Identification of Cre Residues Involved in Synapsis, Isomerization, and Catalysis* □ S

The Cre protein of bacteriophage P1 is a tyrosine recombinase and catalyzes recombination via formation of a covalent protein-DNA complex and a Holliday junction intermediate. Several co-crystal structures of Cre bound to its target lox site have provided novel insights into its biochemical activities. We have used these structures to guide the mutagenesis of several Cre residues that contact the lox spacer region and/or are involved in intersubunit protein-protein interactions. None of the mutant proteins had significant defects in DNA binding, DNA bending, or strand-specific initiation of recombination. We have identified novel functions of several amino acids that are involved in three aspects of the Cre reaction. 1) Single mutation of several NH 2 -terminal ba- sic residues that contact the spacer region of loxP caused the accumulation of Holliday junction (HJ) intermediates but only a modest impairment of recombination. These residues may be involved in the isomerization of the Holliday intermediate. 2) We identified three new residues (Arg-118, Lys-122, and Glu-129) that are involved in synapsis. Cre R118A, K122A, and E129Q were catalytically competent. 3) Mutations E129R, Q133H, and K201A inactivated catalysis by the protein. The function of these Cre residues in recombination is discussed. The

The phage P1 Cre recombinase is a member of the tyrosine recombinase (integrase) family of conservative site-specific recombinases whose members use a conserved tyrosine as a key catalytic nucleophile (1)(2)(3)(4)(5)(6)(7). The publication of several crystal structures has provided remarkable insights into the mechanisms by which this recombinase carries out recombination (8 -13). Cre is of interest not only for the wealth of biochemical and structural information it has offered into the recombination mechanisms, but also because it is one of the most useful tools for engineering mammalian chromosomes in the study of development (14).
The Cre protein catalyzes recombination at its target sequence, the loxP site (Fig. 1a). This 34-bp sequence consists of two identical 13-bp inverted repeats (symmetry elements) that surround an asymmetrical 8-bp spacer region (15). The Cre protein binds specifically to each symmetry element and induces DNA bending (10, 16 -19). The cleavage sites are separated by 6 bp within the spacer region (20). Cleavage occurs by covalent attachment of the protein to the 3Ј-phosphoryl group at the site of the nick via the conserved catalytic Tyr-324 (12,20). We refer to the nucleotides that are immediately 5Ј of the cleavage sites as the scissile nucleotides. The bottom strands are cleaved and exchanged first to form a four-armed Holliday (HJ) 1 intermediate that is then resolved on the top strands to form two reciprocally recombinant molecules (17,(21)(22)(23)(24).
Cre folds into a two domain structure (Fig. 1b): the small NH 2 -terminal domain interacts with the inner portion of the symmetry element and the spacer region, while the COOHterminal catalytic domain contacts the entire symmetry element (8 -13). The two Cre subunits bound to the lox site in slightly different conformations: the "cleaving" subunit is in position to cleave the scissile phosphate, whereas the "noncleaving" subunit is in an inactive mode (Fig. 1c). While the cleaving and non-cleaving Cre subunits make similar contacts with the symmetry elements, they interact differently with the spacer region ( Fig. 1d) (10). In particular, several NH 2 -terminal basic residues in the cleaving subunit directly contact the DNA phosphate backbone opposite the activated scissile phosphate. As a result, the continuous strand (containing the inactive scissile phosphate) is tightly bound in the Cre-DNA interface, while the solvent-exposed crossing strand (containing the activated scissile phosphate) points toward the center cavity of the synapse poised for strand exchange after cleavage (10).
Recombination takes place within a roughly square-planar synaptic complex consisting of four Cre molecules and two lox sites ( Fig. 1c) (8 -13). The Cre-lox synaptic complex is stabilized by an intricate network of cyclic protein-protein interactions between the Cre molecules bound to the same lox site ("crossspacer" interactions) and those bound to two different lox sites ("synaptic" interactions). We hereafter refer to the cross-spacer/synaptic protein-protein interactions as simply "intersubunit" interactions, unless otherwise stated. These intersubunit interactions consist of: 1) an NH 2 -terminal interface primarily between helices A and E, 2) interaction between helix E and the ␤-loop between ␤2 and ␤3 strands (␤2/␤3-loop), and 3) the burying of the COOH-terminal helix N in a hydrophobic pocket of the adjacent subunit (8,12). The flexible ␤2/␤3-loop contains the conserved catalytic Lys-201 residue (Fig. 1b) (6,12,25).
