The role of the conserved Trp330 in Flp-mediated recombination. Functional and structural analysis.

The active site of Flp contains, in addition to a transdonated nucleophilic tyrosine, five other residues that are highly conserved within the lambda-integrase family of site-specific recombinases and the type IB topoisomerases. We have used site-directed mutagenesis and x-ray crystallography to investigate the roles of two such residues, Lys223 and Trp330. Our findings agree with studies on related enzymes showing the importance of Lys223 in catalysis but demonstrate that in Flp-mediated recombination the primary role of Trp330 is architectural rather than catalytic. Eliminating the hydrogen bonding potential of Trp330 by phenylalanine substitution results in surprisingly small changes in reaction rates, compared with dramatic decreases in the activities of W330A, W330H, and W330Q. The structure of a W330F mutant-DNA complex reveals an active site nearly identical to that of the wild type. The phenylalanine side chain preserves most of the van der Waals interactions Trp330 forms with the Tyr343-containing trans helix, which may be particularly important for the docking of this helix. Our studies of Trp330 provide the first detailed examination of this conserved residue in the lambda-integrase family, suggesting that the relative importance of active site residues may differ among Flp and related enzymes.

The active site of Flp contains, in addition to a transdonated nucleophilic tyrosine, five other residues that are highly conserved within the -integrase family of site-specific recombinases and the type IB topoisomerases. We have used site-directed mutagenesis and x-ray crystallography to investigate the roles of two such residues, Lys 223 and Trp 330 . Our findings agree with studies on related enzymes showing the importance of Lys 223 in catalysis but demonstrate that in Flp-mediated recombination the primary role of Trp 330 is architectural rather than catalytic. Eliminating the hydrogen bonding potential of Trp 330 by phenylalanine substitution results in surprisingly small changes in reaction rates, compared with dramatic decreases in the activities of W330A, W330H, and W330Q. The structure of a W330F mutant-DNA complex reveals an active site nearly identical to that of the wild type. The phenylalanine side chain preserves most of the van der Waals interactions Trp 330 forms with the Tyr 343 -containing trans helix, which may be particularly important for the docking of this helix. Our studies of Trp 330 provide the first detailed examination of this conserved residue in the -integrase family, suggesting that the relative importance of active site residues may differ among Flp and related enzymes.
Flp, the best characterized eukaryotic member of the -integrase or tyrosine-based family of site-specific recombinases, is a product of the 2-m plasmid from Saccharomyces cerevisiae and is responsible for maintaining the copy number of this plasmid in vivo (1). Although many site-specific recombinases require co-factors for activity, Flp-mediated recombination is remarkably simple and requires only the enzyme and target DNA, which makes it a valuable genetic tool. The usefulness of Flp has been enhanced by the isolation of a variant of Flp ("Flpe") that displays significantly more robust activity than WT 1 Flp at 37°C (2). Although Flp shares many mechanistic and structural similarities with other recombinases and with type IB topoisomerases, in other aspects it is unique, such as in the assembly and regulation of its active site. It has thus been unclear to what extent the results of biochemical studies on other members of these families could be extrapolated to the Flp system.
The reaction catalyzed by Flp and other tyrosine recombinases minimally requires four protein monomers and two DNA segments containing specific sequences recognized by the enzyme (Fig. 1a). The reaction proceeds in a stepwise fashion, with only one strand of each duplex cleaved at a time. Thus, only two of the four protomers are active in each step of the reaction. For Flp, the required DNA site, named FRT (Flp recognition target), is made up of two 13-bp Flp-binding sequences inverted around an 8-bp spacer region, with a third nonessential Flp-binding site separated by 1 bp from the rest of the sequence. However, whereas most other enzymes in this family assemble the catalytic center in cis, using residues from one protomer, Flp uses a unique in trans mechanism, with its active site consisting of residues from two protomers (3,4). The ␣-helix (helix M) containing the catalytic tyrosine (Tyr 343 ) is donated to an adjacent protomer and cleaves DNA in conjunction with the other active site residues provided by that protomer. Corresponding to half of the sites activity, only one DNA cleavage reaction takes place within an Flp dimer, two of which come together to form a synaptic complex. Each of the two exposed 5Ј-hydroxyl groups then attacks the phosphotyrosyl linkage in the other Flp dimer of the complex. The resulting Holliday junction intermediate can subsequently isomerize and activate the two protomers whose catalytic tyrosines were dormant during the first half of the reaction. These two tyrosines then engage in a second round of DNA cleavage and ligation reactions, ultimately yielding recombined products. Interestingly, structural and biochemical studies have shown significant differences in the mechanisms by which different recombinases control the activity of their catalytic sites (5)(6)(7)(8)(9).
