LPALPVDAT'S DEVRKNLMAM FRDRQAFSEH TWKM[LLSVCR mxoxox2 MSNLL-TVHQN LPALPVDAT'S DEVRKNLMDM FRDRQAE5EFH TWKMLWVCR mxoxox3 MSNLLTVHQN LPALPIDATS DEVRKNLMDM FRDRQAFSD-I TWKMLLSVCR mxoxox4 MSNLLT-VHQN LPALPVDAT'S DEVRKNLMDM FRDRQAFSEH TWKMLLSVCR mxoxox5 MSNLLTrVHQN LPALPVDATS DEVRKNLMAM FRDRQAE5EH TWKMLL-$VCR mxoxox6 MSNLLT-VHQN LPALPVDATS DEVRKNLMDM FRDRQAFSEH TWKMLISVCR A B wtCre SWAAWCKLNN RKWFPAEPED VRDYLLYLQA RGLAVKTIQ-Q [LGQU'4MU{R mxoxox I SWAAWCKLNN RKWFPAEPED VRDYLLYLQA RGLAVKTIQQ MLGQL.NMLH-R MXOXOX2 SWAAWCKLNN RKWrFPAEPED VRDYLLYLJQA RGLAVI IIQQ HLfGQWNMUIR mxoxox3 SWAAWCKLNN RKWFPAEPED VRDYLLYLQA RG!LAVIIQQ HLGQWNMLR mxoxox4 SWAAWCKLNN RKWFPAEPED VPDYLLCLJA RGLAVK11[QQ HLGQWNMWR mxoxox5 SWAAWCKLNN RKWFTPAEPED VRDYLLYLQA aGLAvKTIQQ [ZLGQLNMLE R mxoxox65 SWAAWCKLNN RKWrFPAEPED VRDYLLYLQA RGLAVKTIQQ HLGQLNMLR 5.1 C D wtCre RSGLPRPSDS NAVSLVMRR[ RKENVDAGER & QALARMT DFDQVRSLME xnxoxox 1. RSGLPRPSDS NAVSLVMRI B-ENVDAGER Ai QALAFEPT DFDQVRSLME 'jnxox2: RSGLPRPSDS NAVSLVMRRI RKENVDAGER AKQALAFGRT DFDQVRSLMEff rnxoxox3- BSGLPRPSDS NAVSLVMRRI RI ENVDAGER AKQALAFERT DFD)QVRSLME .moo4 RSGLPRPSDS NAVSLVMRRI RKENVDAGER PEQALAFKRT DFDQVRSLME jTnxoxox'5 QSGLPRPSDS NAVSLVMRR[ RKBNIVDAGER AKQALAFERT DFDQVRSLME .m xoxox 6 RSGLPRPSDS NAVSLVMRRI RKENVDAGER AKQALAFERT DFDQVRSLME 101 E F wtCre NSDRCQDIRN LAFLGIAYNT LLF.V4E[ARI RVKDISRTDG GRMLIHIGRT mxoxoxi NSDRCQDIRN LATLGLAYNT LLF.IAELARI RVKDISRTNG GRMLIHIGRT mxoxox2 NSDRCQDIRN LAFLGIAYNT LLF.LAE[APJ RVKDISRTDG GRMLIHIGRT mxoxox3 NSDRCQDIBN LAXL.GIAYNT LLRLAEIARI RVKDISRTNG GRMLIHISRT mxoxox4 NSDaCQDIBRN LAFLGLAYNT LLF.AEIARI RVKDlSRTDG GRMLIHIGRT mxoxox5 NSDBRCQDIRN LAFLG[AYNT LLF.LAEIARI RVKDLSRTDG GRMLIHIGRT mxoxox6 NSDRCQDIRN LAFLGIAYNT LLFAAEIARI RVKDISRTDG GRMLIHIGRT 151 Poo 10 G H 12 102 WO 00/60091 PCT/US00/09154 In Vitro Evolution of the Cre reconbinase wtCre KTLVSTAGVE KALSLGVTKL VERWISVSGV ADDPNNYLFC RVRKNGVAAP mxoxoxI KTLVSTAGVE KALSLGVTKL VERWISISGV ADDPNNYLFC RVRKNGVAAP mxoxox2 KTILVSTAGVE KALSLGVTKL VERWISVSGV ADDPNNYLFC RVRKNGVAAP mxoxox3 KILVSTAGVE KALSLGVTKL VEQWISVSGV ADDPNNYLFC RVRKNGVAAP mxoxox4 KITLVSTAGVE KALSLGVTKL VERW[SVSGV ADDPNNYLFC RVRKNGVAAP mxoxox5 KITLVSTAGVE KALSLGVTKQ VERWISVSGV ADDPNNYLFC RVRKNGVAAP mxoxox6 KTLVSTAGVE KALSLYTKL VERWISVSGV ADDPNNYLFC RVRKNGVAAP 201 No 00 3 I 4 5 wtCre SATSQLSIBA LEGIFEATHR UYGAKDDSG QRY LAWMGS A:VGAARDMA mxoxoxl SATSQLS'IBA LGGIFEATHR UYGAKDDSG QBYLAWSGHS A VGAARDMA mxoxox2 SATGQLaTBA LGGIFEATHR UYGAKDDSG QBYLAWSGHS A VGAARDMA mxoxox3 SATSRLSTBA LGGIFEATHR UYGAKDDSG QBYLAWSGHS AF VGAARDMA mxoxox4 SATSQLSIBA LEGIFEATHR UYGAKDDSG QBYQAWMHS A VGAARDMA mxoxox5 SATSQLaTBA LGGIFEATHR UYGAKDDSG QBLAWaGhS AFNGAARDMA mxoxox6 SATSQLSTBA LGGIFEATHR LYGAKDDSG QBYQAWSG-HS