In this article we studied the functions of the Cre residues that contact the lox spacer region and/or are involved in intersubunit protein-protein interactions. We find that the mutations of these residues have variable effects on recombination and have divided them into three groups (Table I Holliday intermediates, 2) those that fail to carry out synapsis, and 3) those that are catalytically inactive.

EXPERIMENTAL PROCEDURES
Construction of Cre Mutations-The construction of the His-tagged mutant cre genes is described in Tables S1 and S2 (see "Supplementary Data"). The Q133H mutation was discovered as a secondary mutation during the construction of Cre R101A, and was subsequently cloned as a single mutation (see Table S2). The construction of the non-Histagged Cre A36V was described previously (22). The Cre K201A expression plasmid was a gift from Dr. Greg Van Duyne, University of Pennsylvania.
The mutant Cre proteins were expressed in Escherichia coli BL21 (DE3 pLysS) and purified essentially as described (26). Analysis on SDS-PAGE showed that the proteins were greater than 90% pure. Protein concentration was determined using the Bradford assay (27) with IgG as protein standard (Bio-Rad). The purified proteins were stored at Ϫ70°C.

RESULTS
We are interested in studying the regulation of the order of strand exchange by Cre. We have previously reported that the order of strand exchange and the position of the Cre-induced bends are dictated primarily by the scissile base pairs (Fig. 1a) (19,24,28). Lys-86, which contacts the scissile nucleotides ( Fig.  1d) (8,10,13,19), was found to contribute to the strand specificity in the Cre reaction, but it was not the key determinant of the site of initiation (19,24,28). To identify other amino acids that may regulate the Cre reaction, we changed additional Cre residues that interact with the lox spacer region and/or are involved in intersubunit interactions. Most of the Cre mutant proteins we studied were fused to an NH 2 -terminal His 10 -tag. We showed previously that the His wild-type Cre protein is proficient in recombination and behaves similarly to the untagged wildtype protein in all the assays described here (19,24,28).
The Cre mutant proteins were tested at the same concentration in various aspects of the recombination reactions. All the proteins examined were able to bind to both the loxP site (Fig.  2a) and the lox4 site (Fig. 2b), suggesting that the mutations did not drastically disrupt the overall protein fold. However, we cannot exclude subtle differences in protein stability or conformation that may have contributed to some of the phenotypes. The phenotypes of the Cre variants are summarized in Tables  I and II. The proteins were categorized into three groups based on their ability to synapse, cleave DNA, and form and resolve Holliday junctions. The scissile A and G (bold type) nucleotides are the nucleotides immediately 5Ј to the cleavage sites (vertical arrows, the numbers indicate the order of strand exchange). In the mutated lox4 site, the scissile base pairs are interchanged relative to the wild-type loxP site. b, the tertiary fold of Cre (12). Cre folds into two domains: a small NH 2 -terminal domain (helices A-E) and a COOH-terminal catalytic domain (helices F-N). The active site residues are in black and the Cre residues studied in this paper are in blue. c, schematic diagram of the Cre-lox synaptic complex. The Cre subunits are drawn as colored ovals: non-cleaving (A, AЈ), cleaving (B, B'). The light blue arrows designate the orientation of the loxP site. The crossing strands (thick lines) contain the activated scissile phosphate (green circles), whereas the continuous strands (thin lines) contain the inactive scissile phosphate (magenta circles). The R118-and K122-DNA interactions and the R118-A36 intersubunit interactions occur within a lox site (cross-spacer; green and blue dashed arrows) and between two lox sites in a synaptic complex (synaptic; magenta and red dashed arrows). The boxed section represents the region shown in panel d. d, the Cre contacts with the lox spacer region observed in the Cre R173K/loxS (4CRX) structure (10). The distances between non-hydrogen atoms involved in the contacts shown are less than 3.5 Å. Similar contacts are seen in the other Cre-lox structures. Note that Arg-118 and Lys-122 interact with the distal end of the spacer region from where the Cre molecule is bound (also see panel a). For simplicity, we have excluded most water-(blue dots) mediated contacts (for more details, see Guo et al.,Ref. 10). Also shown are intersubunit protein-protein interactions (7) involving the residues studied in this article. The Cre residues are colored as in a, magenta, non-cleaving subunit A; red, AЈ; green, B. The conserved catalytic residues are indicated with asterisks.