The DNA relaxation reaction catalyzed by type IB topoisomerases also proceeds through a 3Ј-phosphotyrosyl intermediate, and the catalytic similarities between these two families of enzymes correspond to a conserved core structure that harbors the active site (10). Several structures from both families have been determined, including -integrase (11), Cre (12-15), HP1 integrase (16), XerD (17), and Flp (4) from the site-specific recombinase family and vaccinia topoisomerase (18) and human topoisomerase I (19 -21) from the topoisomerase family. Despite low sequence homology among these proteins, the secondary and tertiary structures of their catalytic domains are strikingly similar, and they probably share a common ancestor (10,22). There are, however, clear differences among these enzymes as well. For example, the topoisomerases act as monomers, whereas the recombinases act as tetramers, and Flp differs from other members of both families in the trans assembly of its active site.
The active sites of both families include, in addition to the nucleophilic tyrosine, five other highly conserved amino acid residues (reviewed in Refs. [23][24][25]. In Flp, these are Arg 191 , Arg 308 , His 305 , Lys 223 , and Trp 330 . The two arginines and the histidine, but not the lysine and the tryptophan, have been extensively studied in this system (26 -30). The arginines presumably position the scissile phosphate and stabilize the build-up of negative charge in the transition state. The histidine (replaced by lysine in topoisomerases) is appropriately positioned in the Flp and Cre complex structures to act as the general base that accepts a proton from the attacking tyrosine, although mutagenesis data suggest that it is not essential (27,30,31). 2 The lysine is essential for catalysis in other recombinases and has been shown to act as a general acid in the cleavage step in the vaccinia topoisomerase reaction, protonating the leaving 5Ј-hydroxyl group with the help of one of the arginines (8,(32)(33)(34)(35). The tryptophan is replaced by histidine in most related enzymes, although tryptophan is found at this position in Cre as well. The DNA complex structures of Cre and human topoisomerase I show that the side chain of this tryptophan/histidine forms a hydrogen bond with a nonbridging oxygen of the scissile phosphate, whereas in Flp-DNA structures, Trp 330 sometimes forms a hydrogen bond with the free 5Ј-OH in the active site. Mutation of the equivalent histidines to alanine in vaccinia topoisomerase (His 265 ) and human topoisomerase I (His 632 ) decreased the reaction rates substantially, whereas the effects of glutamine and asparagine replacements were less drastic (36,37). Experiments on vaccinia virus topoisomerase using phosphorothioate-containing substrates showed that this histidine forms a catalytically important hydrogen bond to one of the nonbridging phosphate oxygens (38). All of these results supported a hypothesis that the key feature of the tryptophan/histidine is its ability to hydrogen-bond with DNA. However, this question had not been addressed experimentally in the tyrosine recombinases.
To investigate the roles that Lys 223 and Trp 330 play in Flpmediated recombination, mutagenesis and activity studies have been performed on these two residues in Flpe. Whereas our results confirmed the importance of Lys 223 , they also provided some surprising insights into the function of Trp 330 , indicating that it might be more important in maintaining a viable active site through its van der Waals interactions with neighboring amino acids rather than stabilizing a reaction intermediate through its hydrogen bonding capability. We confirmed this by determining the structure of the W330F protein complexed with DNA.

EXPERIMENTAL PROCEDURES
Protein Cloning and Mutagenesis-In our experiments, Flpe was used as the WT Flp. A Flpe clone was constructed and kindly provided by Dr. Mike Cox. A C-terminal His 6 tag plus a single glycine linker residue was added using PCR, and the tagged protein was recloned into pET24d vector. Point mutagenesis was performed using the QuikChange TM site-directed kit from Stratagene, with the primers ordered from Operon. The results were confirmed by DNA sequencing.