AFVGAARDMA 251 J K wtCre RAGVSIPEIM QAGGWTNVN1 VMN IRNLDS ETGAMVRLLE DGD mxoxoxl RAGVSIPEIM QAGGWTNVNI VMNBRNLDS ETGAMVRLLE DGD mxoxox2 RAGVSIPEIM QAGGWSNVNI VMNYIBNLDS ETGAMVRLLE DGD mxoxox3 RAGVSILEIM QAGGWfTNVNI VMNYIRNLDS EIUAMVRLLE DGD mxoxox4 RAGVSIPEIM QAGGWTNVNI VMNYIBNLDS ETGAMVRLLE DGD mxoxox5 RAGVSIPEIM QAGGWSNVNI VMN 'IBNLDS ETGAMVRLLE DGD mxoxox6 RAGVSIPEIM QAGGWSNVNI VMN IRNLDS ETGAMVRLLE DGD 301 -= L M N G - mutated residue 7 - catalytic residue F - DNA contacting residue - -helix - D sheet 103 WO 00/60091 PCT/USOO/09154 Transition Frequency Transversion Frequency A-G 29%(9) A.- C 6.5 %(2) G-A 26%(8) C-A < 3 % (0) C-T 6.5 %(2) A-T 9.7 %(3) T-C 6.5 %(2) T-A 9.7 %(3) C-G < 3 % (0) G-C < 3 % (0) G-T 6.5 %(2) T-G < 3 % (0) 31 point mutations could be identified in the six analyzed mutants. The mutagenic frequency can therefore be calculated as 0.5%. No frame-shift mutations, or more than one point mutation per codon of the cre sequence were found. The table classifies these mutations into the different types of transition and transversion events. Given are the frequencies and in parenthesis the actual numbers which were found. Interestingly, the A to G and vice versa transisiton occurred much more often than all other possible events. On the other hand, half of the possible transversions were not identified at all. 104 WO 00/60091 PCT/USO0/09154 Table 13 Round Induction % Kan R Nb. Kan R % White 0 4 h < 0.01% - 100% 1 4 h 0.16% 6 84% 2 4 h 0.56% 47 82% 3 2.5 h 0.2% 36 80% 4 2.5 h 4.6% 102 75% 105 WO 00/60091 PCT/USO0/09154 Table 14 % KanR Mutant IoxK22 IoXp2 wt <0.01% 94% R3M1 66% 98% R3M2 54% 98% R3M3 65% 91% R3M4 0.02% 93% R3M5 15% 91% R3M6 3% 94% 106 WO 00/60091 PCT/USO0/091 54 (D )0') r- x CO q C, 1~ C\ C'J C14 Co (N 0)0 (13(N C) 0NJ 1O 0) 0') C6( (6 i) L6 0 mo 'IT I T Co) 0 2- 0) r- Co NJ C CLa r- (NC -~ o(N 4 Co) Co ( (D3 .4-,4 ,< < - ,t - 0 ) 0(D N - O ( 0() LO to) (0 IT 0 0 0) CD (NI 0 (N CD 0 (j) 0 0) I, 00 0+ H- 9 4 ~ ~ ~ - 0 tr)() 0 U)C) E < 0.a C!) 0- St co 0Y) (NJ 'I 0 E- 0 (4 St S 0 S 00 4- U) St (0 co cli 6 < C'i c C) (1 - co Co Co Co Co ZJ (41o c () N Ln 0 Z) St Co) C 00-(1 N o Co C ' _4 co 0C~ (D ~ ~ U Zo ( D ( -0( co C .4107 WO 00/60091 PCT/USOO/091 54 a) CD E >-, 0 Z3 < 0 cz 0 ~( H +0 -1 -H -H H 0 ce 0q0C c 4o2 0 0 0. C N C: 0 =3 E 00 C\ 0 C) 0 -H + -H '- --- 4 T (0 CD C, .0l C) ) (D( 0) .E a o0 00 < (D C) <) w 5- a) 0) C, H) C D C D C m) .: C) C , ) C 0) -- - 0 0 C: a) 0) > -200 Q) 0 H *~ E. L ( a) ~ ~ ~ ~ l C( < 0 00D 0 t!0 m0 CA C C 0 M 00 WO 00/60091 PCT/USO0/091 54 CD D D D D D CD CD H H H H H H H Ln E E U) H H H H H H H E- (9 - ED pD p (9 ElCD E- ~ ~ E-E - -E-- E CD 0D C CD 0D C D C E- H H H E- H - H H Hp HD I'D ko H o Ci2 '-1' 10