Group I Proteins (R100A, R101A, R106A, N111A, and R121A) Are Partially Impaired in Recombination and Accumulate Holliday Intermediates-A cluster of NH 2 -terminal arginine residues (Arg-100, Arg-101, Arg-106, and Arg-121) in the cleaving Cre subunit, but not in the non-cleaving Cre subunit, directly contact the phosphate backbone of the continuous strand opposite the activated scissile phosphate in the lox spacer region (Fig. 1, b-d) (10). To examine the functions of these asymmetric DNA contacts, these residues were mutated to Ala.
Cre R100A, R101A, R106A, and R121A were moderately defective in recombination, strand cleavage and resolution, and are classified together in Group I (Table I). These proteins bound efficiently to loxP and formed higher order (HO) synaptic complex (Fig. 2a). However, the proteins caused the central region of the loxP spacer to become sensitive to OP-Cu, similar to the footprint induced by Cre K86A (data not shown) (19). The enhanced sensitivity to OP-Cu suggests that these Cre mutant proteins may bind the spacer region with slightly lower affinity than wild-type Cre due to the disruption of the DNA contacts. Recombination between loxP sites ranges from a 20% reduction in efficiency for R100A to an 80% reduction for R101A (Fig. 3,  a and b). Similar results were obtained for excisive recombination between directly oriented loxP sites (data not shown). Nevertheless, the proteins were proficient in generating Holliday intermediates and in fact, three of them (R101A, R106A, and R121A) accumulated twice the amount of Holliday intermediates as the wild-type Cre protein. All four of these mutant proteins were also able to cleave loxP linear suicide substrates (Fig. 4) and resolve synthetic loxP Holliday structure ( Fig. 5) with less than 2-fold reduction in efficiency. Note that Cre R121A actually enhanced cleavage on both strands of loxP by about 2-fold relative to the wild-type Cre protein.
Cre R101A was the most defective in recombination (5-fold reduction) among the Group I Cre mutant proteins (Fig. 3). Arg-101 is involved not only in a protein-DNA interaction, but also in a cross-spacer intersubunit interaction with Asn-111 ( Fig. 1d) (8 -13). The side chain of Arg-101 in the cleaving Cre subunit contacts the spacer region, but in the non-cleaving Cre subunit it forms a hydrogen bond with Asn-111 (OD1) in the adjacent cleaving Cre subunit. The mutant phenotype of Cre R101A could result from a disruption of the protein-DNA and/or protein-protein interaction. To distinguish between these two possible functions of Arg-101, we mutated Asn-111 to Ala (N111A). We found that Cre N111A can bind to loxP (Fig.  2, lane 7) and was less than 2-fold impaired in recombination ( Fig. 3), placing it with the Group I mutant proteins. The N111A mutation was less deleterious than the R101A mutation (Figs. [3][4][5], demonstrating that the more severe R101A phenotype is not due solely to the disruption of the Arg-101-N111A interaction, and that the interaction of Arg-101 with the DNA is also important for function. We have shown previously that the wild-type Cre protein initiates recombination adjacent to the scissile G nucleotide, but preferentially resolves HJ and cleaves linear suicide substrates adjacent to the scissile A nucleotide (24,28). To determine whether the asymmetric protein-DNA interactions in the lox spacer region contribute to the order of strand exchange and asymmetric cleavage, we isolated the Holliday intermediates generated in the recombination reaction and analyzed them by denaturing PAGE (Fig. 3c). Like those generated by the wildtype Cre protein, the Holliday intermediates generated by all five Group I Cre mutant proteins had exchanged the bottom strands of the loxP site adjacent to the scissile G nucleotide (Fig. 3c). These Cre mutant proteins also preferentially cleaved linear suicide substrates and resolved Holliday structures on the top strand of loxP (Figs. 4 and 5). Therefore, these NH 2terminal arginine residues (and presumably their asymmetric interactions with the spacer region) and Asn-111 do not dictate the strand preference during strand cleavage and exchange in the loxP sites.