Protein Expression and Purification-The protein was expressed and purified using a new method recently published (5). Two columns were used: a Talon column that bound the His tag and a DNA affinity column containing Flp-binding sequences. The whole purification process took only 1 day (beginning with frozen cell pellets), yielding an average of 1 mg of protein from 1 liter of cell culture. The protein was stored at Ϫ80°C. It was later dialyzed against 50 mM HEPES (pH 7.1), 1 mM EDTA, 1 mM dithiothreitol, and 15% glycerol and 1 M NaCl overnight and aliquoted before usage. The purification and dialysis were performed at 4°C. DNA Recombination Assays-Plasmid pJFS39 was originally provided by Dr. Michael Cox (39). It was amplified in DH5␣ cells and purified with Qiagen Miniprep kits. 15 pmol of SalI-digested pJFS39 plasmid was labeled with [␥-32 P]ATP and T4 polynucleotide kinase. Qiagen nucleotide removal kits were used to remove the enzyme and nucleotide afterward. The recombination assay was performed at 30°C in a 10-l reaction volume with 2 nM radioactively labeled SalI-cleaved pJFS39, 120 nM Flpe protein, 25 mM Taps (pH 8.0), 200 mM NaCl, 12% glycerol, 2 mM EDTA, and 1 mg/ml bovine serum albumin. The concentration of the DNA was calculated based on the assumption that no DNA was lost during the purification procedure after labeling, and the actual concentration might be slightly lower than 2 nM. The reaction was terminated with 0.4% SDS, and the products were separated on a 0.8% agarose gel. The reaction conditions were similar to those used by Gates et al. (39) with minor modifications. Although polyethylene glycol was used in previous studies to enhance activity, it did not increase the reaction rate in our assays and only induced the appearance of reaction intermediates. It was therefore omitted from the reaction buffer.
DNA Cleavage Assays-Synthentic oligonucleotides were purchased from the Keck biotechnology facility at Yale University and purified by urea polyacrylamide gel electrophoresis. S3 (TAA TCC AGT GGA AGT TCC TAT TCT T) was labeled with [␥-32 P]ATP and purified as described under "DNA Recombination Assays." It was then annealed with S4P (5Ј-phospho-CTA GAA GAA TAG GAA CTT CCA CTG GAT TA) to form a suicide substrate containing one Flp-binding site. The final substrate for the DNA cleavage assays consisted of both radioactively labeled and unlabeled suicide substrates at a ratio of 1:9. Each 10-l reaction contained 4 nM DNA substrate, 120 nM Flpe protein, 25 mM Taps (pH 8.0), 200 mM NaCl, 12% glycerol, 2 mM EDTA, and 1 mg/ml bovine serum albumin. The assay was carried out at 30°C and terminated with 0.4% SDS. The products were separated on a 15% SDS-PAGE gel. Results were quantified using an Amersham Biosciences Phosphor-Imager. The data were processed using Origin software, and the curves were fit to a hyperbola equation to attain the apparent rate constant for a second order reaction.
Crystallization and Structure Determination-The complex of W330F Flpe and T4B4 (T4: TAA GTT CCT ATT CT; B4: TTT AAA AGA ATA GGA ACT TC) was prepared with the same procedure as the original Flp-Holliday junction complex (4). A slightly higher protein concentration, 0.48 mM, was used. Crystals appeared in hanging drops containing 20 mM HEPES (pH 7.0), 11% polyethylene glycol 5000 monomethyl ether, 20% xylitol, 100 mM NaCl, 2 mM dithiothreitol, 10 mM CaCl 2 , approximately 1 week after the trays were set up. Data were collected from crystals with a size of 0.2 ϫ 0.2 ϫ 0.4 mm at the BioCARS 14BM-C beamline at the Advanced Photon Source using an ADSC Q4 CCD detector. The crystals were in space group P21, with unit cell dimensions of a ϭ 80 Å, b ϭ 117 Å, c ϭ 142 Å, and ␤ ϭ 97.3°. The diffraction was anisotropic, with ϽI/IϾ falling below ϳ2 at 2.7, 3.0, and 2.9 Å along the three axes. The data were processed with the HKL suite (40). Initial phases were provided by the WT Flpe structure, which crystallized isomorphously (Protein Data Bank number 1M6X) (5). The structure was refined with CNS_SOLVE (41), using a combination of 2and 4-fold noncrystallographic symmetry restraints. Differences in the refined structures of the WT and mutant were in agreement with difference Fourier maps ((F WT Ϫ F W330F ) ⌽ WT and (F WT Ϫ F W330F ) ⌽ W330F ). In addition to changes in the active site discussed under "Results," the DNA strand in the catalytic site of the inactive monomer D could be modeled better as nicked rather than intact. Structure figures were generated with Ribbons (42). Buried surfaces were calculated using CNS_SOLVE with the side chain atoms of the His/Trp residue as molecule 1 and the surrounding protein as molecule 2. The result was the total buried surface area from molecule 1 and molecule 2, and should be divided by 2 when considering only one molecule in the interface.