Claims (63)

1. A method of identifying variant recombinases that mediate recombination at variant recombination sites, the method comprising, (a) bringing into contact a mutant recombinase, a first nucleic acid sequence comprising a first reporter gene and first and second recombination sites, wherein the first and second recombination sites are variant recombination sites, and a second nucleic acid sequence comprising a second reporter gene and third and fourth recombination sites, wherein the third and fourth recombination sites can be recombined by a non-mutant recombinase, (b) determining if recombination occurs between the first and second recombination sites, and determining if recombination occurs between the third and fourth recombination sites, wherein recombination between the first and second recombination sites indicates that the mutant recombinase is a variant recombinase that mediates recombination at variant recombination sites, wherein recombination between the third and fourth recombination sites indicates that the mutant recombinase retains the ability to mediate recombination at non-variant recombination sites.

2. The method of claim 1 wherein the recombination sites comprise recognition sequences and compatibility sequences, wherein the recognition sequences of the first and second recombination sites differ from the recognition sequences of the third and fourth recombination sites, wherein the compatibility sequences of the first and second recombination sites are sufficiently similar to allow recombination between the first and second recombination sites, and wherein the compatibility sequences of the third and fourth recombination sites are sufficiently similar to allow recombination between the third and fourth recombination sites, and wherein the compatibility sequences of the first and second recombination sites differ from the compatibility sequences of the third and fourth recombination 110 WO 00/60091 PCT/USO0/09154 sites such that neither the first nor the second recombination site can be recombined with either the third or the fourth recombination site.

3. The method of claim I wherein the first and second recombination sites cannot be recombined by non-mutant recombinase to a significant extent.

4. The method of claim 1 or 2 wherein the first and second recombination sites have identical sequences, and wherein the third and fourth recombination sites have identical sequences.

5. The method of claim I wherein recombination between the first and second recombination sites alters the expression of the first reporter gene, wherein recombination between the first and second recombination sites is determined by determining if expression of the first reporter gene is altered, and wherein recombination between the third and fourth recombination sites alters the expression of the second reporter gene, wherein recombination between the third and fourth recombination sites is determined by determining if expression of the second reporter gene is altered.

6. The method of claim 5 wherein recombination between the first and second recombination sites allows the first reporter gene to be expressed.

7. The method of claim 6 wherein the first nucleic acid sequence further comprises a spacer sequence flanked by the first and second recombination sites, wherein the spacer sequence interrupts the first reporter gene such that the first reporter gene is not expressed, wherein recombination of the first and second recombination sites excises the spacer sequence which allows the first reporter gene to be expressed.

8. The method of claim 6 wherein a portion of the first reporter gene is inverted, wherein the inverted portion of the first reporter gene is flanked by the first and second recombination sites, wherein recombination of the first and second recombination sites inverts the inverted portion of the first reporter gene which allows the first reporter gene to be expressed.

9. The method of claim 5 wherein recombination between the first and second recombination sites prevents expression of the first reporter gene.

10. The method of claim 9 wherein the first reporter gene is flanked by the first and second recombination sites, wherein recombination of the first and second 111 WO 00/60091 PCT/USOO/09154 recombination sites excises the first reporter gene which prevents expression of the first reporter gene.