We previously found that the order of strand exchange catalyzed by wild-type Cre is reversed when the position of the scissile base pairs is interchanged as in the lox4 site ( Fig. 1a; compare Figs. 3c and 6c, lane 3) (24,28). We therefore examined the activity of the Cre variant proteins on the lox4 site. The Group I Cre mutant proteins bound efficiently to the lox4 site (Fig. 2b). Like the wild-type protein, the Group I Cre mutant proteins generally recombined the lox4 sites less efficiently and accumulated slightly more Holliday intermediates than they did with the loxP sites (compare Figs. 3 and 6, a and b). Interestingly, Cre R101A and N111A formed two lox4 Holliday species (Fig. 6a, lanes 5 and 7); the major 1 species co-migrated with those generated by the other Cre proteins, while the novel faster migrating minor species (2) may repre- lox4 site was incubated with 0.25 M of the indicated Cre protein at room temperature for 30 min as described (24). At this time point, the reactions had reached equilibrium (data not shown). The reaction was then analyzed on a 6% native polyacrylamide gel. S, unbound 82 bp substrate; complexes cI and cII correspond respectively to one and two Cre molecules bound to a single lox site (18); HO (higher order), a synaptic complex of two Cre dimers and two lox sites (17). The upper panels show ϳ10-fold longer exposure of the autoradiogram to better reveal the HO complexes. Lanes 14 and 15 are from a separate gel than lanes 1-13 in this and subsequent figures. sent an alternate HJ isomer. The minor 2 species was also formed by Cre R101A, R106A, N111A, and R121A in the loxP recombination reaction, but at substantially lower amounts than with the lox4 substrates (Fig. 3a, upper panel). The Cre residues in Group I may be involved in the isomerization of the Holliday intermediate (see "Discussion").
Arg-100, Arg-101, Arg-106, Asn-111, and Arg-121 do not influence the order of strand exchange in the lox4 site. Like wild-type Cre, these mutant proteins predominantly initiated strand exchange adjacent to the scissile G nucleotide on the top strand of lox4 (Fig. 6c). They also preferentially cleaved linear suicide substrates (Fig. 7a) and resolved the 4 synthetic Holliday structure (Fig. 7b) near the scissile A nucleotide on the bottom strand of lox4. In conclusion, as with the wild-type Cre protein, the scissile base pairs primarily dictate the order of strand exchange by the Group I Cre mutants.
Group II Cre Proteins (R118A, K122A, E129Q) Are Synapsisdefective-While the NH 2 -terminal Arg residues in the Group I proteins interact with the half of the spacer region adjacent to the bound Cre molecule, we also noticed that the side chains of Arg-118 and Lys-122 in helix E of Cre are positioned across the spacer region and interact with the distal end of the spacer region ("cross-spacer" interactions, Fig. 1, c and d) (8 -10). The latter protein-DNA interactions are also seen within a synaptic complex between a Cre molecule bound to one lox site and the spacer region of an adjacent lox site ("synaptic" interactions). These Arg-118-and Lys-122-DNA interactions form a cyclic network of cross-spacer and synaptic protein-DNA interac- a N, no tag; Y, has an NH 2 -terminal His 10 -tag. b Binding assayed by EMSA (see Fig. 2). cI and cII, complexes I and II. c HO, higher order/ synaptic complex. d Overall efficiency of cleavage on the top strand (CS-Top) and bottom strand (CS-Bottom) suicide substrates (24). e 182 bp ϫ 82 bp intermolecular recombination reactions (24). Similar results were obtained using the excision substrates pRH43 (21); and ReconII (22) (data not shown).
f The resolution of synthetic P (28). g L min , the linker length at minimum relative mobility in the phasing analysis of loxP in the forward orientation (19). Minimum relative mobility is reached when the Cre-induced bend and the sequence-directed A-tract bend are in-phase (in the same direction) (19).