RESULTS
Flpe Purification-One difficulty in studying Flp recombinase is the usually cumbersome protein purification procedure, which is further complicated by the protein's intrinsic instability (43,44). We were able to improve the efficiency of this process with the use of two affinity columns: Talon column for a newly added C-terminal His 6 tag and DNA affinity column containing multimers of Flp-binding sites. The addition of the His 6 tag was based on the observation that the C terminus of the enzyme is exposed and far from the active site and DNAbinding regions in the Flp-DNA complex structure. The DNA column was similar to those in previous studies except that the DNA molecules consisting of multiple Flp-binding sites were prepared with a template-free PCR instead of ligation. Purified protein was obtained within 1 day. Flpe is used in our studies, and the wild type in this paper refers to His 6 -tagged Flpe without further mutations. The four point mutations (P2S, L33S, Y108N, and S294P) found in Flpe are removed from the catalytic center and do not appear to affect the static structure of the enzyme-DNA complex (5).
Trp 330 and Lys 223 Are Important for Activity-Single mutations were introduced at Trp 330 and Lys 223 to study the roles of these two residues in Flp-mediated recombination. Trp 330 was substituted with alanine, histidine, glutamine, and phenylala-nine, whereas Lys 223 was replaced by alanine and arginine. The effects of these mutations were first tested in full recombination assays. The linearized plasmid substrate for the assay, pJFS39, contains a single FRT site with a symmetrical spacer. It can be recombined in two different orientations, one of which regenerates the substrate, whereas the other produces DNA molecules of different lengths (Fig. 1b).
Although recombination catalyzed by the wild type protein reached equilibrium within several hours under our conditions, overnight reactions were carried out for the mutants to allow detection of low activity levels (Fig. 1c). Protein and DNA concentrations were also raised in one experiment to further increase the sensitivity (Fig. 1d). W330A, W330Q, and K223A failed to generate any detectable amount of recombined products. W330H was capable of mediating recombination at a much slower rate, whereas K223R showed an extremely low activity only in the assay where the protein and DNA concentrations were increased. In contrast to the poor performances of the other mutants, W330F displayed an activity level indistinguishable from the wild type in these experiments, a surprising observation in light of the previous hypotheses that the hydrogen bond between the Trp 330 side chain and the DNA would be important for catalysis.
The fact that mutating Lys 223 to alanine abolished activity whereas K223R restored some enzymatic capability is consistent with previous findings that this lysine acts as a general acid in vaccinia topoisomerase. In Flp, the side chain of Lys 223 is also involved in direct contact with a base in the minor groove and therefore may be important for DNA binding. K223A and K223R showed decreased affinities for the DNA column during protein purification (data not shown), suggesting that part of the activity loss might result from impaired DNA binding capabilities.
The long incubation time for the overnight reactions with WT and W330F Flp led to the appearance of some doubly cleaved side products. These double strand cleavage products, like the reaction intermediates, have been observed by others and probably result from slow hydrolysis of the covalent phosphotyrosine intermediate (39,45).
Comparison of Wild Type and W330F in Recombination Assays-A more detailed comparison was carried out between the activities of the wild type and the W330F mutant by measuring the formation of recombined products at time points before equilibrium (Fig. 2). No detectable difference was observed between the reactions catalyzed by the two proteins, suggesting again that the role of Trp 330 in Flp-mediated recombination might be different from previous hypotheses. However, the full recombination reaction requires many steps, and the rate-limiting one may actually be a conformational change in the Holliday junction (46). This could mask a small decrease in the mutant's catalytic activity. A second group of experiments, DNA cleavage assays, was used to determine whether there is any defect in the ability of W330F to catalyze the chemical reactions.
Comparison of Wild Type and W330F in Cleavage Assays-DNA cleavage is the first chemical step in the reactions mediated by this family of enzymes. Our assay used suicide substrates shown in Fig. 3a. The attack of Tyr 343 between the Flp-binding site and the spacer region results in a covalent protein-DNA linkage and the excision of two thymines at the end of the DNA top strand, which diffuse into the bulk solution.

FIG. 2. Time course of recombination assays.