11. The method of claim 9 wherein a portion of the first reporter gene is flanked by the first and second recombination sites, wherein recombination of the first and second recombination sites inverts the flanked portion of the first reporter gene which prevents expression of the first reporter gene.

12. The method of claim 5 wherein recombination between the third and fourth recombination sites allows the second reporter gene to be expressed.

13. The method of claim 12 wherein the second nucleic acid sequence further comprises a spacer sequence flanked by the third and fourth recombination sites, wherein the spacer sequence interrupts the second reporter gene such that the second reporter gene is not expressed, wherein recombination of the third and fourth recombination sites excises the spacer sequence which allows the second reporter gene to be expressed.

14. The method of claim 13 wherein the spacer sequence interrupts the second reporter gene such that the second reporter gene is not transcribed.

15. The method of claim 13 wherein the second reporter gene encodes a protein, wherein the spacer sequence interrupts the second reporter gene such that the protein encoded by the second reporter gene is not translated.

16. The method of claim 13 wherein the spacer sequence interrupts the second reporter gene such that the second reporter gene produces an inactive expression product.

17. The method of claim 12 wherein a portion of the second reporter gene is inverted, wherein the inverted portion of the second reporter gene is flanked by the third and fourth recombination sites, wherein recombination of the third and fourth recombination sites inverts the inverted portion of the second reporter gene which allows the second reporter gene to be expressed.

18. The method of claim 5 wherein recombination between the third and fourth recombination sites prevents expression of the second reporter gene to be expressed.

19. The method of claim 18 wherein the second reporter gene is flanked by the third and fourth recombination sites, wherein recombination of the third and 112 WO 00/60091 PCT/USOO/09154 fourth recombination sites excises the second reporter gene which prevents expression of the second reporter gene.

20. The method of claim 18 wherein a portion of the second reporter gene is flanked by the third and fourth recombination sites, wherein recombination of the third and fourth recombination sites inverts the flanked portion of the second reporter gene which prevents expression of the second reporter gene.

21. The method of claim 1 wherein the first nucleic acid sequence is a first nucleic acid construct and the second nucleic acid sequence is on a second nucleic acid construct.

22. The method of claim 21 wherein the first nucleic acid construct is an extrachromosomal vector and the second nucleic acid construct is in the genome of a host cell.

23. The method of claim 1 wherein the first and second nucleic acid constructs are on the same nucleic acid construct.

24. A method for producing site-specific recombination of DNA, comprising, contacting a variant recombinase identified by the method of claim 1 with first and second DNA sequences, wherein the first DNA sequence comprises a first recombination site and the second DNA sequence comprises a second recombination site, wherein the variant recombinase mediates recombination between the first and second recombination sites thereby producing the site specific recombination.

25. The method of claim 24 wherein the first recombination site, the second recombination site, or both, are variant recombination sites.

26. The method of claim 24, wherein the first and second DNA sequences are connected by a pre-selected DNA segment.

27. The method of claim 26, wherein the first and second recombination sites have the same orientation and the site-specific recombination of DNA is a deletion of the pre-selected DNA segment.

28. The method of claim 27, wherein the pre-selected DNA segment is a gene for a structural protein, an enzyme, or a regulatory molecule.

29. The method of claim 27 further comprising contacting the variant recombinase with a fourth DNA sequence comprising a third recombination site, 113 WO 00/60091 PCT/US0O/09154 wherein the second and fourth DNA sequences are connected by a second pre selected DNA segment.

30. The method of claim 29 wherein the first recombination site is a variant recombination site recognized by the variant recombinase and not by wild type recombinase, and wherein the second and third recombination sites are recombination sites recognized by wild type recombinase and by the variant recombinase.

31. The method of claim 30 further comprising, prior to contacting the variant recombinase with the first, second, and third recombination sites, contacting the recombination sites with wild type recombinase, thereby producing site specific recombination between the second and third recombination sites resulting in a deletion of the second pre-selected DNA segment.

32. The method of claim 29, wherein the second pre-selected DNA segment is a gene for a structural protein, an enzyme, or a regulatory molecule.

33. The method of claim 26, wherein the first and second recombination sites have opposite orientations and the site-specific recombination is an inversion of the nucleotide sequence of the pre-selected DNA segment.

34. The method of claim 33, wherein the first and second recombination sites are variant recombination sites recognized by the variant recombinase.

35. The method of claim 33, wherein the pre-selected DNA segment is a gene for a structural protein, an enzyme, or a regulatory molecule.