tions. Arg118 is also involved in an intersubunit interaction with Ala-36 in the loop following helix A of the adjacent Cre subunit (8,12). Ala-36 has been found to be required for syn-apsis since mutation to Val (A36V) resulted in defects in the formation of higher order complexes, recombination and strand cleavage but not in the resolution Holliday junctions (17, 21,  (24). Following treatment with 0.1% (w/v) SDS and 0.05 mg/ml proteinase K, the reactions were analyzed on a 5% native PAGE. The recombination reaction scheme is illustrated to the left of the autoradiogram (also see Ref. 24). R1 and R2, 149-bp and 115-bp recombinant products. Some of the Cre mutants form two Holliday intermediates: the major 1 species co-migrated with the generated by the wild-type protein, while the minor 2 species may represent an alternate HJ isomer (see "Discussion"). The 2 is more pronounced with the lox4 substrates (see Fig. 6a). The upper panel shows ϳ8-fold longer exposure of the autoradiogram to better reveal the 2 species. b, graph of the percent of recombinant products (R1 ϩ R2; solid bars) and 1 intermediate (striped bars) resulting from loxP x loxP recombination. The Cre proteins are indicated below each bar graph. The level of activity for the His-tagged wild-type ( His Wt) Cre protein is similar to untagged Wt protein (24). In this and subsequent Figs. 4 -7, each reaction was performed at least three times, and the error bars indicate the S.D. c, analysis of the purified species by denaturing PAGE. The intermediates from a preparative loxP x loxP recombination reaction (a) were gel-purified and heat-denatured in formamide prior to loading on a 5% denaturing PAGE (24).  We found that Cre R118A and K122A exhibited similar mutant phenotypes to Cre A36V. They bound efficiently to loxP, but failed to form stable HO complexes that are the presumed synaptic complexes of two loxP sites (17) (Fig. 2a). This suggests that Arg-118 and Lys-122 also contribute to stabilization of the synaptic complex. In addition, Cre R118A and K122A were severely defective in recombining loxP sites and forming Holliday intermediates (Fig. 3). Cre R118A and K122A cleaved linear suicide substrates about 8-fold and 3-fold, respectively, less efficiently than the wild-type Cre protein (Fig. 4). The severe cleavage defect of R118A and A36V relative to K122A implies that the Arg-118 -Ala-36 interaction is likely more important than the interactions made by Lys-122. Like the wild-type Cre protein, both Cre R118A and K122A cleaved the top strand of loxP more efficiently than the bottom strand. The inability of Cre R118A and K122A to mediate recombination and strand cleavage is not due to a defect in catalysis per se, since they both resolved Holliday structures efficiently (Fig. 5) and could recombine lox4 sites (see below). Like wild-type Cre, these synapsis-defective Cre mutants resolved the Holliday junction preferentially on the top strands of loxP. In fact, the bias for Cre A36V in favor of top strand resolution was even more pronounced than the wild-type Cre protein. Therefore, Ala-36, Arg-118, and Lys-122 do not regulate the strand preference during strand cleavage and resolution of loxP. We have also mutated Arg-118 to Gln, Ser, Val, and Trp, and these Cre Arg-118 mutant proteins are also defective in synapsis (data not shown).
Although Cre R118A and K122A could not recombine loxP sites, they were able to recombine lox4 sites (Fig. 6, a and b). The R118A and K122A mutations reduced recombination between lox4 sites only by 2-fold or less compared with the wildtype Cre protein. However, Cre A36V was unable to recombine lox4 sites. Cre R118A and K122A were able to produce some lox4 Holliday intermediates (Fig. 6, a and b) and analysis of these intermediates showed that, like wild-type Cre, the top strands of lox4 had been predominantly exchanged (Fig. 6c). It is unclear how interchanging the scissile base pairs partially suppressed the recombination defect of Cre R118A and K122A, since these proteins still failed to form stable HO complexes with lox4 (Fig. 2b) and cleaved lox4 poorly (Fig. 7a). All three Group II Cre mutants efficiently resolved the 4 Holliday structure (Fig. 7b). Interestingly, the R118A and K122A mutations essentially abolished the strand bias in the resolution of 4 though not of p (compare Figs. 5 and 7b).
Glu-129 is located in the long linker connecting the NH 2terminal and COOH-terminal domains of Cre (E/F-linker; Fig.  1b). In the 3CRX structure (9), Glu-129 contacts the catalytic Lys-201 residue that is located in the ␤2/␤3-loop of the adjacent Cre subunit (Fig. 1d). Substitution of Glu-129 by Gln (E129Q) impaired the formation of HO complex, the initiation of recombination and cleavage of linear suicide substrates (Figs. 2-4,  lane 14). Nonetheless, the E129Q protein is proficient in resolution of synthetic Holliday structures (Fig. 5, lane 14). Therefore, Glu-129 is also important for synapsis. The E129R mutation was more detrimental to recombination than the conservative E129Q mutation (Figs. 2-5, lane 15). As discussed in the section below, Cre E129R is catalytically inactive (Group III) in addition to being defective in synapsis. Another Glu-129 mutation, E129P, is also impaired in synapsis and catalysis like E129R (data not shown).