Top panels, reactions using WT Flpe (a) and W330F Flpe (b) were terminated with SDS at the time points shown, and the products were separated by agarose gel electrophoresis. Conditions were the same as in Fig. 1c. Bottom panels, The percentage of total DNA converted to product amino acid plotted versus time.
The reaction is trapped at this stage due to the absence of the 5Ј-hydroxyl group on the first thymine, which is necessary for the reverse reaction or the subsequent ligation reaction of the recombination process. Because of half of the sites activity, a theoretical 50% of the substrate duplex will be covalently linked to the enzyme when the protein is present in excess. A modest decrease in reaction rate was observed for the W330F mutant, compared with that of the wild type (Fig. 3). The appearance of the covalent protein-DNA product fit better to a second order reaction curve than to a first order one, consistent with the requirement for two DNA-bound Flp protomers to form a single active site. The apparent rate constant for the wild type was estimated to be 1.0 ϫ 10 8 M Ϫ1 min Ϫ1 , whereas that for the mutant was roughly 5-fold lower, 0.18 ϫ 10 8 M Ϫ1 min Ϫ1 . This decrease probably reflects a slight defect in catalysis rather than in DNA binding, because these assays were carried out with a large excess of protein, above the estimated K d (47), and because our preliminary mobility shift assays using a single Flp-binding site indicated no difference in affinity between the two proteins (data not shown).
Radioactive and nonradioactive cleavage assays were also performed on the other mutants. The results were consistent with those of recombination assays, except that W330Q was able to yield covalent protein-DNA adduct at a rate that is slower than the wild type but slightly faster than W330H (Fig.  3b), in contrast to the apparently higher activity level of W330H in the overall recombination assay. This suggests that W330Q is even more defective in the later stages of recombination than in the cleavage step.
When compared with the dramatic activity decrease in W330H and W330Q, the small reaction rate difference between the wild type and W330F mutant suggests that hydrogenbonding capability is not the critical feature of Trp 330 . Examination of the previous Flp-DNA complex structure revealed an extensive network of amino acids contacted by Trp 330 through van der Waals interactions within or close to the catalytic site. The most essential role of Trp 330 in the reaction may be maintaining a stable local architecture. To investigate whether the high activity level of the mutant stemmed from the ability of phenylalanine to retain those van der Waals contacts, W330F mutant protein was crystallized with DNA, and a structure of W330F-Holliday junction complex was determined.
Structure of W330F Mutant Protein Complexed with DNA-W330F Flpe was cocrystallized with the same DNA substrate as used in the WT Flp and Flpe complex structures (4,5). Phases were provided by a newly determined structure of a wild type Flpe-DNA complex (5). The overall structure of the mutant is the same as that of the wild type Flpe. The complex has approximate 2-fold symmetry and consists of a mutant tetramer bound to a DNA Holliday junction (Fig. 4a). Two catalytic centers contain a covalent phosphotyrosyl linkage and a free 5Ј-OH, whereas the other two contain a continuous or nicked DNA strand with Tyr 343 displaced from the scissile phosphate. The mutant structure was refined to R and R free values of 23.6 and 27.5%, respectively (Table I). 87.6% of the nonglycine protein residues are in the most favored conformational regions of the Ramachandran plot (as defined by Procheck (48)). One of the two active protomers ("chain B," with Tyr 343 donated by chain C) was significantly better ordered than the other. Discussions of the active site geometry thus pertain to this protomer. The slightly higher resolution of the W330F Flpe structure, compared with WT Flpe, allows us to model the C-terminal His 6 tag. The His 6 tag of protein chain B is involved in crystal packing. It forms a short ␣-helix, with two histidine side chains stacked against the bases at the end of a DNA arm from the neighboring Holliday junction. It is interesting to note that in some other structures where ordered His 6 tags are seen, they are in extended conformations (Protein Data Bank numbers 1GKK and 1LW6) (49,50). Thus, His 6 is clearly a sequence that can adopt different secondary structures in different environments.