36. The method of claim 24, wherein the second and third DNA sequences are introduced into two different DNA molecules and the site-specific recombination is a reciprocal exchange of DNA segments proximate to the recombination sites.

37. The method of claim 36, wherein the first and second recombination sites are variant recombination sites recognized by the variant recombinase.

38. The method of claim 24 wherein the second DNA sequence includes a label, wherein recombination between the first and second recombination sites associates the label with the first DNA sequence.

39. The method of claim 38 wherein the first DNA sequence is a large circular DNA molecule.

40. The method of claim 24 wherein recombination occurs in a cell. 114 WO 00/60091 PCT/USO0/09154

41. The method of claim 40 wherein the variant recombinase is contacted with the first and second DNA sequences by introducing into the cell a third DNA sequence comprising DNA encoding the variant recombinase.

42. The method of claim 41, wherein the third DNA sequence further comprises a regulatory nucleotide sequence and expression of the variant recombinase is produced by activating the regulatory nucleotide sequence.

43. The method of claim 40, wherein the cell is a eukaryotic cell, a mammalian cell, a yeast cell, a fungal cell, a prokaryotic cell, a bacterial cell, an archae bacterial cell, or a cell in a multicellular organism.

44. The method of claim 43 wherein the multicellular organism is a plant, an animal, or a mammal.

45. The method of claim 40, wherein the first and second DNA sequences are connected by a pre-selected DNA segment, wherein the first and second recombination sites have the same orientation and the site-specific recombination of DNA is a deletion of the pre-selected DNA segment.

46. The method of claim 45 wherein the cell is in a multicellular organism.

47. The method of claim 45, wherein the pre-selected segment is an undesired marker or trait gene.

48. The method of claim 24 wherein the variant recombinase is contacted with the recombination sites in vitro.

49. The method of claim 48 wherein the method further comprises introducing the recombined DNA into a cell.

50. A method for cloning large DNA fragments, the method comprising concatenating DNA fragments to be cloned with vector arms, wherein each vector arm comprises a recombination site, wherein the DNA fragments and vector arms alternate in the concatemer, introducing the concatemer into a cell expressing a variant recombinase identified by the method of claim 1, wherein the recombinase mediates recombination of the recombination sites thereby generating circles each containing a DNA fragment and a vector arm.

51. A method for cloning large DNA fragments, the method comprising ligating a DNA fragment to be cloned to vector arms, wherein each vector arm comprises (i) a blunt end, (ii) another end which is compatible with an end of 115 WO 00/60091 PCT/USOO/09154 the DNA fragment to be cloned, and (iii) a recombination site, wherein concatemers are not formed, and introducing the ligated DNA fragment and vector arms into a cell expressing a variant recombinase identified by the method of claim 1.

52. A method for cloning large DNA fragments, the method comprising ligating a plurality of DNA fragments to be cloned with a plurality of first and second vector arms, wherein each first vector arm comprises two ligatable ends, wherein each second vector arm comprises a recombination site and one non ligatable end, wherein, following ligation, the DNA fragments and first vector arms alternate in concatemers, wherein the concatemers are flanked by second vector arms, introducing the concatemers into a cell expressing a variant recombinase identified by the method of claim 1, wherein the recombinase mediates recombination of the recombination sites thereby generating circles containing the DNA fragments.

53. A variant recombinase identified by the method of claim 1.

54. A nucleic acid molecule encoding a variant recombinase identified by the method of claim 1.

55. The nucleic acid molecule of claim 54 wherein the nucleic acid molecule is a plasmid.

56. A cell containing the nucleic acid molecule of claim 54.

57. The cell of claim 56 wherein the cell is a eukaryotic cell, a mammalian cell, a yeast cell, a fungal cell, a prokaryotic cell, a bacterial cell, an archae bacterial cell, or a cell in a multicellular organism.

58. The cell of claim 57 wherein the multicellular organism is a plant, an animal, or a mammal.

59. A nucleic acid molecule having at least one variant recombination site, wherein the variant recombination site is recognized by a variant recombinase identified by the method of claim 1 and is not recognized by wild type recombinase.

60. The nucleic acid molecule of claim 59 wherein the recombination site is not recognized by wild type recombinase. 116 WO 00/60091 PCT/USO0/09154

61. The nucleic acid molecule of claim 59, wherein a first recombination site and a second recombination site are connected by a pre-selected DNA segment.

62. The nucleic acid molecule of claim 59 wherein the nucleic acid molecule is a plasmid.

63. A cell containing the nucleic acid molecule of claim 59. 117