Group III Proteins (E129R, Q133H, and K201A) Are Defective in Catalysis-The Group III proteins are catalytically inactive, being dramatically defective (Ͼ20-fold reduction) in all the enzymatic reactions assayed including recombination, cleavage, formation, and resolution of Holliday intermediates for both the loxP and lox4 sites (Tables I and II; Figs. 3-7). In addition to E129P/R mentioned above, the Q133H and K201A mutations also disrupt catalysis. Like Glu-129, Gln-133 is located in the E/F-linker of Cre (Fig. 1b). The side chain of Gln-133 is positioned close to the active site (see "Discussion"). Unlike the Glu-129 mutant proteins, however, Cre Q133H formed some higher order complexes (Fig. 2, lane 11). Possible roles for Glu-129 and Gln-133 in catalysis are proposed in the "Discussion." Lys-201, which is located in the ␤2/␤3-loop, is conserved in the Int family members (6,25). Guo et al. (10) have also found that Lys-201 is important for catalysis. We found that Cre K201A was also defective in synapsis (Fig. 2,  lane 12), possibly due to disruption of intersubunit interactions involving the ␤2/␤3-loop (see "Discussion") (8,12).
Lys201 of the cleaving subunit, but not the non-cleaving subunit, makes a minor groove contact with the scissile nucleotide adjacent to the activated scissile phosphate (Fig. 1d) (8 -13). We therefore examined whether Lys201 also contributes to distinguishing the scissile base pairs during the resolution of Holliday structures (Fig. 8). Although this protein gave a very low level of Holliday resolution, we found that Cre K201A exhibited similar strand bias to the wild-type Cre protein in the resolution of the loxP Holliday structure p (Fig. 5a, lane 12 and Fig. 8, lane 9). Preferential resolution by Cre K201A on the top strands of the loxP HJ was also observed for the rev structure in which the orientation of the loxP site was inverted relative to that in the p structure (Fig. 8, lane 10). The strand bias was essentially abolished in symmetric lox Holliday structures such as SA (Fig. 8, lane 11). However, unlike wild-type Cre, interchanging the scissile base pairs did not reverse the strand bias in the resolution of lox4 Holliday structure ( Fig. 7b; Fig. 8, compare lanes 8 and 12). Cre K201A resolved both loxP and lox4 preferentially on the top strands, suggesting that the scissile base pairs do not determine the strand preference in Holliday resolution by Cre K201A. Because of the extremely low efficiency of resolution by Cre K201A, it is difficult to assess the significance of the observed bias in resolution.
The Cre Mutant Proteins Do Not Significantly Alter DNA Bending-The Cre-lox crystal structure revealed that Cre induced an asymmetric DNA bend in the lox site (10). We have confirmed this finding using phasing and circular permutation analyses (19). The Cre-induced asymmetric bend is thought to be due to cross-spacer protein-protein interactions between the Cre subunits (10). Alternatively, the asymmetric protein-DNA interactions in the spacer region may contribute to DNA bending due to asymmetric neutralization of the phosphate charges of the DNA backbone (30 -32).
To examine whether any of the Cre residues studied here contributes to DNA bending, we analyzed the DNA bends in- duced by the Cre mutant proteins using phasing analysis (19,33). The results for loxP and lox4 are summarized in Tables I  and II. We detected minor differences in DNA bending when only one molecule of Cre R121A, R118A, K122A, or Q133H is bound to the loxP site (cI bend). However, because of the small magnitude of the cI bend, it was difficult to assess the significance of these minor changes. None of the Cre mutations dramatically altered the DNA bends in the loxP and lox4 sites when two Cre molecules are bound (cII bend), and this is consistent with the fact that the site of initial strand exchange remained unchanged (Figs. 3c and 7c). The Cre mutant proteins also induced similar DMS methylation protection patterns to the wild-type Cre protein for loxP (data not shown) (19). DISCUSSION We have characterized several Cre residues that interact with the lox spacer region or are involved in intersubunit protein-protein interactions. Although the majority of the Cre mutations did not drastically alter DNA binding, DNA bending or the strand preference, some of them did affect certain aspects of the recombination reactions (Table I). We present in Fig. 9 possible roles for the Cre residues in recombination.