Other than a small shift in the position of the two inactive protomers, which is probably due to minor differences in crystal packing, the only significant differences between the mutant and WT Flpe structures are at Trp 330 and in its immediate vicinity. A prominent peak at the position of the Trp 330 side chain was observed in a difference map comparing the data of the wild type and the mutant, corresponding to the alteration of tryptophan to phenylalanine and a small change in the 1 dihedral angle. Other peaks correspond to a shift in the proximal residues of the trans helix, including Tyr 343 , and the scissile phosphate to which it is covalently attached. There are only rather subtle changes that are difficult to interpret at this resolution in the other active site side chains (Fig. 5). The 5Ј-hydroxyl group formed a hydrogen bond with the side chain of Trp 330 in the Flp structure and was only 3.6 Å away from the tryptophan N⑀ atom in the WT Flpe model. This hydrogen bond is necessarily broken in the W330F structure. The 5Ј-hydroxyl is not well defined in the electron density, but it is possible that it could form an alternative hydrogen bonding interaction with Arg 191 .
The phenylalanine side chain maintained most of the van der Waals interactions between Trp 330 and neighboring residues. Trp 330 interacts with several amino acids residing in its own protomer and on the trans-helix donated from the adjacent one, burying a total of 383 Å 2 in the active site containing the phosphotyrosyl linkage. Phe 330 still contacts the majority of these residues, with a buried area of 340 Å 2 , a slight decrease due to the smaller size of the phenylalanine side chain. Of the amino acids contacted by residue 330, Arg 308 and Tyr 343 are directly involved in catalysis, whereas the others are near the active site. Three (including Tyr 343 ) lie on the trans helix, suggesting that Trp 330 helps position this helix and Tyr 343 for DNA cleavage. The loss of activity in the W330A, W330H, and W330Q mutants may reflect the loss of some or most of these stabilizing hydrophobic interactions.
Docking Helix M-The van der Waals interactions of Trp 330 may be particularly important in Flp because of its trans cleavage mechanism. The correct docking of the Tyr 343 -bearing helix appears to be crucial to Flp-mediated recombination and is clearly regulated by the protein to achieve half of the sites activity. To further investigate the docking of helix M, Asn 325 , whose side chain forms two hydrogen bonds to the trans polypeptide segment (Fig. 6), was mutated to alanine. The N325A mutation caused a significant decrease in both the overall recombination rate and the initial DNA cleavage rate (data not shown).
We also mutated two residues (Thr 224 and Ser 331 ) that lie FIG. 4. W330F-DNA structure. a, the overall structure of W330F mutant-Holliday junction complex. The complex consists of a protein tetramer and a DNA Holliday junction, held in a nearly square planar conformation. The purple and red protomers harbor active catalytic centers, with tyrosines donated by the green and blue protomers, respectively. b, simulated annealing omit map of the active site. The electron density is contoured at 2.5 around residues that were omitted from the structure before removal of phase bias by simulated annealing. Bonds are colored according to the scheme in a. near the conserved pentad and contact the incoming DNA strand near the 5Ј-hydroxyl to see if they were also important in positioning catalytic groups (Fig. 6). However, the activity of the T224A and S331A mutants in recombination assays was not significantly different from that of WT Flpe (data not shown).

DISCUSSION
The biochemical and structural evidence presented here demonstrates that the most important feature of the conserved Trp 330 in Flp-mediated recombination is its ability to form extensive van der Waals interactions rather than its ability to form a hydrogen bond with the DNA. It may be that the hydrogen bond donated by this residue in WT Flp contributes relatively little to the rate-limiting step of catalysis. This is in contrast to the vaccinia virus topoisomerase case, where the equivalent histidine was clearly implicated in forming a hydrogen bond to the nonbridging oxygens of the scissile phosphate during the transition state of the cleavage reaction (38). Structural comparisons show small but intriguing differences in the position of this residue: Trp 315 in Cre and the corresponding histidine (His 632 ) in human topoisomerase I are both hydrogenbonded to a nonbridging oxygen atom on the scissile phosphate, whereas in Flp complexes the N⑀ atom of Trp 330 is a bit farther away from the scissile phosphate (4, 5, 12-14, 19 -21). In the original Flp-DNA complex, Trp 330 forms a hydrogen bond with the free 5Ј-hydroxyl group instead, whereas in the Flpe-DNA structure it is roughly equidistant between the 5Ј-hydroxyl and a nonbridging phosphate oxygen (3.6 versus 3.8 Å). We cannot exclude the possibility that the loss of a hydrogen bond to the 5Ј-hydroxyl in the W330F complex is compensated by formation of an alternate interaction. In the new three-dimensional model of the mutant protein-DNA complex, a small swivel movement of the poorly ordered 5Ј-hydroxyl group could bring it within hydrogen bonding distance of Arg 191 . The slight activity drop in the W330F mutant could therefore be due to the slight deficiency of a smaller phenylalanine side chain in maintaining all of the original van der Waals interactions or some minor defect resulting from the loss or switch of the hydrogen bond acceptor or both. It is not known whether the histidine or glutamine side chain of W330H and W330Q maintained a hydrogen bond with the 5Ј-hydroxyl group, although model building indicated that they could.