Group I Residues May Promote Isomerization of the Holliday Intermediate-The Group I Cre mutants (R100A, R101A, R106A, N111A, and R121A) all formed Holliday intermediates efficiently, and in fact all except R100A accumulated more HJ than wild-type Cre. Arginines 100, 101, 106, and 121 interact asymmetrically with the phosphate backbone of the continuous (non-cleaved) strand in the lox spacer region (Fig. 1d). In addition, Arg-101 and Asn-111 are involved in a cross-spacer intersubunit interaction. Both Cre R101A and N111A accumulated small amounts of a novel species (2) that migrated slightly faster than the predominant 1 species (Figs. 3a and  6a). We propose that 2 may represent an alternate isomer of the Holliday intermediate. This minor species was most obvious with the lox4 substrates (Fig. 6a, lanes 5 and 7) but was also visible with the loxP substrates (Fig. 3a, lanes 5, 6, 7, and  9). The modest reductions in recombination efficiency suggest FIG. 6. lox4 x lox4 Recombination by Cre mutant proteins. a, analysis of the lox4 x lox4 recombination reaction by native PAGE. The reaction was performed as described in Fig. 3 (24). Note that Cre R101A and N111A formed two slow migrating species: the major 1 species co-migrated with the HJ generated by the other Cre proteins, while the minor 2 species may represent an alternate HJ isomer (see "Discussion"). Both Holliday species were also observed for the Cre R101A/N111A double mutant protein (data not shown). The upper panel shows ϳ8-fold longer exposure of the autoradiogram to better reveal the 2 species. (b) Graph of the percent of recombinant products (R1 ϩ R2; solid bars) and 1 intermediate (striped bars) resulting from the lox4 x lox4 recombination assays. c, analysis of the purified species by denaturing PAGE. The intermediates from a preparative lox4 x lox4 recombination reaction (a) were analyzed on a 5% denaturing PAGE (24). The purified intermediates (especially for Cre R101A and N111A) may consist of both the 1 and 2 species, since they migrated so close together. The loxP x loxP recombination reaction mixture (lane 1) was used as size markers (see Fig. 3c). Like wild-type Cre (lane 2), all seven Cre mutant proteins that generated intermediates had predominantly exchanged the top strands of lox4 in the species.
that the individual asymmetric protein-DNA contacts and the Arg-101-Asn-111 interactions are not essential for recombination. Cre R100A may not have shown increased accumulation of HJ because Arg-100 contacts the same phosphate(s) as Arg-101 and Arg-106 (Fig. 1d).
Resolution is believed to be coupled to an isomerization of the HJ in which the continuous and crossing strands in the HJ switch roles (Fig. 9) (9, 34 -36). Isomerization of the HJ would require switching the asymmetric protein-DNA and intersubunit interactions as well. The increased accumulation of the 1 Holliday intermediates and the appearance of the novel 2 species by Cre R101A, R106A, N111A and R121A may arise from a slow isomerization of the Holliday intermediate (Fig. 9) and/or a slight decrease in resolution. We speculate that charge neutralization of the DNA phosphates on the continuous strands by the arginine residues would be ideal mediators of  Fig. 4). b, graph of the results from the resolution of synthetic lox4 ( 4 ) structure. The 87-bp R 1 (solid bars) and 75-bp R 2 (striped bars) products result from resolution on the top and bottom strands of lox4, respectively (28) (also see Fig. 5).
FIG. 8. Resolution of the synthetic structures by the wild-type and K201A Cre. a, the lox spacer sequence in the structures. Note that the rev structure contains the loxP site in the reverse orientation relative to that in the p structure. b, 2 nM synthetic substrate was incubated with 0.25 M of the indicated Cre protein for 1 h at 30°C as described (28) (also see Fig. 5). The 87-bp R 1 and 75-bp R 2 products result from resolution on the top and bottom strands, respectively, of the lox site as illustrated in a (24). The right panel is a ϳ5-fold longer exposure of the autoradiogram showing the resolution products from lanes 9 -12. the isomerization of the Holliday intermediate. The crossspacer protein-protein interactions between Arg-101 and Asn-111 (Fig. 1d) may also contribute to the isomerization. Even though mutations of these residues did not affect the cII bend, they may nevertheless influence the isomeric state of the HJ.
Arg-118 and Lys-122 Are Required for Synapsis-The second group of Cre proteins (R118A and K122A) are defective in synapsis, strand cleavage and the recombination of loxP sites, but they resolve Holliday structures efficiently. These phenotypes are similar to those of the synapsis-defective Cre A36V protein (17,21,22,29), and support the importance of the cyclic Ala-36-Arg-118 intersubunit interactions in synapsis. Arg-118 and Lys-122 also form a cyclic network of protein-DNA interactions (Fig. 1, c and d), which may also contribute to stabilization of the synapse. Despite the defect in synapsis, Cre A36V, R118A, and K122A were proficient in DNA bending, suggesting that synapsis is not essential for DNA bending at least in cII.