The van der Waals interactions involving Trp 330 may be particularly important in the Flp case because of this enzyme's unique trans cleavage mechanism. The importance of the correct docking of the tyrosine-bearing helix M is further illustrated by the experiments on N325A, which is also involved in intermolecular contacts with the trans polypeptide segment. Previous mutational studies also highlighted residues involved in the packing of this helix into the active site cavity of the neighboring protomer. Trp 330 packs against one face of this helix, contacting Ser 336 and Ala 339 in addition to the catalytic Tyr 343 . Mutations at Ser 336 and Ala 339 , which are strictly conserved in Flp-like tyrosine recombinases from other yeast species, caused defects in DNA cleavage and the formation of dimeric complexes (27, 51) (shown in Fig. 6). The roles of His 309 and His 345 , also involved in contacts between helix M and its "host" protomer, have been extensively investigated as well (52). Together with our experiments on Trp 330 and Asn 325 , these results show that the correct positioning of Tyr 343 , which is essential to the activity of Flp, involves multiple residues inside or close to the active site.
These intermolecular contacts must be strong enough to properly position helix M in the two active catalytic centers yet weak enough to be reversible, since they are not seen in the two inactive catalytic centers. In fact, the half of the sites activity of Flp is thought to be regulated by the disruption of these contacts and displacement of Tyr 343 due to the increased distance between the two contributing protomers in the inactive catalytic centers (5). In Cre, where the active site is assembled in cis, Trp 315 and its surrounding residues bury nearly the same amount of interface area in both active site states (ϳ385 Å 2 ), despite changes in the protein backbone and some side chain conformations. It would thus be interesting to know whether or not a W315F mutation in Cre behaves similarly to the W330F mutation in Flp.
Some variability was seen within the topoisomerase family as well, where the histidine that is equivalent to Trp 330 of Flp has been studied extensively. In vaccinia virus topoisomerase, H265A, H265Q, and H265N mutations reduced the activity by 100-, 4-, and 3-fold, respectively. In human topoisomerase I, the mutations H632A, H632Q, H632N, and H632W resulted in decreases of 6000-, 100-, 180-, 12,000-fold in the cleavage reaction rate. The different extents to which the two alanine mutations affected the enzymatic activity suggest that the importance of this histidine's role may vary. For Flp, it is difficult to accurately quantify extremely low activity, because it requires relatively long reaction times. This is complicated by the protein's instability and side reactions such as hydrolysis of the phosphotyrosyl bond. Judging from the sensitivity of our recombination assay, the W330A mutation slowed down the appearance of product by at least 2500-fold, a severe reduction reminiscent of H632A in human topoisomerase I. The topoisomerase histidine buries 294 Å 2 of surface area in the protein-DNA complex (Protein Data Bank number 1EJ9) (21) and thus presumably contributes to the stability of the local active site architecture as well. However, this residue has not been substituted with a hydrophobic residue with comparable side chain size to test the relative importance of its hydrogen bonding and steric roles.
The studies on His 265 and His 632 in the two topoisomerases, respectively, have also ruled out the possibility that this histidine could act as a general acid in the reaction. Instead, during the cleavage reaction, the leaving 5Ј-OH group is protonated by the highly conserved lysine, as shown by experiments with vaccinia virus topoisomerase (33). Our results on the corresponding lysine, Lys 223 , are consistent with the previous findings. K223A showed no detectable activity, whereas K223R was able to produce a minimal amount of products in the recombination assays. The equivalent lysine is essential for catalysis in Cre, XerD, and -integrase as well (8,34,35).
We suggest that the conservation of the histidine/tryptophan residue in the active sites of tyrosine recombinases and type IB topoisomerases reflects two requirements: the need for a hydrogen bond donor and the need for a relatively large hydrophobic group that can help maintain a stable local architecture through van der Waals interactions. The relative importance of these two functionalities may vary greatly in different systems, however, and it appears that in Flp the ability to donate a hydrogen bond is nearly dispensable, whereas the ability to form extensive van der Waals interactions is crucial. This may reflect the unique trans assembly of the Flp active site.