The E/F-linker Is Essential for Recombination-The long E/F-linker connecting the NH 2 -terminal and COOH-terminal domains of Cre is situated close to the active site and interacts with the flexible ␤2/␤3-loop in the adjacent Cre subunit (8,12). We found that Glu-129 and Gln-133, located in the E/F-linker are essential for catalysis and Glu-129 is required for synapsis as well. Mutation of Glu-129 to Arg (or Pro, data not shown) impaired both synapsis and catalysis, whereas the conservative E129Q mutation was able partially to suppress the catalytic defect but not the synaptic defect.
In the 3CRX structure, Glu-129 of the cleaving Cre subunit contacts Val-85 in helix D within the same Cre subunit and Lys-201 of the non-cleaving subunit (Fig. 1d) (9). Disruption of these intra-and/or intersubunit interactions may be responsible for the synaptic/catalytic defects. Unlike the other Cre residues studied, Gln-133 does not directly contact the DNA nor does it contact the adjacent Cre subunit. However, it does make a water-mediated contact with the phosphate 5Ј of the inactive scissile phosphate in the high-resolution 4CRX structure (Fig. 1d) (10). Within the same Cre molecule, the side chain of Gln133 is positioned close to helix M (which contains the nucleophilic Tyr-324) (8,9). Gln-133 may function either directly or indirectly in positioning Tyr-324 and/or other catalytic residues in the active site. Other mutations in helix E and the following E/F-linker (namely, V125F and G128D) have also been reported to impair recombination (29). In addition, an insertion of two amino acids (Val-Asp) at position 182 altered the topology of recombination products, implying that synapsis was affected (37). This insertion is close to Asp-184 that interacts with Arg-130 in the E/F-linker of the adjacent Cre subunit.
It is possible that Glu-129, Gln-133, and other residues in the E/F-linker contribute to synapsis and catalysis either directly or indirectly.
We previously proposed that the NH 2 -terminal domain of Cre allosterically masks the COOH-terminal domain and this inhibitory effect is released upon DNA binding (38,39). The NH 2 -terminal domain of XerD also appears to hinder the activity of the COOH-terminal domain (40,41). Subramanya et al. (41) proposed that a large conformational change may accompany DNA binding and/or synapsis. A conformational change was indeed observed in Int upon DNA binding and cleavage (42)(43)(44)(45). We propose that the E/F-linker in Cre is likely flexible and functions as a hinge to uncover the active site upon activation. The mutations in the E/F-linker may interfere with this allosteric regulation. The E/F-linker may also be important for communication between the two Cre domains as well as with other Cre subunits in the synaptic complex.
Lys201 Is Required for Catalysis-Lys-201 is conserved among the tyrosine recombinases and Type IB topoisomerases (6,46). The equivalent lysine residue in the vaccinia topoisomerase is important for catalysis, acting as a general acid to protonate the leaving 5Ј-hydroxyl group during cleavage (46 -48). The corresponding Lys residue is also essential for catalysis in XerD (25,49) and Flp (50). We have confirmed the finding of Van Duyne and co-workers (10) that Lys-201 is important for catalysis in Cre. We also found that mutation of Lys-201 disrupts the formation of the synaptic complex. This supports the involvement of the intersubunit interactions between the ␤2/␤3-loop and helix E (and the E/F-linker) in synapsis (8 -13).
Martin et al. (13) observed that Lys-201 interacts asymmetrically with the scissile nucleotides in the Cre-loxP HJ crystal structure and proposed that Lys-201 may function to distinguish the scissile nucleotides. However, even with the very low level of resolution that we did detect, we found that Cre K201A exhibited a similar strand bias to the wild-type Cre protein in the resolution of loxP Holliday structures. In addition, since Cre K201A induced similar bends to the wild-type Cre protein and we previously found that DNA bending correlates with the site of initiation (19), we believe that Lys-201 likely does not determine the site of strand initiation. Unfortunately, a direct assay of this function is precluded because Cre K201A failed to form any detectable amount of HJ.
analyzing the Cre-lox crystal structures. We are grateful to Greg Van Duyne (University of Pennsylvania) for generously providing the Cre K201A expression plasmid.